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);
759 * This structure holds an XPS map which can be of variable length. The
760 * map is an array of queues.
764 unsigned int alloc_len;
768 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
769 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
770 - sizeof(struct xps_map)) / sizeof(u16))
773 * This structure holds all XPS maps for device. Maps are indexed by CPU.
775 struct xps_dev_maps {
777 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
780 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
781 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
783 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
784 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
786 #endif /* CONFIG_XPS */
788 #define TC_MAX_QUEUE 16
789 #define TC_BITMASK 15
790 /* HW offloaded queuing disciplines txq count and offset maps */
791 struct netdev_tc_txq {
796 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
798 * This structure is to hold information about the device
799 * configured to run FCoE protocol stack.
801 struct netdev_fcoe_hbainfo {
802 char manufacturer[64];
803 char serial_number[64];
804 char hardware_version[64];
805 char driver_version[64];
806 char optionrom_version[64];
807 char firmware_version[64];
809 char model_description[256];
813 #define MAX_PHYS_ITEM_ID_LEN 32
815 /* This structure holds a unique identifier to identify some
816 * physical item (port for example) used by a netdevice.
818 struct netdev_phys_item_id {
819 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
820 unsigned char id_len;
823 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
824 struct netdev_phys_item_id *b)
826 return a->id_len == b->id_len &&
827 memcmp(a->id, b->id, a->id_len) == 0;
830 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
832 struct net_device *sb_dev);
835 TC_SETUP_QDISC_MQPRIO,
838 TC_SETUP_CLSMATCHALL,
848 TC_SETUP_QDISC_TAPRIO,
856 /* These structures hold the attributes of bpf state that are being passed
857 * to the netdevice through the bpf op.
859 enum bpf_netdev_command {
860 /* Set or clear a bpf program used in the earliest stages of packet
861 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
862 * is responsible for calling bpf_prog_put on any old progs that are
863 * stored. In case of error, the callee need not release the new prog
864 * reference, but on success it takes ownership and must bpf_prog_put
865 * when it is no longer used.
869 /* BPF program for offload callbacks, invoked at program load time. */
870 BPF_OFFLOAD_MAP_ALLOC,
871 BPF_OFFLOAD_MAP_FREE,
875 struct bpf_prog_offload_ops;
876 struct netlink_ext_ack;
878 struct xdp_dev_bulk_queue;
888 struct bpf_xdp_entity {
889 struct bpf_prog *prog;
890 struct bpf_xdp_link *link;
894 enum bpf_netdev_command command;
899 struct bpf_prog *prog;
900 struct netlink_ext_ack *extack;
902 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
904 struct bpf_offloaded_map *offmap;
906 /* XDP_SETUP_XSK_POOL */
908 struct xsk_buff_pool *pool;
914 /* Flags for ndo_xsk_wakeup. */
915 #define XDP_WAKEUP_RX (1 << 0)
916 #define XDP_WAKEUP_TX (1 << 1)
918 #ifdef CONFIG_XFRM_OFFLOAD
920 int (*xdo_dev_state_add) (struct xfrm_state *x);
921 void (*xdo_dev_state_delete) (struct xfrm_state *x);
922 void (*xdo_dev_state_free) (struct xfrm_state *x);
923 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
924 struct xfrm_state *x);
925 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
930 struct rcu_head rcuhead;
937 struct netdev_name_node {
938 struct hlist_node hlist;
939 struct list_head list;
940 struct net_device *dev;
944 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
945 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
947 struct netdev_net_notifier {
948 struct list_head list;
949 struct notifier_block *nb;
953 * This structure defines the management hooks for network devices.
954 * The following hooks can be defined; unless noted otherwise, they are
955 * optional and can be filled with a null pointer.
957 * int (*ndo_init)(struct net_device *dev);
958 * This function is called once when a network device is registered.
959 * The network device can use this for any late stage initialization
960 * or semantic validation. It can fail with an error code which will
961 * be propagated back to register_netdev.
963 * void (*ndo_uninit)(struct net_device *dev);
964 * This function is called when device is unregistered or when registration
965 * fails. It is not called if init fails.
967 * int (*ndo_open)(struct net_device *dev);
968 * This function is called when a network device transitions to the up
971 * int (*ndo_stop)(struct net_device *dev);
972 * This function is called when a network device transitions to the down
975 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
976 * struct net_device *dev);
977 * Called when a packet needs to be transmitted.
978 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
979 * the queue before that can happen; it's for obsolete devices and weird
980 * corner cases, but the stack really does a non-trivial amount
981 * of useless work if you return NETDEV_TX_BUSY.
982 * Required; cannot be NULL.
984 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
985 * struct net_device *dev
986 * netdev_features_t features);
987 * Called by core transmit path to determine if device is capable of
988 * performing offload operations on a given packet. This is to give
989 * the device an opportunity to implement any restrictions that cannot
990 * be otherwise expressed by feature flags. The check is called with
991 * the set of features that the stack has calculated and it returns
992 * those the driver believes to be appropriate.
994 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
995 * struct net_device *sb_dev);
996 * Called to decide which queue to use when device supports multiple
999 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1000 * This function is called to allow device receiver to make
1001 * changes to configuration when multicast or promiscuous is enabled.
1003 * void (*ndo_set_rx_mode)(struct net_device *dev);
1004 * This function is called device changes address list filtering.
1005 * If driver handles unicast address filtering, it should set
1006 * IFF_UNICAST_FLT in its priv_flags.
1008 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1009 * This function is called when the Media Access Control address
1010 * needs to be changed. If this interface is not defined, the
1011 * MAC address can not be changed.
1013 * int (*ndo_validate_addr)(struct net_device *dev);
1014 * Test if Media Access Control address is valid for the device.
1016 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1017 * Called when a user requests an ioctl which can't be handled by
1018 * the generic interface code. If not defined ioctls return
1019 * not supported error code.
1021 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1022 * Used to set network devices bus interface parameters. This interface
1023 * is retained for legacy reasons; new devices should use the bus
1024 * interface (PCI) for low level management.
1026 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1027 * Called when a user wants to change the Maximum Transfer Unit
1030 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1031 * Callback used when the transmitter has not made any progress
1032 * for dev->watchdog ticks.
1034 * void (*ndo_get_stats64)(struct net_device *dev,
1035 * struct rtnl_link_stats64 *storage);
1036 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1037 * Called when a user wants to get the network device usage
1038 * statistics. Drivers must do one of the following:
1039 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1040 * rtnl_link_stats64 structure passed by the caller.
1041 * 2. Define @ndo_get_stats to update a net_device_stats structure
1042 * (which should normally be dev->stats) and return a pointer to
1043 * it. The structure may be changed asynchronously only if each
1044 * field is written atomically.
1045 * 3. Update dev->stats asynchronously and atomically, and define
1046 * neither operation.
1048 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1049 * Return true if this device supports offload stats of this attr_id.
1051 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1053 * Get statistics for offload operations by attr_id. Write it into the
1054 * attr_data pointer.
1056 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1057 * If device supports VLAN filtering this function is called when a
1058 * VLAN id is registered.
1060 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1061 * If device supports VLAN filtering this function is called when a
1062 * VLAN id is unregistered.
1064 * void (*ndo_poll_controller)(struct net_device *dev);
1066 * SR-IOV management functions.
1067 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1068 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1069 * u8 qos, __be16 proto);
1070 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1072 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1073 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1074 * int (*ndo_get_vf_config)(struct net_device *dev,
1075 * int vf, struct ifla_vf_info *ivf);
1076 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1077 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1078 * struct nlattr *port[]);
1080 * Enable or disable the VF ability to query its RSS Redirection Table and
1081 * Hash Key. This is needed since on some devices VF share this information
1082 * with PF and querying it may introduce a theoretical security risk.
1083 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1084 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1085 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1087 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1088 * This is always called from the stack with the rtnl lock held and netif
1089 * tx queues stopped. This allows the netdevice to perform queue
1090 * management safely.
1092 * Fiber Channel over Ethernet (FCoE) offload functions.
1093 * int (*ndo_fcoe_enable)(struct net_device *dev);
1094 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1095 * so the underlying device can perform whatever needed configuration or
1096 * initialization to support acceleration of FCoE traffic.
1098 * int (*ndo_fcoe_disable)(struct net_device *dev);
1099 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1100 * so the underlying device can perform whatever needed clean-ups to
1101 * stop supporting acceleration of FCoE traffic.
1103 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1104 * struct scatterlist *sgl, unsigned int sgc);
1105 * Called when the FCoE Initiator wants to initialize an I/O that
1106 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1107 * perform necessary setup and returns 1 to indicate the device is set up
1108 * successfully to perform DDP on this I/O, otherwise this returns 0.
1110 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1111 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1112 * indicated by the FC exchange id 'xid', so the underlying device can
1113 * clean up and reuse resources for later DDP requests.
1115 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1116 * struct scatterlist *sgl, unsigned int sgc);
1117 * Called when the FCoE Target wants to initialize an I/O that
1118 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1119 * perform necessary setup and returns 1 to indicate the device is set up
1120 * successfully to perform DDP on this I/O, otherwise this returns 0.
1122 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1123 * struct netdev_fcoe_hbainfo *hbainfo);
1124 * Called when the FCoE Protocol stack wants information on the underlying
1125 * device. This information is utilized by the FCoE protocol stack to
1126 * register attributes with Fiber Channel management service as per the
1127 * FC-GS Fabric Device Management Information(FDMI) specification.
1129 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1130 * Called when the underlying device wants to override default World Wide
1131 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1132 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1133 * protocol stack to use.
1136 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1137 * u16 rxq_index, u32 flow_id);
1138 * Set hardware filter for RFS. rxq_index is the target queue index;
1139 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1140 * Return the filter ID on success, or a negative error code.
1142 * Slave management functions (for bridge, bonding, etc).
1143 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1144 * Called to make another netdev an underling.
1146 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1147 * Called to release previously enslaved netdev.
1149 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1150 * struct sk_buff *skb,
1152 * Get the xmit slave of master device. If all_slaves is true, function
1153 * assume all the slaves can transmit.
1155 * Feature/offload setting functions.
1156 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1157 * netdev_features_t features);
1158 * Adjusts the requested feature flags according to device-specific
1159 * constraints, and returns the resulting flags. Must not modify
1162 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1163 * Called to update device configuration to new features. Passed
1164 * feature set might be less than what was returned by ndo_fix_features()).
1165 * Must return >0 or -errno if it changed dev->features itself.
1167 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1168 * struct net_device *dev,
1169 * const unsigned char *addr, u16 vid, u16 flags,
1170 * struct netlink_ext_ack *extack);
1171 * Adds an FDB entry to dev for addr.
1172 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1173 * struct net_device *dev,
1174 * const unsigned char *addr, u16 vid)
1175 * Deletes the FDB entry from dev coresponding to addr.
1176 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1177 * struct net_device *dev, struct net_device *filter_dev,
1179 * Used to add FDB entries to dump requests. Implementers should add
1180 * entries to skb and update idx with the number of entries.
1182 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1183 * u16 flags, struct netlink_ext_ack *extack)
1184 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1185 * struct net_device *dev, u32 filter_mask,
1187 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1190 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1191 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1192 * which do not represent real hardware may define this to allow their
1193 * userspace components to manage their virtual carrier state. Devices
1194 * that determine carrier state from physical hardware properties (eg
1195 * network cables) or protocol-dependent mechanisms (eg
1196 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1198 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1199 * struct netdev_phys_item_id *ppid);
1200 * Called to get ID of physical port of this device. If driver does
1201 * not implement this, it is assumed that the hw is not able to have
1202 * multiple net devices on single physical port.
1204 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1205 * struct netdev_phys_item_id *ppid)
1206 * Called to get the parent ID of the physical port of this device.
1208 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1209 * struct net_device *dev)
1210 * Called by upper layer devices to accelerate switching or other
1211 * station functionality into hardware. 'pdev is the lowerdev
1212 * to use for the offload and 'dev' is the net device that will
1213 * back the offload. Returns a pointer to the private structure
1214 * the upper layer will maintain.
1215 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1216 * Called by upper layer device to delete the station created
1217 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1218 * the station and priv is the structure returned by the add
1220 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1221 * int queue_index, u32 maxrate);
1222 * Called when a user wants to set a max-rate limitation of specific
1224 * int (*ndo_get_iflink)(const struct net_device *dev);
1225 * Called to get the iflink value of this device.
1226 * void (*ndo_change_proto_down)(struct net_device *dev,
1228 * This function is used to pass protocol port error state information
1229 * to the switch driver. The switch driver can react to the proto_down
1230 * by doing a phys down on the associated switch port.
1231 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1232 * This function is used to get egress tunnel information for given skb.
1233 * This is useful for retrieving outer tunnel header parameters while
1235 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1236 * This function is used to specify the headroom that the skb must
1237 * consider when allocation skb during packet reception. Setting
1238 * appropriate rx headroom value allows avoiding skb head copy on
1239 * forward. Setting a negative value resets the rx headroom to the
1241 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1242 * This function is used to set or query state related to XDP on the
1243 * netdevice and manage BPF offload. See definition of
1244 * enum bpf_netdev_command for details.
1245 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1247 * This function is used to submit @n XDP packets for transmit on a
1248 * netdevice. Returns number of frames successfully transmitted, frames
1249 * that got dropped are freed/returned via xdp_return_frame().
1250 * Returns negative number, means general error invoking ndo, meaning
1251 * no frames were xmit'ed and core-caller will free all frames.
1252 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1253 * This function is used to wake up the softirq, ksoftirqd or kthread
1254 * responsible for sending and/or receiving packets on a specific
1255 * queue id bound to an AF_XDP socket. The flags field specifies if
1256 * only RX, only Tx, or both should be woken up using the flags
1257 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1258 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1259 * Get devlink port instance associated with a given netdev.
1260 * Called with a reference on the netdevice and devlink locks only,
1261 * rtnl_lock is not held.
1262 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1264 * Add, change, delete or get information on an IPv4 tunnel.
1265 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1266 * If a device is paired with a peer device, return the peer instance.
1267 * The caller must be under RCU read context.
1269 struct net_device_ops {
1270 int (*ndo_init)(struct net_device *dev);
1271 void (*ndo_uninit)(struct net_device *dev);
1272 int (*ndo_open)(struct net_device *dev);
1273 int (*ndo_stop)(struct net_device *dev);
1274 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1275 struct net_device *dev);
1276 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1277 struct net_device *dev,
1278 netdev_features_t features);
1279 u16 (*ndo_select_queue)(struct net_device *dev,
1280 struct sk_buff *skb,
1281 struct net_device *sb_dev);
1282 void (*ndo_change_rx_flags)(struct net_device *dev,
1284 void (*ndo_set_rx_mode)(struct net_device *dev);
1285 int (*ndo_set_mac_address)(struct net_device *dev,
1287 int (*ndo_validate_addr)(struct net_device *dev);
1288 int (*ndo_do_ioctl)(struct net_device *dev,
1289 struct ifreq *ifr, int cmd);
1290 int (*ndo_set_config)(struct net_device *dev,
1292 int (*ndo_change_mtu)(struct net_device *dev,
1294 int (*ndo_neigh_setup)(struct net_device *dev,
1295 struct neigh_parms *);
1296 void (*ndo_tx_timeout) (struct net_device *dev,
1297 unsigned int txqueue);
1299 void (*ndo_get_stats64)(struct net_device *dev,
1300 struct rtnl_link_stats64 *storage);
1301 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1302 int (*ndo_get_offload_stats)(int attr_id,
1303 const struct net_device *dev,
1305 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1307 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1308 __be16 proto, u16 vid);
1309 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1310 __be16 proto, u16 vid);
1311 #ifdef CONFIG_NET_POLL_CONTROLLER
1312 void (*ndo_poll_controller)(struct net_device *dev);
1313 int (*ndo_netpoll_setup)(struct net_device *dev,
1314 struct netpoll_info *info);
1315 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1317 int (*ndo_set_vf_mac)(struct net_device *dev,
1318 int queue, u8 *mac);
1319 int (*ndo_set_vf_vlan)(struct net_device *dev,
1320 int queue, u16 vlan,
1321 u8 qos, __be16 proto);
1322 int (*ndo_set_vf_rate)(struct net_device *dev,
1323 int vf, int min_tx_rate,
1325 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1326 int vf, bool setting);
1327 int (*ndo_set_vf_trust)(struct net_device *dev,
1328 int vf, bool setting);
1329 int (*ndo_get_vf_config)(struct net_device *dev,
1331 struct ifla_vf_info *ivf);
1332 int (*ndo_set_vf_link_state)(struct net_device *dev,
1333 int vf, int link_state);
1334 int (*ndo_get_vf_stats)(struct net_device *dev,
1336 struct ifla_vf_stats
1338 int (*ndo_set_vf_port)(struct net_device *dev,
1340 struct nlattr *port[]);
1341 int (*ndo_get_vf_port)(struct net_device *dev,
1342 int vf, struct sk_buff *skb);
1343 int (*ndo_get_vf_guid)(struct net_device *dev,
1345 struct ifla_vf_guid *node_guid,
1346 struct ifla_vf_guid *port_guid);
1347 int (*ndo_set_vf_guid)(struct net_device *dev,
1350 int (*ndo_set_vf_rss_query_en)(
1351 struct net_device *dev,
1352 int vf, bool setting);
1353 int (*ndo_setup_tc)(struct net_device *dev,
1354 enum tc_setup_type type,
1356 #if IS_ENABLED(CONFIG_FCOE)
1357 int (*ndo_fcoe_enable)(struct net_device *dev);
1358 int (*ndo_fcoe_disable)(struct net_device *dev);
1359 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1361 struct scatterlist *sgl,
1363 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1365 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1367 struct scatterlist *sgl,
1369 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1370 struct netdev_fcoe_hbainfo *hbainfo);
1373 #if IS_ENABLED(CONFIG_LIBFCOE)
1374 #define NETDEV_FCOE_WWNN 0
1375 #define NETDEV_FCOE_WWPN 1
1376 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1377 u64 *wwn, int type);
1380 #ifdef CONFIG_RFS_ACCEL
1381 int (*ndo_rx_flow_steer)(struct net_device *dev,
1382 const struct sk_buff *skb,
1386 int (*ndo_add_slave)(struct net_device *dev,
1387 struct net_device *slave_dev,
1388 struct netlink_ext_ack *extack);
1389 int (*ndo_del_slave)(struct net_device *dev,
1390 struct net_device *slave_dev);
1391 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1392 struct sk_buff *skb,
1394 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1396 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1397 netdev_features_t features);
1398 int (*ndo_set_features)(struct net_device *dev,
1399 netdev_features_t features);
1400 int (*ndo_neigh_construct)(struct net_device *dev,
1401 struct neighbour *n);
1402 void (*ndo_neigh_destroy)(struct net_device *dev,
1403 struct neighbour *n);
1405 int (*ndo_fdb_add)(struct ndmsg *ndm,
1406 struct nlattr *tb[],
1407 struct net_device *dev,
1408 const unsigned char *addr,
1411 struct netlink_ext_ack *extack);
1412 int (*ndo_fdb_del)(struct ndmsg *ndm,
1413 struct nlattr *tb[],
1414 struct net_device *dev,
1415 const unsigned char *addr,
1417 int (*ndo_fdb_dump)(struct sk_buff *skb,
1418 struct netlink_callback *cb,
1419 struct net_device *dev,
1420 struct net_device *filter_dev,
1422 int (*ndo_fdb_get)(struct sk_buff *skb,
1423 struct nlattr *tb[],
1424 struct net_device *dev,
1425 const unsigned char *addr,
1426 u16 vid, u32 portid, u32 seq,
1427 struct netlink_ext_ack *extack);
1428 int (*ndo_bridge_setlink)(struct net_device *dev,
1429 struct nlmsghdr *nlh,
1431 struct netlink_ext_ack *extack);
1432 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1434 struct net_device *dev,
1437 int (*ndo_bridge_dellink)(struct net_device *dev,
1438 struct nlmsghdr *nlh,
1440 int (*ndo_change_carrier)(struct net_device *dev,
1442 int (*ndo_get_phys_port_id)(struct net_device *dev,
1443 struct netdev_phys_item_id *ppid);
1444 int (*ndo_get_port_parent_id)(struct net_device *dev,
1445 struct netdev_phys_item_id *ppid);
1446 int (*ndo_get_phys_port_name)(struct net_device *dev,
1447 char *name, size_t len);
1448 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1449 struct net_device *dev);
1450 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1453 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1456 int (*ndo_get_iflink)(const struct net_device *dev);
1457 int (*ndo_change_proto_down)(struct net_device *dev,
1459 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1460 struct sk_buff *skb);
1461 void (*ndo_set_rx_headroom)(struct net_device *dev,
1462 int needed_headroom);
1463 int (*ndo_bpf)(struct net_device *dev,
1464 struct netdev_bpf *bpf);
1465 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1466 struct xdp_frame **xdp,
1468 int (*ndo_xsk_wakeup)(struct net_device *dev,
1469 u32 queue_id, u32 flags);
1470 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1471 int (*ndo_tunnel_ctl)(struct net_device *dev,
1472 struct ip_tunnel_parm *p, int cmd);
1473 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1477 * enum netdev_priv_flags - &struct net_device priv_flags
1479 * These are the &struct net_device, they are only set internally
1480 * by drivers and used in the kernel. These flags are invisible to
1481 * userspace; this means that the order of these flags can change
1482 * during any kernel release.
1484 * You should have a pretty good reason to be extending these flags.
1486 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1487 * @IFF_EBRIDGE: Ethernet bridging device
1488 * @IFF_BONDING: bonding master or slave
1489 * @IFF_ISATAP: ISATAP interface (RFC4214)
1490 * @IFF_WAN_HDLC: WAN HDLC device
1491 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1493 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1494 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1495 * @IFF_MACVLAN_PORT: device used as macvlan port
1496 * @IFF_BRIDGE_PORT: device used as bridge port
1497 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1498 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1499 * @IFF_UNICAST_FLT: Supports unicast filtering
1500 * @IFF_TEAM_PORT: device used as team port
1501 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1502 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1503 * change when it's running
1504 * @IFF_MACVLAN: Macvlan device
1505 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1506 * underlying stacked devices
1507 * @IFF_L3MDEV_MASTER: device is an L3 master device
1508 * @IFF_NO_QUEUE: device can run without qdisc attached
1509 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1510 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1511 * @IFF_TEAM: device is a team device
1512 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1513 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1514 * entity (i.e. the master device for bridged veth)
1515 * @IFF_MACSEC: device is a MACsec device
1516 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1517 * @IFF_FAILOVER: device is a failover master device
1518 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1519 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1520 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1522 enum netdev_priv_flags {
1523 IFF_802_1Q_VLAN = 1<<0,
1527 IFF_WAN_HDLC = 1<<4,
1528 IFF_XMIT_DST_RELEASE = 1<<5,
1529 IFF_DONT_BRIDGE = 1<<6,
1530 IFF_DISABLE_NETPOLL = 1<<7,
1531 IFF_MACVLAN_PORT = 1<<8,
1532 IFF_BRIDGE_PORT = 1<<9,
1533 IFF_OVS_DATAPATH = 1<<10,
1534 IFF_TX_SKB_SHARING = 1<<11,
1535 IFF_UNICAST_FLT = 1<<12,
1536 IFF_TEAM_PORT = 1<<13,
1537 IFF_SUPP_NOFCS = 1<<14,
1538 IFF_LIVE_ADDR_CHANGE = 1<<15,
1539 IFF_MACVLAN = 1<<16,
1540 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1541 IFF_L3MDEV_MASTER = 1<<18,
1542 IFF_NO_QUEUE = 1<<19,
1543 IFF_OPENVSWITCH = 1<<20,
1544 IFF_L3MDEV_SLAVE = 1<<21,
1546 IFF_RXFH_CONFIGURED = 1<<23,
1547 IFF_PHONY_HEADROOM = 1<<24,
1549 IFF_NO_RX_HANDLER = 1<<26,
1550 IFF_FAILOVER = 1<<27,
1551 IFF_FAILOVER_SLAVE = 1<<28,
1552 IFF_L3MDEV_RX_HANDLER = 1<<29,
1553 IFF_LIVE_RENAME_OK = 1<<30,
1556 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1557 #define IFF_EBRIDGE IFF_EBRIDGE
1558 #define IFF_BONDING IFF_BONDING
1559 #define IFF_ISATAP IFF_ISATAP
1560 #define IFF_WAN_HDLC IFF_WAN_HDLC
1561 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1562 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1563 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1564 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1565 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1566 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1567 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1568 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1569 #define IFF_TEAM_PORT IFF_TEAM_PORT
1570 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1571 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1572 #define IFF_MACVLAN IFF_MACVLAN
1573 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1574 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1575 #define IFF_NO_QUEUE IFF_NO_QUEUE
1576 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1577 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1578 #define IFF_TEAM IFF_TEAM
1579 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1580 #define IFF_MACSEC IFF_MACSEC
1581 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1582 #define IFF_FAILOVER IFF_FAILOVER
1583 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1584 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1585 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1588 * struct net_device - The DEVICE structure.
1590 * Actually, this whole structure is a big mistake. It mixes I/O
1591 * data with strictly "high-level" data, and it has to know about
1592 * almost every data structure used in the INET module.
1594 * @name: This is the first field of the "visible" part of this structure
1595 * (i.e. as seen by users in the "Space.c" file). It is the name
1598 * @name_node: Name hashlist node
1599 * @ifalias: SNMP alias
1600 * @mem_end: Shared memory end
1601 * @mem_start: Shared memory start
1602 * @base_addr: Device I/O address
1603 * @irq: Device IRQ number
1605 * @state: Generic network queuing layer state, see netdev_state_t
1606 * @dev_list: The global list of network devices
1607 * @napi_list: List entry used for polling NAPI devices
1608 * @unreg_list: List entry when we are unregistering the
1609 * device; see the function unregister_netdev
1610 * @close_list: List entry used when we are closing the device
1611 * @ptype_all: Device-specific packet handlers for all protocols
1612 * @ptype_specific: Device-specific, protocol-specific packet handlers
1614 * @adj_list: Directly linked devices, like slaves for bonding
1615 * @features: Currently active device features
1616 * @hw_features: User-changeable features
1618 * @wanted_features: User-requested features
1619 * @vlan_features: Mask of features inheritable by VLAN devices
1621 * @hw_enc_features: Mask of features inherited by encapsulating devices
1622 * This field indicates what encapsulation
1623 * offloads the hardware is capable of doing,
1624 * and drivers will need to set them appropriately.
1626 * @mpls_features: Mask of features inheritable by MPLS
1627 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1629 * @ifindex: interface index
1630 * @group: The group the device belongs to
1632 * @stats: Statistics struct, which was left as a legacy, use
1633 * rtnl_link_stats64 instead
1635 * @rx_dropped: Dropped packets by core network,
1636 * do not use this in drivers
1637 * @tx_dropped: Dropped packets by core network,
1638 * do not use this in drivers
1639 * @rx_nohandler: nohandler dropped packets by core network on
1640 * inactive devices, do not use this in drivers
1641 * @carrier_up_count: Number of times the carrier has been up
1642 * @carrier_down_count: Number of times the carrier has been down
1644 * @wireless_handlers: List of functions to handle Wireless Extensions,
1646 * see <net/iw_handler.h> for details.
1647 * @wireless_data: Instance data managed by the core of wireless extensions
1649 * @netdev_ops: Includes several pointers to callbacks,
1650 * if one wants to override the ndo_*() functions
1651 * @ethtool_ops: Management operations
1652 * @l3mdev_ops: Layer 3 master device operations
1653 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1654 * discovery handling. Necessary for e.g. 6LoWPAN.
1655 * @xfrmdev_ops: Transformation offload operations
1656 * @tlsdev_ops: Transport Layer Security offload operations
1657 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1658 * of Layer 2 headers.
1660 * @flags: Interface flags (a la BSD)
1661 * @priv_flags: Like 'flags' but invisible to userspace,
1662 * see if.h for the definitions
1663 * @gflags: Global flags ( kept as legacy )
1664 * @padded: How much padding added by alloc_netdev()
1665 * @operstate: RFC2863 operstate
1666 * @link_mode: Mapping policy to operstate
1667 * @if_port: Selectable AUI, TP, ...
1669 * @mtu: Interface MTU value
1670 * @min_mtu: Interface Minimum MTU value
1671 * @max_mtu: Interface Maximum MTU value
1672 * @type: Interface hardware type
1673 * @hard_header_len: Maximum hardware header length.
1674 * @min_header_len: Minimum hardware header length
1676 * @needed_headroom: Extra headroom the hardware may need, but not in all
1677 * cases can this be guaranteed
1678 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1679 * cases can this be guaranteed. Some cases also use
1680 * LL_MAX_HEADER instead to allocate the skb
1682 * interface address info:
1684 * @perm_addr: Permanent hw address
1685 * @addr_assign_type: Hw address assignment type
1686 * @addr_len: Hardware address length
1687 * @upper_level: Maximum depth level of upper devices.
1688 * @lower_level: Maximum depth level of lower devices.
1689 * @neigh_priv_len: Used in neigh_alloc()
1690 * @dev_id: Used to differentiate devices that share
1691 * the same link layer address
1692 * @dev_port: Used to differentiate devices that share
1694 * @addr_list_lock: XXX: need comments on this one
1695 * @name_assign_type: network interface name assignment type
1696 * @uc_promisc: Counter that indicates promiscuous mode
1697 * has been enabled due to the need to listen to
1698 * additional unicast addresses in a device that
1699 * does not implement ndo_set_rx_mode()
1700 * @uc: unicast mac addresses
1701 * @mc: multicast mac addresses
1702 * @dev_addrs: list of device hw addresses
1703 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1704 * @promiscuity: Number of times the NIC is told to work in
1705 * promiscuous mode; if it becomes 0 the NIC will
1706 * exit promiscuous mode
1707 * @allmulti: Counter, enables or disables allmulticast mode
1709 * @vlan_info: VLAN info
1710 * @dsa_ptr: dsa specific data
1711 * @tipc_ptr: TIPC specific data
1712 * @atalk_ptr: AppleTalk link
1713 * @ip_ptr: IPv4 specific data
1714 * @dn_ptr: DECnet specific data
1715 * @ip6_ptr: IPv6 specific data
1716 * @ax25_ptr: AX.25 specific data
1717 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1718 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1720 * @mpls_ptr: mpls_dev struct pointer
1722 * @dev_addr: Hw address (before bcast,
1723 * because most packets are unicast)
1725 * @_rx: Array of RX queues
1726 * @num_rx_queues: Number of RX queues
1727 * allocated at register_netdev() time
1728 * @real_num_rx_queues: Number of RX queues currently active in device
1729 * @xdp_prog: XDP sockets filter program pointer
1730 * @gro_flush_timeout: timeout for GRO layer in NAPI
1731 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1732 * allow to avoid NIC hard IRQ, on busy queues.
1734 * @rx_handler: handler for received packets
1735 * @rx_handler_data: XXX: need comments on this one
1736 * @miniq_ingress: ingress/clsact qdisc specific data for
1737 * ingress processing
1738 * @ingress_queue: XXX: need comments on this one
1739 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1740 * @broadcast: hw bcast address
1742 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1743 * indexed by RX queue number. Assigned by driver.
1744 * This must only be set if the ndo_rx_flow_steer
1745 * operation is defined
1746 * @index_hlist: Device index hash chain
1748 * @_tx: Array of TX queues
1749 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1750 * @real_num_tx_queues: Number of TX queues currently active in device
1751 * @qdisc: Root qdisc from userspace point of view
1752 * @tx_queue_len: Max frames per queue allowed
1753 * @tx_global_lock: XXX: need comments on this one
1754 * @xdp_bulkq: XDP device bulk queue
1755 * @xps_cpus_map: all CPUs map for XPS device
1756 * @xps_rxqs_map: all RXQs map for XPS device
1758 * @xps_maps: XXX: need comments on this one
1759 * @miniq_egress: clsact qdisc specific data for
1761 * @qdisc_hash: qdisc hash table
1762 * @watchdog_timeo: Represents the timeout that is used by
1763 * the watchdog (see dev_watchdog())
1764 * @watchdog_timer: List of timers
1766 * @proto_down_reason: reason a netdev interface is held down
1767 * @pcpu_refcnt: Number of references to this device
1768 * @todo_list: Delayed register/unregister
1769 * @link_watch_list: XXX: need comments on this one
1771 * @reg_state: Register/unregister state machine
1772 * @dismantle: Device is going to be freed
1773 * @rtnl_link_state: This enum represents the phases of creating
1776 * @needs_free_netdev: Should unregister perform free_netdev?
1777 * @priv_destructor: Called from unregister
1778 * @npinfo: XXX: need comments on this one
1779 * @nd_net: Network namespace this network device is inside
1781 * @ml_priv: Mid-layer private
1782 * @lstats: Loopback statistics
1783 * @tstats: Tunnel statistics
1784 * @dstats: Dummy statistics
1785 * @vstats: Virtual ethernet statistics
1790 * @dev: Class/net/name entry
1791 * @sysfs_groups: Space for optional device, statistics and wireless
1794 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1795 * @rtnl_link_ops: Rtnl_link_ops
1797 * @gso_max_size: Maximum size of generic segmentation offload
1798 * @gso_max_segs: Maximum number of segments that can be passed to the
1801 * @dcbnl_ops: Data Center Bridging netlink ops
1802 * @num_tc: Number of traffic classes in the net device
1803 * @tc_to_txq: XXX: need comments on this one
1804 * @prio_tc_map: XXX: need comments on this one
1806 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1808 * @priomap: XXX: need comments on this one
1809 * @phydev: Physical device may attach itself
1810 * for hardware timestamping
1811 * @sfp_bus: attached &struct sfp_bus structure.
1813 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1814 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1816 * @proto_down: protocol port state information can be sent to the
1817 * switch driver and used to set the phys state of the
1820 * @wol_enabled: Wake-on-LAN is enabled
1822 * @threaded: napi threaded mode is enabled
1824 * @net_notifier_list: List of per-net netdev notifier block
1825 * that follow this device when it is moved
1826 * to another network namespace.
1828 * @macsec_ops: MACsec offloading ops
1830 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1831 * offload capabilities of the device
1832 * @udp_tunnel_nic: UDP tunnel offload state
1833 * @xdp_state: stores info on attached XDP BPF programs
1835 * @nested_level: Used as as a parameter of spin_lock_nested() of
1836 * dev->addr_list_lock.
1837 * @unlink_list: As netif_addr_lock() can be called recursively,
1838 * keep a list of interfaces to be deleted.
1840 * FIXME: cleanup struct net_device such that network protocol info
1845 char name[IFNAMSIZ];
1846 struct netdev_name_node *name_node;
1847 struct dev_ifalias __rcu *ifalias;
1849 * I/O specific fields
1850 * FIXME: Merge these and struct ifmap into one
1852 unsigned long mem_end;
1853 unsigned long mem_start;
1854 unsigned long base_addr;
1857 * Some hardware also needs these fields (state,dev_list,
1858 * napi_list,unreg_list,close_list) but they are not
1859 * part of the usual set specified in Space.c.
1862 unsigned long state;
1864 struct list_head dev_list;
1865 struct list_head napi_list;
1866 struct list_head unreg_list;
1867 struct list_head close_list;
1868 struct list_head ptype_all;
1869 struct list_head ptype_specific;
1872 struct list_head upper;
1873 struct list_head lower;
1876 /* Read-mostly cache-line for fast-path access */
1878 unsigned int priv_flags;
1879 const struct net_device_ops *netdev_ops;
1881 unsigned short gflags;
1882 unsigned short hard_header_len;
1884 /* Note : dev->mtu is often read without holding a lock.
1885 * Writers usually hold RTNL.
1886 * It is recommended to use READ_ONCE() to annotate the reads,
1887 * and to use WRITE_ONCE() to annotate the writes.
1890 unsigned short needed_headroom;
1891 unsigned short needed_tailroom;
1893 netdev_features_t features;
1894 netdev_features_t hw_features;
1895 netdev_features_t wanted_features;
1896 netdev_features_t vlan_features;
1897 netdev_features_t hw_enc_features;
1898 netdev_features_t mpls_features;
1899 netdev_features_t gso_partial_features;
1901 unsigned int min_mtu;
1902 unsigned int max_mtu;
1903 unsigned short type;
1904 unsigned char min_header_len;
1905 unsigned char name_assign_type;
1909 struct net_device_stats stats; /* not used by modern drivers */
1911 atomic_long_t rx_dropped;
1912 atomic_long_t tx_dropped;
1913 atomic_long_t rx_nohandler;
1915 /* Stats to monitor link on/off, flapping */
1916 atomic_t carrier_up_count;
1917 atomic_t carrier_down_count;
1919 #ifdef CONFIG_WIRELESS_EXT
1920 const struct iw_handler_def *wireless_handlers;
1921 struct iw_public_data *wireless_data;
1923 const struct ethtool_ops *ethtool_ops;
1924 #ifdef CONFIG_NET_L3_MASTER_DEV
1925 const struct l3mdev_ops *l3mdev_ops;
1927 #if IS_ENABLED(CONFIG_IPV6)
1928 const struct ndisc_ops *ndisc_ops;
1931 #ifdef CONFIG_XFRM_OFFLOAD
1932 const struct xfrmdev_ops *xfrmdev_ops;
1935 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1936 const struct tlsdev_ops *tlsdev_ops;
1939 const struct header_ops *header_ops;
1941 unsigned char operstate;
1942 unsigned char link_mode;
1944 unsigned char if_port;
1947 /* Interface address info. */
1948 unsigned char perm_addr[MAX_ADDR_LEN];
1949 unsigned char addr_assign_type;
1950 unsigned char addr_len;
1951 unsigned char upper_level;
1952 unsigned char lower_level;
1954 unsigned short neigh_priv_len;
1955 unsigned short dev_id;
1956 unsigned short dev_port;
1957 unsigned short padded;
1959 spinlock_t addr_list_lock;
1962 struct netdev_hw_addr_list uc;
1963 struct netdev_hw_addr_list mc;
1964 struct netdev_hw_addr_list dev_addrs;
1967 struct kset *queues_kset;
1969 #ifdef CONFIG_LOCKDEP
1970 struct list_head unlink_list;
1972 unsigned int promiscuity;
1973 unsigned int allmulti;
1975 #ifdef CONFIG_LOCKDEP
1976 unsigned char nested_level;
1980 /* Protocol-specific pointers */
1982 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1983 struct vlan_info __rcu *vlan_info;
1985 #if IS_ENABLED(CONFIG_NET_DSA)
1986 struct dsa_port *dsa_ptr;
1988 #if IS_ENABLED(CONFIG_TIPC)
1989 struct tipc_bearer __rcu *tipc_ptr;
1991 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1994 struct in_device __rcu *ip_ptr;
1995 #if IS_ENABLED(CONFIG_DECNET)
1996 struct dn_dev __rcu *dn_ptr;
1998 struct inet6_dev __rcu *ip6_ptr;
1999 #if IS_ENABLED(CONFIG_AX25)
2002 struct wireless_dev *ieee80211_ptr;
2003 struct wpan_dev *ieee802154_ptr;
2004 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2005 struct mpls_dev __rcu *mpls_ptr;
2009 * Cache lines mostly used on receive path (including eth_type_trans())
2011 /* Interface address info used in eth_type_trans() */
2012 unsigned char *dev_addr;
2014 struct netdev_rx_queue *_rx;
2015 unsigned int num_rx_queues;
2016 unsigned int real_num_rx_queues;
2018 struct bpf_prog __rcu *xdp_prog;
2019 unsigned long gro_flush_timeout;
2020 int napi_defer_hard_irqs;
2021 rx_handler_func_t __rcu *rx_handler;
2022 void __rcu *rx_handler_data;
2024 #ifdef CONFIG_NET_CLS_ACT
2025 struct mini_Qdisc __rcu *miniq_ingress;
2027 struct netdev_queue __rcu *ingress_queue;
2028 #ifdef CONFIG_NETFILTER_INGRESS
2029 struct nf_hook_entries __rcu *nf_hooks_ingress;
2032 unsigned char broadcast[MAX_ADDR_LEN];
2033 #ifdef CONFIG_RFS_ACCEL
2034 struct cpu_rmap *rx_cpu_rmap;
2036 struct hlist_node index_hlist;
2039 * Cache lines mostly used on transmit path
2041 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2042 unsigned int num_tx_queues;
2043 unsigned int real_num_tx_queues;
2044 struct Qdisc *qdisc;
2045 unsigned int tx_queue_len;
2046 spinlock_t tx_global_lock;
2048 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2051 struct xps_dev_maps __rcu *xps_cpus_map;
2052 struct xps_dev_maps __rcu *xps_rxqs_map;
2054 #ifdef CONFIG_NET_CLS_ACT
2055 struct mini_Qdisc __rcu *miniq_egress;
2058 #ifdef CONFIG_NET_SCHED
2059 DECLARE_HASHTABLE (qdisc_hash, 4);
2061 /* These may be needed for future network-power-down code. */
2062 struct timer_list watchdog_timer;
2065 u32 proto_down_reason;
2067 struct list_head todo_list;
2068 int __percpu *pcpu_refcnt;
2070 struct list_head link_watch_list;
2072 enum { NETREG_UNINITIALIZED=0,
2073 NETREG_REGISTERED, /* completed register_netdevice */
2074 NETREG_UNREGISTERING, /* called unregister_netdevice */
2075 NETREG_UNREGISTERED, /* completed unregister todo */
2076 NETREG_RELEASED, /* called free_netdev */
2077 NETREG_DUMMY, /* dummy device for NAPI poll */
2083 RTNL_LINK_INITIALIZED,
2084 RTNL_LINK_INITIALIZING,
2085 } rtnl_link_state:16;
2087 bool needs_free_netdev;
2088 void (*priv_destructor)(struct net_device *dev);
2090 #ifdef CONFIG_NETPOLL
2091 struct netpoll_info __rcu *npinfo;
2094 possible_net_t nd_net;
2096 /* mid-layer private */
2099 struct pcpu_lstats __percpu *lstats;
2100 struct pcpu_sw_netstats __percpu *tstats;
2101 struct pcpu_dstats __percpu *dstats;
2104 #if IS_ENABLED(CONFIG_GARP)
2105 struct garp_port __rcu *garp_port;
2107 #if IS_ENABLED(CONFIG_MRP)
2108 struct mrp_port __rcu *mrp_port;
2112 const struct attribute_group *sysfs_groups[4];
2113 const struct attribute_group *sysfs_rx_queue_group;
2115 const struct rtnl_link_ops *rtnl_link_ops;
2117 /* for setting kernel sock attribute on TCP connection setup */
2118 #define GSO_MAX_SIZE 65536
2119 unsigned int gso_max_size;
2120 #define GSO_MAX_SEGS 65535
2124 const struct dcbnl_rtnl_ops *dcbnl_ops;
2127 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2128 u8 prio_tc_map[TC_BITMASK + 1];
2130 #if IS_ENABLED(CONFIG_FCOE)
2131 unsigned int fcoe_ddp_xid;
2133 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2134 struct netprio_map __rcu *priomap;
2136 struct phy_device *phydev;
2137 struct sfp_bus *sfp_bus;
2138 struct lock_class_key *qdisc_tx_busylock;
2139 struct lock_class_key *qdisc_running_key;
2141 unsigned wol_enabled:1;
2142 unsigned threaded:1;
2144 struct list_head net_notifier_list;
2146 #if IS_ENABLED(CONFIG_MACSEC)
2147 /* MACsec management functions */
2148 const struct macsec_ops *macsec_ops;
2150 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2151 struct udp_tunnel_nic *udp_tunnel_nic;
2153 /* protected by rtnl_lock */
2154 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2156 #define to_net_dev(d) container_of(d, struct net_device, dev)
2158 static inline bool netif_elide_gro(const struct net_device *dev)
2160 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2165 #define NETDEV_ALIGN 32
2168 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2170 return dev->prio_tc_map[prio & TC_BITMASK];
2174 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2176 if (tc >= dev->num_tc)
2179 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2183 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2184 void netdev_reset_tc(struct net_device *dev);
2185 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2186 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2189 int netdev_get_num_tc(struct net_device *dev)
2194 static inline void net_prefetch(void *p)
2197 #if L1_CACHE_BYTES < 128
2198 prefetch((u8 *)p + L1_CACHE_BYTES);
2202 static inline void net_prefetchw(void *p)
2205 #if L1_CACHE_BYTES < 128
2206 prefetchw((u8 *)p + L1_CACHE_BYTES);
2210 void netdev_unbind_sb_channel(struct net_device *dev,
2211 struct net_device *sb_dev);
2212 int netdev_bind_sb_channel_queue(struct net_device *dev,
2213 struct net_device *sb_dev,
2214 u8 tc, u16 count, u16 offset);
2215 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2216 static inline int netdev_get_sb_channel(struct net_device *dev)
2218 return max_t(int, -dev->num_tc, 0);
2222 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2225 return &dev->_tx[index];
2228 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2229 const struct sk_buff *skb)
2231 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2234 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2235 void (*f)(struct net_device *,
2236 struct netdev_queue *,
2242 for (i = 0; i < dev->num_tx_queues; i++)
2243 f(dev, &dev->_tx[i], arg);
2246 #define netdev_lockdep_set_classes(dev) \
2248 static struct lock_class_key qdisc_tx_busylock_key; \
2249 static struct lock_class_key qdisc_running_key; \
2250 static struct lock_class_key qdisc_xmit_lock_key; \
2251 static struct lock_class_key dev_addr_list_lock_key; \
2254 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2255 (dev)->qdisc_running_key = &qdisc_running_key; \
2256 lockdep_set_class(&(dev)->addr_list_lock, \
2257 &dev_addr_list_lock_key); \
2258 for (i = 0; i < (dev)->num_tx_queues; i++) \
2259 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2260 &qdisc_xmit_lock_key); \
2263 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2264 struct net_device *sb_dev);
2265 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2266 struct sk_buff *skb,
2267 struct net_device *sb_dev);
2269 /* returns the headroom that the master device needs to take in account
2270 * when forwarding to this dev
2272 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2274 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2277 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2279 if (dev->netdev_ops->ndo_set_rx_headroom)
2280 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2283 /* set the device rx headroom to the dev's default */
2284 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2286 netdev_set_rx_headroom(dev, -1);
2290 * Net namespace inlines
2293 struct net *dev_net(const struct net_device *dev)
2295 return read_pnet(&dev->nd_net);
2299 void dev_net_set(struct net_device *dev, struct net *net)
2301 write_pnet(&dev->nd_net, net);
2305 * netdev_priv - access network device private data
2306 * @dev: network device
2308 * Get network device private data
2310 static inline void *netdev_priv(const struct net_device *dev)
2312 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2315 /* Set the sysfs physical device reference for the network logical device
2316 * if set prior to registration will cause a symlink during initialization.
2318 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2320 /* Set the sysfs device type for the network logical device to allow
2321 * fine-grained identification of different network device types. For
2322 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2324 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2326 /* Default NAPI poll() weight
2327 * Device drivers are strongly advised to not use bigger value
2329 #define NAPI_POLL_WEIGHT 64
2332 * netif_napi_add - initialize a NAPI context
2333 * @dev: network device
2334 * @napi: NAPI context
2335 * @poll: polling function
2336 * @weight: default weight
2338 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2339 * *any* of the other NAPI-related functions.
2341 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2342 int (*poll)(struct napi_struct *, int), int weight);
2345 * netif_tx_napi_add - initialize a NAPI context
2346 * @dev: network device
2347 * @napi: NAPI context
2348 * @poll: polling function
2349 * @weight: default weight
2351 * This variant of netif_napi_add() should be used from drivers using NAPI
2352 * to exclusively poll a TX queue.
2353 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2355 static inline void netif_tx_napi_add(struct net_device *dev,
2356 struct napi_struct *napi,
2357 int (*poll)(struct napi_struct *, int),
2360 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2361 netif_napi_add(dev, napi, poll, weight);
2365 * __netif_napi_del - remove a NAPI context
2366 * @napi: NAPI context
2368 * Warning: caller must observe RCU grace period before freeing memory
2369 * containing @napi. Drivers might want to call this helper to combine
2370 * all the needed RCU grace periods into a single one.
2372 void __netif_napi_del(struct napi_struct *napi);
2375 * netif_napi_del - remove a NAPI context
2376 * @napi: NAPI context
2378 * netif_napi_del() removes a NAPI context from the network device NAPI list
2380 static inline void netif_napi_del(struct napi_struct *napi)
2382 __netif_napi_del(napi);
2386 struct napi_gro_cb {
2387 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2390 /* Length of frag0. */
2391 unsigned int frag0_len;
2393 /* This indicates where we are processing relative to skb->data. */
2396 /* This is non-zero if the packet cannot be merged with the new skb. */
2399 /* Save the IP ID here and check when we get to the transport layer */
2402 /* Number of segments aggregated. */
2405 /* Start offset for remote checksum offload */
2406 u16 gro_remcsum_start;
2408 /* jiffies when first packet was created/queued */
2411 /* Used in ipv6_gro_receive() and foo-over-udp */
2414 /* This is non-zero if the packet may be of the same flow. */
2417 /* Used in tunnel GRO receive */
2420 /* GRO checksum is valid */
2423 /* Number of checksums via CHECKSUM_UNNECESSARY */
2428 #define NAPI_GRO_FREE 1
2429 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2431 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2434 /* Used in GRE, set in fou/gue_gro_receive */
2437 /* Used to determine if flush_id can be ignored */
2440 /* Number of gro_receive callbacks this packet already went through */
2441 u8 recursion_counter:4;
2443 /* GRO is done by frag_list pointer chaining. */
2446 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2449 /* used in skb_gro_receive() slow path */
2450 struct sk_buff *last;
2453 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2455 #define GRO_RECURSION_LIMIT 15
2456 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2458 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2461 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2462 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2463 struct list_head *head,
2464 struct sk_buff *skb)
2466 if (unlikely(gro_recursion_inc_test(skb))) {
2467 NAPI_GRO_CB(skb)->flush |= 1;
2471 return cb(head, skb);
2474 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2476 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2478 struct list_head *head,
2479 struct sk_buff *skb)
2481 if (unlikely(gro_recursion_inc_test(skb))) {
2482 NAPI_GRO_CB(skb)->flush |= 1;
2486 return cb(sk, head, skb);
2489 struct packet_type {
2490 __be16 type; /* This is really htons(ether_type). */
2491 bool ignore_outgoing;
2492 struct net_device *dev; /* NULL is wildcarded here */
2493 int (*func) (struct sk_buff *,
2494 struct net_device *,
2495 struct packet_type *,
2496 struct net_device *);
2497 void (*list_func) (struct list_head *,
2498 struct packet_type *,
2499 struct net_device *);
2500 bool (*id_match)(struct packet_type *ptype,
2502 void *af_packet_priv;
2503 struct list_head list;
2506 struct offload_callbacks {
2507 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2508 netdev_features_t features);
2509 struct sk_buff *(*gro_receive)(struct list_head *head,
2510 struct sk_buff *skb);
2511 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2514 struct packet_offload {
2515 __be16 type; /* This is really htons(ether_type). */
2517 struct offload_callbacks callbacks;
2518 struct list_head list;
2521 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2522 struct pcpu_sw_netstats {
2527 struct u64_stats_sync syncp;
2528 } __aligned(4 * sizeof(u64));
2530 struct pcpu_lstats {
2531 u64_stats_t packets;
2533 struct u64_stats_sync syncp;
2534 } __aligned(2 * sizeof(u64));
2536 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2538 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2540 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2542 u64_stats_update_begin(&tstats->syncp);
2543 tstats->rx_bytes += len;
2544 tstats->rx_packets++;
2545 u64_stats_update_end(&tstats->syncp);
2548 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2549 unsigned int packets,
2552 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2554 u64_stats_update_begin(&tstats->syncp);
2555 tstats->tx_bytes += len;
2556 tstats->tx_packets += packets;
2557 u64_stats_update_end(&tstats->syncp);
2560 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2562 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2564 u64_stats_update_begin(&lstats->syncp);
2565 u64_stats_add(&lstats->bytes, len);
2566 u64_stats_inc(&lstats->packets);
2567 u64_stats_update_end(&lstats->syncp);
2570 #define __netdev_alloc_pcpu_stats(type, gfp) \
2572 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2575 for_each_possible_cpu(__cpu) { \
2576 typeof(type) *stat; \
2577 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2578 u64_stats_init(&stat->syncp); \
2584 #define netdev_alloc_pcpu_stats(type) \
2585 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2587 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2589 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2592 for_each_possible_cpu(__cpu) { \
2593 typeof(type) *stat; \
2594 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2595 u64_stats_init(&stat->syncp); \
2601 enum netdev_lag_tx_type {
2602 NETDEV_LAG_TX_TYPE_UNKNOWN,
2603 NETDEV_LAG_TX_TYPE_RANDOM,
2604 NETDEV_LAG_TX_TYPE_BROADCAST,
2605 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2606 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2607 NETDEV_LAG_TX_TYPE_HASH,
2610 enum netdev_lag_hash {
2611 NETDEV_LAG_HASH_NONE,
2613 NETDEV_LAG_HASH_L34,
2614 NETDEV_LAG_HASH_L23,
2615 NETDEV_LAG_HASH_E23,
2616 NETDEV_LAG_HASH_E34,
2617 NETDEV_LAG_HASH_VLAN_SRCMAC,
2618 NETDEV_LAG_HASH_UNKNOWN,
2621 struct netdev_lag_upper_info {
2622 enum netdev_lag_tx_type tx_type;
2623 enum netdev_lag_hash hash_type;
2626 struct netdev_lag_lower_state_info {
2631 #include <linux/notifier.h>
2633 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2634 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2638 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2640 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2641 detected a hardware crash and restarted
2642 - we can use this eg to kick tcp sessions
2644 NETDEV_CHANGE, /* Notify device state change */
2647 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2648 NETDEV_CHANGEADDR, /* notify after the address change */
2649 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2653 NETDEV_BONDING_FAILOVER,
2655 NETDEV_PRE_TYPE_CHANGE,
2656 NETDEV_POST_TYPE_CHANGE,
2659 NETDEV_NOTIFY_PEERS,
2663 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2664 NETDEV_CHANGEINFODATA,
2665 NETDEV_BONDING_INFO,
2666 NETDEV_PRECHANGEUPPER,
2667 NETDEV_CHANGELOWERSTATE,
2668 NETDEV_UDP_TUNNEL_PUSH_INFO,
2669 NETDEV_UDP_TUNNEL_DROP_INFO,
2670 NETDEV_CHANGE_TX_QUEUE_LEN,
2671 NETDEV_CVLAN_FILTER_PUSH_INFO,
2672 NETDEV_CVLAN_FILTER_DROP_INFO,
2673 NETDEV_SVLAN_FILTER_PUSH_INFO,
2674 NETDEV_SVLAN_FILTER_DROP_INFO,
2676 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2678 int register_netdevice_notifier(struct notifier_block *nb);
2679 int unregister_netdevice_notifier(struct notifier_block *nb);
2680 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2681 int unregister_netdevice_notifier_net(struct net *net,
2682 struct notifier_block *nb);
2683 int register_netdevice_notifier_dev_net(struct net_device *dev,
2684 struct notifier_block *nb,
2685 struct netdev_net_notifier *nn);
2686 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2687 struct notifier_block *nb,
2688 struct netdev_net_notifier *nn);
2690 struct netdev_notifier_info {
2691 struct net_device *dev;
2692 struct netlink_ext_ack *extack;
2695 struct netdev_notifier_info_ext {
2696 struct netdev_notifier_info info; /* must be first */
2702 struct netdev_notifier_change_info {
2703 struct netdev_notifier_info info; /* must be first */
2704 unsigned int flags_changed;
2707 struct netdev_notifier_changeupper_info {
2708 struct netdev_notifier_info info; /* must be first */
2709 struct net_device *upper_dev; /* new upper dev */
2710 bool master; /* is upper dev master */
2711 bool linking; /* is the notification for link or unlink */
2712 void *upper_info; /* upper dev info */
2715 struct netdev_notifier_changelowerstate_info {
2716 struct netdev_notifier_info info; /* must be first */
2717 void *lower_state_info; /* is lower dev state */
2720 struct netdev_notifier_pre_changeaddr_info {
2721 struct netdev_notifier_info info; /* must be first */
2722 const unsigned char *dev_addr;
2725 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2726 struct net_device *dev)
2729 info->extack = NULL;
2732 static inline struct net_device *
2733 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2738 static inline struct netlink_ext_ack *
2739 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2741 return info->extack;
2744 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2747 extern rwlock_t dev_base_lock; /* Device list lock */
2749 #define for_each_netdev(net, d) \
2750 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2751 #define for_each_netdev_reverse(net, d) \
2752 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2753 #define for_each_netdev_rcu(net, d) \
2754 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2755 #define for_each_netdev_safe(net, d, n) \
2756 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2757 #define for_each_netdev_continue(net, d) \
2758 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2759 #define for_each_netdev_continue_reverse(net, d) \
2760 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2762 #define for_each_netdev_continue_rcu(net, d) \
2763 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2764 #define for_each_netdev_in_bond_rcu(bond, slave) \
2765 for_each_netdev_rcu(&init_net, slave) \
2766 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2767 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2769 static inline struct net_device *next_net_device(struct net_device *dev)
2771 struct list_head *lh;
2775 lh = dev->dev_list.next;
2776 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2779 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2781 struct list_head *lh;
2785 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2786 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2789 static inline struct net_device *first_net_device(struct net *net)
2791 return list_empty(&net->dev_base_head) ? NULL :
2792 net_device_entry(net->dev_base_head.next);
2795 static inline struct net_device *first_net_device_rcu(struct net *net)
2797 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2799 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2802 int netdev_boot_setup_check(struct net_device *dev);
2803 unsigned long netdev_boot_base(const char *prefix, int unit);
2804 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2805 const char *hwaddr);
2806 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2807 void dev_add_pack(struct packet_type *pt);
2808 void dev_remove_pack(struct packet_type *pt);
2809 void __dev_remove_pack(struct packet_type *pt);
2810 void dev_add_offload(struct packet_offload *po);
2811 void dev_remove_offload(struct packet_offload *po);
2813 int dev_get_iflink(const struct net_device *dev);
2814 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2815 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2816 unsigned short mask);
2817 struct net_device *dev_get_by_name(struct net *net, const char *name);
2818 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2819 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2820 int dev_alloc_name(struct net_device *dev, const char *name);
2821 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2822 void dev_close(struct net_device *dev);
2823 void dev_close_many(struct list_head *head, bool unlink);
2824 void dev_disable_lro(struct net_device *dev);
2825 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2826 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2827 struct net_device *sb_dev);
2828 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2829 struct net_device *sb_dev);
2831 int dev_queue_xmit(struct sk_buff *skb);
2832 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2833 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2835 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2839 ret = __dev_direct_xmit(skb, queue_id);
2840 if (!dev_xmit_complete(ret))
2845 int register_netdevice(struct net_device *dev);
2846 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2847 void unregister_netdevice_many(struct list_head *head);
2848 static inline void unregister_netdevice(struct net_device *dev)
2850 unregister_netdevice_queue(dev, NULL);
2853 int netdev_refcnt_read(const struct net_device *dev);
2854 void free_netdev(struct net_device *dev);
2855 void netdev_freemem(struct net_device *dev);
2856 int init_dummy_netdev(struct net_device *dev);
2858 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2859 struct sk_buff *skb,
2861 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2863 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2864 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2865 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2866 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2867 int netdev_get_name(struct net *net, char *name, int ifindex);
2868 int dev_restart(struct net_device *dev);
2869 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2870 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2872 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2874 return NAPI_GRO_CB(skb)->data_offset;
2877 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2879 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2882 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2884 NAPI_GRO_CB(skb)->data_offset += len;
2887 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2888 unsigned int offset)
2890 return NAPI_GRO_CB(skb)->frag0 + offset;
2893 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2895 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2898 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2900 NAPI_GRO_CB(skb)->frag0 = NULL;
2901 NAPI_GRO_CB(skb)->frag0_len = 0;
2904 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2905 unsigned int offset)
2907 if (!pskb_may_pull(skb, hlen))
2910 skb_gro_frag0_invalidate(skb);
2911 return skb->data + offset;
2914 static inline void *skb_gro_network_header(struct sk_buff *skb)
2916 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2917 skb_network_offset(skb);
2920 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2921 const void *start, unsigned int len)
2923 if (NAPI_GRO_CB(skb)->csum_valid)
2924 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2925 csum_partial(start, len, 0));
2928 /* GRO checksum functions. These are logical equivalents of the normal
2929 * checksum functions (in skbuff.h) except that they operate on the GRO
2930 * offsets and fields in sk_buff.
2933 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2935 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2937 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2940 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2944 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2945 skb_checksum_start_offset(skb) <
2946 skb_gro_offset(skb)) &&
2947 !skb_at_gro_remcsum_start(skb) &&
2948 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2949 (!zero_okay || check));
2952 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2955 if (NAPI_GRO_CB(skb)->csum_valid &&
2956 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2959 NAPI_GRO_CB(skb)->csum = psum;
2961 return __skb_gro_checksum_complete(skb);
2964 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2966 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2967 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2968 NAPI_GRO_CB(skb)->csum_cnt--;
2970 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2971 * verified a new top level checksum or an encapsulated one
2972 * during GRO. This saves work if we fallback to normal path.
2974 __skb_incr_checksum_unnecessary(skb);
2978 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2981 __sum16 __ret = 0; \
2982 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2983 __ret = __skb_gro_checksum_validate_complete(skb, \
2984 compute_pseudo(skb, proto)); \
2986 skb_gro_incr_csum_unnecessary(skb); \
2990 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2991 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2993 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2995 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2997 #define skb_gro_checksum_simple_validate(skb) \
2998 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3000 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3002 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3003 !NAPI_GRO_CB(skb)->csum_valid);
3006 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3009 NAPI_GRO_CB(skb)->csum = ~pseudo;
3010 NAPI_GRO_CB(skb)->csum_valid = 1;
3013 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3015 if (__skb_gro_checksum_convert_check(skb)) \
3016 __skb_gro_checksum_convert(skb, \
3017 compute_pseudo(skb, proto)); \
3020 struct gro_remcsum {
3025 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3031 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3032 unsigned int off, size_t hdrlen,
3033 int start, int offset,
3034 struct gro_remcsum *grc,
3038 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3040 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3043 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3047 ptr = skb_gro_header_fast(skb, off);
3048 if (skb_gro_header_hard(skb, off + plen)) {
3049 ptr = skb_gro_header_slow(skb, off + plen, off);
3054 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3057 /* Adjust skb->csum since we changed the packet */
3058 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3060 grc->offset = off + hdrlen + offset;
3066 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3067 struct gro_remcsum *grc)
3070 size_t plen = grc->offset + sizeof(u16);
3075 ptr = skb_gro_header_fast(skb, grc->offset);
3076 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3077 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3082 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3085 #ifdef CONFIG_XFRM_OFFLOAD
3086 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3088 if (PTR_ERR(pp) != -EINPROGRESS)
3089 NAPI_GRO_CB(skb)->flush |= flush;
3091 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3094 struct gro_remcsum *grc)
3096 if (PTR_ERR(pp) != -EINPROGRESS) {
3097 NAPI_GRO_CB(skb)->flush |= flush;
3098 skb_gro_remcsum_cleanup(skb, grc);
3099 skb->remcsum_offload = 0;
3103 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3105 NAPI_GRO_CB(skb)->flush |= flush;
3107 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3110 struct gro_remcsum *grc)
3112 NAPI_GRO_CB(skb)->flush |= flush;
3113 skb_gro_remcsum_cleanup(skb, grc);
3114 skb->remcsum_offload = 0;
3118 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3119 unsigned short type,
3120 const void *daddr, const void *saddr,
3123 if (!dev->header_ops || !dev->header_ops->create)
3126 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3129 static inline int dev_parse_header(const struct sk_buff *skb,
3130 unsigned char *haddr)
3132 const struct net_device *dev = skb->dev;
3134 if (!dev->header_ops || !dev->header_ops->parse)
3136 return dev->header_ops->parse(skb, haddr);
3139 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3141 const struct net_device *dev = skb->dev;
3143 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3145 return dev->header_ops->parse_protocol(skb);
3148 /* ll_header must have at least hard_header_len allocated */
3149 static inline bool dev_validate_header(const struct net_device *dev,
3150 char *ll_header, int len)
3152 if (likely(len >= dev->hard_header_len))
3154 if (len < dev->min_header_len)
3157 if (capable(CAP_SYS_RAWIO)) {
3158 memset(ll_header + len, 0, dev->hard_header_len - len);
3162 if (dev->header_ops && dev->header_ops->validate)
3163 return dev->header_ops->validate(ll_header, len);
3168 static inline bool dev_has_header(const struct net_device *dev)
3170 return dev->header_ops && dev->header_ops->create;
3173 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3175 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3176 static inline int unregister_gifconf(unsigned int family)
3178 return register_gifconf(family, NULL);
3181 #ifdef CONFIG_NET_FLOW_LIMIT
3182 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3183 struct sd_flow_limit {
3185 unsigned int num_buckets;
3186 unsigned int history_head;
3187 u16 history[FLOW_LIMIT_HISTORY];
3191 extern int netdev_flow_limit_table_len;
3192 #endif /* CONFIG_NET_FLOW_LIMIT */
3195 * Incoming packets are placed on per-CPU queues
3197 struct softnet_data {
3198 struct list_head poll_list;
3199 struct sk_buff_head process_queue;
3202 unsigned int processed;
3203 unsigned int time_squeeze;
3204 unsigned int received_rps;
3206 struct softnet_data *rps_ipi_list;
3208 #ifdef CONFIG_NET_FLOW_LIMIT
3209 struct sd_flow_limit __rcu *flow_limit;
3211 struct Qdisc *output_queue;
3212 struct Qdisc **output_queue_tailp;
3213 struct sk_buff *completion_queue;
3214 #ifdef CONFIG_XFRM_OFFLOAD
3215 struct sk_buff_head xfrm_backlog;
3217 /* written and read only by owning cpu: */
3223 /* input_queue_head should be written by cpu owning this struct,
3224 * and only read by other cpus. Worth using a cache line.
3226 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3228 /* Elements below can be accessed between CPUs for RPS/RFS */
3229 call_single_data_t csd ____cacheline_aligned_in_smp;
3230 struct softnet_data *rps_ipi_next;
3232 unsigned int input_queue_tail;
3234 unsigned int dropped;
3235 struct sk_buff_head input_pkt_queue;
3236 struct napi_struct backlog;
3240 static inline void input_queue_head_incr(struct softnet_data *sd)
3243 sd->input_queue_head++;
3247 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3248 unsigned int *qtail)
3251 *qtail = ++sd->input_queue_tail;
3255 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3257 static inline int dev_recursion_level(void)
3259 return this_cpu_read(softnet_data.xmit.recursion);
3262 #define XMIT_RECURSION_LIMIT 8
3263 static inline bool dev_xmit_recursion(void)
3265 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3266 XMIT_RECURSION_LIMIT);
3269 static inline void dev_xmit_recursion_inc(void)
3271 __this_cpu_inc(softnet_data.xmit.recursion);
3274 static inline void dev_xmit_recursion_dec(void)
3276 __this_cpu_dec(softnet_data.xmit.recursion);
3279 void __netif_schedule(struct Qdisc *q);
3280 void netif_schedule_queue(struct netdev_queue *txq);
3282 static inline void netif_tx_schedule_all(struct net_device *dev)
3286 for (i = 0; i < dev->num_tx_queues; i++)
3287 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3290 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3292 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3296 * netif_start_queue - allow transmit
3297 * @dev: network device
3299 * Allow upper layers to call the device hard_start_xmit routine.
3301 static inline void netif_start_queue(struct net_device *dev)
3303 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3306 static inline void netif_tx_start_all_queues(struct net_device *dev)
3310 for (i = 0; i < dev->num_tx_queues; i++) {
3311 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3312 netif_tx_start_queue(txq);
3316 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3319 * netif_wake_queue - restart transmit
3320 * @dev: network device
3322 * Allow upper layers to call the device hard_start_xmit routine.
3323 * Used for flow control when transmit resources are available.
3325 static inline void netif_wake_queue(struct net_device *dev)
3327 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3330 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3334 for (i = 0; i < dev->num_tx_queues; i++) {
3335 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3336 netif_tx_wake_queue(txq);
3340 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3342 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3346 * netif_stop_queue - stop transmitted packets
3347 * @dev: network device
3349 * Stop upper layers calling the device hard_start_xmit routine.
3350 * Used for flow control when transmit resources are unavailable.
3352 static inline void netif_stop_queue(struct net_device *dev)
3354 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3357 void netif_tx_stop_all_queues(struct net_device *dev);
3359 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3361 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3365 * netif_queue_stopped - test if transmit queue is flowblocked
3366 * @dev: network device
3368 * Test if transmit queue on device is currently unable to send.
3370 static inline bool netif_queue_stopped(const struct net_device *dev)
3372 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3375 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3377 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3381 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3383 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3387 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3389 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3393 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3394 * @dev_queue: pointer to transmit queue
3396 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3397 * to give appropriate hint to the CPU.
3399 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3402 prefetchw(&dev_queue->dql.num_queued);
3407 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3408 * @dev_queue: pointer to transmit queue
3410 * BQL enabled drivers might use this helper in their TX completion path,
3411 * to give appropriate hint to the CPU.
3413 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3416 prefetchw(&dev_queue->dql.limit);
3420 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3424 dql_queued(&dev_queue->dql, bytes);
3426 if (likely(dql_avail(&dev_queue->dql) >= 0))
3429 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3432 * The XOFF flag must be set before checking the dql_avail below,
3433 * because in netdev_tx_completed_queue we update the dql_completed
3434 * before checking the XOFF flag.
3438 /* check again in case another CPU has just made room avail */
3439 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3440 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3444 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3445 * that they should not test BQL status themselves.
3446 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3448 * Returns true if the doorbell must be used to kick the NIC.
3450 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3456 dql_queued(&dev_queue->dql, bytes);
3458 return netif_tx_queue_stopped(dev_queue);
3460 netdev_tx_sent_queue(dev_queue, bytes);
3465 * netdev_sent_queue - report the number of bytes queued to hardware
3466 * @dev: network device
3467 * @bytes: number of bytes queued to the hardware device queue
3469 * Report the number of bytes queued for sending/completion to the network
3470 * device hardware queue. @bytes should be a good approximation and should
3471 * exactly match netdev_completed_queue() @bytes
3473 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3475 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3478 static inline bool __netdev_sent_queue(struct net_device *dev,
3482 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3486 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3487 unsigned int pkts, unsigned int bytes)
3490 if (unlikely(!bytes))
3493 dql_completed(&dev_queue->dql, bytes);
3496 * Without the memory barrier there is a small possiblity that
3497 * netdev_tx_sent_queue will miss the update and cause the queue to
3498 * be stopped forever
3502 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3505 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3506 netif_schedule_queue(dev_queue);
3511 * netdev_completed_queue - report bytes and packets completed by device
3512 * @dev: network device
3513 * @pkts: actual number of packets sent over the medium
3514 * @bytes: actual number of bytes sent over the medium
3516 * Report the number of bytes and packets transmitted by the network device
3517 * hardware queue over the physical medium, @bytes must exactly match the
3518 * @bytes amount passed to netdev_sent_queue()
3520 static inline void netdev_completed_queue(struct net_device *dev,
3521 unsigned int pkts, unsigned int bytes)
3523 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3526 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3529 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3535 * netdev_reset_queue - reset the packets and bytes count of a network device
3536 * @dev_queue: network device
3538 * Reset the bytes and packet count of a network device and clear the
3539 * software flow control OFF bit for this network device
3541 static inline void netdev_reset_queue(struct net_device *dev_queue)
3543 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3547 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3548 * @dev: network device
3549 * @queue_index: given tx queue index
3551 * Returns 0 if given tx queue index >= number of device tx queues,
3552 * otherwise returns the originally passed tx queue index.
3554 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3556 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3557 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3558 dev->name, queue_index,
3559 dev->real_num_tx_queues);
3567 * netif_running - test if up
3568 * @dev: network device
3570 * Test if the device has been brought up.
3572 static inline bool netif_running(const struct net_device *dev)
3574 return test_bit(__LINK_STATE_START, &dev->state);
3578 * Routines to manage the subqueues on a device. We only need start,
3579 * stop, and a check if it's stopped. All other device management is
3580 * done at the overall netdevice level.
3581 * Also test the device if we're multiqueue.
3585 * netif_start_subqueue - allow sending packets on subqueue
3586 * @dev: network device
3587 * @queue_index: sub queue index
3589 * Start individual transmit queue of a device with multiple transmit queues.
3591 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3593 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3595 netif_tx_start_queue(txq);
3599 * netif_stop_subqueue - stop sending packets on subqueue
3600 * @dev: network device
3601 * @queue_index: sub queue index
3603 * Stop individual transmit queue of a device with multiple transmit queues.
3605 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3607 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3608 netif_tx_stop_queue(txq);
3612 * __netif_subqueue_stopped - test status of subqueue
3613 * @dev: network device
3614 * @queue_index: sub queue index
3616 * Check individual transmit queue of a device with multiple transmit queues.
3618 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3621 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3623 return netif_tx_queue_stopped(txq);
3627 * netif_subqueue_stopped - test status of subqueue
3628 * @dev: network device
3629 * @skb: sub queue buffer pointer
3631 * Check individual transmit queue of a device with multiple transmit queues.
3633 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3634 struct sk_buff *skb)
3636 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3640 * netif_wake_subqueue - allow sending packets on subqueue
3641 * @dev: network device
3642 * @queue_index: sub queue index
3644 * Resume individual transmit queue of a device with multiple transmit queues.
3646 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3648 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3650 netif_tx_wake_queue(txq);
3654 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3656 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3657 u16 index, bool is_rxqs_map);
3660 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3661 * @j: CPU/Rx queue index
3662 * @mask: bitmask of all cpus/rx queues
3663 * @nr_bits: number of bits in the bitmask
3665 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3667 static inline bool netif_attr_test_mask(unsigned long j,
3668 const unsigned long *mask,
3669 unsigned int nr_bits)
3671 cpu_max_bits_warn(j, nr_bits);
3672 return test_bit(j, mask);
3676 * netif_attr_test_online - Test for online CPU/Rx queue
3677 * @j: CPU/Rx queue index
3678 * @online_mask: bitmask for CPUs/Rx queues that are online
3679 * @nr_bits: number of bits in the bitmask
3681 * Returns true if a CPU/Rx queue is online.
3683 static inline bool netif_attr_test_online(unsigned long j,
3684 const unsigned long *online_mask,
3685 unsigned int nr_bits)
3687 cpu_max_bits_warn(j, nr_bits);
3690 return test_bit(j, online_mask);
3692 return (j < nr_bits);
3696 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3697 * @n: CPU/Rx queue index
3698 * @srcp: the cpumask/Rx queue mask pointer
3699 * @nr_bits: number of bits in the bitmask
3701 * Returns >= nr_bits if no further CPUs/Rx queues set.
3703 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3704 unsigned int nr_bits)
3706 /* -1 is a legal arg here. */
3708 cpu_max_bits_warn(n, nr_bits);
3711 return find_next_bit(srcp, nr_bits, n + 1);
3717 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3718 * @n: CPU/Rx queue index
3719 * @src1p: the first CPUs/Rx queues mask pointer
3720 * @src2p: the second CPUs/Rx queues mask pointer
3721 * @nr_bits: number of bits in the bitmask
3723 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3725 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3726 const unsigned long *src2p,
3727 unsigned int nr_bits)
3729 /* -1 is a legal arg here. */
3731 cpu_max_bits_warn(n, nr_bits);
3734 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3736 return find_next_bit(src1p, nr_bits, n + 1);
3738 return find_next_bit(src2p, nr_bits, n + 1);
3743 static inline int netif_set_xps_queue(struct net_device *dev,
3744 const struct cpumask *mask,
3750 static inline int __netif_set_xps_queue(struct net_device *dev,
3751 const unsigned long *mask,
3752 u16 index, bool is_rxqs_map)
3759 * netif_is_multiqueue - test if device has multiple transmit queues
3760 * @dev: network device
3762 * Check if device has multiple transmit queues
3764 static inline bool netif_is_multiqueue(const struct net_device *dev)
3766 return dev->num_tx_queues > 1;
3769 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3772 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3774 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3777 dev->real_num_rx_queues = rxqs;
3782 static inline struct netdev_rx_queue *
3783 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3785 return dev->_rx + rxq;
3789 static inline unsigned int get_netdev_rx_queue_index(
3790 struct netdev_rx_queue *queue)
3792 struct net_device *dev = queue->dev;
3793 int index = queue - dev->_rx;
3795 BUG_ON(index >= dev->num_rx_queues);
3800 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3801 int netif_get_num_default_rss_queues(void);
3803 enum skb_free_reason {
3804 SKB_REASON_CONSUMED,
3808 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3809 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3812 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3813 * interrupt context or with hardware interrupts being disabled.
3814 * (in_irq() || irqs_disabled())
3816 * We provide four helpers that can be used in following contexts :
3818 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3819 * replacing kfree_skb(skb)
3821 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3822 * Typically used in place of consume_skb(skb) in TX completion path
3824 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3825 * replacing kfree_skb(skb)
3827 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3828 * and consumed a packet. Used in place of consume_skb(skb)
3830 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3832 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3835 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3837 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3840 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3842 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3845 static inline void dev_consume_skb_any(struct sk_buff *skb)
3847 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3850 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3851 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3852 int netif_rx(struct sk_buff *skb);
3853 int netif_rx_ni(struct sk_buff *skb);
3854 int netif_rx_any_context(struct sk_buff *skb);
3855 int netif_receive_skb(struct sk_buff *skb);
3856 int netif_receive_skb_core(struct sk_buff *skb);
3857 void netif_receive_skb_list(struct list_head *head);
3858 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3859 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3860 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3861 gro_result_t napi_gro_frags(struct napi_struct *napi);
3862 struct packet_offload *gro_find_receive_by_type(__be16 type);
3863 struct packet_offload *gro_find_complete_by_type(__be16 type);
3865 static inline void napi_free_frags(struct napi_struct *napi)
3867 kfree_skb(napi->skb);
3871 bool netdev_is_rx_handler_busy(struct net_device *dev);
3872 int netdev_rx_handler_register(struct net_device *dev,
3873 rx_handler_func_t *rx_handler,
3874 void *rx_handler_data);
3875 void netdev_rx_handler_unregister(struct net_device *dev);
3877 bool dev_valid_name(const char *name);
3878 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3879 bool *need_copyout);
3880 int dev_ifconf(struct net *net, struct ifconf *, int);
3881 int dev_ethtool(struct net *net, struct ifreq *);
3882 unsigned int dev_get_flags(const struct net_device *);
3883 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3884 struct netlink_ext_ack *extack);
3885 int dev_change_flags(struct net_device *dev, unsigned int flags,
3886 struct netlink_ext_ack *extack);
3887 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3888 unsigned int gchanges);
3889 int dev_change_name(struct net_device *, const char *);
3890 int dev_set_alias(struct net_device *, const char *, size_t);
3891 int dev_get_alias(const struct net_device *, char *, size_t);
3892 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3893 int __dev_set_mtu(struct net_device *, int);
3894 int dev_validate_mtu(struct net_device *dev, int mtu,
3895 struct netlink_ext_ack *extack);
3896 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3897 struct netlink_ext_ack *extack);
3898 int dev_set_mtu(struct net_device *, int);
3899 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3900 void dev_set_group(struct net_device *, int);
3901 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3902 struct netlink_ext_ack *extack);
3903 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3904 struct netlink_ext_ack *extack);
3905 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3906 struct netlink_ext_ack *extack);
3907 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3908 int dev_change_carrier(struct net_device *, bool new_carrier);
3909 int dev_get_phys_port_id(struct net_device *dev,
3910 struct netdev_phys_item_id *ppid);
3911 int dev_get_phys_port_name(struct net_device *dev,
3912 char *name, size_t len);
3913 int dev_get_port_parent_id(struct net_device *dev,
3914 struct netdev_phys_item_id *ppid, bool recurse);
3915 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3916 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3917 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3918 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3920 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3921 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3922 struct netdev_queue *txq, int *ret);
3924 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3925 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3926 int fd, int expected_fd, u32 flags);
3927 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3928 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3930 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3932 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3933 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3934 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3935 bool is_skb_forwardable(const struct net_device *dev,
3936 const struct sk_buff *skb);
3938 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3939 const struct sk_buff *skb,
3940 const bool check_mtu)
3942 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3945 if (!(dev->flags & IFF_UP))
3951 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3952 if (skb->len <= len)
3955 /* if TSO is enabled, we don't care about the length as the packet
3956 * could be forwarded without being segmented before
3958 if (skb_is_gso(skb))
3964 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3965 struct sk_buff *skb,
3966 const bool check_mtu)
3968 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3969 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3970 atomic_long_inc(&dev->rx_dropped);
3975 skb_scrub_packet(skb, true);
3980 bool dev_nit_active(struct net_device *dev);
3981 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3983 extern int netdev_budget;
3984 extern unsigned int netdev_budget_usecs;
3986 /* Called by rtnetlink.c:rtnl_unlock() */
3987 void netdev_run_todo(void);
3990 * dev_put - release reference to device
3991 * @dev: network device
3993 * Release reference to device to allow it to be freed.
3995 static inline void dev_put(struct net_device *dev)
3997 this_cpu_dec(*dev->pcpu_refcnt);
4001 * dev_hold - get reference to device
4002 * @dev: network device
4004 * Hold reference to device to keep it from being freed.
4006 static inline void dev_hold(struct net_device *dev)
4008 this_cpu_inc(*dev->pcpu_refcnt);
4011 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4012 * and _off may be called from IRQ context, but it is caller
4013 * who is responsible for serialization of these calls.
4015 * The name carrier is inappropriate, these functions should really be
4016 * called netif_lowerlayer_*() because they represent the state of any
4017 * kind of lower layer not just hardware media.
4020 void linkwatch_init_dev(struct net_device *dev);
4021 void linkwatch_fire_event(struct net_device *dev);
4022 void linkwatch_forget_dev(struct net_device *dev);
4025 * netif_carrier_ok - test if carrier present
4026 * @dev: network device
4028 * Check if carrier is present on device
4030 static inline bool netif_carrier_ok(const struct net_device *dev)
4032 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4035 unsigned long dev_trans_start(struct net_device *dev);
4037 void __netdev_watchdog_up(struct net_device *dev);
4039 void netif_carrier_on(struct net_device *dev);
4041 void netif_carrier_off(struct net_device *dev);
4044 * netif_dormant_on - mark device as dormant.
4045 * @dev: network device
4047 * Mark device as dormant (as per RFC2863).
4049 * The dormant state indicates that the relevant interface is not
4050 * actually in a condition to pass packets (i.e., it is not 'up') but is
4051 * in a "pending" state, waiting for some external event. For "on-
4052 * demand" interfaces, this new state identifies the situation where the
4053 * interface is waiting for events to place it in the up state.
4055 static inline void netif_dormant_on(struct net_device *dev)
4057 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4058 linkwatch_fire_event(dev);
4062 * netif_dormant_off - set device as not dormant.
4063 * @dev: network device
4065 * Device is not in dormant state.
4067 static inline void netif_dormant_off(struct net_device *dev)
4069 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4070 linkwatch_fire_event(dev);
4074 * netif_dormant - test if device is dormant
4075 * @dev: network device
4077 * Check if device is dormant.
4079 static inline bool netif_dormant(const struct net_device *dev)
4081 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4086 * netif_testing_on - mark device as under test.
4087 * @dev: network device
4089 * Mark device as under test (as per RFC2863).
4091 * The testing state indicates that some test(s) must be performed on
4092 * the interface. After completion, of the test, the interface state
4093 * will change to up, dormant, or down, as appropriate.
4095 static inline void netif_testing_on(struct net_device *dev)
4097 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4098 linkwatch_fire_event(dev);
4102 * netif_testing_off - set device as not under test.
4103 * @dev: network device
4105 * Device is not in testing state.
4107 static inline void netif_testing_off(struct net_device *dev)
4109 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4110 linkwatch_fire_event(dev);
4114 * netif_testing - test if device is under test
4115 * @dev: network device
4117 * Check if device is under test
4119 static inline bool netif_testing(const struct net_device *dev)
4121 return test_bit(__LINK_STATE_TESTING, &dev->state);
4126 * netif_oper_up - test if device is operational
4127 * @dev: network device
4129 * Check if carrier is operational
4131 static inline bool netif_oper_up(const struct net_device *dev)
4133 return (dev->operstate == IF_OPER_UP ||
4134 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4138 * netif_device_present - is device available or removed
4139 * @dev: network device
4141 * Check if device has not been removed from system.
4143 static inline bool netif_device_present(struct net_device *dev)
4145 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4148 void netif_device_detach(struct net_device *dev);
4150 void netif_device_attach(struct net_device *dev);
4153 * Network interface message level settings
4158 NETIF_MSG_PROBE_BIT,
4160 NETIF_MSG_TIMER_BIT,
4161 NETIF_MSG_IFDOWN_BIT,
4163 NETIF_MSG_RX_ERR_BIT,
4164 NETIF_MSG_TX_ERR_BIT,
4165 NETIF_MSG_TX_QUEUED_BIT,
4167 NETIF_MSG_TX_DONE_BIT,
4168 NETIF_MSG_RX_STATUS_BIT,
4169 NETIF_MSG_PKTDATA_BIT,
4173 /* When you add a new bit above, update netif_msg_class_names array
4174 * in net/ethtool/common.c
4176 NETIF_MSG_CLASS_COUNT,
4178 /* Both ethtool_ops interface and internal driver implementation use u32 */
4179 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4181 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4182 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4184 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4185 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4186 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4187 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4188 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4189 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4190 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4191 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4192 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4193 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4194 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4195 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4196 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4197 #define NETIF_MSG_HW __NETIF_MSG(HW)
4198 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4200 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4201 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4202 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4203 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4204 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4205 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4206 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4207 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4208 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4209 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4210 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4211 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4212 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4213 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4214 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4216 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4219 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4220 return default_msg_enable_bits;
4221 if (debug_value == 0) /* no output */
4223 /* set low N bits */
4224 return (1U << debug_value) - 1;
4227 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4229 spin_lock(&txq->_xmit_lock);
4230 txq->xmit_lock_owner = cpu;
4233 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4235 __acquire(&txq->_xmit_lock);
4239 static inline void __netif_tx_release(struct netdev_queue *txq)
4241 __release(&txq->_xmit_lock);
4244 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4246 spin_lock_bh(&txq->_xmit_lock);
4247 txq->xmit_lock_owner = smp_processor_id();
4250 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4252 bool ok = spin_trylock(&txq->_xmit_lock);
4254 txq->xmit_lock_owner = smp_processor_id();
4258 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4260 txq->xmit_lock_owner = -1;
4261 spin_unlock(&txq->_xmit_lock);
4264 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4266 txq->xmit_lock_owner = -1;
4267 spin_unlock_bh(&txq->_xmit_lock);
4270 static inline void txq_trans_update(struct netdev_queue *txq)
4272 if (txq->xmit_lock_owner != -1)
4273 txq->trans_start = jiffies;
4276 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4277 static inline void netif_trans_update(struct net_device *dev)
4279 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4281 if (txq->trans_start != jiffies)
4282 txq->trans_start = jiffies;
4286 * netif_tx_lock - grab network device transmit lock
4287 * @dev: network device
4289 * Get network device transmit lock
4291 static inline void netif_tx_lock(struct net_device *dev)
4296 spin_lock(&dev->tx_global_lock);
4297 cpu = smp_processor_id();
4298 for (i = 0; i < dev->num_tx_queues; i++) {
4299 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4301 /* We are the only thread of execution doing a
4302 * freeze, but we have to grab the _xmit_lock in
4303 * order to synchronize with threads which are in
4304 * the ->hard_start_xmit() handler and already
4305 * checked the frozen bit.
4307 __netif_tx_lock(txq, cpu);
4308 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4309 __netif_tx_unlock(txq);
4313 static inline void netif_tx_lock_bh(struct net_device *dev)
4319 static inline void netif_tx_unlock(struct net_device *dev)
4323 for (i = 0; i < dev->num_tx_queues; i++) {
4324 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4326 /* No need to grab the _xmit_lock here. If the
4327 * queue is not stopped for another reason, we
4330 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4331 netif_schedule_queue(txq);
4333 spin_unlock(&dev->tx_global_lock);
4336 static inline void netif_tx_unlock_bh(struct net_device *dev)
4338 netif_tx_unlock(dev);
4342 #define HARD_TX_LOCK(dev, txq, cpu) { \
4343 if ((dev->features & NETIF_F_LLTX) == 0) { \
4344 __netif_tx_lock(txq, cpu); \
4346 __netif_tx_acquire(txq); \
4350 #define HARD_TX_TRYLOCK(dev, txq) \
4351 (((dev->features & NETIF_F_LLTX) == 0) ? \
4352 __netif_tx_trylock(txq) : \
4353 __netif_tx_acquire(txq))
4355 #define HARD_TX_UNLOCK(dev, txq) { \
4356 if ((dev->features & NETIF_F_LLTX) == 0) { \
4357 __netif_tx_unlock(txq); \
4359 __netif_tx_release(txq); \
4363 static inline void netif_tx_disable(struct net_device *dev)
4369 cpu = smp_processor_id();
4370 spin_lock(&dev->tx_global_lock);
4371 for (i = 0; i < dev->num_tx_queues; i++) {
4372 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4374 __netif_tx_lock(txq, cpu);
4375 netif_tx_stop_queue(txq);
4376 __netif_tx_unlock(txq);
4378 spin_unlock(&dev->tx_global_lock);
4382 static inline void netif_addr_lock(struct net_device *dev)
4384 unsigned char nest_level = 0;
4386 #ifdef CONFIG_LOCKDEP
4387 nest_level = dev->nested_level;
4389 spin_lock_nested(&dev->addr_list_lock, nest_level);
4392 static inline void netif_addr_lock_bh(struct net_device *dev)
4394 unsigned char nest_level = 0;
4396 #ifdef CONFIG_LOCKDEP
4397 nest_level = dev->nested_level;
4400 spin_lock_nested(&dev->addr_list_lock, nest_level);
4403 static inline void netif_addr_unlock(struct net_device *dev)
4405 spin_unlock(&dev->addr_list_lock);
4408 static inline void netif_addr_unlock_bh(struct net_device *dev)
4410 spin_unlock_bh(&dev->addr_list_lock);
4414 * dev_addrs walker. Should be used only for read access. Call with
4415 * rcu_read_lock held.
4417 #define for_each_dev_addr(dev, ha) \
4418 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4420 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4422 void ether_setup(struct net_device *dev);
4424 /* Support for loadable net-drivers */
4425 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4426 unsigned char name_assign_type,
4427 void (*setup)(struct net_device *),
4428 unsigned int txqs, unsigned int rxqs);
4429 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4430 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4432 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4433 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4436 int register_netdev(struct net_device *dev);
4437 void unregister_netdev(struct net_device *dev);
4439 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4441 /* General hardware address lists handling functions */
4442 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4443 struct netdev_hw_addr_list *from_list, int addr_len);
4444 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4445 struct netdev_hw_addr_list *from_list, int addr_len);
4446 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4447 struct net_device *dev,
4448 int (*sync)(struct net_device *, const unsigned char *),
4449 int (*unsync)(struct net_device *,
4450 const unsigned char *));
4451 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4452 struct net_device *dev,
4453 int (*sync)(struct net_device *,
4454 const unsigned char *, int),
4455 int (*unsync)(struct net_device *,
4456 const unsigned char *, int));
4457 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4458 struct net_device *dev,
4459 int (*unsync)(struct net_device *,
4460 const unsigned char *, int));
4461 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4462 struct net_device *dev,
4463 int (*unsync)(struct net_device *,
4464 const unsigned char *));
4465 void __hw_addr_init(struct netdev_hw_addr_list *list);
4467 /* Functions used for device addresses handling */
4468 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4469 unsigned char addr_type);
4470 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4471 unsigned char addr_type);
4472 void dev_addr_flush(struct net_device *dev);
4473 int dev_addr_init(struct net_device *dev);
4475 /* Functions used for unicast addresses handling */
4476 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4477 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4478 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4479 int dev_uc_sync(struct net_device *to, struct net_device *from);
4480 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4481 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4482 void dev_uc_flush(struct net_device *dev);
4483 void dev_uc_init(struct net_device *dev);
4486 * __dev_uc_sync - Synchonize device's unicast list
4487 * @dev: device to sync
4488 * @sync: function to call if address should be added
4489 * @unsync: function to call if address should be removed
4491 * Add newly added addresses to the interface, and release
4492 * addresses that have been deleted.
4494 static inline int __dev_uc_sync(struct net_device *dev,
4495 int (*sync)(struct net_device *,
4496 const unsigned char *),
4497 int (*unsync)(struct net_device *,
4498 const unsigned char *))
4500 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4504 * __dev_uc_unsync - Remove synchronized addresses from device
4505 * @dev: device to sync
4506 * @unsync: function to call if address should be removed
4508 * Remove all addresses that were added to the device by dev_uc_sync().
4510 static inline void __dev_uc_unsync(struct net_device *dev,
4511 int (*unsync)(struct net_device *,
4512 const unsigned char *))
4514 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4517 /* Functions used for multicast addresses handling */
4518 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4519 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4520 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4521 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4522 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4523 int dev_mc_sync(struct net_device *to, struct net_device *from);
4524 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4525 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4526 void dev_mc_flush(struct net_device *dev);
4527 void dev_mc_init(struct net_device *dev);
4530 * __dev_mc_sync - Synchonize device's multicast list
4531 * @dev: device to sync
4532 * @sync: function to call if address should be added
4533 * @unsync: function to call if address should be removed
4535 * Add newly added addresses to the interface, and release
4536 * addresses that have been deleted.
4538 static inline int __dev_mc_sync(struct net_device *dev,
4539 int (*sync)(struct net_device *,
4540 const unsigned char *),
4541 int (*unsync)(struct net_device *,
4542 const unsigned char *))
4544 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4548 * __dev_mc_unsync - Remove synchronized addresses from device
4549 * @dev: device to sync
4550 * @unsync: function to call if address should be removed
4552 * Remove all addresses that were added to the device by dev_mc_sync().
4554 static inline void __dev_mc_unsync(struct net_device *dev,
4555 int (*unsync)(struct net_device *,
4556 const unsigned char *))
4558 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4561 /* Functions used for secondary unicast and multicast support */
4562 void dev_set_rx_mode(struct net_device *dev);
4563 void __dev_set_rx_mode(struct net_device *dev);
4564 int dev_set_promiscuity(struct net_device *dev, int inc);
4565 int dev_set_allmulti(struct net_device *dev, int inc);
4566 void netdev_state_change(struct net_device *dev);
4567 void __netdev_notify_peers(struct net_device *dev);
4568 void netdev_notify_peers(struct net_device *dev);
4569 void netdev_features_change(struct net_device *dev);
4570 /* Load a device via the kmod */
4571 void dev_load(struct net *net, const char *name);
4572 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4573 struct rtnl_link_stats64 *storage);
4574 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4575 const struct net_device_stats *netdev_stats);
4576 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4577 const struct pcpu_sw_netstats __percpu *netstats);
4578 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4580 extern int netdev_max_backlog;
4581 extern int netdev_tstamp_prequeue;
4582 extern int weight_p;
4583 extern int dev_weight_rx_bias;
4584 extern int dev_weight_tx_bias;
4585 extern int dev_rx_weight;
4586 extern int dev_tx_weight;
4587 extern int gro_normal_batch;
4590 NESTED_SYNC_IMM_BIT,
4591 NESTED_SYNC_TODO_BIT,
4594 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4595 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4597 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4598 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4600 struct netdev_nested_priv {
4601 unsigned char flags;
4605 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4606 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4607 struct list_head **iter);
4608 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4609 struct list_head **iter);
4611 #ifdef CONFIG_LOCKDEP
4612 static LIST_HEAD(net_unlink_list);
4614 static inline void net_unlink_todo(struct net_device *dev)
4616 if (list_empty(&dev->unlink_list))
4617 list_add_tail(&dev->unlink_list, &net_unlink_list);
4621 /* iterate through upper list, must be called under RCU read lock */
4622 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4623 for (iter = &(dev)->adj_list.upper, \
4624 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4626 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4628 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4629 int (*fn)(struct net_device *upper_dev,
4630 struct netdev_nested_priv *priv),
4631 struct netdev_nested_priv *priv);
4633 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4634 struct net_device *upper_dev);
4636 bool netdev_has_any_upper_dev(struct net_device *dev);
4638 void *netdev_lower_get_next_private(struct net_device *dev,
4639 struct list_head **iter);
4640 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4641 struct list_head **iter);
4643 #define netdev_for_each_lower_private(dev, priv, iter) \
4644 for (iter = (dev)->adj_list.lower.next, \
4645 priv = netdev_lower_get_next_private(dev, &(iter)); \
4647 priv = netdev_lower_get_next_private(dev, &(iter)))
4649 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4650 for (iter = &(dev)->adj_list.lower, \
4651 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4653 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4655 void *netdev_lower_get_next(struct net_device *dev,
4656 struct list_head **iter);
4658 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4659 for (iter = (dev)->adj_list.lower.next, \
4660 ldev = netdev_lower_get_next(dev, &(iter)); \
4662 ldev = netdev_lower_get_next(dev, &(iter)))
4664 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4665 struct list_head **iter);
4666 int netdev_walk_all_lower_dev(struct net_device *dev,
4667 int (*fn)(struct net_device *lower_dev,
4668 struct netdev_nested_priv *priv),
4669 struct netdev_nested_priv *priv);
4670 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4671 int (*fn)(struct net_device *lower_dev,
4672 struct netdev_nested_priv *priv),
4673 struct netdev_nested_priv *priv);
4675 void *netdev_adjacent_get_private(struct list_head *adj_list);
4676 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4677 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4678 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4679 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4680 struct netlink_ext_ack *extack);
4681 int netdev_master_upper_dev_link(struct net_device *dev,
4682 struct net_device *upper_dev,
4683 void *upper_priv, void *upper_info,
4684 struct netlink_ext_ack *extack);
4685 void netdev_upper_dev_unlink(struct net_device *dev,
4686 struct net_device *upper_dev);
4687 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4688 struct net_device *new_dev,
4689 struct net_device *dev,
4690 struct netlink_ext_ack *extack);
4691 void netdev_adjacent_change_commit(struct net_device *old_dev,
4692 struct net_device *new_dev,
4693 struct net_device *dev);
4694 void netdev_adjacent_change_abort(struct net_device *old_dev,
4695 struct net_device *new_dev,
4696 struct net_device *dev);
4697 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4698 void *netdev_lower_dev_get_private(struct net_device *dev,
4699 struct net_device *lower_dev);
4700 void netdev_lower_state_changed(struct net_device *lower_dev,
4701 void *lower_state_info);
4703 /* RSS keys are 40 or 52 bytes long */
4704 #define NETDEV_RSS_KEY_LEN 52
4705 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4706 void netdev_rss_key_fill(void *buffer, size_t len);
4708 int skb_checksum_help(struct sk_buff *skb);
4709 int skb_crc32c_csum_help(struct sk_buff *skb);
4710 int skb_csum_hwoffload_help(struct sk_buff *skb,
4711 const netdev_features_t features);
4713 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4714 netdev_features_t features, bool tx_path);
4715 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4716 netdev_features_t features);
4718 struct netdev_bonding_info {
4723 struct netdev_notifier_bonding_info {
4724 struct netdev_notifier_info info; /* must be first */
4725 struct netdev_bonding_info bonding_info;
4728 void netdev_bonding_info_change(struct net_device *dev,
4729 struct netdev_bonding_info *bonding_info);
4731 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4732 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4734 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4741 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4743 return __skb_gso_segment(skb, features, true);
4745 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4747 static inline bool can_checksum_protocol(netdev_features_t features,
4750 if (protocol == htons(ETH_P_FCOE))
4751 return !!(features & NETIF_F_FCOE_CRC);
4753 /* Assume this is an IP checksum (not SCTP CRC) */
4755 if (features & NETIF_F_HW_CSUM) {
4756 /* Can checksum everything */
4761 case htons(ETH_P_IP):
4762 return !!(features & NETIF_F_IP_CSUM);
4763 case htons(ETH_P_IPV6):
4764 return !!(features & NETIF_F_IPV6_CSUM);
4771 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4773 static inline void netdev_rx_csum_fault(struct net_device *dev,
4774 struct sk_buff *skb)
4778 /* rx skb timestamps */
4779 void net_enable_timestamp(void);
4780 void net_disable_timestamp(void);
4782 #ifdef CONFIG_PROC_FS
4783 int __init dev_proc_init(void);
4785 #define dev_proc_init() 0
4788 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4789 struct sk_buff *skb, struct net_device *dev,
4792 __this_cpu_write(softnet_data.xmit.more, more);
4793 return ops->ndo_start_xmit(skb, dev);
4796 static inline bool netdev_xmit_more(void)
4798 return __this_cpu_read(softnet_data.xmit.more);
4801 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4802 struct netdev_queue *txq, bool more)
4804 const struct net_device_ops *ops = dev->netdev_ops;
4807 rc = __netdev_start_xmit(ops, skb, dev, more);
4808 if (rc == NETDEV_TX_OK)
4809 txq_trans_update(txq);
4814 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4816 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4819 extern const struct kobj_ns_type_operations net_ns_type_operations;
4821 const char *netdev_drivername(const struct net_device *dev);
4823 void linkwatch_run_queue(void);
4825 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4826 netdev_features_t f2)
4828 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4829 if (f1 & NETIF_F_HW_CSUM)
4830 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4832 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4838 static inline netdev_features_t netdev_get_wanted_features(
4839 struct net_device *dev)
4841 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4843 netdev_features_t netdev_increment_features(netdev_features_t all,
4844 netdev_features_t one, netdev_features_t mask);
4846 /* Allow TSO being used on stacked device :
4847 * Performing the GSO segmentation before last device
4848 * is a performance improvement.
4850 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4851 netdev_features_t mask)
4853 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4856 int __netdev_update_features(struct net_device *dev);
4857 void netdev_update_features(struct net_device *dev);
4858 void netdev_change_features(struct net_device *dev);
4860 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4861 struct net_device *dev);
4863 netdev_features_t passthru_features_check(struct sk_buff *skb,
4864 struct net_device *dev,
4865 netdev_features_t features);
4866 netdev_features_t netif_skb_features(struct sk_buff *skb);
4868 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4870 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4872 /* check flags correspondence */
4873 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4874 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4875 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4876 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4877 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4878 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4879 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4880 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4881 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4882 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4883 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4884 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4885 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4886 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4887 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4888 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4889 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4890 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4891 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4893 return (features & feature) == feature;
4896 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4898 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4899 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4902 static inline bool netif_needs_gso(struct sk_buff *skb,
4903 netdev_features_t features)
4905 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4906 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4907 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4910 static inline void netif_set_gso_max_size(struct net_device *dev,
4913 dev->gso_max_size = size;
4916 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4917 int pulled_hlen, u16 mac_offset,
4920 skb->protocol = protocol;
4921 skb->encapsulation = 1;
4922 skb_push(skb, pulled_hlen);
4923 skb_reset_transport_header(skb);
4924 skb->mac_header = mac_offset;
4925 skb->network_header = skb->mac_header + mac_len;
4926 skb->mac_len = mac_len;
4929 static inline bool netif_is_macsec(const struct net_device *dev)
4931 return dev->priv_flags & IFF_MACSEC;
4934 static inline bool netif_is_macvlan(const struct net_device *dev)
4936 return dev->priv_flags & IFF_MACVLAN;
4939 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4941 return dev->priv_flags & IFF_MACVLAN_PORT;
4944 static inline bool netif_is_bond_master(const struct net_device *dev)
4946 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4949 static inline bool netif_is_bond_slave(const struct net_device *dev)
4951 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4954 static inline bool netif_supports_nofcs(struct net_device *dev)
4956 return dev->priv_flags & IFF_SUPP_NOFCS;
4959 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4961 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4964 static inline bool netif_is_l3_master(const struct net_device *dev)
4966 return dev->priv_flags & IFF_L3MDEV_MASTER;
4969 static inline bool netif_is_l3_slave(const struct net_device *dev)
4971 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4974 static inline bool netif_is_bridge_master(const struct net_device *dev)
4976 return dev->priv_flags & IFF_EBRIDGE;
4979 static inline bool netif_is_bridge_port(const struct net_device *dev)
4981 return dev->priv_flags & IFF_BRIDGE_PORT;
4984 static inline bool netif_is_ovs_master(const struct net_device *dev)
4986 return dev->priv_flags & IFF_OPENVSWITCH;
4989 static inline bool netif_is_ovs_port(const struct net_device *dev)
4991 return dev->priv_flags & IFF_OVS_DATAPATH;
4994 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4996 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4999 static inline bool netif_is_team_master(const struct net_device *dev)
5001 return dev->priv_flags & IFF_TEAM;
5004 static inline bool netif_is_team_port(const struct net_device *dev)
5006 return dev->priv_flags & IFF_TEAM_PORT;
5009 static inline bool netif_is_lag_master(const struct net_device *dev)
5011 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5014 static inline bool netif_is_lag_port(const struct net_device *dev)
5016 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5019 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5021 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5024 static inline bool netif_is_failover(const struct net_device *dev)
5026 return dev->priv_flags & IFF_FAILOVER;
5029 static inline bool netif_is_failover_slave(const struct net_device *dev)
5031 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5034 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5035 static inline void netif_keep_dst(struct net_device *dev)
5037 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5040 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5041 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5043 /* TODO: reserve and use an additional IFF bit, if we get more users */
5044 return dev->priv_flags & IFF_MACSEC;
5047 extern struct pernet_operations __net_initdata loopback_net_ops;
5049 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5051 /* netdev_printk helpers, similar to dev_printk */
5053 static inline const char *netdev_name(const struct net_device *dev)
5055 if (!dev->name[0] || strchr(dev->name, '%'))
5056 return "(unnamed net_device)";
5060 static inline bool netdev_unregistering(const struct net_device *dev)
5062 return dev->reg_state == NETREG_UNREGISTERING;
5065 static inline const char *netdev_reg_state(const struct net_device *dev)
5067 switch (dev->reg_state) {
5068 case NETREG_UNINITIALIZED: return " (uninitialized)";
5069 case NETREG_REGISTERED: return "";
5070 case NETREG_UNREGISTERING: return " (unregistering)";
5071 case NETREG_UNREGISTERED: return " (unregistered)";
5072 case NETREG_RELEASED: return " (released)";
5073 case NETREG_DUMMY: return " (dummy)";
5076 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5077 return " (unknown)";
5080 __printf(3, 4) __cold
5081 void netdev_printk(const char *level, const struct net_device *dev,
5082 const char *format, ...);
5083 __printf(2, 3) __cold
5084 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5085 __printf(2, 3) __cold
5086 void netdev_alert(const struct net_device *dev, const char *format, ...);
5087 __printf(2, 3) __cold
5088 void netdev_crit(const struct net_device *dev, const char *format, ...);
5089 __printf(2, 3) __cold
5090 void netdev_err(const struct net_device *dev, const char *format, ...);
5091 __printf(2, 3) __cold
5092 void netdev_warn(const struct net_device *dev, const char *format, ...);
5093 __printf(2, 3) __cold
5094 void netdev_notice(const struct net_device *dev, const char *format, ...);
5095 __printf(2, 3) __cold
5096 void netdev_info(const struct net_device *dev, const char *format, ...);
5098 #define netdev_level_once(level, dev, fmt, ...) \
5100 static bool __print_once __read_mostly; \
5102 if (!__print_once) { \
5103 __print_once = true; \
5104 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5108 #define netdev_emerg_once(dev, fmt, ...) \
5109 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5110 #define netdev_alert_once(dev, fmt, ...) \
5111 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5112 #define netdev_crit_once(dev, fmt, ...) \
5113 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5114 #define netdev_err_once(dev, fmt, ...) \
5115 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5116 #define netdev_warn_once(dev, fmt, ...) \
5117 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5118 #define netdev_notice_once(dev, fmt, ...) \
5119 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5120 #define netdev_info_once(dev, fmt, ...) \
5121 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5123 #define MODULE_ALIAS_NETDEV(device) \
5124 MODULE_ALIAS("netdev-" device)
5126 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5127 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5128 #define netdev_dbg(__dev, format, args...) \
5130 dynamic_netdev_dbg(__dev, format, ##args); \
5132 #elif defined(DEBUG)
5133 #define netdev_dbg(__dev, format, args...) \
5134 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5136 #define netdev_dbg(__dev, format, args...) \
5139 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5143 #if defined(VERBOSE_DEBUG)
5144 #define netdev_vdbg netdev_dbg
5147 #define netdev_vdbg(dev, format, args...) \
5150 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5156 * netdev_WARN() acts like dev_printk(), but with the key difference
5157 * of using a WARN/WARN_ON to get the message out, including the
5158 * file/line information and a backtrace.
5160 #define netdev_WARN(dev, format, args...) \
5161 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5162 netdev_reg_state(dev), ##args)
5164 #define netdev_WARN_ONCE(dev, format, args...) \
5165 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5166 netdev_reg_state(dev), ##args)
5168 /* netif printk helpers, similar to netdev_printk */
5170 #define netif_printk(priv, type, level, dev, fmt, args...) \
5172 if (netif_msg_##type(priv)) \
5173 netdev_printk(level, (dev), fmt, ##args); \
5176 #define netif_level(level, priv, type, dev, fmt, args...) \
5178 if (netif_msg_##type(priv)) \
5179 netdev_##level(dev, fmt, ##args); \
5182 #define netif_emerg(priv, type, dev, fmt, args...) \
5183 netif_level(emerg, priv, type, dev, fmt, ##args)
5184 #define netif_alert(priv, type, dev, fmt, args...) \
5185 netif_level(alert, priv, type, dev, fmt, ##args)
5186 #define netif_crit(priv, type, dev, fmt, args...) \
5187 netif_level(crit, priv, type, dev, fmt, ##args)
5188 #define netif_err(priv, type, dev, fmt, args...) \
5189 netif_level(err, priv, type, dev, fmt, ##args)
5190 #define netif_warn(priv, type, dev, fmt, args...) \
5191 netif_level(warn, priv, type, dev, fmt, ##args)
5192 #define netif_notice(priv, type, dev, fmt, args...) \
5193 netif_level(notice, priv, type, dev, fmt, ##args)
5194 #define netif_info(priv, type, dev, fmt, args...) \
5195 netif_level(info, priv, type, dev, fmt, ##args)
5197 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5198 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5199 #define netif_dbg(priv, type, netdev, format, args...) \
5201 if (netif_msg_##type(priv)) \
5202 dynamic_netdev_dbg(netdev, format, ##args); \
5204 #elif defined(DEBUG)
5205 #define netif_dbg(priv, type, dev, format, args...) \
5206 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5208 #define netif_dbg(priv, type, dev, format, args...) \
5211 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5216 /* if @cond then downgrade to debug, else print at @level */
5217 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5220 netif_dbg(priv, type, netdev, fmt, ##args); \
5222 netif_ ## level(priv, type, netdev, fmt, ##args); \
5225 #if defined(VERBOSE_DEBUG)
5226 #define netif_vdbg netif_dbg
5228 #define netif_vdbg(priv, type, dev, format, args...) \
5231 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5237 * The list of packet types we will receive (as opposed to discard)
5238 * and the routines to invoke.
5240 * Why 16. Because with 16 the only overlap we get on a hash of the
5241 * low nibble of the protocol value is RARP/SNAP/X.25.
5255 #define PTYPE_HASH_SIZE (16)
5256 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5258 extern struct net_device *blackhole_netdev;
5260 #endif /* _LINUX_NETDEVICE_H */