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 * We keep track of the number of traffic classes used when the struct is
776 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
777 * not crossing its upper bound, as the original dev->num_tc can be updated in
780 struct xps_dev_maps {
783 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
786 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
787 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
789 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
790 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
792 #endif /* CONFIG_XPS */
794 #define TC_MAX_QUEUE 16
795 #define TC_BITMASK 15
796 /* HW offloaded queuing disciplines txq count and offset maps */
797 struct netdev_tc_txq {
802 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
804 * This structure is to hold information about the device
805 * configured to run FCoE protocol stack.
807 struct netdev_fcoe_hbainfo {
808 char manufacturer[64];
809 char serial_number[64];
810 char hardware_version[64];
811 char driver_version[64];
812 char optionrom_version[64];
813 char firmware_version[64];
815 char model_description[256];
819 #define MAX_PHYS_ITEM_ID_LEN 32
821 /* This structure holds a unique identifier to identify some
822 * physical item (port for example) used by a netdevice.
824 struct netdev_phys_item_id {
825 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
826 unsigned char id_len;
829 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
830 struct netdev_phys_item_id *b)
832 return a->id_len == b->id_len &&
833 memcmp(a->id, b->id, a->id_len) == 0;
836 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
838 struct net_device *sb_dev);
841 TC_SETUP_QDISC_MQPRIO,
844 TC_SETUP_CLSMATCHALL,
854 TC_SETUP_QDISC_TAPRIO,
862 /* These structures hold the attributes of bpf state that are being passed
863 * to the netdevice through the bpf op.
865 enum bpf_netdev_command {
866 /* Set or clear a bpf program used in the earliest stages of packet
867 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
868 * is responsible for calling bpf_prog_put on any old progs that are
869 * stored. In case of error, the callee need not release the new prog
870 * reference, but on success it takes ownership and must bpf_prog_put
871 * when it is no longer used.
875 /* BPF program for offload callbacks, invoked at program load time. */
876 BPF_OFFLOAD_MAP_ALLOC,
877 BPF_OFFLOAD_MAP_FREE,
881 struct bpf_prog_offload_ops;
882 struct netlink_ext_ack;
884 struct xdp_dev_bulk_queue;
894 struct bpf_xdp_entity {
895 struct bpf_prog *prog;
896 struct bpf_xdp_link *link;
900 enum bpf_netdev_command command;
905 struct bpf_prog *prog;
906 struct netlink_ext_ack *extack;
908 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
910 struct bpf_offloaded_map *offmap;
912 /* XDP_SETUP_XSK_POOL */
914 struct xsk_buff_pool *pool;
920 /* Flags for ndo_xsk_wakeup. */
921 #define XDP_WAKEUP_RX (1 << 0)
922 #define XDP_WAKEUP_TX (1 << 1)
924 #ifdef CONFIG_XFRM_OFFLOAD
926 int (*xdo_dev_state_add) (struct xfrm_state *x);
927 void (*xdo_dev_state_delete) (struct xfrm_state *x);
928 void (*xdo_dev_state_free) (struct xfrm_state *x);
929 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
930 struct xfrm_state *x);
931 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
936 struct rcu_head rcuhead;
943 struct netdev_name_node {
944 struct hlist_node hlist;
945 struct list_head list;
946 struct net_device *dev;
950 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
951 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
953 struct netdev_net_notifier {
954 struct list_head list;
955 struct notifier_block *nb;
959 * This structure defines the management hooks for network devices.
960 * The following hooks can be defined; unless noted otherwise, they are
961 * optional and can be filled with a null pointer.
963 * int (*ndo_init)(struct net_device *dev);
964 * This function is called once when a network device is registered.
965 * The network device can use this for any late stage initialization
966 * or semantic validation. It can fail with an error code which will
967 * be propagated back to register_netdev.
969 * void (*ndo_uninit)(struct net_device *dev);
970 * This function is called when device is unregistered or when registration
971 * fails. It is not called if init fails.
973 * int (*ndo_open)(struct net_device *dev);
974 * This function is called when a network device transitions to the up
977 * int (*ndo_stop)(struct net_device *dev);
978 * This function is called when a network device transitions to the down
981 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
982 * struct net_device *dev);
983 * Called when a packet needs to be transmitted.
984 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
985 * the queue before that can happen; it's for obsolete devices and weird
986 * corner cases, but the stack really does a non-trivial amount
987 * of useless work if you return NETDEV_TX_BUSY.
988 * Required; cannot be NULL.
990 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
991 * struct net_device *dev
992 * netdev_features_t features);
993 * Called by core transmit path to determine if device is capable of
994 * performing offload operations on a given packet. This is to give
995 * the device an opportunity to implement any restrictions that cannot
996 * be otherwise expressed by feature flags. The check is called with
997 * the set of features that the stack has calculated and it returns
998 * those the driver believes to be appropriate.
1000 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1001 * struct net_device *sb_dev);
1002 * Called to decide which queue to use when device supports multiple
1005 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1006 * This function is called to allow device receiver to make
1007 * changes to configuration when multicast or promiscuous is enabled.
1009 * void (*ndo_set_rx_mode)(struct net_device *dev);
1010 * This function is called device changes address list filtering.
1011 * If driver handles unicast address filtering, it should set
1012 * IFF_UNICAST_FLT in its priv_flags.
1014 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1015 * This function is called when the Media Access Control address
1016 * needs to be changed. If this interface is not defined, the
1017 * MAC address can not be changed.
1019 * int (*ndo_validate_addr)(struct net_device *dev);
1020 * Test if Media Access Control address is valid for the device.
1022 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1023 * Called when a user requests an ioctl which can't be handled by
1024 * the generic interface code. If not defined ioctls return
1025 * not supported error code.
1027 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1028 * Used to set network devices bus interface parameters. This interface
1029 * is retained for legacy reasons; new devices should use the bus
1030 * interface (PCI) for low level management.
1032 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1033 * Called when a user wants to change the Maximum Transfer Unit
1036 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1037 * Callback used when the transmitter has not made any progress
1038 * for dev->watchdog ticks.
1040 * void (*ndo_get_stats64)(struct net_device *dev,
1041 * struct rtnl_link_stats64 *storage);
1042 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1043 * Called when a user wants to get the network device usage
1044 * statistics. Drivers must do one of the following:
1045 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1046 * rtnl_link_stats64 structure passed by the caller.
1047 * 2. Define @ndo_get_stats to update a net_device_stats structure
1048 * (which should normally be dev->stats) and return a pointer to
1049 * it. The structure may be changed asynchronously only if each
1050 * field is written atomically.
1051 * 3. Update dev->stats asynchronously and atomically, and define
1052 * neither operation.
1054 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1055 * Return true if this device supports offload stats of this attr_id.
1057 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1059 * Get statistics for offload operations by attr_id. Write it into the
1060 * attr_data pointer.
1062 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1063 * If device supports VLAN filtering this function is called when a
1064 * VLAN id is registered.
1066 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1067 * If device supports VLAN filtering this function is called when a
1068 * VLAN id is unregistered.
1070 * void (*ndo_poll_controller)(struct net_device *dev);
1072 * SR-IOV management functions.
1073 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1074 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1075 * u8 qos, __be16 proto);
1076 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1078 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1079 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1080 * int (*ndo_get_vf_config)(struct net_device *dev,
1081 * int vf, struct ifla_vf_info *ivf);
1082 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1083 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1084 * struct nlattr *port[]);
1086 * Enable or disable the VF ability to query its RSS Redirection Table and
1087 * Hash Key. This is needed since on some devices VF share this information
1088 * with PF and querying it may introduce a theoretical security risk.
1089 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1090 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1091 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1093 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1094 * This is always called from the stack with the rtnl lock held and netif
1095 * tx queues stopped. This allows the netdevice to perform queue
1096 * management safely.
1098 * Fiber Channel over Ethernet (FCoE) offload functions.
1099 * int (*ndo_fcoe_enable)(struct net_device *dev);
1100 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1101 * so the underlying device can perform whatever needed configuration or
1102 * initialization to support acceleration of FCoE traffic.
1104 * int (*ndo_fcoe_disable)(struct net_device *dev);
1105 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1106 * so the underlying device can perform whatever needed clean-ups to
1107 * stop supporting acceleration of FCoE traffic.
1109 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1110 * struct scatterlist *sgl, unsigned int sgc);
1111 * Called when the FCoE Initiator wants to initialize an I/O that
1112 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1113 * perform necessary setup and returns 1 to indicate the device is set up
1114 * successfully to perform DDP on this I/O, otherwise this returns 0.
1116 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1117 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1118 * indicated by the FC exchange id 'xid', so the underlying device can
1119 * clean up and reuse resources for later DDP requests.
1121 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1122 * struct scatterlist *sgl, unsigned int sgc);
1123 * Called when the FCoE Target wants to initialize an I/O that
1124 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1125 * perform necessary setup and returns 1 to indicate the device is set up
1126 * successfully to perform DDP on this I/O, otherwise this returns 0.
1128 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1129 * struct netdev_fcoe_hbainfo *hbainfo);
1130 * Called when the FCoE Protocol stack wants information on the underlying
1131 * device. This information is utilized by the FCoE protocol stack to
1132 * register attributes with Fiber Channel management service as per the
1133 * FC-GS Fabric Device Management Information(FDMI) specification.
1135 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1136 * Called when the underlying device wants to override default World Wide
1137 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1138 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1139 * protocol stack to use.
1142 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1143 * u16 rxq_index, u32 flow_id);
1144 * Set hardware filter for RFS. rxq_index is the target queue index;
1145 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1146 * Return the filter ID on success, or a negative error code.
1148 * Slave management functions (for bridge, bonding, etc).
1149 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1150 * Called to make another netdev an underling.
1152 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1153 * Called to release previously enslaved netdev.
1155 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1156 * struct sk_buff *skb,
1158 * Get the xmit slave of master device. If all_slaves is true, function
1159 * assume all the slaves can transmit.
1161 * Feature/offload setting functions.
1162 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1163 * netdev_features_t features);
1164 * Adjusts the requested feature flags according to device-specific
1165 * constraints, and returns the resulting flags. Must not modify
1168 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1169 * Called to update device configuration to new features. Passed
1170 * feature set might be less than what was returned by ndo_fix_features()).
1171 * Must return >0 or -errno if it changed dev->features itself.
1173 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1174 * struct net_device *dev,
1175 * const unsigned char *addr, u16 vid, u16 flags,
1176 * struct netlink_ext_ack *extack);
1177 * Adds an FDB entry to dev for addr.
1178 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1179 * struct net_device *dev,
1180 * const unsigned char *addr, u16 vid)
1181 * Deletes the FDB entry from dev coresponding to addr.
1182 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1183 * struct net_device *dev, struct net_device *filter_dev,
1185 * Used to add FDB entries to dump requests. Implementers should add
1186 * entries to skb and update idx with the number of entries.
1188 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1189 * u16 flags, struct netlink_ext_ack *extack)
1190 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1191 * struct net_device *dev, u32 filter_mask,
1193 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1196 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1197 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1198 * which do not represent real hardware may define this to allow their
1199 * userspace components to manage their virtual carrier state. Devices
1200 * that determine carrier state from physical hardware properties (eg
1201 * network cables) or protocol-dependent mechanisms (eg
1202 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1204 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1205 * struct netdev_phys_item_id *ppid);
1206 * Called to get ID of physical port of this device. If driver does
1207 * not implement this, it is assumed that the hw is not able to have
1208 * multiple net devices on single physical port.
1210 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1211 * struct netdev_phys_item_id *ppid)
1212 * Called to get the parent ID of the physical port of this device.
1214 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1215 * struct net_device *dev)
1216 * Called by upper layer devices to accelerate switching or other
1217 * station functionality into hardware. 'pdev is the lowerdev
1218 * to use for the offload and 'dev' is the net device that will
1219 * back the offload. Returns a pointer to the private structure
1220 * the upper layer will maintain.
1221 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1222 * Called by upper layer device to delete the station created
1223 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1224 * the station and priv is the structure returned by the add
1226 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1227 * int queue_index, u32 maxrate);
1228 * Called when a user wants to set a max-rate limitation of specific
1230 * int (*ndo_get_iflink)(const struct net_device *dev);
1231 * Called to get the iflink value of this device.
1232 * void (*ndo_change_proto_down)(struct net_device *dev,
1234 * This function is used to pass protocol port error state information
1235 * to the switch driver. The switch driver can react to the proto_down
1236 * by doing a phys down on the associated switch port.
1237 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1238 * This function is used to get egress tunnel information for given skb.
1239 * This is useful for retrieving outer tunnel header parameters while
1241 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1242 * This function is used to specify the headroom that the skb must
1243 * consider when allocation skb during packet reception. Setting
1244 * appropriate rx headroom value allows avoiding skb head copy on
1245 * forward. Setting a negative value resets the rx headroom to the
1247 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1248 * This function is used to set or query state related to XDP on the
1249 * netdevice and manage BPF offload. See definition of
1250 * enum bpf_netdev_command for details.
1251 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1253 * This function is used to submit @n XDP packets for transmit on a
1254 * netdevice. Returns number of frames successfully transmitted, frames
1255 * that got dropped are freed/returned via xdp_return_frame().
1256 * Returns negative number, means general error invoking ndo, meaning
1257 * no frames were xmit'ed and core-caller will free all frames.
1258 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1259 * This function is used to wake up the softirq, ksoftirqd or kthread
1260 * responsible for sending and/or receiving packets on a specific
1261 * queue id bound to an AF_XDP socket. The flags field specifies if
1262 * only RX, only Tx, or both should be woken up using the flags
1263 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1264 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1265 * Get devlink port instance associated with a given netdev.
1266 * Called with a reference on the netdevice and devlink locks only,
1267 * rtnl_lock is not held.
1268 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1270 * Add, change, delete or get information on an IPv4 tunnel.
1271 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1272 * If a device is paired with a peer device, return the peer instance.
1273 * The caller must be under RCU read context.
1275 struct net_device_ops {
1276 int (*ndo_init)(struct net_device *dev);
1277 void (*ndo_uninit)(struct net_device *dev);
1278 int (*ndo_open)(struct net_device *dev);
1279 int (*ndo_stop)(struct net_device *dev);
1280 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1281 struct net_device *dev);
1282 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1283 struct net_device *dev,
1284 netdev_features_t features);
1285 u16 (*ndo_select_queue)(struct net_device *dev,
1286 struct sk_buff *skb,
1287 struct net_device *sb_dev);
1288 void (*ndo_change_rx_flags)(struct net_device *dev,
1290 void (*ndo_set_rx_mode)(struct net_device *dev);
1291 int (*ndo_set_mac_address)(struct net_device *dev,
1293 int (*ndo_validate_addr)(struct net_device *dev);
1294 int (*ndo_do_ioctl)(struct net_device *dev,
1295 struct ifreq *ifr, int cmd);
1296 int (*ndo_set_config)(struct net_device *dev,
1298 int (*ndo_change_mtu)(struct net_device *dev,
1300 int (*ndo_neigh_setup)(struct net_device *dev,
1301 struct neigh_parms *);
1302 void (*ndo_tx_timeout) (struct net_device *dev,
1303 unsigned int txqueue);
1305 void (*ndo_get_stats64)(struct net_device *dev,
1306 struct rtnl_link_stats64 *storage);
1307 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1308 int (*ndo_get_offload_stats)(int attr_id,
1309 const struct net_device *dev,
1311 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1313 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1314 __be16 proto, u16 vid);
1315 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1316 __be16 proto, u16 vid);
1317 #ifdef CONFIG_NET_POLL_CONTROLLER
1318 void (*ndo_poll_controller)(struct net_device *dev);
1319 int (*ndo_netpoll_setup)(struct net_device *dev,
1320 struct netpoll_info *info);
1321 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1323 int (*ndo_set_vf_mac)(struct net_device *dev,
1324 int queue, u8 *mac);
1325 int (*ndo_set_vf_vlan)(struct net_device *dev,
1326 int queue, u16 vlan,
1327 u8 qos, __be16 proto);
1328 int (*ndo_set_vf_rate)(struct net_device *dev,
1329 int vf, int min_tx_rate,
1331 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1332 int vf, bool setting);
1333 int (*ndo_set_vf_trust)(struct net_device *dev,
1334 int vf, bool setting);
1335 int (*ndo_get_vf_config)(struct net_device *dev,
1337 struct ifla_vf_info *ivf);
1338 int (*ndo_set_vf_link_state)(struct net_device *dev,
1339 int vf, int link_state);
1340 int (*ndo_get_vf_stats)(struct net_device *dev,
1342 struct ifla_vf_stats
1344 int (*ndo_set_vf_port)(struct net_device *dev,
1346 struct nlattr *port[]);
1347 int (*ndo_get_vf_port)(struct net_device *dev,
1348 int vf, struct sk_buff *skb);
1349 int (*ndo_get_vf_guid)(struct net_device *dev,
1351 struct ifla_vf_guid *node_guid,
1352 struct ifla_vf_guid *port_guid);
1353 int (*ndo_set_vf_guid)(struct net_device *dev,
1356 int (*ndo_set_vf_rss_query_en)(
1357 struct net_device *dev,
1358 int vf, bool setting);
1359 int (*ndo_setup_tc)(struct net_device *dev,
1360 enum tc_setup_type type,
1362 #if IS_ENABLED(CONFIG_FCOE)
1363 int (*ndo_fcoe_enable)(struct net_device *dev);
1364 int (*ndo_fcoe_disable)(struct net_device *dev);
1365 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1367 struct scatterlist *sgl,
1369 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1371 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1373 struct scatterlist *sgl,
1375 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1376 struct netdev_fcoe_hbainfo *hbainfo);
1379 #if IS_ENABLED(CONFIG_LIBFCOE)
1380 #define NETDEV_FCOE_WWNN 0
1381 #define NETDEV_FCOE_WWPN 1
1382 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1383 u64 *wwn, int type);
1386 #ifdef CONFIG_RFS_ACCEL
1387 int (*ndo_rx_flow_steer)(struct net_device *dev,
1388 const struct sk_buff *skb,
1392 int (*ndo_add_slave)(struct net_device *dev,
1393 struct net_device *slave_dev,
1394 struct netlink_ext_ack *extack);
1395 int (*ndo_del_slave)(struct net_device *dev,
1396 struct net_device *slave_dev);
1397 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1398 struct sk_buff *skb,
1400 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1402 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1403 netdev_features_t features);
1404 int (*ndo_set_features)(struct net_device *dev,
1405 netdev_features_t features);
1406 int (*ndo_neigh_construct)(struct net_device *dev,
1407 struct neighbour *n);
1408 void (*ndo_neigh_destroy)(struct net_device *dev,
1409 struct neighbour *n);
1411 int (*ndo_fdb_add)(struct ndmsg *ndm,
1412 struct nlattr *tb[],
1413 struct net_device *dev,
1414 const unsigned char *addr,
1417 struct netlink_ext_ack *extack);
1418 int (*ndo_fdb_del)(struct ndmsg *ndm,
1419 struct nlattr *tb[],
1420 struct net_device *dev,
1421 const unsigned char *addr,
1423 int (*ndo_fdb_dump)(struct sk_buff *skb,
1424 struct netlink_callback *cb,
1425 struct net_device *dev,
1426 struct net_device *filter_dev,
1428 int (*ndo_fdb_get)(struct sk_buff *skb,
1429 struct nlattr *tb[],
1430 struct net_device *dev,
1431 const unsigned char *addr,
1432 u16 vid, u32 portid, u32 seq,
1433 struct netlink_ext_ack *extack);
1434 int (*ndo_bridge_setlink)(struct net_device *dev,
1435 struct nlmsghdr *nlh,
1437 struct netlink_ext_ack *extack);
1438 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1440 struct net_device *dev,
1443 int (*ndo_bridge_dellink)(struct net_device *dev,
1444 struct nlmsghdr *nlh,
1446 int (*ndo_change_carrier)(struct net_device *dev,
1448 int (*ndo_get_phys_port_id)(struct net_device *dev,
1449 struct netdev_phys_item_id *ppid);
1450 int (*ndo_get_port_parent_id)(struct net_device *dev,
1451 struct netdev_phys_item_id *ppid);
1452 int (*ndo_get_phys_port_name)(struct net_device *dev,
1453 char *name, size_t len);
1454 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1455 struct net_device *dev);
1456 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1459 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1462 int (*ndo_get_iflink)(const struct net_device *dev);
1463 int (*ndo_change_proto_down)(struct net_device *dev,
1465 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1466 struct sk_buff *skb);
1467 void (*ndo_set_rx_headroom)(struct net_device *dev,
1468 int needed_headroom);
1469 int (*ndo_bpf)(struct net_device *dev,
1470 struct netdev_bpf *bpf);
1471 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1472 struct xdp_frame **xdp,
1474 int (*ndo_xsk_wakeup)(struct net_device *dev,
1475 u32 queue_id, u32 flags);
1476 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1477 int (*ndo_tunnel_ctl)(struct net_device *dev,
1478 struct ip_tunnel_parm *p, int cmd);
1479 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1483 * enum netdev_priv_flags - &struct net_device priv_flags
1485 * These are the &struct net_device, they are only set internally
1486 * by drivers and used in the kernel. These flags are invisible to
1487 * userspace; this means that the order of these flags can change
1488 * during any kernel release.
1490 * You should have a pretty good reason to be extending these flags.
1492 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1493 * @IFF_EBRIDGE: Ethernet bridging device
1494 * @IFF_BONDING: bonding master or slave
1495 * @IFF_ISATAP: ISATAP interface (RFC4214)
1496 * @IFF_WAN_HDLC: WAN HDLC device
1497 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1499 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1500 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1501 * @IFF_MACVLAN_PORT: device used as macvlan port
1502 * @IFF_BRIDGE_PORT: device used as bridge port
1503 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1504 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1505 * @IFF_UNICAST_FLT: Supports unicast filtering
1506 * @IFF_TEAM_PORT: device used as team port
1507 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1508 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1509 * change when it's running
1510 * @IFF_MACVLAN: Macvlan device
1511 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1512 * underlying stacked devices
1513 * @IFF_L3MDEV_MASTER: device is an L3 master device
1514 * @IFF_NO_QUEUE: device can run without qdisc attached
1515 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1516 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1517 * @IFF_TEAM: device is a team device
1518 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1519 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1520 * entity (i.e. the master device for bridged veth)
1521 * @IFF_MACSEC: device is a MACsec device
1522 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1523 * @IFF_FAILOVER: device is a failover master device
1524 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1525 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1526 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1527 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1528 * skb_headlen(skb) == 0 (data starts from frag0)
1530 enum netdev_priv_flags {
1531 IFF_802_1Q_VLAN = 1<<0,
1535 IFF_WAN_HDLC = 1<<4,
1536 IFF_XMIT_DST_RELEASE = 1<<5,
1537 IFF_DONT_BRIDGE = 1<<6,
1538 IFF_DISABLE_NETPOLL = 1<<7,
1539 IFF_MACVLAN_PORT = 1<<8,
1540 IFF_BRIDGE_PORT = 1<<9,
1541 IFF_OVS_DATAPATH = 1<<10,
1542 IFF_TX_SKB_SHARING = 1<<11,
1543 IFF_UNICAST_FLT = 1<<12,
1544 IFF_TEAM_PORT = 1<<13,
1545 IFF_SUPP_NOFCS = 1<<14,
1546 IFF_LIVE_ADDR_CHANGE = 1<<15,
1547 IFF_MACVLAN = 1<<16,
1548 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1549 IFF_L3MDEV_MASTER = 1<<18,
1550 IFF_NO_QUEUE = 1<<19,
1551 IFF_OPENVSWITCH = 1<<20,
1552 IFF_L3MDEV_SLAVE = 1<<21,
1554 IFF_RXFH_CONFIGURED = 1<<23,
1555 IFF_PHONY_HEADROOM = 1<<24,
1557 IFF_NO_RX_HANDLER = 1<<26,
1558 IFF_FAILOVER = 1<<27,
1559 IFF_FAILOVER_SLAVE = 1<<28,
1560 IFF_L3MDEV_RX_HANDLER = 1<<29,
1561 IFF_LIVE_RENAME_OK = 1<<30,
1562 IFF_TX_SKB_NO_LINEAR = 1<<31,
1565 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1566 #define IFF_EBRIDGE IFF_EBRIDGE
1567 #define IFF_BONDING IFF_BONDING
1568 #define IFF_ISATAP IFF_ISATAP
1569 #define IFF_WAN_HDLC IFF_WAN_HDLC
1570 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1571 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1572 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1573 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1574 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1575 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1576 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1577 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1578 #define IFF_TEAM_PORT IFF_TEAM_PORT
1579 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1580 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1581 #define IFF_MACVLAN IFF_MACVLAN
1582 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1583 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1584 #define IFF_NO_QUEUE IFF_NO_QUEUE
1585 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1586 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1587 #define IFF_TEAM IFF_TEAM
1588 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1589 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1590 #define IFF_MACSEC IFF_MACSEC
1591 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1592 #define IFF_FAILOVER IFF_FAILOVER
1593 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1594 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1595 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1596 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1598 /* Specifies the type of the struct net_device::ml_priv pointer */
1599 enum netdev_ml_priv_type {
1605 * struct net_device - The DEVICE structure.
1607 * Actually, this whole structure is a big mistake. It mixes I/O
1608 * data with strictly "high-level" data, and it has to know about
1609 * almost every data structure used in the INET module.
1611 * @name: This is the first field of the "visible" part of this structure
1612 * (i.e. as seen by users in the "Space.c" file). It is the name
1615 * @name_node: Name hashlist node
1616 * @ifalias: SNMP alias
1617 * @mem_end: Shared memory end
1618 * @mem_start: Shared memory start
1619 * @base_addr: Device I/O address
1620 * @irq: Device IRQ number
1622 * @state: Generic network queuing layer state, see netdev_state_t
1623 * @dev_list: The global list of network devices
1624 * @napi_list: List entry used for polling NAPI devices
1625 * @unreg_list: List entry when we are unregistering the
1626 * device; see the function unregister_netdev
1627 * @close_list: List entry used when we are closing the device
1628 * @ptype_all: Device-specific packet handlers for all protocols
1629 * @ptype_specific: Device-specific, protocol-specific packet handlers
1631 * @adj_list: Directly linked devices, like slaves for bonding
1632 * @features: Currently active device features
1633 * @hw_features: User-changeable features
1635 * @wanted_features: User-requested features
1636 * @vlan_features: Mask of features inheritable by VLAN devices
1638 * @hw_enc_features: Mask of features inherited by encapsulating devices
1639 * This field indicates what encapsulation
1640 * offloads the hardware is capable of doing,
1641 * and drivers will need to set them appropriately.
1643 * @mpls_features: Mask of features inheritable by MPLS
1644 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1646 * @ifindex: interface index
1647 * @group: The group the device belongs to
1649 * @stats: Statistics struct, which was left as a legacy, use
1650 * rtnl_link_stats64 instead
1652 * @rx_dropped: Dropped packets by core network,
1653 * do not use this in drivers
1654 * @tx_dropped: Dropped packets by core network,
1655 * do not use this in drivers
1656 * @rx_nohandler: nohandler dropped packets by core network on
1657 * inactive devices, do not use this in drivers
1658 * @carrier_up_count: Number of times the carrier has been up
1659 * @carrier_down_count: Number of times the carrier has been down
1661 * @wireless_handlers: List of functions to handle Wireless Extensions,
1663 * see <net/iw_handler.h> for details.
1664 * @wireless_data: Instance data managed by the core of wireless extensions
1666 * @netdev_ops: Includes several pointers to callbacks,
1667 * if one wants to override the ndo_*() functions
1668 * @ethtool_ops: Management operations
1669 * @l3mdev_ops: Layer 3 master device operations
1670 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1671 * discovery handling. Necessary for e.g. 6LoWPAN.
1672 * @xfrmdev_ops: Transformation offload operations
1673 * @tlsdev_ops: Transport Layer Security offload operations
1674 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1675 * of Layer 2 headers.
1677 * @flags: Interface flags (a la BSD)
1678 * @priv_flags: Like 'flags' but invisible to userspace,
1679 * see if.h for the definitions
1680 * @gflags: Global flags ( kept as legacy )
1681 * @padded: How much padding added by alloc_netdev()
1682 * @operstate: RFC2863 operstate
1683 * @link_mode: Mapping policy to operstate
1684 * @if_port: Selectable AUI, TP, ...
1686 * @mtu: Interface MTU value
1687 * @min_mtu: Interface Minimum MTU value
1688 * @max_mtu: Interface Maximum MTU value
1689 * @type: Interface hardware type
1690 * @hard_header_len: Maximum hardware header length.
1691 * @min_header_len: Minimum hardware header length
1693 * @needed_headroom: Extra headroom the hardware may need, but not in all
1694 * cases can this be guaranteed
1695 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1696 * cases can this be guaranteed. Some cases also use
1697 * LL_MAX_HEADER instead to allocate the skb
1699 * interface address info:
1701 * @perm_addr: Permanent hw address
1702 * @addr_assign_type: Hw address assignment type
1703 * @addr_len: Hardware address length
1704 * @upper_level: Maximum depth level of upper devices.
1705 * @lower_level: Maximum depth level of lower devices.
1706 * @neigh_priv_len: Used in neigh_alloc()
1707 * @dev_id: Used to differentiate devices that share
1708 * the same link layer address
1709 * @dev_port: Used to differentiate devices that share
1711 * @addr_list_lock: XXX: need comments on this one
1712 * @name_assign_type: network interface name assignment type
1713 * @uc_promisc: Counter that indicates promiscuous mode
1714 * has been enabled due to the need to listen to
1715 * additional unicast addresses in a device that
1716 * does not implement ndo_set_rx_mode()
1717 * @uc: unicast mac addresses
1718 * @mc: multicast mac addresses
1719 * @dev_addrs: list of device hw addresses
1720 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1721 * @promiscuity: Number of times the NIC is told to work in
1722 * promiscuous mode; if it becomes 0 the NIC will
1723 * exit promiscuous mode
1724 * @allmulti: Counter, enables or disables allmulticast mode
1726 * @vlan_info: VLAN info
1727 * @dsa_ptr: dsa specific data
1728 * @tipc_ptr: TIPC specific data
1729 * @atalk_ptr: AppleTalk link
1730 * @ip_ptr: IPv4 specific data
1731 * @dn_ptr: DECnet specific data
1732 * @ip6_ptr: IPv6 specific data
1733 * @ax25_ptr: AX.25 specific data
1734 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1735 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1737 * @mpls_ptr: mpls_dev struct pointer
1739 * @dev_addr: Hw address (before bcast,
1740 * because most packets are unicast)
1742 * @_rx: Array of RX queues
1743 * @num_rx_queues: Number of RX queues
1744 * allocated at register_netdev() time
1745 * @real_num_rx_queues: Number of RX queues currently active in device
1746 * @xdp_prog: XDP sockets filter program pointer
1747 * @gro_flush_timeout: timeout for GRO layer in NAPI
1748 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1749 * allow to avoid NIC hard IRQ, on busy queues.
1751 * @rx_handler: handler for received packets
1752 * @rx_handler_data: XXX: need comments on this one
1753 * @miniq_ingress: ingress/clsact qdisc specific data for
1754 * ingress processing
1755 * @ingress_queue: XXX: need comments on this one
1756 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1757 * @broadcast: hw bcast address
1759 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1760 * indexed by RX queue number. Assigned by driver.
1761 * This must only be set if the ndo_rx_flow_steer
1762 * operation is defined
1763 * @index_hlist: Device index hash chain
1765 * @_tx: Array of TX queues
1766 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1767 * @real_num_tx_queues: Number of TX queues currently active in device
1768 * @qdisc: Root qdisc from userspace point of view
1769 * @tx_queue_len: Max frames per queue allowed
1770 * @tx_global_lock: XXX: need comments on this one
1771 * @xdp_bulkq: XDP device bulk queue
1772 * @xps_cpus_map: all CPUs map for XPS device
1773 * @xps_rxqs_map: all RXQs map for XPS device
1775 * @xps_maps: XXX: need comments on this one
1776 * @miniq_egress: clsact qdisc specific data for
1778 * @qdisc_hash: qdisc hash table
1779 * @watchdog_timeo: Represents the timeout that is used by
1780 * the watchdog (see dev_watchdog())
1781 * @watchdog_timer: List of timers
1783 * @proto_down_reason: reason a netdev interface is held down
1784 * @pcpu_refcnt: Number of references to this device
1785 * @todo_list: Delayed register/unregister
1786 * @link_watch_list: XXX: need comments on this one
1788 * @reg_state: Register/unregister state machine
1789 * @dismantle: Device is going to be freed
1790 * @rtnl_link_state: This enum represents the phases of creating
1793 * @needs_free_netdev: Should unregister perform free_netdev?
1794 * @priv_destructor: Called from unregister
1795 * @npinfo: XXX: need comments on this one
1796 * @nd_net: Network namespace this network device is inside
1798 * @ml_priv: Mid-layer private
1799 * @ml_priv_type: Mid-layer private type
1800 * @lstats: Loopback statistics
1801 * @tstats: Tunnel statistics
1802 * @dstats: Dummy statistics
1803 * @vstats: Virtual ethernet statistics
1808 * @dev: Class/net/name entry
1809 * @sysfs_groups: Space for optional device, statistics and wireless
1812 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1813 * @rtnl_link_ops: Rtnl_link_ops
1815 * @gso_max_size: Maximum size of generic segmentation offload
1816 * @gso_max_segs: Maximum number of segments that can be passed to the
1819 * @dcbnl_ops: Data Center Bridging netlink ops
1820 * @num_tc: Number of traffic classes in the net device
1821 * @tc_to_txq: XXX: need comments on this one
1822 * @prio_tc_map: XXX: need comments on this one
1824 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1826 * @priomap: XXX: need comments on this one
1827 * @phydev: Physical device may attach itself
1828 * for hardware timestamping
1829 * @sfp_bus: attached &struct sfp_bus structure.
1831 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1832 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1834 * @proto_down: protocol port state information can be sent to the
1835 * switch driver and used to set the phys state of the
1838 * @wol_enabled: Wake-on-LAN is enabled
1840 * @threaded: napi threaded mode is enabled
1842 * @net_notifier_list: List of per-net netdev notifier block
1843 * that follow this device when it is moved
1844 * to another network namespace.
1846 * @macsec_ops: MACsec offloading ops
1848 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1849 * offload capabilities of the device
1850 * @udp_tunnel_nic: UDP tunnel offload state
1851 * @xdp_state: stores info on attached XDP BPF programs
1853 * @nested_level: Used as as a parameter of spin_lock_nested() of
1854 * dev->addr_list_lock.
1855 * @unlink_list: As netif_addr_lock() can be called recursively,
1856 * keep a list of interfaces to be deleted.
1858 * FIXME: cleanup struct net_device such that network protocol info
1863 char name[IFNAMSIZ];
1864 struct netdev_name_node *name_node;
1865 struct dev_ifalias __rcu *ifalias;
1867 * I/O specific fields
1868 * FIXME: Merge these and struct ifmap into one
1870 unsigned long mem_end;
1871 unsigned long mem_start;
1872 unsigned long base_addr;
1875 * Some hardware also needs these fields (state,dev_list,
1876 * napi_list,unreg_list,close_list) but they are not
1877 * part of the usual set specified in Space.c.
1880 unsigned long state;
1882 struct list_head dev_list;
1883 struct list_head napi_list;
1884 struct list_head unreg_list;
1885 struct list_head close_list;
1886 struct list_head ptype_all;
1887 struct list_head ptype_specific;
1890 struct list_head upper;
1891 struct list_head lower;
1894 /* Read-mostly cache-line for fast-path access */
1896 unsigned int priv_flags;
1897 const struct net_device_ops *netdev_ops;
1899 unsigned short gflags;
1900 unsigned short hard_header_len;
1902 /* Note : dev->mtu is often read without holding a lock.
1903 * Writers usually hold RTNL.
1904 * It is recommended to use READ_ONCE() to annotate the reads,
1905 * and to use WRITE_ONCE() to annotate the writes.
1908 unsigned short needed_headroom;
1909 unsigned short needed_tailroom;
1911 netdev_features_t features;
1912 netdev_features_t hw_features;
1913 netdev_features_t wanted_features;
1914 netdev_features_t vlan_features;
1915 netdev_features_t hw_enc_features;
1916 netdev_features_t mpls_features;
1917 netdev_features_t gso_partial_features;
1919 unsigned int min_mtu;
1920 unsigned int max_mtu;
1921 unsigned short type;
1922 unsigned char min_header_len;
1923 unsigned char name_assign_type;
1927 struct net_device_stats stats; /* not used by modern drivers */
1929 atomic_long_t rx_dropped;
1930 atomic_long_t tx_dropped;
1931 atomic_long_t rx_nohandler;
1933 /* Stats to monitor link on/off, flapping */
1934 atomic_t carrier_up_count;
1935 atomic_t carrier_down_count;
1937 #ifdef CONFIG_WIRELESS_EXT
1938 const struct iw_handler_def *wireless_handlers;
1939 struct iw_public_data *wireless_data;
1941 const struct ethtool_ops *ethtool_ops;
1942 #ifdef CONFIG_NET_L3_MASTER_DEV
1943 const struct l3mdev_ops *l3mdev_ops;
1945 #if IS_ENABLED(CONFIG_IPV6)
1946 const struct ndisc_ops *ndisc_ops;
1949 #ifdef CONFIG_XFRM_OFFLOAD
1950 const struct xfrmdev_ops *xfrmdev_ops;
1953 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1954 const struct tlsdev_ops *tlsdev_ops;
1957 const struct header_ops *header_ops;
1959 unsigned char operstate;
1960 unsigned char link_mode;
1962 unsigned char if_port;
1965 /* Interface address info. */
1966 unsigned char perm_addr[MAX_ADDR_LEN];
1967 unsigned char addr_assign_type;
1968 unsigned char addr_len;
1969 unsigned char upper_level;
1970 unsigned char lower_level;
1972 unsigned short neigh_priv_len;
1973 unsigned short dev_id;
1974 unsigned short dev_port;
1975 unsigned short padded;
1977 spinlock_t addr_list_lock;
1980 struct netdev_hw_addr_list uc;
1981 struct netdev_hw_addr_list mc;
1982 struct netdev_hw_addr_list dev_addrs;
1985 struct kset *queues_kset;
1987 #ifdef CONFIG_LOCKDEP
1988 struct list_head unlink_list;
1990 unsigned int promiscuity;
1991 unsigned int allmulti;
1993 #ifdef CONFIG_LOCKDEP
1994 unsigned char nested_level;
1998 /* Protocol-specific pointers */
2000 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2001 struct vlan_info __rcu *vlan_info;
2003 #if IS_ENABLED(CONFIG_NET_DSA)
2004 struct dsa_port *dsa_ptr;
2006 #if IS_ENABLED(CONFIG_TIPC)
2007 struct tipc_bearer __rcu *tipc_ptr;
2009 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2012 struct in_device __rcu *ip_ptr;
2013 #if IS_ENABLED(CONFIG_DECNET)
2014 struct dn_dev __rcu *dn_ptr;
2016 struct inet6_dev __rcu *ip6_ptr;
2017 #if IS_ENABLED(CONFIG_AX25)
2020 struct wireless_dev *ieee80211_ptr;
2021 struct wpan_dev *ieee802154_ptr;
2022 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2023 struct mpls_dev __rcu *mpls_ptr;
2027 * Cache lines mostly used on receive path (including eth_type_trans())
2029 /* Interface address info used in eth_type_trans() */
2030 unsigned char *dev_addr;
2032 struct netdev_rx_queue *_rx;
2033 unsigned int num_rx_queues;
2034 unsigned int real_num_rx_queues;
2036 struct bpf_prog __rcu *xdp_prog;
2037 unsigned long gro_flush_timeout;
2038 int napi_defer_hard_irqs;
2039 rx_handler_func_t __rcu *rx_handler;
2040 void __rcu *rx_handler_data;
2042 #ifdef CONFIG_NET_CLS_ACT
2043 struct mini_Qdisc __rcu *miniq_ingress;
2045 struct netdev_queue __rcu *ingress_queue;
2046 #ifdef CONFIG_NETFILTER_INGRESS
2047 struct nf_hook_entries __rcu *nf_hooks_ingress;
2050 unsigned char broadcast[MAX_ADDR_LEN];
2051 #ifdef CONFIG_RFS_ACCEL
2052 struct cpu_rmap *rx_cpu_rmap;
2054 struct hlist_node index_hlist;
2057 * Cache lines mostly used on transmit path
2059 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2060 unsigned int num_tx_queues;
2061 unsigned int real_num_tx_queues;
2062 struct Qdisc *qdisc;
2063 unsigned int tx_queue_len;
2064 spinlock_t tx_global_lock;
2066 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2069 struct xps_dev_maps __rcu *xps_cpus_map;
2070 struct xps_dev_maps __rcu *xps_rxqs_map;
2072 #ifdef CONFIG_NET_CLS_ACT
2073 struct mini_Qdisc __rcu *miniq_egress;
2076 #ifdef CONFIG_NET_SCHED
2077 DECLARE_HASHTABLE (qdisc_hash, 4);
2079 /* These may be needed for future network-power-down code. */
2080 struct timer_list watchdog_timer;
2083 u32 proto_down_reason;
2085 struct list_head todo_list;
2086 int __percpu *pcpu_refcnt;
2088 struct list_head link_watch_list;
2090 enum { NETREG_UNINITIALIZED=0,
2091 NETREG_REGISTERED, /* completed register_netdevice */
2092 NETREG_UNREGISTERING, /* called unregister_netdevice */
2093 NETREG_UNREGISTERED, /* completed unregister todo */
2094 NETREG_RELEASED, /* called free_netdev */
2095 NETREG_DUMMY, /* dummy device for NAPI poll */
2101 RTNL_LINK_INITIALIZED,
2102 RTNL_LINK_INITIALIZING,
2103 } rtnl_link_state:16;
2105 bool needs_free_netdev;
2106 void (*priv_destructor)(struct net_device *dev);
2108 #ifdef CONFIG_NETPOLL
2109 struct netpoll_info __rcu *npinfo;
2112 possible_net_t nd_net;
2114 /* mid-layer private */
2116 enum netdev_ml_priv_type ml_priv_type;
2119 struct pcpu_lstats __percpu *lstats;
2120 struct pcpu_sw_netstats __percpu *tstats;
2121 struct pcpu_dstats __percpu *dstats;
2124 #if IS_ENABLED(CONFIG_GARP)
2125 struct garp_port __rcu *garp_port;
2127 #if IS_ENABLED(CONFIG_MRP)
2128 struct mrp_port __rcu *mrp_port;
2132 const struct attribute_group *sysfs_groups[4];
2133 const struct attribute_group *sysfs_rx_queue_group;
2135 const struct rtnl_link_ops *rtnl_link_ops;
2137 /* for setting kernel sock attribute on TCP connection setup */
2138 #define GSO_MAX_SIZE 65536
2139 unsigned int gso_max_size;
2140 #define GSO_MAX_SEGS 65535
2144 const struct dcbnl_rtnl_ops *dcbnl_ops;
2147 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2148 u8 prio_tc_map[TC_BITMASK + 1];
2150 #if IS_ENABLED(CONFIG_FCOE)
2151 unsigned int fcoe_ddp_xid;
2153 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2154 struct netprio_map __rcu *priomap;
2156 struct phy_device *phydev;
2157 struct sfp_bus *sfp_bus;
2158 struct lock_class_key *qdisc_tx_busylock;
2159 struct lock_class_key *qdisc_running_key;
2161 unsigned wol_enabled:1;
2162 unsigned threaded:1;
2164 struct list_head net_notifier_list;
2166 #if IS_ENABLED(CONFIG_MACSEC)
2167 /* MACsec management functions */
2168 const struct macsec_ops *macsec_ops;
2170 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2171 struct udp_tunnel_nic *udp_tunnel_nic;
2173 /* protected by rtnl_lock */
2174 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2176 #define to_net_dev(d) container_of(d, struct net_device, dev)
2178 static inline bool netif_elide_gro(const struct net_device *dev)
2180 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2185 #define NETDEV_ALIGN 32
2188 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2190 return dev->prio_tc_map[prio & TC_BITMASK];
2194 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2196 if (tc >= dev->num_tc)
2199 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2203 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2204 void netdev_reset_tc(struct net_device *dev);
2205 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2206 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2209 int netdev_get_num_tc(struct net_device *dev)
2214 static inline void net_prefetch(void *p)
2217 #if L1_CACHE_BYTES < 128
2218 prefetch((u8 *)p + L1_CACHE_BYTES);
2222 static inline void net_prefetchw(void *p)
2225 #if L1_CACHE_BYTES < 128
2226 prefetchw((u8 *)p + L1_CACHE_BYTES);
2230 void netdev_unbind_sb_channel(struct net_device *dev,
2231 struct net_device *sb_dev);
2232 int netdev_bind_sb_channel_queue(struct net_device *dev,
2233 struct net_device *sb_dev,
2234 u8 tc, u16 count, u16 offset);
2235 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2236 static inline int netdev_get_sb_channel(struct net_device *dev)
2238 return max_t(int, -dev->num_tc, 0);
2242 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2245 return &dev->_tx[index];
2248 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2249 const struct sk_buff *skb)
2251 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2254 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2255 void (*f)(struct net_device *,
2256 struct netdev_queue *,
2262 for (i = 0; i < dev->num_tx_queues; i++)
2263 f(dev, &dev->_tx[i], arg);
2266 #define netdev_lockdep_set_classes(dev) \
2268 static struct lock_class_key qdisc_tx_busylock_key; \
2269 static struct lock_class_key qdisc_running_key; \
2270 static struct lock_class_key qdisc_xmit_lock_key; \
2271 static struct lock_class_key dev_addr_list_lock_key; \
2274 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2275 (dev)->qdisc_running_key = &qdisc_running_key; \
2276 lockdep_set_class(&(dev)->addr_list_lock, \
2277 &dev_addr_list_lock_key); \
2278 for (i = 0; i < (dev)->num_tx_queues; i++) \
2279 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2280 &qdisc_xmit_lock_key); \
2283 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2284 struct net_device *sb_dev);
2285 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2286 struct sk_buff *skb,
2287 struct net_device *sb_dev);
2289 /* returns the headroom that the master device needs to take in account
2290 * when forwarding to this dev
2292 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2294 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2297 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2299 if (dev->netdev_ops->ndo_set_rx_headroom)
2300 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2303 /* set the device rx headroom to the dev's default */
2304 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2306 netdev_set_rx_headroom(dev, -1);
2309 static inline void *netdev_get_ml_priv(struct net_device *dev,
2310 enum netdev_ml_priv_type type)
2312 if (dev->ml_priv_type != type)
2315 return dev->ml_priv;
2318 static inline void netdev_set_ml_priv(struct net_device *dev,
2320 enum netdev_ml_priv_type type)
2322 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2323 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2324 dev->ml_priv_type, type);
2325 WARN(!dev->ml_priv_type && dev->ml_priv,
2326 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2328 dev->ml_priv = ml_priv;
2329 dev->ml_priv_type = type;
2333 * Net namespace inlines
2336 struct net *dev_net(const struct net_device *dev)
2338 return read_pnet(&dev->nd_net);
2342 void dev_net_set(struct net_device *dev, struct net *net)
2344 write_pnet(&dev->nd_net, net);
2348 * netdev_priv - access network device private data
2349 * @dev: network device
2351 * Get network device private data
2353 static inline void *netdev_priv(const struct net_device *dev)
2355 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2358 /* Set the sysfs physical device reference for the network logical device
2359 * if set prior to registration will cause a symlink during initialization.
2361 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2363 /* Set the sysfs device type for the network logical device to allow
2364 * fine-grained identification of different network device types. For
2365 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2367 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2369 /* Default NAPI poll() weight
2370 * Device drivers are strongly advised to not use bigger value
2372 #define NAPI_POLL_WEIGHT 64
2375 * netif_napi_add - initialize a NAPI context
2376 * @dev: network device
2377 * @napi: NAPI context
2378 * @poll: polling function
2379 * @weight: default weight
2381 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2382 * *any* of the other NAPI-related functions.
2384 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2385 int (*poll)(struct napi_struct *, int), int weight);
2388 * netif_tx_napi_add - initialize a NAPI context
2389 * @dev: network device
2390 * @napi: NAPI context
2391 * @poll: polling function
2392 * @weight: default weight
2394 * This variant of netif_napi_add() should be used from drivers using NAPI
2395 * to exclusively poll a TX queue.
2396 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2398 static inline void netif_tx_napi_add(struct net_device *dev,
2399 struct napi_struct *napi,
2400 int (*poll)(struct napi_struct *, int),
2403 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2404 netif_napi_add(dev, napi, poll, weight);
2408 * __netif_napi_del - remove a NAPI context
2409 * @napi: NAPI context
2411 * Warning: caller must observe RCU grace period before freeing memory
2412 * containing @napi. Drivers might want to call this helper to combine
2413 * all the needed RCU grace periods into a single one.
2415 void __netif_napi_del(struct napi_struct *napi);
2418 * netif_napi_del - remove a NAPI context
2419 * @napi: NAPI context
2421 * netif_napi_del() removes a NAPI context from the network device NAPI list
2423 static inline void netif_napi_del(struct napi_struct *napi)
2425 __netif_napi_del(napi);
2429 struct napi_gro_cb {
2430 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2433 /* Length of frag0. */
2434 unsigned int frag0_len;
2436 /* This indicates where we are processing relative to skb->data. */
2439 /* This is non-zero if the packet cannot be merged with the new skb. */
2442 /* Save the IP ID here and check when we get to the transport layer */
2445 /* Number of segments aggregated. */
2448 /* Start offset for remote checksum offload */
2449 u16 gro_remcsum_start;
2451 /* jiffies when first packet was created/queued */
2454 /* Used in ipv6_gro_receive() and foo-over-udp */
2457 /* This is non-zero if the packet may be of the same flow. */
2460 /* Used in tunnel GRO receive */
2463 /* GRO checksum is valid */
2466 /* Number of checksums via CHECKSUM_UNNECESSARY */
2471 #define NAPI_GRO_FREE 1
2472 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2474 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2477 /* Used in GRE, set in fou/gue_gro_receive */
2480 /* Used to determine if flush_id can be ignored */
2483 /* Number of gro_receive callbacks this packet already went through */
2484 u8 recursion_counter:4;
2486 /* GRO is done by frag_list pointer chaining. */
2489 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2492 /* used in skb_gro_receive() slow path */
2493 struct sk_buff *last;
2496 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2498 #define GRO_RECURSION_LIMIT 15
2499 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2501 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2504 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2505 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2506 struct list_head *head,
2507 struct sk_buff *skb)
2509 if (unlikely(gro_recursion_inc_test(skb))) {
2510 NAPI_GRO_CB(skb)->flush |= 1;
2514 return cb(head, skb);
2517 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2519 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2521 struct list_head *head,
2522 struct sk_buff *skb)
2524 if (unlikely(gro_recursion_inc_test(skb))) {
2525 NAPI_GRO_CB(skb)->flush |= 1;
2529 return cb(sk, head, skb);
2532 struct packet_type {
2533 __be16 type; /* This is really htons(ether_type). */
2534 bool ignore_outgoing;
2535 struct net_device *dev; /* NULL is wildcarded here */
2536 int (*func) (struct sk_buff *,
2537 struct net_device *,
2538 struct packet_type *,
2539 struct net_device *);
2540 void (*list_func) (struct list_head *,
2541 struct packet_type *,
2542 struct net_device *);
2543 bool (*id_match)(struct packet_type *ptype,
2545 void *af_packet_priv;
2546 struct list_head list;
2549 struct offload_callbacks {
2550 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2551 netdev_features_t features);
2552 struct sk_buff *(*gro_receive)(struct list_head *head,
2553 struct sk_buff *skb);
2554 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2557 struct packet_offload {
2558 __be16 type; /* This is really htons(ether_type). */
2560 struct offload_callbacks callbacks;
2561 struct list_head list;
2564 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2565 struct pcpu_sw_netstats {
2570 struct u64_stats_sync syncp;
2571 } __aligned(4 * sizeof(u64));
2573 struct pcpu_lstats {
2574 u64_stats_t packets;
2576 struct u64_stats_sync syncp;
2577 } __aligned(2 * sizeof(u64));
2579 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2581 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2583 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2585 u64_stats_update_begin(&tstats->syncp);
2586 tstats->rx_bytes += len;
2587 tstats->rx_packets++;
2588 u64_stats_update_end(&tstats->syncp);
2591 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2592 unsigned int packets,
2595 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2597 u64_stats_update_begin(&tstats->syncp);
2598 tstats->tx_bytes += len;
2599 tstats->tx_packets += packets;
2600 u64_stats_update_end(&tstats->syncp);
2603 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2605 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2607 u64_stats_update_begin(&lstats->syncp);
2608 u64_stats_add(&lstats->bytes, len);
2609 u64_stats_inc(&lstats->packets);
2610 u64_stats_update_end(&lstats->syncp);
2613 #define __netdev_alloc_pcpu_stats(type, gfp) \
2615 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2618 for_each_possible_cpu(__cpu) { \
2619 typeof(type) *stat; \
2620 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2621 u64_stats_init(&stat->syncp); \
2627 #define netdev_alloc_pcpu_stats(type) \
2628 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2630 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2632 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2635 for_each_possible_cpu(__cpu) { \
2636 typeof(type) *stat; \
2637 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2638 u64_stats_init(&stat->syncp); \
2644 enum netdev_lag_tx_type {
2645 NETDEV_LAG_TX_TYPE_UNKNOWN,
2646 NETDEV_LAG_TX_TYPE_RANDOM,
2647 NETDEV_LAG_TX_TYPE_BROADCAST,
2648 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2649 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2650 NETDEV_LAG_TX_TYPE_HASH,
2653 enum netdev_lag_hash {
2654 NETDEV_LAG_HASH_NONE,
2656 NETDEV_LAG_HASH_L34,
2657 NETDEV_LAG_HASH_L23,
2658 NETDEV_LAG_HASH_E23,
2659 NETDEV_LAG_HASH_E34,
2660 NETDEV_LAG_HASH_VLAN_SRCMAC,
2661 NETDEV_LAG_HASH_UNKNOWN,
2664 struct netdev_lag_upper_info {
2665 enum netdev_lag_tx_type tx_type;
2666 enum netdev_lag_hash hash_type;
2669 struct netdev_lag_lower_state_info {
2674 #include <linux/notifier.h>
2676 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2677 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2681 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2683 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2684 detected a hardware crash and restarted
2685 - we can use this eg to kick tcp sessions
2687 NETDEV_CHANGE, /* Notify device state change */
2690 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2691 NETDEV_CHANGEADDR, /* notify after the address change */
2692 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2696 NETDEV_BONDING_FAILOVER,
2698 NETDEV_PRE_TYPE_CHANGE,
2699 NETDEV_POST_TYPE_CHANGE,
2702 NETDEV_NOTIFY_PEERS,
2706 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2707 NETDEV_CHANGEINFODATA,
2708 NETDEV_BONDING_INFO,
2709 NETDEV_PRECHANGEUPPER,
2710 NETDEV_CHANGELOWERSTATE,
2711 NETDEV_UDP_TUNNEL_PUSH_INFO,
2712 NETDEV_UDP_TUNNEL_DROP_INFO,
2713 NETDEV_CHANGE_TX_QUEUE_LEN,
2714 NETDEV_CVLAN_FILTER_PUSH_INFO,
2715 NETDEV_CVLAN_FILTER_DROP_INFO,
2716 NETDEV_SVLAN_FILTER_PUSH_INFO,
2717 NETDEV_SVLAN_FILTER_DROP_INFO,
2719 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2721 int register_netdevice_notifier(struct notifier_block *nb);
2722 int unregister_netdevice_notifier(struct notifier_block *nb);
2723 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2724 int unregister_netdevice_notifier_net(struct net *net,
2725 struct notifier_block *nb);
2726 int register_netdevice_notifier_dev_net(struct net_device *dev,
2727 struct notifier_block *nb,
2728 struct netdev_net_notifier *nn);
2729 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2730 struct notifier_block *nb,
2731 struct netdev_net_notifier *nn);
2733 struct netdev_notifier_info {
2734 struct net_device *dev;
2735 struct netlink_ext_ack *extack;
2738 struct netdev_notifier_info_ext {
2739 struct netdev_notifier_info info; /* must be first */
2745 struct netdev_notifier_change_info {
2746 struct netdev_notifier_info info; /* must be first */
2747 unsigned int flags_changed;
2750 struct netdev_notifier_changeupper_info {
2751 struct netdev_notifier_info info; /* must be first */
2752 struct net_device *upper_dev; /* new upper dev */
2753 bool master; /* is upper dev master */
2754 bool linking; /* is the notification for link or unlink */
2755 void *upper_info; /* upper dev info */
2758 struct netdev_notifier_changelowerstate_info {
2759 struct netdev_notifier_info info; /* must be first */
2760 void *lower_state_info; /* is lower dev state */
2763 struct netdev_notifier_pre_changeaddr_info {
2764 struct netdev_notifier_info info; /* must be first */
2765 const unsigned char *dev_addr;
2768 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2769 struct net_device *dev)
2772 info->extack = NULL;
2775 static inline struct net_device *
2776 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2781 static inline struct netlink_ext_ack *
2782 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2784 return info->extack;
2787 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2790 extern rwlock_t dev_base_lock; /* Device list lock */
2792 #define for_each_netdev(net, d) \
2793 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2794 #define for_each_netdev_reverse(net, d) \
2795 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2796 #define for_each_netdev_rcu(net, d) \
2797 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2798 #define for_each_netdev_safe(net, d, n) \
2799 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2800 #define for_each_netdev_continue(net, d) \
2801 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2802 #define for_each_netdev_continue_reverse(net, d) \
2803 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2805 #define for_each_netdev_continue_rcu(net, d) \
2806 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2807 #define for_each_netdev_in_bond_rcu(bond, slave) \
2808 for_each_netdev_rcu(&init_net, slave) \
2809 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2810 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2812 static inline struct net_device *next_net_device(struct net_device *dev)
2814 struct list_head *lh;
2818 lh = dev->dev_list.next;
2819 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2822 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2824 struct list_head *lh;
2828 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2829 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2832 static inline struct net_device *first_net_device(struct net *net)
2834 return list_empty(&net->dev_base_head) ? NULL :
2835 net_device_entry(net->dev_base_head.next);
2838 static inline struct net_device *first_net_device_rcu(struct net *net)
2840 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2842 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2845 int netdev_boot_setup_check(struct net_device *dev);
2846 unsigned long netdev_boot_base(const char *prefix, int unit);
2847 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2848 const char *hwaddr);
2849 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2850 void dev_add_pack(struct packet_type *pt);
2851 void dev_remove_pack(struct packet_type *pt);
2852 void __dev_remove_pack(struct packet_type *pt);
2853 void dev_add_offload(struct packet_offload *po);
2854 void dev_remove_offload(struct packet_offload *po);
2856 int dev_get_iflink(const struct net_device *dev);
2857 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2858 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2859 unsigned short mask);
2860 struct net_device *dev_get_by_name(struct net *net, const char *name);
2861 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2862 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2863 int dev_alloc_name(struct net_device *dev, const char *name);
2864 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2865 void dev_close(struct net_device *dev);
2866 void dev_close_many(struct list_head *head, bool unlink);
2867 void dev_disable_lro(struct net_device *dev);
2868 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2869 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2870 struct net_device *sb_dev);
2871 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2872 struct net_device *sb_dev);
2874 int dev_queue_xmit(struct sk_buff *skb);
2875 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2876 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2878 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2882 ret = __dev_direct_xmit(skb, queue_id);
2883 if (!dev_xmit_complete(ret))
2888 int register_netdevice(struct net_device *dev);
2889 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2890 void unregister_netdevice_many(struct list_head *head);
2891 static inline void unregister_netdevice(struct net_device *dev)
2893 unregister_netdevice_queue(dev, NULL);
2896 int netdev_refcnt_read(const struct net_device *dev);
2897 void free_netdev(struct net_device *dev);
2898 void netdev_freemem(struct net_device *dev);
2899 int init_dummy_netdev(struct net_device *dev);
2901 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2902 struct sk_buff *skb,
2904 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2906 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2907 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2908 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2909 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2910 int netdev_get_name(struct net *net, char *name, int ifindex);
2911 int dev_restart(struct net_device *dev);
2912 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2913 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2915 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2917 return NAPI_GRO_CB(skb)->data_offset;
2920 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2922 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2925 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2927 NAPI_GRO_CB(skb)->data_offset += len;
2930 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2931 unsigned int offset)
2933 return NAPI_GRO_CB(skb)->frag0 + offset;
2936 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2938 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2941 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2943 NAPI_GRO_CB(skb)->frag0 = NULL;
2944 NAPI_GRO_CB(skb)->frag0_len = 0;
2947 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2948 unsigned int offset)
2950 if (!pskb_may_pull(skb, hlen))
2953 skb_gro_frag0_invalidate(skb);
2954 return skb->data + offset;
2957 static inline void *skb_gro_network_header(struct sk_buff *skb)
2959 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2960 skb_network_offset(skb);
2963 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2964 const void *start, unsigned int len)
2966 if (NAPI_GRO_CB(skb)->csum_valid)
2967 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2968 csum_partial(start, len, 0));
2971 /* GRO checksum functions. These are logical equivalents of the normal
2972 * checksum functions (in skbuff.h) except that they operate on the GRO
2973 * offsets and fields in sk_buff.
2976 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2978 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2980 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2983 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2987 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2988 skb_checksum_start_offset(skb) <
2989 skb_gro_offset(skb)) &&
2990 !skb_at_gro_remcsum_start(skb) &&
2991 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2992 (!zero_okay || check));
2995 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2998 if (NAPI_GRO_CB(skb)->csum_valid &&
2999 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3002 NAPI_GRO_CB(skb)->csum = psum;
3004 return __skb_gro_checksum_complete(skb);
3007 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3009 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3010 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3011 NAPI_GRO_CB(skb)->csum_cnt--;
3013 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3014 * verified a new top level checksum or an encapsulated one
3015 * during GRO. This saves work if we fallback to normal path.
3017 __skb_incr_checksum_unnecessary(skb);
3021 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3024 __sum16 __ret = 0; \
3025 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3026 __ret = __skb_gro_checksum_validate_complete(skb, \
3027 compute_pseudo(skb, proto)); \
3029 skb_gro_incr_csum_unnecessary(skb); \
3033 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3034 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3036 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3038 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3040 #define skb_gro_checksum_simple_validate(skb) \
3041 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3043 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3045 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3046 !NAPI_GRO_CB(skb)->csum_valid);
3049 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3052 NAPI_GRO_CB(skb)->csum = ~pseudo;
3053 NAPI_GRO_CB(skb)->csum_valid = 1;
3056 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3058 if (__skb_gro_checksum_convert_check(skb)) \
3059 __skb_gro_checksum_convert(skb, \
3060 compute_pseudo(skb, proto)); \
3063 struct gro_remcsum {
3068 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3074 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3075 unsigned int off, size_t hdrlen,
3076 int start, int offset,
3077 struct gro_remcsum *grc,
3081 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3083 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3086 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3090 ptr = skb_gro_header_fast(skb, off);
3091 if (skb_gro_header_hard(skb, off + plen)) {
3092 ptr = skb_gro_header_slow(skb, off + plen, off);
3097 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3100 /* Adjust skb->csum since we changed the packet */
3101 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3103 grc->offset = off + hdrlen + offset;
3109 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3110 struct gro_remcsum *grc)
3113 size_t plen = grc->offset + sizeof(u16);
3118 ptr = skb_gro_header_fast(skb, grc->offset);
3119 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3120 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3125 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3128 #ifdef CONFIG_XFRM_OFFLOAD
3129 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3131 if (PTR_ERR(pp) != -EINPROGRESS)
3132 NAPI_GRO_CB(skb)->flush |= flush;
3134 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3137 struct gro_remcsum *grc)
3139 if (PTR_ERR(pp) != -EINPROGRESS) {
3140 NAPI_GRO_CB(skb)->flush |= flush;
3141 skb_gro_remcsum_cleanup(skb, grc);
3142 skb->remcsum_offload = 0;
3146 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3148 NAPI_GRO_CB(skb)->flush |= flush;
3150 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3153 struct gro_remcsum *grc)
3155 NAPI_GRO_CB(skb)->flush |= flush;
3156 skb_gro_remcsum_cleanup(skb, grc);
3157 skb->remcsum_offload = 0;
3161 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3162 unsigned short type,
3163 const void *daddr, const void *saddr,
3166 if (!dev->header_ops || !dev->header_ops->create)
3169 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3172 static inline int dev_parse_header(const struct sk_buff *skb,
3173 unsigned char *haddr)
3175 const struct net_device *dev = skb->dev;
3177 if (!dev->header_ops || !dev->header_ops->parse)
3179 return dev->header_ops->parse(skb, haddr);
3182 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3184 const struct net_device *dev = skb->dev;
3186 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3188 return dev->header_ops->parse_protocol(skb);
3191 /* ll_header must have at least hard_header_len allocated */
3192 static inline bool dev_validate_header(const struct net_device *dev,
3193 char *ll_header, int len)
3195 if (likely(len >= dev->hard_header_len))
3197 if (len < dev->min_header_len)
3200 if (capable(CAP_SYS_RAWIO)) {
3201 memset(ll_header + len, 0, dev->hard_header_len - len);
3205 if (dev->header_ops && dev->header_ops->validate)
3206 return dev->header_ops->validate(ll_header, len);
3211 static inline bool dev_has_header(const struct net_device *dev)
3213 return dev->header_ops && dev->header_ops->create;
3216 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3218 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3219 static inline int unregister_gifconf(unsigned int family)
3221 return register_gifconf(family, NULL);
3224 #ifdef CONFIG_NET_FLOW_LIMIT
3225 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3226 struct sd_flow_limit {
3228 unsigned int num_buckets;
3229 unsigned int history_head;
3230 u16 history[FLOW_LIMIT_HISTORY];
3234 extern int netdev_flow_limit_table_len;
3235 #endif /* CONFIG_NET_FLOW_LIMIT */
3238 * Incoming packets are placed on per-CPU queues
3240 struct softnet_data {
3241 struct list_head poll_list;
3242 struct sk_buff_head process_queue;
3245 unsigned int processed;
3246 unsigned int time_squeeze;
3247 unsigned int received_rps;
3249 struct softnet_data *rps_ipi_list;
3251 #ifdef CONFIG_NET_FLOW_LIMIT
3252 struct sd_flow_limit __rcu *flow_limit;
3254 struct Qdisc *output_queue;
3255 struct Qdisc **output_queue_tailp;
3256 struct sk_buff *completion_queue;
3257 #ifdef CONFIG_XFRM_OFFLOAD
3258 struct sk_buff_head xfrm_backlog;
3260 /* written and read only by owning cpu: */
3266 /* input_queue_head should be written by cpu owning this struct,
3267 * and only read by other cpus. Worth using a cache line.
3269 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3271 /* Elements below can be accessed between CPUs for RPS/RFS */
3272 call_single_data_t csd ____cacheline_aligned_in_smp;
3273 struct softnet_data *rps_ipi_next;
3275 unsigned int input_queue_tail;
3277 unsigned int dropped;
3278 struct sk_buff_head input_pkt_queue;
3279 struct napi_struct backlog;
3283 static inline void input_queue_head_incr(struct softnet_data *sd)
3286 sd->input_queue_head++;
3290 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3291 unsigned int *qtail)
3294 *qtail = ++sd->input_queue_tail;
3298 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3300 static inline int dev_recursion_level(void)
3302 return this_cpu_read(softnet_data.xmit.recursion);
3305 #define XMIT_RECURSION_LIMIT 8
3306 static inline bool dev_xmit_recursion(void)
3308 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3309 XMIT_RECURSION_LIMIT);
3312 static inline void dev_xmit_recursion_inc(void)
3314 __this_cpu_inc(softnet_data.xmit.recursion);
3317 static inline void dev_xmit_recursion_dec(void)
3319 __this_cpu_dec(softnet_data.xmit.recursion);
3322 void __netif_schedule(struct Qdisc *q);
3323 void netif_schedule_queue(struct netdev_queue *txq);
3325 static inline void netif_tx_schedule_all(struct net_device *dev)
3329 for (i = 0; i < dev->num_tx_queues; i++)
3330 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3333 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3335 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3339 * netif_start_queue - allow transmit
3340 * @dev: network device
3342 * Allow upper layers to call the device hard_start_xmit routine.
3344 static inline void netif_start_queue(struct net_device *dev)
3346 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3349 static inline void netif_tx_start_all_queues(struct net_device *dev)
3353 for (i = 0; i < dev->num_tx_queues; i++) {
3354 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3355 netif_tx_start_queue(txq);
3359 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3362 * netif_wake_queue - restart transmit
3363 * @dev: network device
3365 * Allow upper layers to call the device hard_start_xmit routine.
3366 * Used for flow control when transmit resources are available.
3368 static inline void netif_wake_queue(struct net_device *dev)
3370 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3373 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3377 for (i = 0; i < dev->num_tx_queues; i++) {
3378 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3379 netif_tx_wake_queue(txq);
3383 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3385 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3389 * netif_stop_queue - stop transmitted packets
3390 * @dev: network device
3392 * Stop upper layers calling the device hard_start_xmit routine.
3393 * Used for flow control when transmit resources are unavailable.
3395 static inline void netif_stop_queue(struct net_device *dev)
3397 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3400 void netif_tx_stop_all_queues(struct net_device *dev);
3402 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3404 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3408 * netif_queue_stopped - test if transmit queue is flowblocked
3409 * @dev: network device
3411 * Test if transmit queue on device is currently unable to send.
3413 static inline bool netif_queue_stopped(const struct net_device *dev)
3415 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3418 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3420 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3424 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3426 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3430 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3432 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3436 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3437 * @dev_queue: pointer to transmit queue
3439 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3440 * to give appropriate hint to the CPU.
3442 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3445 prefetchw(&dev_queue->dql.num_queued);
3450 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3451 * @dev_queue: pointer to transmit queue
3453 * BQL enabled drivers might use this helper in their TX completion path,
3454 * to give appropriate hint to the CPU.
3456 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3459 prefetchw(&dev_queue->dql.limit);
3463 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3467 dql_queued(&dev_queue->dql, bytes);
3469 if (likely(dql_avail(&dev_queue->dql) >= 0))
3472 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3475 * The XOFF flag must be set before checking the dql_avail below,
3476 * because in netdev_tx_completed_queue we update the dql_completed
3477 * before checking the XOFF flag.
3481 /* check again in case another CPU has just made room avail */
3482 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3483 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3487 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3488 * that they should not test BQL status themselves.
3489 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3491 * Returns true if the doorbell must be used to kick the NIC.
3493 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3499 dql_queued(&dev_queue->dql, bytes);
3501 return netif_tx_queue_stopped(dev_queue);
3503 netdev_tx_sent_queue(dev_queue, bytes);
3508 * netdev_sent_queue - report the number of bytes queued to hardware
3509 * @dev: network device
3510 * @bytes: number of bytes queued to the hardware device queue
3512 * Report the number of bytes queued for sending/completion to the network
3513 * device hardware queue. @bytes should be a good approximation and should
3514 * exactly match netdev_completed_queue() @bytes
3516 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3518 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3521 static inline bool __netdev_sent_queue(struct net_device *dev,
3525 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3529 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3530 unsigned int pkts, unsigned int bytes)
3533 if (unlikely(!bytes))
3536 dql_completed(&dev_queue->dql, bytes);
3539 * Without the memory barrier there is a small possiblity that
3540 * netdev_tx_sent_queue will miss the update and cause the queue to
3541 * be stopped forever
3545 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3548 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3549 netif_schedule_queue(dev_queue);
3554 * netdev_completed_queue - report bytes and packets completed by device
3555 * @dev: network device
3556 * @pkts: actual number of packets sent over the medium
3557 * @bytes: actual number of bytes sent over the medium
3559 * Report the number of bytes and packets transmitted by the network device
3560 * hardware queue over the physical medium, @bytes must exactly match the
3561 * @bytes amount passed to netdev_sent_queue()
3563 static inline void netdev_completed_queue(struct net_device *dev,
3564 unsigned int pkts, unsigned int bytes)
3566 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3569 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3572 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3578 * netdev_reset_queue - reset the packets and bytes count of a network device
3579 * @dev_queue: network device
3581 * Reset the bytes and packet count of a network device and clear the
3582 * software flow control OFF bit for this network device
3584 static inline void netdev_reset_queue(struct net_device *dev_queue)
3586 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3590 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3591 * @dev: network device
3592 * @queue_index: given tx queue index
3594 * Returns 0 if given tx queue index >= number of device tx queues,
3595 * otherwise returns the originally passed tx queue index.
3597 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3599 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3600 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3601 dev->name, queue_index,
3602 dev->real_num_tx_queues);
3610 * netif_running - test if up
3611 * @dev: network device
3613 * Test if the device has been brought up.
3615 static inline bool netif_running(const struct net_device *dev)
3617 return test_bit(__LINK_STATE_START, &dev->state);
3621 * Routines to manage the subqueues on a device. We only need start,
3622 * stop, and a check if it's stopped. All other device management is
3623 * done at the overall netdevice level.
3624 * Also test the device if we're multiqueue.
3628 * netif_start_subqueue - allow sending packets on subqueue
3629 * @dev: network device
3630 * @queue_index: sub queue index
3632 * Start individual transmit queue of a device with multiple transmit queues.
3634 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3636 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3638 netif_tx_start_queue(txq);
3642 * netif_stop_subqueue - stop sending packets on subqueue
3643 * @dev: network device
3644 * @queue_index: sub queue index
3646 * Stop individual transmit queue of a device with multiple transmit queues.
3648 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3650 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3651 netif_tx_stop_queue(txq);
3655 * __netif_subqueue_stopped - test status of subqueue
3656 * @dev: network device
3657 * @queue_index: sub queue index
3659 * Check individual transmit queue of a device with multiple transmit queues.
3661 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3664 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3666 return netif_tx_queue_stopped(txq);
3670 * netif_subqueue_stopped - test status of subqueue
3671 * @dev: network device
3672 * @skb: sub queue buffer pointer
3674 * Check individual transmit queue of a device with multiple transmit queues.
3676 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3677 struct sk_buff *skb)
3679 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3683 * netif_wake_subqueue - allow sending packets on subqueue
3684 * @dev: network device
3685 * @queue_index: sub queue index
3687 * Resume individual transmit queue of a device with multiple transmit queues.
3689 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3691 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3693 netif_tx_wake_queue(txq);
3697 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3699 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3700 u16 index, bool is_rxqs_map);
3703 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3704 * @j: CPU/Rx queue index
3705 * @mask: bitmask of all cpus/rx queues
3706 * @nr_bits: number of bits in the bitmask
3708 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3710 static inline bool netif_attr_test_mask(unsigned long j,
3711 const unsigned long *mask,
3712 unsigned int nr_bits)
3714 cpu_max_bits_warn(j, nr_bits);
3715 return test_bit(j, mask);
3719 * netif_attr_test_online - Test for online CPU/Rx queue
3720 * @j: CPU/Rx queue index
3721 * @online_mask: bitmask for CPUs/Rx queues that are online
3722 * @nr_bits: number of bits in the bitmask
3724 * Returns true if a CPU/Rx queue is online.
3726 static inline bool netif_attr_test_online(unsigned long j,
3727 const unsigned long *online_mask,
3728 unsigned int nr_bits)
3730 cpu_max_bits_warn(j, nr_bits);
3733 return test_bit(j, online_mask);
3735 return (j < nr_bits);
3739 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3740 * @n: CPU/Rx queue index
3741 * @srcp: the cpumask/Rx queue mask pointer
3742 * @nr_bits: number of bits in the bitmask
3744 * Returns >= nr_bits if no further CPUs/Rx queues set.
3746 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3747 unsigned int nr_bits)
3749 /* -1 is a legal arg here. */
3751 cpu_max_bits_warn(n, nr_bits);
3754 return find_next_bit(srcp, nr_bits, n + 1);
3760 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3761 * @n: CPU/Rx queue index
3762 * @src1p: the first CPUs/Rx queues mask pointer
3763 * @src2p: the second CPUs/Rx queues mask pointer
3764 * @nr_bits: number of bits in the bitmask
3766 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3768 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3769 const unsigned long *src2p,
3770 unsigned int nr_bits)
3772 /* -1 is a legal arg here. */
3774 cpu_max_bits_warn(n, nr_bits);
3777 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3779 return find_next_bit(src1p, nr_bits, n + 1);
3781 return find_next_bit(src2p, nr_bits, n + 1);
3786 static inline int netif_set_xps_queue(struct net_device *dev,
3787 const struct cpumask *mask,
3793 static inline int __netif_set_xps_queue(struct net_device *dev,
3794 const unsigned long *mask,
3795 u16 index, bool is_rxqs_map)
3802 * netif_is_multiqueue - test if device has multiple transmit queues
3803 * @dev: network device
3805 * Check if device has multiple transmit queues
3807 static inline bool netif_is_multiqueue(const struct net_device *dev)
3809 return dev->num_tx_queues > 1;
3812 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3815 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3817 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3820 dev->real_num_rx_queues = rxqs;
3825 static inline struct netdev_rx_queue *
3826 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3828 return dev->_rx + rxq;
3832 static inline unsigned int get_netdev_rx_queue_index(
3833 struct netdev_rx_queue *queue)
3835 struct net_device *dev = queue->dev;
3836 int index = queue - dev->_rx;
3838 BUG_ON(index >= dev->num_rx_queues);
3843 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3844 int netif_get_num_default_rss_queues(void);
3846 enum skb_free_reason {
3847 SKB_REASON_CONSUMED,
3851 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3852 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3855 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3856 * interrupt context or with hardware interrupts being disabled.
3857 * (in_irq() || irqs_disabled())
3859 * We provide four helpers that can be used in following contexts :
3861 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3862 * replacing kfree_skb(skb)
3864 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3865 * Typically used in place of consume_skb(skb) in TX completion path
3867 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3868 * replacing kfree_skb(skb)
3870 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3871 * and consumed a packet. Used in place of consume_skb(skb)
3873 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3875 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3878 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3880 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3883 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3885 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3888 static inline void dev_consume_skb_any(struct sk_buff *skb)
3890 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3893 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3894 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3895 int netif_rx(struct sk_buff *skb);
3896 int netif_rx_ni(struct sk_buff *skb);
3897 int netif_rx_any_context(struct sk_buff *skb);
3898 int netif_receive_skb(struct sk_buff *skb);
3899 int netif_receive_skb_core(struct sk_buff *skb);
3900 void netif_receive_skb_list(struct list_head *head);
3901 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3902 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3903 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3904 gro_result_t napi_gro_frags(struct napi_struct *napi);
3905 struct packet_offload *gro_find_receive_by_type(__be16 type);
3906 struct packet_offload *gro_find_complete_by_type(__be16 type);
3908 static inline void napi_free_frags(struct napi_struct *napi)
3910 kfree_skb(napi->skb);
3914 bool netdev_is_rx_handler_busy(struct net_device *dev);
3915 int netdev_rx_handler_register(struct net_device *dev,
3916 rx_handler_func_t *rx_handler,
3917 void *rx_handler_data);
3918 void netdev_rx_handler_unregister(struct net_device *dev);
3920 bool dev_valid_name(const char *name);
3921 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3922 bool *need_copyout);
3923 int dev_ifconf(struct net *net, struct ifconf *, int);
3924 int dev_ethtool(struct net *net, struct ifreq *);
3925 unsigned int dev_get_flags(const struct net_device *);
3926 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3927 struct netlink_ext_ack *extack);
3928 int dev_change_flags(struct net_device *dev, unsigned int flags,
3929 struct netlink_ext_ack *extack);
3930 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3931 unsigned int gchanges);
3932 int dev_change_name(struct net_device *, const char *);
3933 int dev_set_alias(struct net_device *, const char *, size_t);
3934 int dev_get_alias(const struct net_device *, char *, size_t);
3935 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3936 int __dev_set_mtu(struct net_device *, int);
3937 int dev_validate_mtu(struct net_device *dev, int mtu,
3938 struct netlink_ext_ack *extack);
3939 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3940 struct netlink_ext_ack *extack);
3941 int dev_set_mtu(struct net_device *, int);
3942 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3943 void dev_set_group(struct net_device *, int);
3944 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3945 struct netlink_ext_ack *extack);
3946 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3947 struct netlink_ext_ack *extack);
3948 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3949 struct netlink_ext_ack *extack);
3950 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3951 int dev_change_carrier(struct net_device *, bool new_carrier);
3952 int dev_get_phys_port_id(struct net_device *dev,
3953 struct netdev_phys_item_id *ppid);
3954 int dev_get_phys_port_name(struct net_device *dev,
3955 char *name, size_t len);
3956 int dev_get_port_parent_id(struct net_device *dev,
3957 struct netdev_phys_item_id *ppid, bool recurse);
3958 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3959 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3960 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3961 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3963 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3964 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3965 struct netdev_queue *txq, int *ret);
3967 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3968 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3969 int fd, int expected_fd, u32 flags);
3970 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3971 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3973 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3974 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3975 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3976 bool is_skb_forwardable(const struct net_device *dev,
3977 const struct sk_buff *skb);
3979 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3980 const struct sk_buff *skb,
3981 const bool check_mtu)
3983 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3986 if (!(dev->flags & IFF_UP))
3992 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3993 if (skb->len <= len)
3996 /* if TSO is enabled, we don't care about the length as the packet
3997 * could be forwarded without being segmented before
3999 if (skb_is_gso(skb))
4005 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4006 struct sk_buff *skb,
4007 const bool check_mtu)
4009 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4010 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4011 atomic_long_inc(&dev->rx_dropped);
4016 skb_scrub_packet(skb, true);
4021 bool dev_nit_active(struct net_device *dev);
4022 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4024 extern int netdev_budget;
4025 extern unsigned int netdev_budget_usecs;
4027 /* Called by rtnetlink.c:rtnl_unlock() */
4028 void netdev_run_todo(void);
4031 * dev_put - release reference to device
4032 * @dev: network device
4034 * Release reference to device to allow it to be freed.
4036 static inline void dev_put(struct net_device *dev)
4038 this_cpu_dec(*dev->pcpu_refcnt);
4042 * dev_hold - get reference to device
4043 * @dev: network device
4045 * Hold reference to device to keep it from being freed.
4047 static inline void dev_hold(struct net_device *dev)
4049 this_cpu_inc(*dev->pcpu_refcnt);
4052 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4053 * and _off may be called from IRQ context, but it is caller
4054 * who is responsible for serialization of these calls.
4056 * The name carrier is inappropriate, these functions should really be
4057 * called netif_lowerlayer_*() because they represent the state of any
4058 * kind of lower layer not just hardware media.
4061 void linkwatch_init_dev(struct net_device *dev);
4062 void linkwatch_fire_event(struct net_device *dev);
4063 void linkwatch_forget_dev(struct net_device *dev);
4066 * netif_carrier_ok - test if carrier present
4067 * @dev: network device
4069 * Check if carrier is present on device
4071 static inline bool netif_carrier_ok(const struct net_device *dev)
4073 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4076 unsigned long dev_trans_start(struct net_device *dev);
4078 void __netdev_watchdog_up(struct net_device *dev);
4080 void netif_carrier_on(struct net_device *dev);
4082 void netif_carrier_off(struct net_device *dev);
4085 * netif_dormant_on - mark device as dormant.
4086 * @dev: network device
4088 * Mark device as dormant (as per RFC2863).
4090 * The dormant state indicates that the relevant interface is not
4091 * actually in a condition to pass packets (i.e., it is not 'up') but is
4092 * in a "pending" state, waiting for some external event. For "on-
4093 * demand" interfaces, this new state identifies the situation where the
4094 * interface is waiting for events to place it in the up state.
4096 static inline void netif_dormant_on(struct net_device *dev)
4098 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4099 linkwatch_fire_event(dev);
4103 * netif_dormant_off - set device as not dormant.
4104 * @dev: network device
4106 * Device is not in dormant state.
4108 static inline void netif_dormant_off(struct net_device *dev)
4110 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4111 linkwatch_fire_event(dev);
4115 * netif_dormant - test if device is dormant
4116 * @dev: network device
4118 * Check if device is dormant.
4120 static inline bool netif_dormant(const struct net_device *dev)
4122 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4127 * netif_testing_on - mark device as under test.
4128 * @dev: network device
4130 * Mark device as under test (as per RFC2863).
4132 * The testing state indicates that some test(s) must be performed on
4133 * the interface. After completion, of the test, the interface state
4134 * will change to up, dormant, or down, as appropriate.
4136 static inline void netif_testing_on(struct net_device *dev)
4138 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4139 linkwatch_fire_event(dev);
4143 * netif_testing_off - set device as not under test.
4144 * @dev: network device
4146 * Device is not in testing state.
4148 static inline void netif_testing_off(struct net_device *dev)
4150 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4151 linkwatch_fire_event(dev);
4155 * netif_testing - test if device is under test
4156 * @dev: network device
4158 * Check if device is under test
4160 static inline bool netif_testing(const struct net_device *dev)
4162 return test_bit(__LINK_STATE_TESTING, &dev->state);
4167 * netif_oper_up - test if device is operational
4168 * @dev: network device
4170 * Check if carrier is operational
4172 static inline bool netif_oper_up(const struct net_device *dev)
4174 return (dev->operstate == IF_OPER_UP ||
4175 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4179 * netif_device_present - is device available or removed
4180 * @dev: network device
4182 * Check if device has not been removed from system.
4184 static inline bool netif_device_present(const struct net_device *dev)
4186 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4189 void netif_device_detach(struct net_device *dev);
4191 void netif_device_attach(struct net_device *dev);
4194 * Network interface message level settings
4199 NETIF_MSG_PROBE_BIT,
4201 NETIF_MSG_TIMER_BIT,
4202 NETIF_MSG_IFDOWN_BIT,
4204 NETIF_MSG_RX_ERR_BIT,
4205 NETIF_MSG_TX_ERR_BIT,
4206 NETIF_MSG_TX_QUEUED_BIT,
4208 NETIF_MSG_TX_DONE_BIT,
4209 NETIF_MSG_RX_STATUS_BIT,
4210 NETIF_MSG_PKTDATA_BIT,
4214 /* When you add a new bit above, update netif_msg_class_names array
4215 * in net/ethtool/common.c
4217 NETIF_MSG_CLASS_COUNT,
4219 /* Both ethtool_ops interface and internal driver implementation use u32 */
4220 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4222 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4223 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4225 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4226 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4227 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4228 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4229 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4230 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4231 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4232 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4233 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4234 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4235 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4236 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4237 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4238 #define NETIF_MSG_HW __NETIF_MSG(HW)
4239 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4241 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4242 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4243 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4244 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4245 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4246 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4247 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4248 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4249 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4250 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4251 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4252 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4253 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4254 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4255 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4257 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4260 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4261 return default_msg_enable_bits;
4262 if (debug_value == 0) /* no output */
4264 /* set low N bits */
4265 return (1U << debug_value) - 1;
4268 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4270 spin_lock(&txq->_xmit_lock);
4271 txq->xmit_lock_owner = cpu;
4274 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4276 __acquire(&txq->_xmit_lock);
4280 static inline void __netif_tx_release(struct netdev_queue *txq)
4282 __release(&txq->_xmit_lock);
4285 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4287 spin_lock_bh(&txq->_xmit_lock);
4288 txq->xmit_lock_owner = smp_processor_id();
4291 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4293 bool ok = spin_trylock(&txq->_xmit_lock);
4295 txq->xmit_lock_owner = smp_processor_id();
4299 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4301 txq->xmit_lock_owner = -1;
4302 spin_unlock(&txq->_xmit_lock);
4305 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4307 txq->xmit_lock_owner = -1;
4308 spin_unlock_bh(&txq->_xmit_lock);
4311 static inline void txq_trans_update(struct netdev_queue *txq)
4313 if (txq->xmit_lock_owner != -1)
4314 txq->trans_start = jiffies;
4317 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4318 static inline void netif_trans_update(struct net_device *dev)
4320 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4322 if (txq->trans_start != jiffies)
4323 txq->trans_start = jiffies;
4327 * netif_tx_lock - grab network device transmit lock
4328 * @dev: network device
4330 * Get network device transmit lock
4332 static inline void netif_tx_lock(struct net_device *dev)
4337 spin_lock(&dev->tx_global_lock);
4338 cpu = smp_processor_id();
4339 for (i = 0; i < dev->num_tx_queues; i++) {
4340 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4342 /* We are the only thread of execution doing a
4343 * freeze, but we have to grab the _xmit_lock in
4344 * order to synchronize with threads which are in
4345 * the ->hard_start_xmit() handler and already
4346 * checked the frozen bit.
4348 __netif_tx_lock(txq, cpu);
4349 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4350 __netif_tx_unlock(txq);
4354 static inline void netif_tx_lock_bh(struct net_device *dev)
4360 static inline void netif_tx_unlock(struct net_device *dev)
4364 for (i = 0; i < dev->num_tx_queues; i++) {
4365 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4367 /* No need to grab the _xmit_lock here. If the
4368 * queue is not stopped for another reason, we
4371 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4372 netif_schedule_queue(txq);
4374 spin_unlock(&dev->tx_global_lock);
4377 static inline void netif_tx_unlock_bh(struct net_device *dev)
4379 netif_tx_unlock(dev);
4383 #define HARD_TX_LOCK(dev, txq, cpu) { \
4384 if ((dev->features & NETIF_F_LLTX) == 0) { \
4385 __netif_tx_lock(txq, cpu); \
4387 __netif_tx_acquire(txq); \
4391 #define HARD_TX_TRYLOCK(dev, txq) \
4392 (((dev->features & NETIF_F_LLTX) == 0) ? \
4393 __netif_tx_trylock(txq) : \
4394 __netif_tx_acquire(txq))
4396 #define HARD_TX_UNLOCK(dev, txq) { \
4397 if ((dev->features & NETIF_F_LLTX) == 0) { \
4398 __netif_tx_unlock(txq); \
4400 __netif_tx_release(txq); \
4404 static inline void netif_tx_disable(struct net_device *dev)
4410 cpu = smp_processor_id();
4411 spin_lock(&dev->tx_global_lock);
4412 for (i = 0; i < dev->num_tx_queues; i++) {
4413 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4415 __netif_tx_lock(txq, cpu);
4416 netif_tx_stop_queue(txq);
4417 __netif_tx_unlock(txq);
4419 spin_unlock(&dev->tx_global_lock);
4423 static inline void netif_addr_lock(struct net_device *dev)
4425 unsigned char nest_level = 0;
4427 #ifdef CONFIG_LOCKDEP
4428 nest_level = dev->nested_level;
4430 spin_lock_nested(&dev->addr_list_lock, nest_level);
4433 static inline void netif_addr_lock_bh(struct net_device *dev)
4435 unsigned char nest_level = 0;
4437 #ifdef CONFIG_LOCKDEP
4438 nest_level = dev->nested_level;
4441 spin_lock_nested(&dev->addr_list_lock, nest_level);
4444 static inline void netif_addr_unlock(struct net_device *dev)
4446 spin_unlock(&dev->addr_list_lock);
4449 static inline void netif_addr_unlock_bh(struct net_device *dev)
4451 spin_unlock_bh(&dev->addr_list_lock);
4455 * dev_addrs walker. Should be used only for read access. Call with
4456 * rcu_read_lock held.
4458 #define for_each_dev_addr(dev, ha) \
4459 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4461 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4463 void ether_setup(struct net_device *dev);
4465 /* Support for loadable net-drivers */
4466 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4467 unsigned char name_assign_type,
4468 void (*setup)(struct net_device *),
4469 unsigned int txqs, unsigned int rxqs);
4470 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4471 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4473 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4474 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4477 int register_netdev(struct net_device *dev);
4478 void unregister_netdev(struct net_device *dev);
4480 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4482 /* General hardware address lists handling functions */
4483 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4484 struct netdev_hw_addr_list *from_list, int addr_len);
4485 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4486 struct netdev_hw_addr_list *from_list, int addr_len);
4487 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4488 struct net_device *dev,
4489 int (*sync)(struct net_device *, const unsigned char *),
4490 int (*unsync)(struct net_device *,
4491 const unsigned char *));
4492 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4493 struct net_device *dev,
4494 int (*sync)(struct net_device *,
4495 const unsigned char *, int),
4496 int (*unsync)(struct net_device *,
4497 const unsigned char *, int));
4498 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4499 struct net_device *dev,
4500 int (*unsync)(struct net_device *,
4501 const unsigned char *, int));
4502 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4503 struct net_device *dev,
4504 int (*unsync)(struct net_device *,
4505 const unsigned char *));
4506 void __hw_addr_init(struct netdev_hw_addr_list *list);
4508 /* Functions used for device addresses handling */
4509 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4510 unsigned char addr_type);
4511 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4512 unsigned char addr_type);
4513 void dev_addr_flush(struct net_device *dev);
4514 int dev_addr_init(struct net_device *dev);
4516 /* Functions used for unicast addresses handling */
4517 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4518 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4519 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4520 int dev_uc_sync(struct net_device *to, struct net_device *from);
4521 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4522 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4523 void dev_uc_flush(struct net_device *dev);
4524 void dev_uc_init(struct net_device *dev);
4527 * __dev_uc_sync - Synchonize device's unicast list
4528 * @dev: device to sync
4529 * @sync: function to call if address should be added
4530 * @unsync: function to call if address should be removed
4532 * Add newly added addresses to the interface, and release
4533 * addresses that have been deleted.
4535 static inline int __dev_uc_sync(struct net_device *dev,
4536 int (*sync)(struct net_device *,
4537 const unsigned char *),
4538 int (*unsync)(struct net_device *,
4539 const unsigned char *))
4541 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4545 * __dev_uc_unsync - Remove synchronized addresses from device
4546 * @dev: device to sync
4547 * @unsync: function to call if address should be removed
4549 * Remove all addresses that were added to the device by dev_uc_sync().
4551 static inline void __dev_uc_unsync(struct net_device *dev,
4552 int (*unsync)(struct net_device *,
4553 const unsigned char *))
4555 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4558 /* Functions used for multicast addresses handling */
4559 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4560 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4561 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4562 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4563 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4564 int dev_mc_sync(struct net_device *to, struct net_device *from);
4565 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4566 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4567 void dev_mc_flush(struct net_device *dev);
4568 void dev_mc_init(struct net_device *dev);
4571 * __dev_mc_sync - Synchonize device's multicast list
4572 * @dev: device to sync
4573 * @sync: function to call if address should be added
4574 * @unsync: function to call if address should be removed
4576 * Add newly added addresses to the interface, and release
4577 * addresses that have been deleted.
4579 static inline int __dev_mc_sync(struct net_device *dev,
4580 int (*sync)(struct net_device *,
4581 const unsigned char *),
4582 int (*unsync)(struct net_device *,
4583 const unsigned char *))
4585 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4589 * __dev_mc_unsync - Remove synchronized addresses from device
4590 * @dev: device to sync
4591 * @unsync: function to call if address should be removed
4593 * Remove all addresses that were added to the device by dev_mc_sync().
4595 static inline void __dev_mc_unsync(struct net_device *dev,
4596 int (*unsync)(struct net_device *,
4597 const unsigned char *))
4599 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4602 /* Functions used for secondary unicast and multicast support */
4603 void dev_set_rx_mode(struct net_device *dev);
4604 void __dev_set_rx_mode(struct net_device *dev);
4605 int dev_set_promiscuity(struct net_device *dev, int inc);
4606 int dev_set_allmulti(struct net_device *dev, int inc);
4607 void netdev_state_change(struct net_device *dev);
4608 void __netdev_notify_peers(struct net_device *dev);
4609 void netdev_notify_peers(struct net_device *dev);
4610 void netdev_features_change(struct net_device *dev);
4611 /* Load a device via the kmod */
4612 void dev_load(struct net *net, const char *name);
4613 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4614 struct rtnl_link_stats64 *storage);
4615 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4616 const struct net_device_stats *netdev_stats);
4617 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4618 const struct pcpu_sw_netstats __percpu *netstats);
4619 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4621 extern int netdev_max_backlog;
4622 extern int netdev_tstamp_prequeue;
4623 extern int weight_p;
4624 extern int dev_weight_rx_bias;
4625 extern int dev_weight_tx_bias;
4626 extern int dev_rx_weight;
4627 extern int dev_tx_weight;
4628 extern int gro_normal_batch;
4631 NESTED_SYNC_IMM_BIT,
4632 NESTED_SYNC_TODO_BIT,
4635 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4636 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4638 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4639 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4641 struct netdev_nested_priv {
4642 unsigned char flags;
4646 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4647 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4648 struct list_head **iter);
4649 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4650 struct list_head **iter);
4652 #ifdef CONFIG_LOCKDEP
4653 static LIST_HEAD(net_unlink_list);
4655 static inline void net_unlink_todo(struct net_device *dev)
4657 if (list_empty(&dev->unlink_list))
4658 list_add_tail(&dev->unlink_list, &net_unlink_list);
4662 /* iterate through upper list, must be called under RCU read lock */
4663 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4664 for (iter = &(dev)->adj_list.upper, \
4665 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4667 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4669 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4670 int (*fn)(struct net_device *upper_dev,
4671 struct netdev_nested_priv *priv),
4672 struct netdev_nested_priv *priv);
4674 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4675 struct net_device *upper_dev);
4677 bool netdev_has_any_upper_dev(struct net_device *dev);
4679 void *netdev_lower_get_next_private(struct net_device *dev,
4680 struct list_head **iter);
4681 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4682 struct list_head **iter);
4684 #define netdev_for_each_lower_private(dev, priv, iter) \
4685 for (iter = (dev)->adj_list.lower.next, \
4686 priv = netdev_lower_get_next_private(dev, &(iter)); \
4688 priv = netdev_lower_get_next_private(dev, &(iter)))
4690 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4691 for (iter = &(dev)->adj_list.lower, \
4692 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4694 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4696 void *netdev_lower_get_next(struct net_device *dev,
4697 struct list_head **iter);
4699 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4700 for (iter = (dev)->adj_list.lower.next, \
4701 ldev = netdev_lower_get_next(dev, &(iter)); \
4703 ldev = netdev_lower_get_next(dev, &(iter)))
4705 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4706 struct list_head **iter);
4707 int netdev_walk_all_lower_dev(struct net_device *dev,
4708 int (*fn)(struct net_device *lower_dev,
4709 struct netdev_nested_priv *priv),
4710 struct netdev_nested_priv *priv);
4711 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4712 int (*fn)(struct net_device *lower_dev,
4713 struct netdev_nested_priv *priv),
4714 struct netdev_nested_priv *priv);
4716 void *netdev_adjacent_get_private(struct list_head *adj_list);
4717 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4718 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4719 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4720 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4721 struct netlink_ext_ack *extack);
4722 int netdev_master_upper_dev_link(struct net_device *dev,
4723 struct net_device *upper_dev,
4724 void *upper_priv, void *upper_info,
4725 struct netlink_ext_ack *extack);
4726 void netdev_upper_dev_unlink(struct net_device *dev,
4727 struct net_device *upper_dev);
4728 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4729 struct net_device *new_dev,
4730 struct net_device *dev,
4731 struct netlink_ext_ack *extack);
4732 void netdev_adjacent_change_commit(struct net_device *old_dev,
4733 struct net_device *new_dev,
4734 struct net_device *dev);
4735 void netdev_adjacent_change_abort(struct net_device *old_dev,
4736 struct net_device *new_dev,
4737 struct net_device *dev);
4738 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4739 void *netdev_lower_dev_get_private(struct net_device *dev,
4740 struct net_device *lower_dev);
4741 void netdev_lower_state_changed(struct net_device *lower_dev,
4742 void *lower_state_info);
4744 /* RSS keys are 40 or 52 bytes long */
4745 #define NETDEV_RSS_KEY_LEN 52
4746 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4747 void netdev_rss_key_fill(void *buffer, size_t len);
4749 int skb_checksum_help(struct sk_buff *skb);
4750 int skb_crc32c_csum_help(struct sk_buff *skb);
4751 int skb_csum_hwoffload_help(struct sk_buff *skb,
4752 const netdev_features_t features);
4754 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4755 netdev_features_t features, bool tx_path);
4756 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4757 netdev_features_t features);
4759 struct netdev_bonding_info {
4764 struct netdev_notifier_bonding_info {
4765 struct netdev_notifier_info info; /* must be first */
4766 struct netdev_bonding_info bonding_info;
4769 void netdev_bonding_info_change(struct net_device *dev,
4770 struct netdev_bonding_info *bonding_info);
4772 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4773 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4775 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4782 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4784 return __skb_gso_segment(skb, features, true);
4786 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4788 static inline bool can_checksum_protocol(netdev_features_t features,
4791 if (protocol == htons(ETH_P_FCOE))
4792 return !!(features & NETIF_F_FCOE_CRC);
4794 /* Assume this is an IP checksum (not SCTP CRC) */
4796 if (features & NETIF_F_HW_CSUM) {
4797 /* Can checksum everything */
4802 case htons(ETH_P_IP):
4803 return !!(features & NETIF_F_IP_CSUM);
4804 case htons(ETH_P_IPV6):
4805 return !!(features & NETIF_F_IPV6_CSUM);
4812 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4814 static inline void netdev_rx_csum_fault(struct net_device *dev,
4815 struct sk_buff *skb)
4819 /* rx skb timestamps */
4820 void net_enable_timestamp(void);
4821 void net_disable_timestamp(void);
4823 #ifdef CONFIG_PROC_FS
4824 int __init dev_proc_init(void);
4826 #define dev_proc_init() 0
4829 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4830 struct sk_buff *skb, struct net_device *dev,
4833 __this_cpu_write(softnet_data.xmit.more, more);
4834 return ops->ndo_start_xmit(skb, dev);
4837 static inline bool netdev_xmit_more(void)
4839 return __this_cpu_read(softnet_data.xmit.more);
4842 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4843 struct netdev_queue *txq, bool more)
4845 const struct net_device_ops *ops = dev->netdev_ops;
4848 rc = __netdev_start_xmit(ops, skb, dev, more);
4849 if (rc == NETDEV_TX_OK)
4850 txq_trans_update(txq);
4855 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4857 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4860 extern const struct kobj_ns_type_operations net_ns_type_operations;
4862 const char *netdev_drivername(const struct net_device *dev);
4864 void linkwatch_run_queue(void);
4866 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4867 netdev_features_t f2)
4869 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4870 if (f1 & NETIF_F_HW_CSUM)
4871 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4873 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4879 static inline netdev_features_t netdev_get_wanted_features(
4880 struct net_device *dev)
4882 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4884 netdev_features_t netdev_increment_features(netdev_features_t all,
4885 netdev_features_t one, netdev_features_t mask);
4887 /* Allow TSO being used on stacked device :
4888 * Performing the GSO segmentation before last device
4889 * is a performance improvement.
4891 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4892 netdev_features_t mask)
4894 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4897 int __netdev_update_features(struct net_device *dev);
4898 void netdev_update_features(struct net_device *dev);
4899 void netdev_change_features(struct net_device *dev);
4901 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4902 struct net_device *dev);
4904 netdev_features_t passthru_features_check(struct sk_buff *skb,
4905 struct net_device *dev,
4906 netdev_features_t features);
4907 netdev_features_t netif_skb_features(struct sk_buff *skb);
4909 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4911 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4913 /* check flags correspondence */
4914 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4915 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4916 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4917 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4918 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4919 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4920 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4921 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4922 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4923 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4924 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4925 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4926 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4927 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4928 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4929 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4930 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4931 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4932 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4934 return (features & feature) == feature;
4937 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4939 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4940 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4943 static inline bool netif_needs_gso(struct sk_buff *skb,
4944 netdev_features_t features)
4946 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4947 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4948 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4951 static inline void netif_set_gso_max_size(struct net_device *dev,
4954 dev->gso_max_size = size;
4957 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4958 int pulled_hlen, u16 mac_offset,
4961 skb->protocol = protocol;
4962 skb->encapsulation = 1;
4963 skb_push(skb, pulled_hlen);
4964 skb_reset_transport_header(skb);
4965 skb->mac_header = mac_offset;
4966 skb->network_header = skb->mac_header + mac_len;
4967 skb->mac_len = mac_len;
4970 static inline bool netif_is_macsec(const struct net_device *dev)
4972 return dev->priv_flags & IFF_MACSEC;
4975 static inline bool netif_is_macvlan(const struct net_device *dev)
4977 return dev->priv_flags & IFF_MACVLAN;
4980 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4982 return dev->priv_flags & IFF_MACVLAN_PORT;
4985 static inline bool netif_is_bond_master(const struct net_device *dev)
4987 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4990 static inline bool netif_is_bond_slave(const struct net_device *dev)
4992 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4995 static inline bool netif_supports_nofcs(struct net_device *dev)
4997 return dev->priv_flags & IFF_SUPP_NOFCS;
5000 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5002 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5005 static inline bool netif_is_l3_master(const struct net_device *dev)
5007 return dev->priv_flags & IFF_L3MDEV_MASTER;
5010 static inline bool netif_is_l3_slave(const struct net_device *dev)
5012 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5015 static inline bool netif_is_bridge_master(const struct net_device *dev)
5017 return dev->priv_flags & IFF_EBRIDGE;
5020 static inline bool netif_is_bridge_port(const struct net_device *dev)
5022 return dev->priv_flags & IFF_BRIDGE_PORT;
5025 static inline bool netif_is_ovs_master(const struct net_device *dev)
5027 return dev->priv_flags & IFF_OPENVSWITCH;
5030 static inline bool netif_is_ovs_port(const struct net_device *dev)
5032 return dev->priv_flags & IFF_OVS_DATAPATH;
5035 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5037 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5040 static inline bool netif_is_team_master(const struct net_device *dev)
5042 return dev->priv_flags & IFF_TEAM;
5045 static inline bool netif_is_team_port(const struct net_device *dev)
5047 return dev->priv_flags & IFF_TEAM_PORT;
5050 static inline bool netif_is_lag_master(const struct net_device *dev)
5052 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5055 static inline bool netif_is_lag_port(const struct net_device *dev)
5057 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5060 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5062 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5065 static inline bool netif_is_failover(const struct net_device *dev)
5067 return dev->priv_flags & IFF_FAILOVER;
5070 static inline bool netif_is_failover_slave(const struct net_device *dev)
5072 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5075 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5076 static inline void netif_keep_dst(struct net_device *dev)
5078 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5081 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5082 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5084 /* TODO: reserve and use an additional IFF bit, if we get more users */
5085 return dev->priv_flags & IFF_MACSEC;
5088 extern struct pernet_operations __net_initdata loopback_net_ops;
5090 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5092 /* netdev_printk helpers, similar to dev_printk */
5094 static inline const char *netdev_name(const struct net_device *dev)
5096 if (!dev->name[0] || strchr(dev->name, '%'))
5097 return "(unnamed net_device)";
5101 static inline bool netdev_unregistering(const struct net_device *dev)
5103 return dev->reg_state == NETREG_UNREGISTERING;
5106 static inline const char *netdev_reg_state(const struct net_device *dev)
5108 switch (dev->reg_state) {
5109 case NETREG_UNINITIALIZED: return " (uninitialized)";
5110 case NETREG_REGISTERED: return "";
5111 case NETREG_UNREGISTERING: return " (unregistering)";
5112 case NETREG_UNREGISTERED: return " (unregistered)";
5113 case NETREG_RELEASED: return " (released)";
5114 case NETREG_DUMMY: return " (dummy)";
5117 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5118 return " (unknown)";
5121 __printf(3, 4) __cold
5122 void netdev_printk(const char *level, const struct net_device *dev,
5123 const char *format, ...);
5124 __printf(2, 3) __cold
5125 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5126 __printf(2, 3) __cold
5127 void netdev_alert(const struct net_device *dev, const char *format, ...);
5128 __printf(2, 3) __cold
5129 void netdev_crit(const struct net_device *dev, const char *format, ...);
5130 __printf(2, 3) __cold
5131 void netdev_err(const struct net_device *dev, const char *format, ...);
5132 __printf(2, 3) __cold
5133 void netdev_warn(const struct net_device *dev, const char *format, ...);
5134 __printf(2, 3) __cold
5135 void netdev_notice(const struct net_device *dev, const char *format, ...);
5136 __printf(2, 3) __cold
5137 void netdev_info(const struct net_device *dev, const char *format, ...);
5139 #define netdev_level_once(level, dev, fmt, ...) \
5141 static bool __print_once __read_mostly; \
5143 if (!__print_once) { \
5144 __print_once = true; \
5145 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5149 #define netdev_emerg_once(dev, fmt, ...) \
5150 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5151 #define netdev_alert_once(dev, fmt, ...) \
5152 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5153 #define netdev_crit_once(dev, fmt, ...) \
5154 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5155 #define netdev_err_once(dev, fmt, ...) \
5156 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5157 #define netdev_warn_once(dev, fmt, ...) \
5158 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5159 #define netdev_notice_once(dev, fmt, ...) \
5160 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5161 #define netdev_info_once(dev, fmt, ...) \
5162 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5164 #define MODULE_ALIAS_NETDEV(device) \
5165 MODULE_ALIAS("netdev-" device)
5167 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5168 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5169 #define netdev_dbg(__dev, format, args...) \
5171 dynamic_netdev_dbg(__dev, format, ##args); \
5173 #elif defined(DEBUG)
5174 #define netdev_dbg(__dev, format, args...) \
5175 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5177 #define netdev_dbg(__dev, format, args...) \
5180 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5184 #if defined(VERBOSE_DEBUG)
5185 #define netdev_vdbg netdev_dbg
5188 #define netdev_vdbg(dev, format, args...) \
5191 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5197 * netdev_WARN() acts like dev_printk(), but with the key difference
5198 * of using a WARN/WARN_ON to get the message out, including the
5199 * file/line information and a backtrace.
5201 #define netdev_WARN(dev, format, args...) \
5202 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5203 netdev_reg_state(dev), ##args)
5205 #define netdev_WARN_ONCE(dev, format, args...) \
5206 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5207 netdev_reg_state(dev), ##args)
5209 /* netif printk helpers, similar to netdev_printk */
5211 #define netif_printk(priv, type, level, dev, fmt, args...) \
5213 if (netif_msg_##type(priv)) \
5214 netdev_printk(level, (dev), fmt, ##args); \
5217 #define netif_level(level, priv, type, dev, fmt, args...) \
5219 if (netif_msg_##type(priv)) \
5220 netdev_##level(dev, fmt, ##args); \
5223 #define netif_emerg(priv, type, dev, fmt, args...) \
5224 netif_level(emerg, priv, type, dev, fmt, ##args)
5225 #define netif_alert(priv, type, dev, fmt, args...) \
5226 netif_level(alert, priv, type, dev, fmt, ##args)
5227 #define netif_crit(priv, type, dev, fmt, args...) \
5228 netif_level(crit, priv, type, dev, fmt, ##args)
5229 #define netif_err(priv, type, dev, fmt, args...) \
5230 netif_level(err, priv, type, dev, fmt, ##args)
5231 #define netif_warn(priv, type, dev, fmt, args...) \
5232 netif_level(warn, priv, type, dev, fmt, ##args)
5233 #define netif_notice(priv, type, dev, fmt, args...) \
5234 netif_level(notice, priv, type, dev, fmt, ##args)
5235 #define netif_info(priv, type, dev, fmt, args...) \
5236 netif_level(info, priv, type, dev, fmt, ##args)
5238 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5239 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5240 #define netif_dbg(priv, type, netdev, format, args...) \
5242 if (netif_msg_##type(priv)) \
5243 dynamic_netdev_dbg(netdev, format, ##args); \
5245 #elif defined(DEBUG)
5246 #define netif_dbg(priv, type, dev, format, args...) \
5247 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5249 #define netif_dbg(priv, type, dev, format, args...) \
5252 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5257 /* if @cond then downgrade to debug, else print at @level */
5258 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5261 netif_dbg(priv, type, netdev, fmt, ##args); \
5263 netif_ ## level(priv, type, netdev, fmt, ##args); \
5266 #if defined(VERBOSE_DEBUG)
5267 #define netif_vdbg netif_dbg
5269 #define netif_vdbg(priv, type, dev, format, args...) \
5272 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5278 * The list of packet types we will receive (as opposed to discard)
5279 * and the routines to invoke.
5281 * Why 16. Because with 16 the only overlap we get on a hash of the
5282 * low nibble of the protocol value is RARP/SNAP/X.25.
5296 #define PTYPE_HASH_SIZE (16)
5297 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5299 extern struct net_device *blackhole_netdev;
5301 #endif /* _LINUX_NETDEVICE_H */