2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
44 #include <net/dcbnl.h>
46 #include <net/netprio_cgroup.h>
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
64 /* 802.15.4 specific */
67 /* UDP Tunnel offloads */
68 struct udp_tunnel_info;
72 void netdev_set_default_ethtool_ops(struct net_device *dev,
73 const struct ethtool_ops *ops);
75 /* Backlog congestion levels */
76 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
77 #define NET_RX_DROP 1 /* packet dropped */
80 * Transmit return codes: transmit return codes originate from three different
83 * - qdisc return codes
84 * - driver transmit return codes
87 * Drivers are allowed to return any one of those in their hard_start_xmit()
88 * function. Real network devices commonly used with qdiscs should only return
89 * the driver transmit return codes though - when qdiscs are used, the actual
90 * transmission happens asynchronously, so the value is not propagated to
91 * higher layers. Virtual network devices transmit synchronously; in this case
92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
93 * others are propagated to higher layers.
96 /* qdisc ->enqueue() return codes. */
97 #define NET_XMIT_SUCCESS 0x00
98 #define NET_XMIT_DROP 0x01 /* skb dropped */
99 #define NET_XMIT_CN 0x02 /* congestion notification */
100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103 * indicates that the device will soon be dropping packets, or already drops
104 * some packets of the same priority; prompting us to send less aggressively. */
105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
108 /* Driver transmit return codes */
109 #define NETDEV_TX_MASK 0xf0
112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
113 NETDEV_TX_OK = 0x00, /* driver took care of packet */
114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
116 typedef enum netdev_tx netdev_tx_t;
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 static inline bool dev_xmit_complete(int rc)
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 if (likely(rc < NET_XMIT_MASK))
137 * Compute the worst-case header length according to the protocols
141 #if defined(CONFIG_HYPERV_NET)
142 # define LL_MAX_HEADER 128
143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
144 # if defined(CONFIG_MAC80211_MESH)
145 # define LL_MAX_HEADER 128
147 # define LL_MAX_HEADER 96
150 # define LL_MAX_HEADER 32
153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
155 #define MAX_HEADER LL_MAX_HEADER
157 #define MAX_HEADER (LL_MAX_HEADER + 48)
161 * Old network device statistics. Fields are native words
162 * (unsigned long) so they can be read and written atomically.
165 struct net_device_stats {
166 unsigned long rx_packets;
167 unsigned long tx_packets;
168 unsigned long rx_bytes;
169 unsigned long tx_bytes;
170 unsigned long rx_errors;
171 unsigned long tx_errors;
172 unsigned long rx_dropped;
173 unsigned long tx_dropped;
174 unsigned long multicast;
175 unsigned long collisions;
176 unsigned long rx_length_errors;
177 unsigned long rx_over_errors;
178 unsigned long rx_crc_errors;
179 unsigned long rx_frame_errors;
180 unsigned long rx_fifo_errors;
181 unsigned long rx_missed_errors;
182 unsigned long tx_aborted_errors;
183 unsigned long tx_carrier_errors;
184 unsigned long tx_fifo_errors;
185 unsigned long tx_heartbeat_errors;
186 unsigned long tx_window_errors;
187 unsigned long rx_compressed;
188 unsigned long tx_compressed;
192 #include <linux/cache.h>
193 #include <linux/skbuff.h>
196 #include <linux/static_key.h>
197 extern struct static_key rps_needed;
198 extern struct static_key rfs_needed;
205 struct netdev_hw_addr {
206 struct list_head list;
207 unsigned char addr[MAX_ADDR_LEN];
209 #define NETDEV_HW_ADDR_T_LAN 1
210 #define NETDEV_HW_ADDR_T_SAN 2
211 #define NETDEV_HW_ADDR_T_SLAVE 3
212 #define NETDEV_HW_ADDR_T_UNICAST 4
213 #define NETDEV_HW_ADDR_T_MULTICAST 5
218 struct rcu_head rcu_head;
221 struct netdev_hw_addr_list {
222 struct list_head list;
226 #define netdev_hw_addr_list_count(l) ((l)->count)
227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228 #define netdev_hw_addr_list_for_each(ha, l) \
229 list_for_each_entry(ha, &(l)->list, list)
231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233 #define netdev_for_each_uc_addr(ha, dev) \
234 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
238 #define netdev_for_each_mc_addr(ha, dev) \
239 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
245 /* cached hardware header; allow for machine alignment needs. */
246 #define HH_DATA_MOD 16
247 #define HH_DATA_OFF(__len) \
248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
249 #define HH_DATA_ALIGN(__len) \
250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
256 * dev->hard_header_len ? (dev->hard_header_len +
257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
259 * We could use other alignment values, but we must maintain the
260 * relationship HH alignment <= LL alignment.
262 #define LL_RESERVED_SPACE(dev) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
270 const void *saddr, unsigned int len);
271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
276 bool (*validate)(const char *ll_header, unsigned int len);
279 /* These flag bits are private to the generic network queueing
280 * layer; they may not be explicitly referenced by any other
284 enum netdev_state_t {
286 __LINK_STATE_PRESENT,
287 __LINK_STATE_NOCARRIER,
288 __LINK_STATE_LINKWATCH_PENDING,
289 __LINK_STATE_DORMANT,
294 * This structure holds boot-time configured netdevice settings. They
295 * are then used in the device probing.
297 struct netdev_boot_setup {
301 #define NETDEV_BOOT_SETUP_MAX 8
303 int __init netdev_boot_setup(char *str);
306 struct list_head list;
311 * Structure for NAPI scheduling similar to tasklet but with weighting
313 #define GRO_HASH_BUCKETS 8
315 /* The poll_list must only be managed by the entity which
316 * changes the state of the NAPI_STATE_SCHED bit. This means
317 * whoever atomically sets that bit can add this napi_struct
318 * to the per-CPU poll_list, and whoever clears that bit
319 * can remove from the list right before clearing the bit.
321 struct list_head poll_list;
325 unsigned int gro_count;
326 int (*poll)(struct napi_struct *, int);
327 #ifdef CONFIG_NETPOLL
330 struct net_device *dev;
331 struct gro_list gro_hash[GRO_HASH_BUCKETS];
333 struct hrtimer timer;
334 struct list_head dev_list;
335 struct hlist_node napi_hash_node;
336 unsigned int napi_id;
340 NAPI_STATE_SCHED, /* Poll is scheduled */
341 NAPI_STATE_MISSED, /* reschedule a napi */
342 NAPI_STATE_DISABLE, /* Disable pending */
343 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
344 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
345 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
346 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
350 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
351 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
352 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
353 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
354 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
355 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
356 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
367 typedef enum gro_result gro_result_t;
370 * enum rx_handler_result - Possible return values for rx_handlers.
371 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
373 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
374 * case skb->dev was changed by rx_handler.
375 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
376 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
378 * rx_handlers are functions called from inside __netif_receive_skb(), to do
379 * special processing of the skb, prior to delivery to protocol handlers.
381 * Currently, a net_device can only have a single rx_handler registered. Trying
382 * to register a second rx_handler will return -EBUSY.
384 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
385 * To unregister a rx_handler on a net_device, use
386 * netdev_rx_handler_unregister().
388 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
391 * If the rx_handler consumed the skb in some way, it should return
392 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
393 * the skb to be delivered in some other way.
395 * If the rx_handler changed skb->dev, to divert the skb to another
396 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
397 * new device will be called if it exists.
399 * If the rx_handler decides the skb should be ignored, it should return
400 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
401 * are registered on exact device (ptype->dev == skb->dev).
403 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
404 * delivered, it should return RX_HANDLER_PASS.
406 * A device without a registered rx_handler will behave as if rx_handler
407 * returned RX_HANDLER_PASS.
410 enum rx_handler_result {
416 typedef enum rx_handler_result rx_handler_result_t;
417 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
419 void __napi_schedule(struct napi_struct *n);
420 void __napi_schedule_irqoff(struct napi_struct *n);
422 static inline bool napi_disable_pending(struct napi_struct *n)
424 return test_bit(NAPI_STATE_DISABLE, &n->state);
427 bool napi_schedule_prep(struct napi_struct *n);
430 * napi_schedule - schedule NAPI poll
433 * Schedule NAPI poll routine to be called if it is not already
436 static inline void napi_schedule(struct napi_struct *n)
438 if (napi_schedule_prep(n))
443 * napi_schedule_irqoff - schedule NAPI poll
446 * Variant of napi_schedule(), assuming hard irqs are masked.
448 static inline void napi_schedule_irqoff(struct napi_struct *n)
450 if (napi_schedule_prep(n))
451 __napi_schedule_irqoff(n);
454 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
455 static inline bool napi_reschedule(struct napi_struct *napi)
457 if (napi_schedule_prep(napi)) {
458 __napi_schedule(napi);
464 bool napi_complete_done(struct napi_struct *n, int work_done);
466 * napi_complete - NAPI processing complete
469 * Mark NAPI processing as complete.
470 * Consider using napi_complete_done() instead.
471 * Return false if device should avoid rearming interrupts.
473 static inline bool napi_complete(struct napi_struct *n)
475 return napi_complete_done(n, 0);
479 * napi_hash_del - remove a NAPI from global table
480 * @napi: NAPI context
482 * Warning: caller must observe RCU grace period
483 * before freeing memory containing @napi, if
484 * this function returns true.
485 * Note: core networking stack automatically calls it
486 * from netif_napi_del().
487 * Drivers might want to call this helper to combine all
488 * the needed RCU grace periods into a single one.
490 bool napi_hash_del(struct napi_struct *napi);
493 * napi_disable - prevent NAPI from scheduling
496 * Stop NAPI from being scheduled on this context.
497 * Waits till any outstanding processing completes.
499 void napi_disable(struct napi_struct *n);
502 * napi_enable - enable NAPI scheduling
505 * Resume NAPI from being scheduled on this context.
506 * Must be paired with napi_disable.
508 static inline void napi_enable(struct napi_struct *n)
510 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
511 smp_mb__before_atomic();
512 clear_bit(NAPI_STATE_SCHED, &n->state);
513 clear_bit(NAPI_STATE_NPSVC, &n->state);
517 * napi_synchronize - wait until NAPI is not running
520 * Wait until NAPI is done being scheduled on this context.
521 * Waits till any outstanding processing completes but
522 * does not disable future activations.
524 static inline void napi_synchronize(const struct napi_struct *n)
526 if (IS_ENABLED(CONFIG_SMP))
527 while (test_bit(NAPI_STATE_SCHED, &n->state))
533 enum netdev_queue_state_t {
534 __QUEUE_STATE_DRV_XOFF,
535 __QUEUE_STATE_STACK_XOFF,
536 __QUEUE_STATE_FROZEN,
539 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
540 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
543 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
546 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
551 * netif_tx_* functions below are used to manipulate this flag. The
552 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
553 * queue independently. The netif_xmit_*stopped functions below are called
554 * to check if the queue has been stopped by the driver or stack (either
555 * of the XOFF bits are set in the state). Drivers should not need to call
556 * netif_xmit*stopped functions, they should only be using netif_tx_*.
559 struct netdev_queue {
563 struct net_device *dev;
564 struct Qdisc __rcu *qdisc;
565 struct Qdisc *qdisc_sleeping;
569 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 unsigned long tx_maxrate;
574 * Number of TX timeouts for this queue
575 * (/sys/class/net/DEV/Q/trans_timeout)
577 unsigned long trans_timeout;
579 /* Subordinate device that the queue has been assigned to */
580 struct net_device *sb_dev;
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
587 * Time (in jiffies) of last Tx
589 unsigned long trans_start;
596 } ____cacheline_aligned_in_smp;
598 extern int sysctl_fb_tunnels_only_for_init_net;
600 static inline bool net_has_fallback_tunnels(const struct net *net)
602 return net == &init_net ||
603 !IS_ENABLED(CONFIG_SYSCTL) ||
604 !sysctl_fb_tunnels_only_for_init_net;
607 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
616 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
618 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
625 * This structure holds an RPS map which can be of variable length. The
626 * map is an array of CPUs.
633 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
636 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
637 * tail pointer for that CPU's input queue at the time of last enqueue, and
638 * a hardware filter index.
640 struct rps_dev_flow {
643 unsigned int last_qtail;
645 #define RPS_NO_FILTER 0xffff
648 * The rps_dev_flow_table structure contains a table of flow mappings.
650 struct rps_dev_flow_table {
653 struct rps_dev_flow flows[0];
655 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
656 ((_num) * sizeof(struct rps_dev_flow)))
659 * The rps_sock_flow_table contains mappings of flows to the last CPU
660 * on which they were processed by the application (set in recvmsg).
661 * Each entry is a 32bit value. Upper part is the high-order bits
662 * of flow hash, lower part is CPU number.
663 * rps_cpu_mask is used to partition the space, depending on number of
664 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
665 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
666 * meaning we use 32-6=26 bits for the hash.
668 struct rps_sock_flow_table {
671 u32 ents[0] ____cacheline_aligned_in_smp;
673 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
675 #define RPS_NO_CPU 0xffff
677 extern u32 rps_cpu_mask;
678 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
680 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
684 unsigned int index = hash & table->mask;
685 u32 val = hash & ~rps_cpu_mask;
687 /* We only give a hint, preemption can change CPU under us */
688 val |= raw_smp_processor_id();
690 if (table->ents[index] != val)
691 table->ents[index] = val;
695 #ifdef CONFIG_RFS_ACCEL
696 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
699 #endif /* CONFIG_RPS */
701 /* This structure contains an instance of an RX queue. */
702 struct netdev_rx_queue {
704 struct rps_map __rcu *rps_map;
705 struct rps_dev_flow_table __rcu *rps_flow_table;
708 struct net_device *dev;
709 struct xdp_rxq_info xdp_rxq;
710 } ____cacheline_aligned_in_smp;
713 * RX queue sysfs structures and functions.
715 struct rx_queue_attribute {
716 struct attribute attr;
717 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
718 ssize_t (*store)(struct netdev_rx_queue *queue,
719 const char *buf, size_t len);
724 * This structure holds an XPS map which can be of variable length. The
725 * map is an array of queues.
729 unsigned int alloc_len;
733 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
734 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
735 - sizeof(struct xps_map)) / sizeof(u16))
738 * This structure holds all XPS maps for device. Maps are indexed by CPU.
740 struct xps_dev_maps {
742 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
745 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
746 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
748 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
749 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
751 #endif /* CONFIG_XPS */
753 #define TC_MAX_QUEUE 16
754 #define TC_BITMASK 15
755 /* HW offloaded queuing disciplines txq count and offset maps */
756 struct netdev_tc_txq {
761 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
763 * This structure is to hold information about the device
764 * configured to run FCoE protocol stack.
766 struct netdev_fcoe_hbainfo {
767 char manufacturer[64];
768 char serial_number[64];
769 char hardware_version[64];
770 char driver_version[64];
771 char optionrom_version[64];
772 char firmware_version[64];
774 char model_description[256];
778 #define MAX_PHYS_ITEM_ID_LEN 32
780 /* This structure holds a unique identifier to identify some
781 * physical item (port for example) used by a netdevice.
783 struct netdev_phys_item_id {
784 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
785 unsigned char id_len;
788 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
789 struct netdev_phys_item_id *b)
791 return a->id_len == b->id_len &&
792 memcmp(a->id, b->id, a->id_len) == 0;
795 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
796 struct sk_buff *skb);
799 TC_SETUP_QDISC_MQPRIO,
802 TC_SETUP_CLSMATCHALL,
812 /* These structures hold the attributes of bpf state that are being passed
813 * to the netdevice through the bpf op.
815 enum bpf_netdev_command {
816 /* Set or clear a bpf program used in the earliest stages of packet
817 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
818 * is responsible for calling bpf_prog_put on any old progs that are
819 * stored. In case of error, the callee need not release the new prog
820 * reference, but on success it takes ownership and must bpf_prog_put
821 * when it is no longer used.
825 /* Check if a bpf program is set on the device. The callee should
826 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
827 * is equivalent to XDP_ATTACHED_DRV.
830 /* BPF program for offload callbacks, invoked at program load time. */
831 BPF_OFFLOAD_VERIFIER_PREP,
832 BPF_OFFLOAD_TRANSLATE,
834 BPF_OFFLOAD_MAP_ALLOC,
835 BPF_OFFLOAD_MAP_FREE,
840 struct bpf_prog_offload_ops;
841 struct netlink_ext_ack;
845 enum bpf_netdev_command command;
850 struct bpf_prog *prog;
851 struct netlink_ext_ack *extack;
857 /* flags with which program was installed */
860 /* BPF_OFFLOAD_VERIFIER_PREP */
862 struct bpf_prog *prog;
863 const struct bpf_prog_offload_ops *ops; /* callee set */
865 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
867 struct bpf_prog *prog;
869 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
871 struct bpf_offloaded_map *offmap;
873 /* XDP_SETUP_XSK_UMEM */
875 struct xdp_umem *umem;
881 #ifdef CONFIG_XFRM_OFFLOAD
883 int (*xdo_dev_state_add) (struct xfrm_state *x);
884 void (*xdo_dev_state_delete) (struct xfrm_state *x);
885 void (*xdo_dev_state_free) (struct xfrm_state *x);
886 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
887 struct xfrm_state *x);
888 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
892 #if IS_ENABLED(CONFIG_TLS_DEVICE)
893 enum tls_offload_ctx_dir {
894 TLS_OFFLOAD_CTX_DIR_RX,
895 TLS_OFFLOAD_CTX_DIR_TX,
898 struct tls_crypto_info;
902 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk,
903 enum tls_offload_ctx_dir direction,
904 struct tls_crypto_info *crypto_info,
905 u32 start_offload_tcp_sn);
906 void (*tls_dev_del)(struct net_device *netdev,
907 struct tls_context *ctx,
908 enum tls_offload_ctx_dir direction);
913 struct rcu_head rcuhead;
918 * This structure defines the management hooks for network devices.
919 * The following hooks can be defined; unless noted otherwise, they are
920 * optional and can be filled with a null pointer.
922 * int (*ndo_init)(struct net_device *dev);
923 * This function is called once when a network device is registered.
924 * The network device can use this for any late stage initialization
925 * or semantic validation. It can fail with an error code which will
926 * be propagated back to register_netdev.
928 * void (*ndo_uninit)(struct net_device *dev);
929 * This function is called when device is unregistered or when registration
930 * fails. It is not called if init fails.
932 * int (*ndo_open)(struct net_device *dev);
933 * This function is called when a network device transitions to the up
936 * int (*ndo_stop)(struct net_device *dev);
937 * This function is called when a network device transitions to the down
940 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
941 * struct net_device *dev);
942 * Called when a packet needs to be transmitted.
943 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
944 * the queue before that can happen; it's for obsolete devices and weird
945 * corner cases, but the stack really does a non-trivial amount
946 * of useless work if you return NETDEV_TX_BUSY.
947 * Required; cannot be NULL.
949 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
950 * struct net_device *dev
951 * netdev_features_t features);
952 * Called by core transmit path to determine if device is capable of
953 * performing offload operations on a given packet. This is to give
954 * the device an opportunity to implement any restrictions that cannot
955 * be otherwise expressed by feature flags. The check is called with
956 * the set of features that the stack has calculated and it returns
957 * those the driver believes to be appropriate.
959 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
960 * void *accel_priv, select_queue_fallback_t fallback);
961 * Called to decide which queue to use when device supports multiple
964 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
965 * This function is called to allow device receiver to make
966 * changes to configuration when multicast or promiscuous is enabled.
968 * void (*ndo_set_rx_mode)(struct net_device *dev);
969 * This function is called device changes address list filtering.
970 * If driver handles unicast address filtering, it should set
971 * IFF_UNICAST_FLT in its priv_flags.
973 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
974 * This function is called when the Media Access Control address
975 * needs to be changed. If this interface is not defined, the
976 * MAC address can not be changed.
978 * int (*ndo_validate_addr)(struct net_device *dev);
979 * Test if Media Access Control address is valid for the device.
981 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
982 * Called when a user requests an ioctl which can't be handled by
983 * the generic interface code. If not defined ioctls return
984 * not supported error code.
986 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
987 * Used to set network devices bus interface parameters. This interface
988 * is retained for legacy reasons; new devices should use the bus
989 * interface (PCI) for low level management.
991 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
992 * Called when a user wants to change the Maximum Transfer Unit
995 * void (*ndo_tx_timeout)(struct net_device *dev);
996 * Callback used when the transmitter has not made any progress
997 * for dev->watchdog ticks.
999 * void (*ndo_get_stats64)(struct net_device *dev,
1000 * struct rtnl_link_stats64 *storage);
1001 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1002 * Called when a user wants to get the network device usage
1003 * statistics. Drivers must do one of the following:
1004 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1005 * rtnl_link_stats64 structure passed by the caller.
1006 * 2. Define @ndo_get_stats to update a net_device_stats structure
1007 * (which should normally be dev->stats) and return a pointer to
1008 * it. The structure may be changed asynchronously only if each
1009 * field is written atomically.
1010 * 3. Update dev->stats asynchronously and atomically, and define
1011 * neither operation.
1013 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1014 * Return true if this device supports offload stats of this attr_id.
1016 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1018 * Get statistics for offload operations by attr_id. Write it into the
1019 * attr_data pointer.
1021 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1022 * If device supports VLAN filtering this function is called when a
1023 * VLAN id is registered.
1025 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1026 * If device supports VLAN filtering this function is called when a
1027 * VLAN id is unregistered.
1029 * void (*ndo_poll_controller)(struct net_device *dev);
1031 * SR-IOV management functions.
1032 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1033 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1034 * u8 qos, __be16 proto);
1035 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1037 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1038 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1039 * int (*ndo_get_vf_config)(struct net_device *dev,
1040 * int vf, struct ifla_vf_info *ivf);
1041 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1042 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1043 * struct nlattr *port[]);
1045 * Enable or disable the VF ability to query its RSS Redirection Table and
1046 * Hash Key. This is needed since on some devices VF share this information
1047 * with PF and querying it may introduce a theoretical security risk.
1048 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1049 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1050 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1052 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1053 * This is always called from the stack with the rtnl lock held and netif
1054 * tx queues stopped. This allows the netdevice to perform queue
1055 * management safely.
1057 * Fiber Channel over Ethernet (FCoE) offload functions.
1058 * int (*ndo_fcoe_enable)(struct net_device *dev);
1059 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1060 * so the underlying device can perform whatever needed configuration or
1061 * initialization to support acceleration of FCoE traffic.
1063 * int (*ndo_fcoe_disable)(struct net_device *dev);
1064 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1065 * so the underlying device can perform whatever needed clean-ups to
1066 * stop supporting acceleration of FCoE traffic.
1068 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1069 * struct scatterlist *sgl, unsigned int sgc);
1070 * Called when the FCoE Initiator wants to initialize an I/O that
1071 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1072 * perform necessary setup and returns 1 to indicate the device is set up
1073 * successfully to perform DDP on this I/O, otherwise this returns 0.
1075 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1076 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1077 * indicated by the FC exchange id 'xid', so the underlying device can
1078 * clean up and reuse resources for later DDP requests.
1080 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1081 * struct scatterlist *sgl, unsigned int sgc);
1082 * Called when the FCoE Target wants to initialize an I/O that
1083 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1084 * perform necessary setup and returns 1 to indicate the device is set up
1085 * successfully to perform DDP on this I/O, otherwise this returns 0.
1087 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1088 * struct netdev_fcoe_hbainfo *hbainfo);
1089 * Called when the FCoE Protocol stack wants information on the underlying
1090 * device. This information is utilized by the FCoE protocol stack to
1091 * register attributes with Fiber Channel management service as per the
1092 * FC-GS Fabric Device Management Information(FDMI) specification.
1094 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1095 * Called when the underlying device wants to override default World Wide
1096 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1097 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1098 * protocol stack to use.
1101 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1102 * u16 rxq_index, u32 flow_id);
1103 * Set hardware filter for RFS. rxq_index is the target queue index;
1104 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1105 * Return the filter ID on success, or a negative error code.
1107 * Slave management functions (for bridge, bonding, etc).
1108 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1109 * Called to make another netdev an underling.
1111 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1112 * Called to release previously enslaved netdev.
1114 * Feature/offload setting functions.
1115 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1116 * netdev_features_t features);
1117 * Adjusts the requested feature flags according to device-specific
1118 * constraints, and returns the resulting flags. Must not modify
1121 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1122 * Called to update device configuration to new features. Passed
1123 * feature set might be less than what was returned by ndo_fix_features()).
1124 * Must return >0 or -errno if it changed dev->features itself.
1126 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1127 * struct net_device *dev,
1128 * const unsigned char *addr, u16 vid, u16 flags)
1129 * Adds an FDB entry to dev for addr.
1130 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1131 * struct net_device *dev,
1132 * const unsigned char *addr, u16 vid)
1133 * Deletes the FDB entry from dev coresponding to addr.
1134 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1135 * struct net_device *dev, struct net_device *filter_dev,
1137 * Used to add FDB entries to dump requests. Implementers should add
1138 * entries to skb and update idx with the number of entries.
1140 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1142 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1143 * struct net_device *dev, u32 filter_mask,
1145 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1148 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1149 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1150 * which do not represent real hardware may define this to allow their
1151 * userspace components to manage their virtual carrier state. Devices
1152 * that determine carrier state from physical hardware properties (eg
1153 * network cables) or protocol-dependent mechanisms (eg
1154 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1156 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1157 * struct netdev_phys_item_id *ppid);
1158 * Called to get ID of physical port of this device. If driver does
1159 * not implement this, it is assumed that the hw is not able to have
1160 * multiple net devices on single physical port.
1162 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1163 * struct udp_tunnel_info *ti);
1164 * Called by UDP tunnel to notify a driver about the UDP port and socket
1165 * address family that a UDP tunnel is listnening to. It is called only
1166 * when a new port starts listening. The operation is protected by the
1169 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1170 * struct udp_tunnel_info *ti);
1171 * Called by UDP tunnel to notify the driver about a UDP port and socket
1172 * address family that the UDP tunnel is not listening to anymore. The
1173 * operation is protected by the RTNL.
1175 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1176 * struct net_device *dev)
1177 * Called by upper layer devices to accelerate switching or other
1178 * station functionality into hardware. 'pdev is the lowerdev
1179 * to use for the offload and 'dev' is the net device that will
1180 * back the offload. Returns a pointer to the private structure
1181 * the upper layer will maintain.
1182 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1183 * Called by upper layer device to delete the station created
1184 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1185 * the station and priv is the structure returned by the add
1187 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1188 * int queue_index, u32 maxrate);
1189 * Called when a user wants to set a max-rate limitation of specific
1191 * int (*ndo_get_iflink)(const struct net_device *dev);
1192 * Called to get the iflink value of this device.
1193 * void (*ndo_change_proto_down)(struct net_device *dev,
1195 * This function is used to pass protocol port error state information
1196 * to the switch driver. The switch driver can react to the proto_down
1197 * by doing a phys down on the associated switch port.
1198 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1199 * This function is used to get egress tunnel information for given skb.
1200 * This is useful for retrieving outer tunnel header parameters while
1202 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1203 * This function is used to specify the headroom that the skb must
1204 * consider when allocation skb during packet reception. Setting
1205 * appropriate rx headroom value allows avoiding skb head copy on
1206 * forward. Setting a negative value resets the rx headroom to the
1208 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1209 * This function is used to set or query state related to XDP on the
1210 * netdevice and manage BPF offload. See definition of
1211 * enum bpf_netdev_command for details.
1212 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1214 * This function is used to submit @n XDP packets for transmit on a
1215 * netdevice. Returns number of frames successfully transmitted, frames
1216 * that got dropped are freed/returned via xdp_return_frame().
1217 * Returns negative number, means general error invoking ndo, meaning
1218 * no frames were xmit'ed and core-caller will free all frames.
1220 struct net_device_ops {
1221 int (*ndo_init)(struct net_device *dev);
1222 void (*ndo_uninit)(struct net_device *dev);
1223 int (*ndo_open)(struct net_device *dev);
1224 int (*ndo_stop)(struct net_device *dev);
1225 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1226 struct net_device *dev);
1227 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1228 struct net_device *dev,
1229 netdev_features_t features);
1230 u16 (*ndo_select_queue)(struct net_device *dev,
1231 struct sk_buff *skb,
1233 select_queue_fallback_t fallback);
1234 void (*ndo_change_rx_flags)(struct net_device *dev,
1236 void (*ndo_set_rx_mode)(struct net_device *dev);
1237 int (*ndo_set_mac_address)(struct net_device *dev,
1239 int (*ndo_validate_addr)(struct net_device *dev);
1240 int (*ndo_do_ioctl)(struct net_device *dev,
1241 struct ifreq *ifr, int cmd);
1242 int (*ndo_set_config)(struct net_device *dev,
1244 int (*ndo_change_mtu)(struct net_device *dev,
1246 int (*ndo_neigh_setup)(struct net_device *dev,
1247 struct neigh_parms *);
1248 void (*ndo_tx_timeout) (struct net_device *dev);
1250 void (*ndo_get_stats64)(struct net_device *dev,
1251 struct rtnl_link_stats64 *storage);
1252 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1253 int (*ndo_get_offload_stats)(int attr_id,
1254 const struct net_device *dev,
1256 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1258 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1259 __be16 proto, u16 vid);
1260 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1261 __be16 proto, u16 vid);
1262 #ifdef CONFIG_NET_POLL_CONTROLLER
1263 void (*ndo_poll_controller)(struct net_device *dev);
1264 int (*ndo_netpoll_setup)(struct net_device *dev,
1265 struct netpoll_info *info);
1266 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1268 int (*ndo_set_vf_mac)(struct net_device *dev,
1269 int queue, u8 *mac);
1270 int (*ndo_set_vf_vlan)(struct net_device *dev,
1271 int queue, u16 vlan,
1272 u8 qos, __be16 proto);
1273 int (*ndo_set_vf_rate)(struct net_device *dev,
1274 int vf, int min_tx_rate,
1276 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1277 int vf, bool setting);
1278 int (*ndo_set_vf_trust)(struct net_device *dev,
1279 int vf, bool setting);
1280 int (*ndo_get_vf_config)(struct net_device *dev,
1282 struct ifla_vf_info *ivf);
1283 int (*ndo_set_vf_link_state)(struct net_device *dev,
1284 int vf, int link_state);
1285 int (*ndo_get_vf_stats)(struct net_device *dev,
1287 struct ifla_vf_stats
1289 int (*ndo_set_vf_port)(struct net_device *dev,
1291 struct nlattr *port[]);
1292 int (*ndo_get_vf_port)(struct net_device *dev,
1293 int vf, struct sk_buff *skb);
1294 int (*ndo_set_vf_guid)(struct net_device *dev,
1297 int (*ndo_set_vf_rss_query_en)(
1298 struct net_device *dev,
1299 int vf, bool setting);
1300 int (*ndo_setup_tc)(struct net_device *dev,
1301 enum tc_setup_type type,
1303 #if IS_ENABLED(CONFIG_FCOE)
1304 int (*ndo_fcoe_enable)(struct net_device *dev);
1305 int (*ndo_fcoe_disable)(struct net_device *dev);
1306 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1308 struct scatterlist *sgl,
1310 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1312 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1314 struct scatterlist *sgl,
1316 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1317 struct netdev_fcoe_hbainfo *hbainfo);
1320 #if IS_ENABLED(CONFIG_LIBFCOE)
1321 #define NETDEV_FCOE_WWNN 0
1322 #define NETDEV_FCOE_WWPN 1
1323 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1324 u64 *wwn, int type);
1327 #ifdef CONFIG_RFS_ACCEL
1328 int (*ndo_rx_flow_steer)(struct net_device *dev,
1329 const struct sk_buff *skb,
1333 int (*ndo_add_slave)(struct net_device *dev,
1334 struct net_device *slave_dev,
1335 struct netlink_ext_ack *extack);
1336 int (*ndo_del_slave)(struct net_device *dev,
1337 struct net_device *slave_dev);
1338 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1339 netdev_features_t features);
1340 int (*ndo_set_features)(struct net_device *dev,
1341 netdev_features_t features);
1342 int (*ndo_neigh_construct)(struct net_device *dev,
1343 struct neighbour *n);
1344 void (*ndo_neigh_destroy)(struct net_device *dev,
1345 struct neighbour *n);
1347 int (*ndo_fdb_add)(struct ndmsg *ndm,
1348 struct nlattr *tb[],
1349 struct net_device *dev,
1350 const unsigned char *addr,
1353 int (*ndo_fdb_del)(struct ndmsg *ndm,
1354 struct nlattr *tb[],
1355 struct net_device *dev,
1356 const unsigned char *addr,
1358 int (*ndo_fdb_dump)(struct sk_buff *skb,
1359 struct netlink_callback *cb,
1360 struct net_device *dev,
1361 struct net_device *filter_dev,
1364 int (*ndo_bridge_setlink)(struct net_device *dev,
1365 struct nlmsghdr *nlh,
1367 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1369 struct net_device *dev,
1372 int (*ndo_bridge_dellink)(struct net_device *dev,
1373 struct nlmsghdr *nlh,
1375 int (*ndo_change_carrier)(struct net_device *dev,
1377 int (*ndo_get_phys_port_id)(struct net_device *dev,
1378 struct netdev_phys_item_id *ppid);
1379 int (*ndo_get_phys_port_name)(struct net_device *dev,
1380 char *name, size_t len);
1381 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1382 struct udp_tunnel_info *ti);
1383 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1384 struct udp_tunnel_info *ti);
1385 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1386 struct net_device *dev);
1387 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1390 int (*ndo_get_lock_subclass)(struct net_device *dev);
1391 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1394 int (*ndo_get_iflink)(const struct net_device *dev);
1395 int (*ndo_change_proto_down)(struct net_device *dev,
1397 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1398 struct sk_buff *skb);
1399 void (*ndo_set_rx_headroom)(struct net_device *dev,
1400 int needed_headroom);
1401 int (*ndo_bpf)(struct net_device *dev,
1402 struct netdev_bpf *bpf);
1403 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1404 struct xdp_frame **xdp,
1406 int (*ndo_xsk_async_xmit)(struct net_device *dev,
1411 * enum net_device_priv_flags - &struct net_device priv_flags
1413 * These are the &struct net_device, they are only set internally
1414 * by drivers and used in the kernel. These flags are invisible to
1415 * userspace; this means that the order of these flags can change
1416 * during any kernel release.
1418 * You should have a pretty good reason to be extending these flags.
1420 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1421 * @IFF_EBRIDGE: Ethernet bridging device
1422 * @IFF_BONDING: bonding master or slave
1423 * @IFF_ISATAP: ISATAP interface (RFC4214)
1424 * @IFF_WAN_HDLC: WAN HDLC device
1425 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1427 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1428 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1429 * @IFF_MACVLAN_PORT: device used as macvlan port
1430 * @IFF_BRIDGE_PORT: device used as bridge port
1431 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1432 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1433 * @IFF_UNICAST_FLT: Supports unicast filtering
1434 * @IFF_TEAM_PORT: device used as team port
1435 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1436 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1437 * change when it's running
1438 * @IFF_MACVLAN: Macvlan device
1439 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1440 * underlying stacked devices
1441 * @IFF_L3MDEV_MASTER: device is an L3 master device
1442 * @IFF_NO_QUEUE: device can run without qdisc attached
1443 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1444 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1445 * @IFF_TEAM: device is a team device
1446 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1447 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1448 * entity (i.e. the master device for bridged veth)
1449 * @IFF_MACSEC: device is a MACsec device
1450 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1451 * @IFF_FAILOVER: device is a failover master device
1452 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1454 enum netdev_priv_flags {
1455 IFF_802_1Q_VLAN = 1<<0,
1459 IFF_WAN_HDLC = 1<<4,
1460 IFF_XMIT_DST_RELEASE = 1<<5,
1461 IFF_DONT_BRIDGE = 1<<6,
1462 IFF_DISABLE_NETPOLL = 1<<7,
1463 IFF_MACVLAN_PORT = 1<<8,
1464 IFF_BRIDGE_PORT = 1<<9,
1465 IFF_OVS_DATAPATH = 1<<10,
1466 IFF_TX_SKB_SHARING = 1<<11,
1467 IFF_UNICAST_FLT = 1<<12,
1468 IFF_TEAM_PORT = 1<<13,
1469 IFF_SUPP_NOFCS = 1<<14,
1470 IFF_LIVE_ADDR_CHANGE = 1<<15,
1471 IFF_MACVLAN = 1<<16,
1472 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1473 IFF_L3MDEV_MASTER = 1<<18,
1474 IFF_NO_QUEUE = 1<<19,
1475 IFF_OPENVSWITCH = 1<<20,
1476 IFF_L3MDEV_SLAVE = 1<<21,
1478 IFF_RXFH_CONFIGURED = 1<<23,
1479 IFF_PHONY_HEADROOM = 1<<24,
1481 IFF_NO_RX_HANDLER = 1<<26,
1482 IFF_FAILOVER = 1<<27,
1483 IFF_FAILOVER_SLAVE = 1<<28,
1486 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1487 #define IFF_EBRIDGE IFF_EBRIDGE
1488 #define IFF_BONDING IFF_BONDING
1489 #define IFF_ISATAP IFF_ISATAP
1490 #define IFF_WAN_HDLC IFF_WAN_HDLC
1491 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1492 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1493 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1494 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1495 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1496 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1497 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1498 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1499 #define IFF_TEAM_PORT IFF_TEAM_PORT
1500 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1501 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1502 #define IFF_MACVLAN IFF_MACVLAN
1503 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1504 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1505 #define IFF_NO_QUEUE IFF_NO_QUEUE
1506 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1507 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1508 #define IFF_TEAM IFF_TEAM
1509 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1510 #define IFF_MACSEC IFF_MACSEC
1511 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1512 #define IFF_FAILOVER IFF_FAILOVER
1513 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1516 * struct net_device - The DEVICE structure.
1518 * Actually, this whole structure is a big mistake. It mixes I/O
1519 * data with strictly "high-level" data, and it has to know about
1520 * almost every data structure used in the INET module.
1522 * @name: This is the first field of the "visible" part of this structure
1523 * (i.e. as seen by users in the "Space.c" file). It is the name
1526 * @name_hlist: Device name hash chain, please keep it close to name[]
1527 * @ifalias: SNMP alias
1528 * @mem_end: Shared memory end
1529 * @mem_start: Shared memory start
1530 * @base_addr: Device I/O address
1531 * @irq: Device IRQ number
1533 * @state: Generic network queuing layer state, see netdev_state_t
1534 * @dev_list: The global list of network devices
1535 * @napi_list: List entry used for polling NAPI devices
1536 * @unreg_list: List entry when we are unregistering the
1537 * device; see the function unregister_netdev
1538 * @close_list: List entry used when we are closing the device
1539 * @ptype_all: Device-specific packet handlers for all protocols
1540 * @ptype_specific: Device-specific, protocol-specific packet handlers
1542 * @adj_list: Directly linked devices, like slaves for bonding
1543 * @features: Currently active device features
1544 * @hw_features: User-changeable features
1546 * @wanted_features: User-requested features
1547 * @vlan_features: Mask of features inheritable by VLAN devices
1549 * @hw_enc_features: Mask of features inherited by encapsulating devices
1550 * This field indicates what encapsulation
1551 * offloads the hardware is capable of doing,
1552 * and drivers will need to set them appropriately.
1554 * @mpls_features: Mask of features inheritable by MPLS
1556 * @ifindex: interface index
1557 * @group: The group the device belongs to
1559 * @stats: Statistics struct, which was left as a legacy, use
1560 * rtnl_link_stats64 instead
1562 * @rx_dropped: Dropped packets by core network,
1563 * do not use this in drivers
1564 * @tx_dropped: Dropped packets by core network,
1565 * do not use this in drivers
1566 * @rx_nohandler: nohandler dropped packets by core network on
1567 * inactive devices, do not use this in drivers
1568 * @carrier_up_count: Number of times the carrier has been up
1569 * @carrier_down_count: Number of times the carrier has been down
1571 * @wireless_handlers: List of functions to handle Wireless Extensions,
1573 * see <net/iw_handler.h> for details.
1574 * @wireless_data: Instance data managed by the core of wireless extensions
1576 * @netdev_ops: Includes several pointers to callbacks,
1577 * if one wants to override the ndo_*() functions
1578 * @ethtool_ops: Management operations
1579 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1580 * discovery handling. Necessary for e.g. 6LoWPAN.
1581 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1582 * of Layer 2 headers.
1584 * @flags: Interface flags (a la BSD)
1585 * @priv_flags: Like 'flags' but invisible to userspace,
1586 * see if.h for the definitions
1587 * @gflags: Global flags ( kept as legacy )
1588 * @padded: How much padding added by alloc_netdev()
1589 * @operstate: RFC2863 operstate
1590 * @link_mode: Mapping policy to operstate
1591 * @if_port: Selectable AUI, TP, ...
1593 * @mtu: Interface MTU value
1594 * @min_mtu: Interface Minimum MTU value
1595 * @max_mtu: Interface Maximum MTU value
1596 * @type: Interface hardware type
1597 * @hard_header_len: Maximum hardware header length.
1598 * @min_header_len: Minimum hardware header length
1600 * @needed_headroom: Extra headroom the hardware may need, but not in all
1601 * cases can this be guaranteed
1602 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1603 * cases can this be guaranteed. Some cases also use
1604 * LL_MAX_HEADER instead to allocate the skb
1606 * interface address info:
1608 * @perm_addr: Permanent hw address
1609 * @addr_assign_type: Hw address assignment type
1610 * @addr_len: Hardware address length
1611 * @neigh_priv_len: Used in neigh_alloc()
1612 * @dev_id: Used to differentiate devices that share
1613 * the same link layer address
1614 * @dev_port: Used to differentiate devices that share
1616 * @addr_list_lock: XXX: need comments on this one
1617 * @uc_promisc: Counter that indicates promiscuous mode
1618 * has been enabled due to the need to listen to
1619 * additional unicast addresses in a device that
1620 * does not implement ndo_set_rx_mode()
1621 * @uc: unicast mac addresses
1622 * @mc: multicast mac addresses
1623 * @dev_addrs: list of device hw addresses
1624 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1625 * @promiscuity: Number of times the NIC is told to work in
1626 * promiscuous mode; if it becomes 0 the NIC will
1627 * exit promiscuous mode
1628 * @allmulti: Counter, enables or disables allmulticast mode
1630 * @vlan_info: VLAN info
1631 * @dsa_ptr: dsa specific data
1632 * @tipc_ptr: TIPC specific data
1633 * @atalk_ptr: AppleTalk link
1634 * @ip_ptr: IPv4 specific data
1635 * @dn_ptr: DECnet specific data
1636 * @ip6_ptr: IPv6 specific data
1637 * @ax25_ptr: AX.25 specific data
1638 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1640 * @dev_addr: Hw address (before bcast,
1641 * because most packets are unicast)
1643 * @_rx: Array of RX queues
1644 * @num_rx_queues: Number of RX queues
1645 * allocated at register_netdev() time
1646 * @real_num_rx_queues: Number of RX queues currently active in device
1648 * @rx_handler: handler for received packets
1649 * @rx_handler_data: XXX: need comments on this one
1650 * @miniq_ingress: ingress/clsact qdisc specific data for
1651 * ingress processing
1652 * @ingress_queue: XXX: need comments on this one
1653 * @broadcast: hw bcast address
1655 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1656 * indexed by RX queue number. Assigned by driver.
1657 * This must only be set if the ndo_rx_flow_steer
1658 * operation is defined
1659 * @index_hlist: Device index hash chain
1661 * @_tx: Array of TX queues
1662 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1663 * @real_num_tx_queues: Number of TX queues currently active in device
1664 * @qdisc: Root qdisc from userspace point of view
1665 * @tx_queue_len: Max frames per queue allowed
1666 * @tx_global_lock: XXX: need comments on this one
1668 * @xps_maps: XXX: need comments on this one
1669 * @miniq_egress: clsact qdisc specific data for
1671 * @watchdog_timeo: Represents the timeout that is used by
1672 * the watchdog (see dev_watchdog())
1673 * @watchdog_timer: List of timers
1675 * @pcpu_refcnt: Number of references to this device
1676 * @todo_list: Delayed register/unregister
1677 * @link_watch_list: XXX: need comments on this one
1679 * @reg_state: Register/unregister state machine
1680 * @dismantle: Device is going to be freed
1681 * @rtnl_link_state: This enum represents the phases of creating
1684 * @needs_free_netdev: Should unregister perform free_netdev?
1685 * @priv_destructor: Called from unregister
1686 * @npinfo: XXX: need comments on this one
1687 * @nd_net: Network namespace this network device is inside
1689 * @ml_priv: Mid-layer private
1690 * @lstats: Loopback statistics
1691 * @tstats: Tunnel statistics
1692 * @dstats: Dummy statistics
1693 * @vstats: Virtual ethernet statistics
1698 * @dev: Class/net/name entry
1699 * @sysfs_groups: Space for optional device, statistics and wireless
1702 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1703 * @rtnl_link_ops: Rtnl_link_ops
1705 * @gso_max_size: Maximum size of generic segmentation offload
1706 * @gso_max_segs: Maximum number of segments that can be passed to the
1709 * @dcbnl_ops: Data Center Bridging netlink ops
1710 * @num_tc: Number of traffic classes in the net device
1711 * @tc_to_txq: XXX: need comments on this one
1712 * @prio_tc_map: XXX: need comments on this one
1714 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1716 * @priomap: XXX: need comments on this one
1717 * @phydev: Physical device may attach itself
1718 * for hardware timestamping
1719 * @sfp_bus: attached &struct sfp_bus structure.
1721 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1722 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1724 * @proto_down: protocol port state information can be sent to the
1725 * switch driver and used to set the phys state of the
1728 * FIXME: cleanup struct net_device such that network protocol info
1733 char name[IFNAMSIZ];
1734 struct hlist_node name_hlist;
1735 struct dev_ifalias __rcu *ifalias;
1737 * I/O specific fields
1738 * FIXME: Merge these and struct ifmap into one
1740 unsigned long mem_end;
1741 unsigned long mem_start;
1742 unsigned long base_addr;
1746 * Some hardware also needs these fields (state,dev_list,
1747 * napi_list,unreg_list,close_list) but they are not
1748 * part of the usual set specified in Space.c.
1751 unsigned long state;
1753 struct list_head dev_list;
1754 struct list_head napi_list;
1755 struct list_head unreg_list;
1756 struct list_head close_list;
1757 struct list_head ptype_all;
1758 struct list_head ptype_specific;
1761 struct list_head upper;
1762 struct list_head lower;
1765 netdev_features_t features;
1766 netdev_features_t hw_features;
1767 netdev_features_t wanted_features;
1768 netdev_features_t vlan_features;
1769 netdev_features_t hw_enc_features;
1770 netdev_features_t mpls_features;
1771 netdev_features_t gso_partial_features;
1776 struct net_device_stats stats;
1778 atomic_long_t rx_dropped;
1779 atomic_long_t tx_dropped;
1780 atomic_long_t rx_nohandler;
1782 /* Stats to monitor link on/off, flapping */
1783 atomic_t carrier_up_count;
1784 atomic_t carrier_down_count;
1786 #ifdef CONFIG_WIRELESS_EXT
1787 const struct iw_handler_def *wireless_handlers;
1788 struct iw_public_data *wireless_data;
1790 const struct net_device_ops *netdev_ops;
1791 const struct ethtool_ops *ethtool_ops;
1792 #ifdef CONFIG_NET_SWITCHDEV
1793 const struct switchdev_ops *switchdev_ops;
1795 #ifdef CONFIG_NET_L3_MASTER_DEV
1796 const struct l3mdev_ops *l3mdev_ops;
1798 #if IS_ENABLED(CONFIG_IPV6)
1799 const struct ndisc_ops *ndisc_ops;
1802 #ifdef CONFIG_XFRM_OFFLOAD
1803 const struct xfrmdev_ops *xfrmdev_ops;
1806 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1807 const struct tlsdev_ops *tlsdev_ops;
1810 const struct header_ops *header_ops;
1813 unsigned int priv_flags;
1815 unsigned short gflags;
1816 unsigned short padded;
1818 unsigned char operstate;
1819 unsigned char link_mode;
1821 unsigned char if_port;
1825 unsigned int min_mtu;
1826 unsigned int max_mtu;
1827 unsigned short type;
1828 unsigned short hard_header_len;
1829 unsigned char min_header_len;
1831 unsigned short needed_headroom;
1832 unsigned short needed_tailroom;
1834 /* Interface address info. */
1835 unsigned char perm_addr[MAX_ADDR_LEN];
1836 unsigned char addr_assign_type;
1837 unsigned char addr_len;
1838 unsigned short neigh_priv_len;
1839 unsigned short dev_id;
1840 unsigned short dev_port;
1841 spinlock_t addr_list_lock;
1842 unsigned char name_assign_type;
1844 struct netdev_hw_addr_list uc;
1845 struct netdev_hw_addr_list mc;
1846 struct netdev_hw_addr_list dev_addrs;
1849 struct kset *queues_kset;
1851 unsigned int promiscuity;
1852 unsigned int allmulti;
1855 /* Protocol-specific pointers */
1857 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1858 struct vlan_info __rcu *vlan_info;
1860 #if IS_ENABLED(CONFIG_NET_DSA)
1861 struct dsa_port *dsa_ptr;
1863 #if IS_ENABLED(CONFIG_TIPC)
1864 struct tipc_bearer __rcu *tipc_ptr;
1866 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1869 struct in_device __rcu *ip_ptr;
1870 #if IS_ENABLED(CONFIG_DECNET)
1871 struct dn_dev __rcu *dn_ptr;
1873 struct inet6_dev __rcu *ip6_ptr;
1874 #if IS_ENABLED(CONFIG_AX25)
1877 struct wireless_dev *ieee80211_ptr;
1878 struct wpan_dev *ieee802154_ptr;
1879 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1880 struct mpls_dev __rcu *mpls_ptr;
1884 * Cache lines mostly used on receive path (including eth_type_trans())
1886 /* Interface address info used in eth_type_trans() */
1887 unsigned char *dev_addr;
1889 struct netdev_rx_queue *_rx;
1890 unsigned int num_rx_queues;
1891 unsigned int real_num_rx_queues;
1893 struct bpf_prog __rcu *xdp_prog;
1894 unsigned long gro_flush_timeout;
1895 rx_handler_func_t __rcu *rx_handler;
1896 void __rcu *rx_handler_data;
1898 #ifdef CONFIG_NET_CLS_ACT
1899 struct mini_Qdisc __rcu *miniq_ingress;
1901 struct netdev_queue __rcu *ingress_queue;
1902 #ifdef CONFIG_NETFILTER_INGRESS
1903 struct nf_hook_entries __rcu *nf_hooks_ingress;
1906 unsigned char broadcast[MAX_ADDR_LEN];
1907 #ifdef CONFIG_RFS_ACCEL
1908 struct cpu_rmap *rx_cpu_rmap;
1910 struct hlist_node index_hlist;
1913 * Cache lines mostly used on transmit path
1915 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1916 unsigned int num_tx_queues;
1917 unsigned int real_num_tx_queues;
1918 struct Qdisc *qdisc;
1919 #ifdef CONFIG_NET_SCHED
1920 DECLARE_HASHTABLE (qdisc_hash, 4);
1922 unsigned int tx_queue_len;
1923 spinlock_t tx_global_lock;
1927 struct xps_dev_maps __rcu *xps_cpus_map;
1928 struct xps_dev_maps __rcu *xps_rxqs_map;
1930 #ifdef CONFIG_NET_CLS_ACT
1931 struct mini_Qdisc __rcu *miniq_egress;
1934 /* These may be needed for future network-power-down code. */
1935 struct timer_list watchdog_timer;
1937 int __percpu *pcpu_refcnt;
1938 struct list_head todo_list;
1940 struct list_head link_watch_list;
1942 enum { NETREG_UNINITIALIZED=0,
1943 NETREG_REGISTERED, /* completed register_netdevice */
1944 NETREG_UNREGISTERING, /* called unregister_netdevice */
1945 NETREG_UNREGISTERED, /* completed unregister todo */
1946 NETREG_RELEASED, /* called free_netdev */
1947 NETREG_DUMMY, /* dummy device for NAPI poll */
1953 RTNL_LINK_INITIALIZED,
1954 RTNL_LINK_INITIALIZING,
1955 } rtnl_link_state:16;
1957 bool needs_free_netdev;
1958 void (*priv_destructor)(struct net_device *dev);
1960 #ifdef CONFIG_NETPOLL
1961 struct netpoll_info __rcu *npinfo;
1964 possible_net_t nd_net;
1966 /* mid-layer private */
1969 struct pcpu_lstats __percpu *lstats;
1970 struct pcpu_sw_netstats __percpu *tstats;
1971 struct pcpu_dstats __percpu *dstats;
1972 struct pcpu_vstats __percpu *vstats;
1975 #if IS_ENABLED(CONFIG_GARP)
1976 struct garp_port __rcu *garp_port;
1978 #if IS_ENABLED(CONFIG_MRP)
1979 struct mrp_port __rcu *mrp_port;
1983 const struct attribute_group *sysfs_groups[4];
1984 const struct attribute_group *sysfs_rx_queue_group;
1986 const struct rtnl_link_ops *rtnl_link_ops;
1988 /* for setting kernel sock attribute on TCP connection setup */
1989 #define GSO_MAX_SIZE 65536
1990 unsigned int gso_max_size;
1991 #define GSO_MAX_SEGS 65535
1995 const struct dcbnl_rtnl_ops *dcbnl_ops;
1998 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1999 u8 prio_tc_map[TC_BITMASK + 1];
2001 #if IS_ENABLED(CONFIG_FCOE)
2002 unsigned int fcoe_ddp_xid;
2004 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2005 struct netprio_map __rcu *priomap;
2007 struct phy_device *phydev;
2008 struct sfp_bus *sfp_bus;
2009 struct lock_class_key *qdisc_tx_busylock;
2010 struct lock_class_key *qdisc_running_key;
2013 #define to_net_dev(d) container_of(d, struct net_device, dev)
2015 static inline bool netif_elide_gro(const struct net_device *dev)
2017 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2022 #define NETDEV_ALIGN 32
2025 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2027 return dev->prio_tc_map[prio & TC_BITMASK];
2031 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2033 if (tc >= dev->num_tc)
2036 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2040 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2041 void netdev_reset_tc(struct net_device *dev);
2042 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2043 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2046 int netdev_get_num_tc(struct net_device *dev)
2051 void netdev_unbind_sb_channel(struct net_device *dev,
2052 struct net_device *sb_dev);
2053 int netdev_bind_sb_channel_queue(struct net_device *dev,
2054 struct net_device *sb_dev,
2055 u8 tc, u16 count, u16 offset);
2056 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2057 static inline int netdev_get_sb_channel(struct net_device *dev)
2059 return max_t(int, -dev->num_tc, 0);
2063 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2066 return &dev->_tx[index];
2069 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2070 const struct sk_buff *skb)
2072 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2075 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2076 void (*f)(struct net_device *,
2077 struct netdev_queue *,
2083 for (i = 0; i < dev->num_tx_queues; i++)
2084 f(dev, &dev->_tx[i], arg);
2087 #define netdev_lockdep_set_classes(dev) \
2089 static struct lock_class_key qdisc_tx_busylock_key; \
2090 static struct lock_class_key qdisc_running_key; \
2091 static struct lock_class_key qdisc_xmit_lock_key; \
2092 static struct lock_class_key dev_addr_list_lock_key; \
2095 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2096 (dev)->qdisc_running_key = &qdisc_running_key; \
2097 lockdep_set_class(&(dev)->addr_list_lock, \
2098 &dev_addr_list_lock_key); \
2099 for (i = 0; i < (dev)->num_tx_queues; i++) \
2100 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2101 &qdisc_xmit_lock_key); \
2104 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2105 struct sk_buff *skb,
2106 struct net_device *sb_dev);
2108 /* returns the headroom that the master device needs to take in account
2109 * when forwarding to this dev
2111 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2113 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2116 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2118 if (dev->netdev_ops->ndo_set_rx_headroom)
2119 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2122 /* set the device rx headroom to the dev's default */
2123 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2125 netdev_set_rx_headroom(dev, -1);
2129 * Net namespace inlines
2132 struct net *dev_net(const struct net_device *dev)
2134 return read_pnet(&dev->nd_net);
2138 void dev_net_set(struct net_device *dev, struct net *net)
2140 write_pnet(&dev->nd_net, net);
2144 * netdev_priv - access network device private data
2145 * @dev: network device
2147 * Get network device private data
2149 static inline void *netdev_priv(const struct net_device *dev)
2151 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2154 /* Set the sysfs physical device reference for the network logical device
2155 * if set prior to registration will cause a symlink during initialization.
2157 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2159 /* Set the sysfs device type for the network logical device to allow
2160 * fine-grained identification of different network device types. For
2161 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2163 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2165 /* Default NAPI poll() weight
2166 * Device drivers are strongly advised to not use bigger value
2168 #define NAPI_POLL_WEIGHT 64
2171 * netif_napi_add - initialize a NAPI context
2172 * @dev: network device
2173 * @napi: NAPI context
2174 * @poll: polling function
2175 * @weight: default weight
2177 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2178 * *any* of the other NAPI-related functions.
2180 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2181 int (*poll)(struct napi_struct *, int), int weight);
2184 * netif_tx_napi_add - initialize a NAPI context
2185 * @dev: network device
2186 * @napi: NAPI context
2187 * @poll: polling function
2188 * @weight: default weight
2190 * This variant of netif_napi_add() should be used from drivers using NAPI
2191 * to exclusively poll a TX queue.
2192 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2194 static inline void netif_tx_napi_add(struct net_device *dev,
2195 struct napi_struct *napi,
2196 int (*poll)(struct napi_struct *, int),
2199 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2200 netif_napi_add(dev, napi, poll, weight);
2204 * netif_napi_del - remove a NAPI context
2205 * @napi: NAPI context
2207 * netif_napi_del() removes a NAPI context from the network device NAPI list
2209 void netif_napi_del(struct napi_struct *napi);
2211 struct napi_gro_cb {
2212 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2215 /* Length of frag0. */
2216 unsigned int frag0_len;
2218 /* This indicates where we are processing relative to skb->data. */
2221 /* This is non-zero if the packet cannot be merged with the new skb. */
2224 /* Save the IP ID here and check when we get to the transport layer */
2227 /* Number of segments aggregated. */
2230 /* Start offset for remote checksum offload */
2231 u16 gro_remcsum_start;
2233 /* jiffies when first packet was created/queued */
2236 /* Used in ipv6_gro_receive() and foo-over-udp */
2239 /* This is non-zero if the packet may be of the same flow. */
2242 /* Used in tunnel GRO receive */
2245 /* GRO checksum is valid */
2248 /* Number of checksums via CHECKSUM_UNNECESSARY */
2253 #define NAPI_GRO_FREE 1
2254 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2256 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2259 /* Used in GRE, set in fou/gue_gro_receive */
2262 /* Used to determine if flush_id can be ignored */
2265 /* Number of gro_receive callbacks this packet already went through */
2266 u8 recursion_counter:4;
2270 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2273 /* used in skb_gro_receive() slow path */
2274 struct sk_buff *last;
2277 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2279 #define GRO_RECURSION_LIMIT 15
2280 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2282 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2285 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2286 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2287 struct list_head *head,
2288 struct sk_buff *skb)
2290 if (unlikely(gro_recursion_inc_test(skb))) {
2291 NAPI_GRO_CB(skb)->flush |= 1;
2295 return cb(head, skb);
2298 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2300 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2302 struct list_head *head,
2303 struct sk_buff *skb)
2305 if (unlikely(gro_recursion_inc_test(skb))) {
2306 NAPI_GRO_CB(skb)->flush |= 1;
2310 return cb(sk, head, skb);
2313 struct packet_type {
2314 __be16 type; /* This is really htons(ether_type). */
2315 struct net_device *dev; /* NULL is wildcarded here */
2316 int (*func) (struct sk_buff *,
2317 struct net_device *,
2318 struct packet_type *,
2319 struct net_device *);
2320 void (*list_func) (struct list_head *,
2321 struct packet_type *,
2322 struct net_device *);
2323 bool (*id_match)(struct packet_type *ptype,
2325 void *af_packet_priv;
2326 struct list_head list;
2329 struct offload_callbacks {
2330 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2331 netdev_features_t features);
2332 struct sk_buff *(*gro_receive)(struct list_head *head,
2333 struct sk_buff *skb);
2334 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2337 struct packet_offload {
2338 __be16 type; /* This is really htons(ether_type). */
2340 struct offload_callbacks callbacks;
2341 struct list_head list;
2344 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2345 struct pcpu_sw_netstats {
2350 struct u64_stats_sync syncp;
2353 #define __netdev_alloc_pcpu_stats(type, gfp) \
2355 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2358 for_each_possible_cpu(__cpu) { \
2359 typeof(type) *stat; \
2360 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2361 u64_stats_init(&stat->syncp); \
2367 #define netdev_alloc_pcpu_stats(type) \
2368 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2370 enum netdev_lag_tx_type {
2371 NETDEV_LAG_TX_TYPE_UNKNOWN,
2372 NETDEV_LAG_TX_TYPE_RANDOM,
2373 NETDEV_LAG_TX_TYPE_BROADCAST,
2374 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2375 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2376 NETDEV_LAG_TX_TYPE_HASH,
2379 enum netdev_lag_hash {
2380 NETDEV_LAG_HASH_NONE,
2382 NETDEV_LAG_HASH_L34,
2383 NETDEV_LAG_HASH_L23,
2384 NETDEV_LAG_HASH_E23,
2385 NETDEV_LAG_HASH_E34,
2386 NETDEV_LAG_HASH_UNKNOWN,
2389 struct netdev_lag_upper_info {
2390 enum netdev_lag_tx_type tx_type;
2391 enum netdev_lag_hash hash_type;
2394 struct netdev_lag_lower_state_info {
2399 #include <linux/notifier.h>
2401 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2402 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2406 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2408 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2409 detected a hardware crash and restarted
2410 - we can use this eg to kick tcp sessions
2412 NETDEV_CHANGE, /* Notify device state change */
2415 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2420 NETDEV_BONDING_FAILOVER,
2422 NETDEV_PRE_TYPE_CHANGE,
2423 NETDEV_POST_TYPE_CHANGE,
2426 NETDEV_NOTIFY_PEERS,
2430 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2431 NETDEV_CHANGEINFODATA,
2432 NETDEV_BONDING_INFO,
2433 NETDEV_PRECHANGEUPPER,
2434 NETDEV_CHANGELOWERSTATE,
2435 NETDEV_UDP_TUNNEL_PUSH_INFO,
2436 NETDEV_UDP_TUNNEL_DROP_INFO,
2437 NETDEV_CHANGE_TX_QUEUE_LEN,
2438 NETDEV_CVLAN_FILTER_PUSH_INFO,
2439 NETDEV_CVLAN_FILTER_DROP_INFO,
2440 NETDEV_SVLAN_FILTER_PUSH_INFO,
2441 NETDEV_SVLAN_FILTER_DROP_INFO,
2443 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2445 int register_netdevice_notifier(struct notifier_block *nb);
2446 int unregister_netdevice_notifier(struct notifier_block *nb);
2448 struct netdev_notifier_info {
2449 struct net_device *dev;
2450 struct netlink_ext_ack *extack;
2453 struct netdev_notifier_change_info {
2454 struct netdev_notifier_info info; /* must be first */
2455 unsigned int flags_changed;
2458 struct netdev_notifier_changeupper_info {
2459 struct netdev_notifier_info info; /* must be first */
2460 struct net_device *upper_dev; /* new upper dev */
2461 bool master; /* is upper dev master */
2462 bool linking; /* is the notification for link or unlink */
2463 void *upper_info; /* upper dev info */
2466 struct netdev_notifier_changelowerstate_info {
2467 struct netdev_notifier_info info; /* must be first */
2468 void *lower_state_info; /* is lower dev state */
2471 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2472 struct net_device *dev)
2475 info->extack = NULL;
2478 static inline struct net_device *
2479 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2484 static inline struct netlink_ext_ack *
2485 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2487 return info->extack;
2490 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2493 extern rwlock_t dev_base_lock; /* Device list lock */
2495 #define for_each_netdev(net, d) \
2496 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2497 #define for_each_netdev_reverse(net, d) \
2498 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2499 #define for_each_netdev_rcu(net, d) \
2500 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2501 #define for_each_netdev_safe(net, d, n) \
2502 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2503 #define for_each_netdev_continue(net, d) \
2504 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2505 #define for_each_netdev_continue_rcu(net, d) \
2506 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2507 #define for_each_netdev_in_bond_rcu(bond, slave) \
2508 for_each_netdev_rcu(&init_net, slave) \
2509 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2510 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2512 static inline struct net_device *next_net_device(struct net_device *dev)
2514 struct list_head *lh;
2518 lh = dev->dev_list.next;
2519 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2522 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2524 struct list_head *lh;
2528 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2529 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2532 static inline struct net_device *first_net_device(struct net *net)
2534 return list_empty(&net->dev_base_head) ? NULL :
2535 net_device_entry(net->dev_base_head.next);
2538 static inline struct net_device *first_net_device_rcu(struct net *net)
2540 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2542 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2545 int netdev_boot_setup_check(struct net_device *dev);
2546 unsigned long netdev_boot_base(const char *prefix, int unit);
2547 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2548 const char *hwaddr);
2549 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2550 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2551 void dev_add_pack(struct packet_type *pt);
2552 void dev_remove_pack(struct packet_type *pt);
2553 void __dev_remove_pack(struct packet_type *pt);
2554 void dev_add_offload(struct packet_offload *po);
2555 void dev_remove_offload(struct packet_offload *po);
2557 int dev_get_iflink(const struct net_device *dev);
2558 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2559 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2560 unsigned short mask);
2561 struct net_device *dev_get_by_name(struct net *net, const char *name);
2562 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2563 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2564 int dev_alloc_name(struct net_device *dev, const char *name);
2565 int dev_open(struct net_device *dev);
2566 void dev_close(struct net_device *dev);
2567 void dev_close_many(struct list_head *head, bool unlink);
2568 void dev_disable_lro(struct net_device *dev);
2569 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2570 int dev_queue_xmit(struct sk_buff *skb);
2571 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2572 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2573 int register_netdevice(struct net_device *dev);
2574 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2575 void unregister_netdevice_many(struct list_head *head);
2576 static inline void unregister_netdevice(struct net_device *dev)
2578 unregister_netdevice_queue(dev, NULL);
2581 int netdev_refcnt_read(const struct net_device *dev);
2582 void free_netdev(struct net_device *dev);
2583 void netdev_freemem(struct net_device *dev);
2584 void synchronize_net(void);
2585 int init_dummy_netdev(struct net_device *dev);
2587 DECLARE_PER_CPU(int, xmit_recursion);
2588 #define XMIT_RECURSION_LIMIT 10
2590 static inline int dev_recursion_level(void)
2592 return this_cpu_read(xmit_recursion);
2595 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2596 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2597 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2598 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2599 int netdev_get_name(struct net *net, char *name, int ifindex);
2600 int dev_restart(struct net_device *dev);
2601 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2603 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2605 return NAPI_GRO_CB(skb)->data_offset;
2608 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2610 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2613 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2615 NAPI_GRO_CB(skb)->data_offset += len;
2618 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2619 unsigned int offset)
2621 return NAPI_GRO_CB(skb)->frag0 + offset;
2624 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2626 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2629 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2631 NAPI_GRO_CB(skb)->frag0 = NULL;
2632 NAPI_GRO_CB(skb)->frag0_len = 0;
2635 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2636 unsigned int offset)
2638 if (!pskb_may_pull(skb, hlen))
2641 skb_gro_frag0_invalidate(skb);
2642 return skb->data + offset;
2645 static inline void *skb_gro_network_header(struct sk_buff *skb)
2647 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2648 skb_network_offset(skb);
2651 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2652 const void *start, unsigned int len)
2654 if (NAPI_GRO_CB(skb)->csum_valid)
2655 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2656 csum_partial(start, len, 0));
2659 /* GRO checksum functions. These are logical equivalents of the normal
2660 * checksum functions (in skbuff.h) except that they operate on the GRO
2661 * offsets and fields in sk_buff.
2664 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2666 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2668 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2671 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2675 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2676 skb_checksum_start_offset(skb) <
2677 skb_gro_offset(skb)) &&
2678 !skb_at_gro_remcsum_start(skb) &&
2679 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2680 (!zero_okay || check));
2683 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2686 if (NAPI_GRO_CB(skb)->csum_valid &&
2687 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2690 NAPI_GRO_CB(skb)->csum = psum;
2692 return __skb_gro_checksum_complete(skb);
2695 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2697 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2698 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2699 NAPI_GRO_CB(skb)->csum_cnt--;
2701 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2702 * verified a new top level checksum or an encapsulated one
2703 * during GRO. This saves work if we fallback to normal path.
2705 __skb_incr_checksum_unnecessary(skb);
2709 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2712 __sum16 __ret = 0; \
2713 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2714 __ret = __skb_gro_checksum_validate_complete(skb, \
2715 compute_pseudo(skb, proto)); \
2717 skb_gro_incr_csum_unnecessary(skb); \
2721 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2722 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2724 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2726 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2728 #define skb_gro_checksum_simple_validate(skb) \
2729 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2731 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2733 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2734 !NAPI_GRO_CB(skb)->csum_valid);
2737 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2738 __sum16 check, __wsum pseudo)
2740 NAPI_GRO_CB(skb)->csum = ~pseudo;
2741 NAPI_GRO_CB(skb)->csum_valid = 1;
2744 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2746 if (__skb_gro_checksum_convert_check(skb)) \
2747 __skb_gro_checksum_convert(skb, check, \
2748 compute_pseudo(skb, proto)); \
2751 struct gro_remcsum {
2756 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2762 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2763 unsigned int off, size_t hdrlen,
2764 int start, int offset,
2765 struct gro_remcsum *grc,
2769 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2771 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2774 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2778 ptr = skb_gro_header_fast(skb, off);
2779 if (skb_gro_header_hard(skb, off + plen)) {
2780 ptr = skb_gro_header_slow(skb, off + plen, off);
2785 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2788 /* Adjust skb->csum since we changed the packet */
2789 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2791 grc->offset = off + hdrlen + offset;
2797 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2798 struct gro_remcsum *grc)
2801 size_t plen = grc->offset + sizeof(u16);
2806 ptr = skb_gro_header_fast(skb, grc->offset);
2807 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2808 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2813 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2816 #ifdef CONFIG_XFRM_OFFLOAD
2817 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2819 if (PTR_ERR(pp) != -EINPROGRESS)
2820 NAPI_GRO_CB(skb)->flush |= flush;
2822 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2825 struct gro_remcsum *grc)
2827 if (PTR_ERR(pp) != -EINPROGRESS) {
2828 NAPI_GRO_CB(skb)->flush |= flush;
2829 skb_gro_remcsum_cleanup(skb, grc);
2830 skb->remcsum_offload = 0;
2834 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2836 NAPI_GRO_CB(skb)->flush |= flush;
2838 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2841 struct gro_remcsum *grc)
2843 NAPI_GRO_CB(skb)->flush |= flush;
2844 skb_gro_remcsum_cleanup(skb, grc);
2845 skb->remcsum_offload = 0;
2849 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2850 unsigned short type,
2851 const void *daddr, const void *saddr,
2854 if (!dev->header_ops || !dev->header_ops->create)
2857 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2860 static inline int dev_parse_header(const struct sk_buff *skb,
2861 unsigned char *haddr)
2863 const struct net_device *dev = skb->dev;
2865 if (!dev->header_ops || !dev->header_ops->parse)
2867 return dev->header_ops->parse(skb, haddr);
2870 /* ll_header must have at least hard_header_len allocated */
2871 static inline bool dev_validate_header(const struct net_device *dev,
2872 char *ll_header, int len)
2874 if (likely(len >= dev->hard_header_len))
2876 if (len < dev->min_header_len)
2879 if (capable(CAP_SYS_RAWIO)) {
2880 memset(ll_header + len, 0, dev->hard_header_len - len);
2884 if (dev->header_ops && dev->header_ops->validate)
2885 return dev->header_ops->validate(ll_header, len);
2890 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2892 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2893 static inline int unregister_gifconf(unsigned int family)
2895 return register_gifconf(family, NULL);
2898 #ifdef CONFIG_NET_FLOW_LIMIT
2899 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2900 struct sd_flow_limit {
2902 unsigned int num_buckets;
2903 unsigned int history_head;
2904 u16 history[FLOW_LIMIT_HISTORY];
2908 extern int netdev_flow_limit_table_len;
2909 #endif /* CONFIG_NET_FLOW_LIMIT */
2912 * Incoming packets are placed on per-CPU queues
2914 struct softnet_data {
2915 struct list_head poll_list;
2916 struct sk_buff_head process_queue;
2919 unsigned int processed;
2920 unsigned int time_squeeze;
2921 unsigned int received_rps;
2923 struct softnet_data *rps_ipi_list;
2925 #ifdef CONFIG_NET_FLOW_LIMIT
2926 struct sd_flow_limit __rcu *flow_limit;
2928 struct Qdisc *output_queue;
2929 struct Qdisc **output_queue_tailp;
2930 struct sk_buff *completion_queue;
2931 #ifdef CONFIG_XFRM_OFFLOAD
2932 struct sk_buff_head xfrm_backlog;
2935 /* input_queue_head should be written by cpu owning this struct,
2936 * and only read by other cpus. Worth using a cache line.
2938 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2940 /* Elements below can be accessed between CPUs for RPS/RFS */
2941 call_single_data_t csd ____cacheline_aligned_in_smp;
2942 struct softnet_data *rps_ipi_next;
2944 unsigned int input_queue_tail;
2946 unsigned int dropped;
2947 struct sk_buff_head input_pkt_queue;
2948 struct napi_struct backlog;
2952 static inline void input_queue_head_incr(struct softnet_data *sd)
2955 sd->input_queue_head++;
2959 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2960 unsigned int *qtail)
2963 *qtail = ++sd->input_queue_tail;
2967 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2969 void __netif_schedule(struct Qdisc *q);
2970 void netif_schedule_queue(struct netdev_queue *txq);
2972 static inline void netif_tx_schedule_all(struct net_device *dev)
2976 for (i = 0; i < dev->num_tx_queues; i++)
2977 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2980 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2982 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2986 * netif_start_queue - allow transmit
2987 * @dev: network device
2989 * Allow upper layers to call the device hard_start_xmit routine.
2991 static inline void netif_start_queue(struct net_device *dev)
2993 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2996 static inline void netif_tx_start_all_queues(struct net_device *dev)
3000 for (i = 0; i < dev->num_tx_queues; i++) {
3001 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3002 netif_tx_start_queue(txq);
3006 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3009 * netif_wake_queue - restart transmit
3010 * @dev: network device
3012 * Allow upper layers to call the device hard_start_xmit routine.
3013 * Used for flow control when transmit resources are available.
3015 static inline void netif_wake_queue(struct net_device *dev)
3017 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3020 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3024 for (i = 0; i < dev->num_tx_queues; i++) {
3025 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3026 netif_tx_wake_queue(txq);
3030 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3032 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3036 * netif_stop_queue - stop transmitted packets
3037 * @dev: network device
3039 * Stop upper layers calling the device hard_start_xmit routine.
3040 * Used for flow control when transmit resources are unavailable.
3042 static inline void netif_stop_queue(struct net_device *dev)
3044 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3047 void netif_tx_stop_all_queues(struct net_device *dev);
3049 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3051 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3055 * netif_queue_stopped - test if transmit queue is flowblocked
3056 * @dev: network device
3058 * Test if transmit queue on device is currently unable to send.
3060 static inline bool netif_queue_stopped(const struct net_device *dev)
3062 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3065 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3067 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3071 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3073 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3077 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3079 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3083 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3084 * @dev_queue: pointer to transmit queue
3086 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3087 * to give appropriate hint to the CPU.
3089 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3092 prefetchw(&dev_queue->dql.num_queued);
3097 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3098 * @dev_queue: pointer to transmit queue
3100 * BQL enabled drivers might use this helper in their TX completion path,
3101 * to give appropriate hint to the CPU.
3103 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3106 prefetchw(&dev_queue->dql.limit);
3110 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3114 dql_queued(&dev_queue->dql, bytes);
3116 if (likely(dql_avail(&dev_queue->dql) >= 0))
3119 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3122 * The XOFF flag must be set before checking the dql_avail below,
3123 * because in netdev_tx_completed_queue we update the dql_completed
3124 * before checking the XOFF flag.
3128 /* check again in case another CPU has just made room avail */
3129 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3130 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3135 * netdev_sent_queue - report the number of bytes queued to hardware
3136 * @dev: network device
3137 * @bytes: number of bytes queued to the hardware device queue
3139 * Report the number of bytes queued for sending/completion to the network
3140 * device hardware queue. @bytes should be a good approximation and should
3141 * exactly match netdev_completed_queue() @bytes
3143 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3145 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3148 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3149 unsigned int pkts, unsigned int bytes)
3152 if (unlikely(!bytes))
3155 dql_completed(&dev_queue->dql, bytes);
3158 * Without the memory barrier there is a small possiblity that
3159 * netdev_tx_sent_queue will miss the update and cause the queue to
3160 * be stopped forever
3164 if (dql_avail(&dev_queue->dql) < 0)
3167 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3168 netif_schedule_queue(dev_queue);
3173 * netdev_completed_queue - report bytes and packets completed by device
3174 * @dev: network device
3175 * @pkts: actual number of packets sent over the medium
3176 * @bytes: actual number of bytes sent over the medium
3178 * Report the number of bytes and packets transmitted by the network device
3179 * hardware queue over the physical medium, @bytes must exactly match the
3180 * @bytes amount passed to netdev_sent_queue()
3182 static inline void netdev_completed_queue(struct net_device *dev,
3183 unsigned int pkts, unsigned int bytes)
3185 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3188 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3191 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3197 * netdev_reset_queue - reset the packets and bytes count of a network device
3198 * @dev_queue: network device
3200 * Reset the bytes and packet count of a network device and clear the
3201 * software flow control OFF bit for this network device
3203 static inline void netdev_reset_queue(struct net_device *dev_queue)
3205 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3209 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3210 * @dev: network device
3211 * @queue_index: given tx queue index
3213 * Returns 0 if given tx queue index >= number of device tx queues,
3214 * otherwise returns the originally passed tx queue index.
3216 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3218 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3219 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3220 dev->name, queue_index,
3221 dev->real_num_tx_queues);
3229 * netif_running - test if up
3230 * @dev: network device
3232 * Test if the device has been brought up.
3234 static inline bool netif_running(const struct net_device *dev)
3236 return test_bit(__LINK_STATE_START, &dev->state);
3240 * Routines to manage the subqueues on a device. We only need start,
3241 * stop, and a check if it's stopped. All other device management is
3242 * done at the overall netdevice level.
3243 * Also test the device if we're multiqueue.
3247 * netif_start_subqueue - allow sending packets on subqueue
3248 * @dev: network device
3249 * @queue_index: sub queue index
3251 * Start individual transmit queue of a device with multiple transmit queues.
3253 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3255 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3257 netif_tx_start_queue(txq);
3261 * netif_stop_subqueue - stop sending packets on subqueue
3262 * @dev: network device
3263 * @queue_index: sub queue index
3265 * Stop individual transmit queue of a device with multiple transmit queues.
3267 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3269 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3270 netif_tx_stop_queue(txq);
3274 * netif_subqueue_stopped - test status of subqueue
3275 * @dev: network device
3276 * @queue_index: sub queue index
3278 * Check individual transmit queue of a device with multiple transmit queues.
3280 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3283 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3285 return netif_tx_queue_stopped(txq);
3288 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3289 struct sk_buff *skb)
3291 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3295 * netif_wake_subqueue - allow sending packets on subqueue
3296 * @dev: network device
3297 * @queue_index: sub queue index
3299 * Resume individual transmit queue of a device with multiple transmit queues.
3301 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3303 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3305 netif_tx_wake_queue(txq);
3309 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3311 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3312 u16 index, bool is_rxqs_map);
3315 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3316 * @j: CPU/Rx queue index
3317 * @mask: bitmask of all cpus/rx queues
3318 * @nr_bits: number of bits in the bitmask
3320 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3322 static inline bool netif_attr_test_mask(unsigned long j,
3323 const unsigned long *mask,
3324 unsigned int nr_bits)
3326 cpu_max_bits_warn(j, nr_bits);
3327 return test_bit(j, mask);
3331 * netif_attr_test_online - Test for online CPU/Rx queue
3332 * @j: CPU/Rx queue index
3333 * @online_mask: bitmask for CPUs/Rx queues that are online
3334 * @nr_bits: number of bits in the bitmask
3336 * Returns true if a CPU/Rx queue is online.
3338 static inline bool netif_attr_test_online(unsigned long j,
3339 const unsigned long *online_mask,
3340 unsigned int nr_bits)
3342 cpu_max_bits_warn(j, nr_bits);
3345 return test_bit(j, online_mask);
3347 return (j < nr_bits);
3351 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3352 * @n: CPU/Rx queue index
3353 * @srcp: the cpumask/Rx queue mask pointer
3354 * @nr_bits: number of bits in the bitmask
3356 * Returns >= nr_bits if no further CPUs/Rx queues set.
3358 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3359 unsigned int nr_bits)
3361 /* -1 is a legal arg here. */
3363 cpu_max_bits_warn(n, nr_bits);
3366 return find_next_bit(srcp, nr_bits, n + 1);
3372 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3373 * @n: CPU/Rx queue index
3374 * @src1p: the first CPUs/Rx queues mask pointer
3375 * @src2p: the second CPUs/Rx queues mask pointer
3376 * @nr_bits: number of bits in the bitmask
3378 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3380 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3381 const unsigned long *src2p,
3382 unsigned int nr_bits)
3384 /* -1 is a legal arg here. */
3386 cpu_max_bits_warn(n, nr_bits);
3389 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3391 return find_next_bit(src1p, nr_bits, n + 1);
3393 return find_next_bit(src2p, nr_bits, n + 1);
3398 static inline int netif_set_xps_queue(struct net_device *dev,
3399 const struct cpumask *mask,
3407 * netif_is_multiqueue - test if device has multiple transmit queues
3408 * @dev: network device
3410 * Check if device has multiple transmit queues
3412 static inline bool netif_is_multiqueue(const struct net_device *dev)
3414 return dev->num_tx_queues > 1;
3417 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3420 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3422 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3429 static inline struct netdev_rx_queue *
3430 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3432 return dev->_rx + rxq;
3436 static inline unsigned int get_netdev_rx_queue_index(
3437 struct netdev_rx_queue *queue)
3439 struct net_device *dev = queue->dev;
3440 int index = queue - dev->_rx;
3442 BUG_ON(index >= dev->num_rx_queues);
3447 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3448 int netif_get_num_default_rss_queues(void);
3450 enum skb_free_reason {
3451 SKB_REASON_CONSUMED,
3455 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3456 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3459 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3460 * interrupt context or with hardware interrupts being disabled.
3461 * (in_irq() || irqs_disabled())
3463 * We provide four helpers that can be used in following contexts :
3465 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3466 * replacing kfree_skb(skb)
3468 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3469 * Typically used in place of consume_skb(skb) in TX completion path
3471 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3472 * replacing kfree_skb(skb)
3474 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3475 * and consumed a packet. Used in place of consume_skb(skb)
3477 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3479 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3482 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3484 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3487 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3489 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3492 static inline void dev_consume_skb_any(struct sk_buff *skb)
3494 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3497 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3498 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3499 int netif_rx(struct sk_buff *skb);
3500 int netif_rx_ni(struct sk_buff *skb);
3501 int netif_receive_skb(struct sk_buff *skb);
3502 int netif_receive_skb_core(struct sk_buff *skb);
3503 void netif_receive_skb_list(struct list_head *head);
3504 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3505 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3506 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3507 gro_result_t napi_gro_frags(struct napi_struct *napi);
3508 struct packet_offload *gro_find_receive_by_type(__be16 type);
3509 struct packet_offload *gro_find_complete_by_type(__be16 type);
3511 static inline void napi_free_frags(struct napi_struct *napi)
3513 kfree_skb(napi->skb);
3517 bool netdev_is_rx_handler_busy(struct net_device *dev);
3518 int netdev_rx_handler_register(struct net_device *dev,
3519 rx_handler_func_t *rx_handler,
3520 void *rx_handler_data);
3521 void netdev_rx_handler_unregister(struct net_device *dev);
3523 bool dev_valid_name(const char *name);
3524 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3525 bool *need_copyout);
3526 int dev_ifconf(struct net *net, struct ifconf *, int);
3527 int dev_ethtool(struct net *net, struct ifreq *);
3528 unsigned int dev_get_flags(const struct net_device *);
3529 int __dev_change_flags(struct net_device *, unsigned int flags);
3530 int dev_change_flags(struct net_device *, unsigned int);
3531 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3532 unsigned int gchanges);
3533 int dev_change_name(struct net_device *, const char *);
3534 int dev_set_alias(struct net_device *, const char *, size_t);
3535 int dev_get_alias(const struct net_device *, char *, size_t);
3536 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3537 int __dev_set_mtu(struct net_device *, int);
3538 int dev_set_mtu(struct net_device *, int);
3539 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3540 void dev_set_group(struct net_device *, int);
3541 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3542 int dev_change_carrier(struct net_device *, bool new_carrier);
3543 int dev_get_phys_port_id(struct net_device *dev,
3544 struct netdev_phys_item_id *ppid);
3545 int dev_get_phys_port_name(struct net_device *dev,
3546 char *name, size_t len);
3547 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3548 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3549 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3550 struct netdev_queue *txq, int *ret);
3552 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3553 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3555 void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3556 struct netdev_bpf *xdp);
3558 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3559 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3560 bool is_skb_forwardable(const struct net_device *dev,
3561 const struct sk_buff *skb);
3563 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3564 struct sk_buff *skb)
3566 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3567 unlikely(!is_skb_forwardable(dev, skb))) {
3568 atomic_long_inc(&dev->rx_dropped);
3573 skb_scrub_packet(skb, true);
3578 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3580 extern int netdev_budget;
3581 extern unsigned int netdev_budget_usecs;
3583 /* Called by rtnetlink.c:rtnl_unlock() */
3584 void netdev_run_todo(void);
3587 * dev_put - release reference to device
3588 * @dev: network device
3590 * Release reference to device to allow it to be freed.
3592 static inline void dev_put(struct net_device *dev)
3594 this_cpu_dec(*dev->pcpu_refcnt);
3598 * dev_hold - get reference to device
3599 * @dev: network device
3601 * Hold reference to device to keep it from being freed.
3603 static inline void dev_hold(struct net_device *dev)
3605 this_cpu_inc(*dev->pcpu_refcnt);
3608 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3609 * and _off may be called from IRQ context, but it is caller
3610 * who is responsible for serialization of these calls.
3612 * The name carrier is inappropriate, these functions should really be
3613 * called netif_lowerlayer_*() because they represent the state of any
3614 * kind of lower layer not just hardware media.
3617 void linkwatch_init_dev(struct net_device *dev);
3618 void linkwatch_fire_event(struct net_device *dev);
3619 void linkwatch_forget_dev(struct net_device *dev);
3622 * netif_carrier_ok - test if carrier present
3623 * @dev: network device
3625 * Check if carrier is present on device
3627 static inline bool netif_carrier_ok(const struct net_device *dev)
3629 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3632 unsigned long dev_trans_start(struct net_device *dev);
3634 void __netdev_watchdog_up(struct net_device *dev);
3636 void netif_carrier_on(struct net_device *dev);
3638 void netif_carrier_off(struct net_device *dev);
3641 * netif_dormant_on - mark device as dormant.
3642 * @dev: network device
3644 * Mark device as dormant (as per RFC2863).
3646 * The dormant state indicates that the relevant interface is not
3647 * actually in a condition to pass packets (i.e., it is not 'up') but is
3648 * in a "pending" state, waiting for some external event. For "on-
3649 * demand" interfaces, this new state identifies the situation where the
3650 * interface is waiting for events to place it in the up state.
3652 static inline void netif_dormant_on(struct net_device *dev)
3654 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3655 linkwatch_fire_event(dev);
3659 * netif_dormant_off - set device as not dormant.
3660 * @dev: network device
3662 * Device is not in dormant state.
3664 static inline void netif_dormant_off(struct net_device *dev)
3666 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3667 linkwatch_fire_event(dev);
3671 * netif_dormant - test if device is dormant
3672 * @dev: network device
3674 * Check if device is dormant.
3676 static inline bool netif_dormant(const struct net_device *dev)
3678 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3683 * netif_oper_up - test if device is operational
3684 * @dev: network device
3686 * Check if carrier is operational
3688 static inline bool netif_oper_up(const struct net_device *dev)
3690 return (dev->operstate == IF_OPER_UP ||
3691 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3695 * netif_device_present - is device available or removed
3696 * @dev: network device
3698 * Check if device has not been removed from system.
3700 static inline bool netif_device_present(struct net_device *dev)
3702 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3705 void netif_device_detach(struct net_device *dev);
3707 void netif_device_attach(struct net_device *dev);
3710 * Network interface message level settings
3714 NETIF_MSG_DRV = 0x0001,
3715 NETIF_MSG_PROBE = 0x0002,
3716 NETIF_MSG_LINK = 0x0004,
3717 NETIF_MSG_TIMER = 0x0008,
3718 NETIF_MSG_IFDOWN = 0x0010,
3719 NETIF_MSG_IFUP = 0x0020,
3720 NETIF_MSG_RX_ERR = 0x0040,
3721 NETIF_MSG_TX_ERR = 0x0080,
3722 NETIF_MSG_TX_QUEUED = 0x0100,
3723 NETIF_MSG_INTR = 0x0200,
3724 NETIF_MSG_TX_DONE = 0x0400,
3725 NETIF_MSG_RX_STATUS = 0x0800,
3726 NETIF_MSG_PKTDATA = 0x1000,
3727 NETIF_MSG_HW = 0x2000,
3728 NETIF_MSG_WOL = 0x4000,
3731 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3732 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3733 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3734 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3735 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3736 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3737 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3738 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3739 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3740 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3741 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3742 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3743 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3744 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3745 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3747 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3750 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3751 return default_msg_enable_bits;
3752 if (debug_value == 0) /* no output */
3754 /* set low N bits */
3755 return (1 << debug_value) - 1;
3758 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3760 spin_lock(&txq->_xmit_lock);
3761 txq->xmit_lock_owner = cpu;
3764 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3766 __acquire(&txq->_xmit_lock);
3770 static inline void __netif_tx_release(struct netdev_queue *txq)
3772 __release(&txq->_xmit_lock);
3775 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3777 spin_lock_bh(&txq->_xmit_lock);
3778 txq->xmit_lock_owner = smp_processor_id();
3781 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3783 bool ok = spin_trylock(&txq->_xmit_lock);
3785 txq->xmit_lock_owner = smp_processor_id();
3789 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3791 txq->xmit_lock_owner = -1;
3792 spin_unlock(&txq->_xmit_lock);
3795 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3797 txq->xmit_lock_owner = -1;
3798 spin_unlock_bh(&txq->_xmit_lock);
3801 static inline void txq_trans_update(struct netdev_queue *txq)
3803 if (txq->xmit_lock_owner != -1)
3804 txq->trans_start = jiffies;
3807 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3808 static inline void netif_trans_update(struct net_device *dev)
3810 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3812 if (txq->trans_start != jiffies)
3813 txq->trans_start = jiffies;
3817 * netif_tx_lock - grab network device transmit lock
3818 * @dev: network device
3820 * Get network device transmit lock
3822 static inline void netif_tx_lock(struct net_device *dev)
3827 spin_lock(&dev->tx_global_lock);
3828 cpu = smp_processor_id();
3829 for (i = 0; i < dev->num_tx_queues; i++) {
3830 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3832 /* We are the only thread of execution doing a
3833 * freeze, but we have to grab the _xmit_lock in
3834 * order to synchronize with threads which are in
3835 * the ->hard_start_xmit() handler and already
3836 * checked the frozen bit.
3838 __netif_tx_lock(txq, cpu);
3839 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3840 __netif_tx_unlock(txq);
3844 static inline void netif_tx_lock_bh(struct net_device *dev)
3850 static inline void netif_tx_unlock(struct net_device *dev)
3854 for (i = 0; i < dev->num_tx_queues; i++) {
3855 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3857 /* No need to grab the _xmit_lock here. If the
3858 * queue is not stopped for another reason, we
3861 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3862 netif_schedule_queue(txq);
3864 spin_unlock(&dev->tx_global_lock);
3867 static inline void netif_tx_unlock_bh(struct net_device *dev)
3869 netif_tx_unlock(dev);
3873 #define HARD_TX_LOCK(dev, txq, cpu) { \
3874 if ((dev->features & NETIF_F_LLTX) == 0) { \
3875 __netif_tx_lock(txq, cpu); \
3877 __netif_tx_acquire(txq); \
3881 #define HARD_TX_TRYLOCK(dev, txq) \
3882 (((dev->features & NETIF_F_LLTX) == 0) ? \
3883 __netif_tx_trylock(txq) : \
3884 __netif_tx_acquire(txq))
3886 #define HARD_TX_UNLOCK(dev, txq) { \
3887 if ((dev->features & NETIF_F_LLTX) == 0) { \
3888 __netif_tx_unlock(txq); \
3890 __netif_tx_release(txq); \
3894 static inline void netif_tx_disable(struct net_device *dev)
3900 cpu = smp_processor_id();
3901 for (i = 0; i < dev->num_tx_queues; i++) {
3902 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3904 __netif_tx_lock(txq, cpu);
3905 netif_tx_stop_queue(txq);
3906 __netif_tx_unlock(txq);
3911 static inline void netif_addr_lock(struct net_device *dev)
3913 spin_lock(&dev->addr_list_lock);
3916 static inline void netif_addr_lock_nested(struct net_device *dev)
3918 int subclass = SINGLE_DEPTH_NESTING;
3920 if (dev->netdev_ops->ndo_get_lock_subclass)
3921 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3923 spin_lock_nested(&dev->addr_list_lock, subclass);
3926 static inline void netif_addr_lock_bh(struct net_device *dev)
3928 spin_lock_bh(&dev->addr_list_lock);
3931 static inline void netif_addr_unlock(struct net_device *dev)
3933 spin_unlock(&dev->addr_list_lock);
3936 static inline void netif_addr_unlock_bh(struct net_device *dev)
3938 spin_unlock_bh(&dev->addr_list_lock);
3942 * dev_addrs walker. Should be used only for read access. Call with
3943 * rcu_read_lock held.
3945 #define for_each_dev_addr(dev, ha) \
3946 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3948 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3950 void ether_setup(struct net_device *dev);
3952 /* Support for loadable net-drivers */
3953 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3954 unsigned char name_assign_type,
3955 void (*setup)(struct net_device *),
3956 unsigned int txqs, unsigned int rxqs);
3957 int dev_get_valid_name(struct net *net, struct net_device *dev,
3960 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3961 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3963 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3964 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3967 int register_netdev(struct net_device *dev);
3968 void unregister_netdev(struct net_device *dev);
3970 /* General hardware address lists handling functions */
3971 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3972 struct netdev_hw_addr_list *from_list, int addr_len);
3973 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3974 struct netdev_hw_addr_list *from_list, int addr_len);
3975 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3976 struct net_device *dev,
3977 int (*sync)(struct net_device *, const unsigned char *),
3978 int (*unsync)(struct net_device *,
3979 const unsigned char *));
3980 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3981 struct net_device *dev,
3982 int (*unsync)(struct net_device *,
3983 const unsigned char *));
3984 void __hw_addr_init(struct netdev_hw_addr_list *list);
3986 /* Functions used for device addresses handling */
3987 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3988 unsigned char addr_type);
3989 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3990 unsigned char addr_type);
3991 void dev_addr_flush(struct net_device *dev);
3992 int dev_addr_init(struct net_device *dev);
3994 /* Functions used for unicast addresses handling */
3995 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3996 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3997 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3998 int dev_uc_sync(struct net_device *to, struct net_device *from);
3999 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4000 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4001 void dev_uc_flush(struct net_device *dev);
4002 void dev_uc_init(struct net_device *dev);
4005 * __dev_uc_sync - Synchonize device's unicast list
4006 * @dev: device to sync
4007 * @sync: function to call if address should be added
4008 * @unsync: function to call if address should be removed
4010 * Add newly added addresses to the interface, and release
4011 * addresses that have been deleted.
4013 static inline int __dev_uc_sync(struct net_device *dev,
4014 int (*sync)(struct net_device *,
4015 const unsigned char *),
4016 int (*unsync)(struct net_device *,
4017 const unsigned char *))
4019 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4023 * __dev_uc_unsync - Remove synchronized addresses from device
4024 * @dev: device to sync
4025 * @unsync: function to call if address should be removed
4027 * Remove all addresses that were added to the device by dev_uc_sync().
4029 static inline void __dev_uc_unsync(struct net_device *dev,
4030 int (*unsync)(struct net_device *,
4031 const unsigned char *))
4033 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4036 /* Functions used for multicast addresses handling */
4037 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4038 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4039 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4040 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4041 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4042 int dev_mc_sync(struct net_device *to, struct net_device *from);
4043 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4044 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4045 void dev_mc_flush(struct net_device *dev);
4046 void dev_mc_init(struct net_device *dev);
4049 * __dev_mc_sync - Synchonize device's multicast list
4050 * @dev: device to sync
4051 * @sync: function to call if address should be added
4052 * @unsync: function to call if address should be removed
4054 * Add newly added addresses to the interface, and release
4055 * addresses that have been deleted.
4057 static inline int __dev_mc_sync(struct net_device *dev,
4058 int (*sync)(struct net_device *,
4059 const unsigned char *),
4060 int (*unsync)(struct net_device *,
4061 const unsigned char *))
4063 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4067 * __dev_mc_unsync - Remove synchronized addresses from device
4068 * @dev: device to sync
4069 * @unsync: function to call if address should be removed
4071 * Remove all addresses that were added to the device by dev_mc_sync().
4073 static inline void __dev_mc_unsync(struct net_device *dev,
4074 int (*unsync)(struct net_device *,
4075 const unsigned char *))
4077 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4080 /* Functions used for secondary unicast and multicast support */
4081 void dev_set_rx_mode(struct net_device *dev);
4082 void __dev_set_rx_mode(struct net_device *dev);
4083 int dev_set_promiscuity(struct net_device *dev, int inc);
4084 int dev_set_allmulti(struct net_device *dev, int inc);
4085 void netdev_state_change(struct net_device *dev);
4086 void netdev_notify_peers(struct net_device *dev);
4087 void netdev_features_change(struct net_device *dev);
4088 /* Load a device via the kmod */
4089 void dev_load(struct net *net, const char *name);
4090 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4091 struct rtnl_link_stats64 *storage);
4092 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4093 const struct net_device_stats *netdev_stats);
4095 extern int netdev_max_backlog;
4096 extern int netdev_tstamp_prequeue;
4097 extern int weight_p;
4098 extern int dev_weight_rx_bias;
4099 extern int dev_weight_tx_bias;
4100 extern int dev_rx_weight;
4101 extern int dev_tx_weight;
4103 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4104 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4105 struct list_head **iter);
4106 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4107 struct list_head **iter);
4109 /* iterate through upper list, must be called under RCU read lock */
4110 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4111 for (iter = &(dev)->adj_list.upper, \
4112 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4114 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4116 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4117 int (*fn)(struct net_device *upper_dev,
4121 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4122 struct net_device *upper_dev);
4124 bool netdev_has_any_upper_dev(struct net_device *dev);
4126 void *netdev_lower_get_next_private(struct net_device *dev,
4127 struct list_head **iter);
4128 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4129 struct list_head **iter);
4131 #define netdev_for_each_lower_private(dev, priv, iter) \
4132 for (iter = (dev)->adj_list.lower.next, \
4133 priv = netdev_lower_get_next_private(dev, &(iter)); \
4135 priv = netdev_lower_get_next_private(dev, &(iter)))
4137 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4138 for (iter = &(dev)->adj_list.lower, \
4139 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4141 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4143 void *netdev_lower_get_next(struct net_device *dev,
4144 struct list_head **iter);
4146 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4147 for (iter = (dev)->adj_list.lower.next, \
4148 ldev = netdev_lower_get_next(dev, &(iter)); \
4150 ldev = netdev_lower_get_next(dev, &(iter)))
4152 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4153 struct list_head **iter);
4154 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4155 struct list_head **iter);
4157 int netdev_walk_all_lower_dev(struct net_device *dev,
4158 int (*fn)(struct net_device *lower_dev,
4161 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4162 int (*fn)(struct net_device *lower_dev,
4166 void *netdev_adjacent_get_private(struct list_head *adj_list);
4167 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4168 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4169 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4170 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4171 struct netlink_ext_ack *extack);
4172 int netdev_master_upper_dev_link(struct net_device *dev,
4173 struct net_device *upper_dev,
4174 void *upper_priv, void *upper_info,
4175 struct netlink_ext_ack *extack);
4176 void netdev_upper_dev_unlink(struct net_device *dev,
4177 struct net_device *upper_dev);
4178 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4179 void *netdev_lower_dev_get_private(struct net_device *dev,
4180 struct net_device *lower_dev);
4181 void netdev_lower_state_changed(struct net_device *lower_dev,
4182 void *lower_state_info);
4184 /* RSS keys are 40 or 52 bytes long */
4185 #define NETDEV_RSS_KEY_LEN 52
4186 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4187 void netdev_rss_key_fill(void *buffer, size_t len);
4189 int dev_get_nest_level(struct net_device *dev);
4190 int skb_checksum_help(struct sk_buff *skb);
4191 int skb_crc32c_csum_help(struct sk_buff *skb);
4192 int skb_csum_hwoffload_help(struct sk_buff *skb,
4193 const netdev_features_t features);
4195 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4196 netdev_features_t features, bool tx_path);
4197 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4198 netdev_features_t features);
4200 struct netdev_bonding_info {
4205 struct netdev_notifier_bonding_info {
4206 struct netdev_notifier_info info; /* must be first */
4207 struct netdev_bonding_info bonding_info;
4210 void netdev_bonding_info_change(struct net_device *dev,
4211 struct netdev_bonding_info *bonding_info);
4214 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4216 return __skb_gso_segment(skb, features, true);
4218 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4220 static inline bool can_checksum_protocol(netdev_features_t features,
4223 if (protocol == htons(ETH_P_FCOE))
4224 return !!(features & NETIF_F_FCOE_CRC);
4226 /* Assume this is an IP checksum (not SCTP CRC) */
4228 if (features & NETIF_F_HW_CSUM) {
4229 /* Can checksum everything */
4234 case htons(ETH_P_IP):
4235 return !!(features & NETIF_F_IP_CSUM);
4236 case htons(ETH_P_IPV6):
4237 return !!(features & NETIF_F_IPV6_CSUM);
4244 void netdev_rx_csum_fault(struct net_device *dev);
4246 static inline void netdev_rx_csum_fault(struct net_device *dev)
4250 /* rx skb timestamps */
4251 void net_enable_timestamp(void);
4252 void net_disable_timestamp(void);
4254 #ifdef CONFIG_PROC_FS
4255 int __init dev_proc_init(void);
4257 #define dev_proc_init() 0
4260 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4261 struct sk_buff *skb, struct net_device *dev,
4264 skb->xmit_more = more ? 1 : 0;
4265 return ops->ndo_start_xmit(skb, dev);
4268 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4269 struct netdev_queue *txq, bool more)
4271 const struct net_device_ops *ops = dev->netdev_ops;
4274 rc = __netdev_start_xmit(ops, skb, dev, more);
4275 if (rc == NETDEV_TX_OK)
4276 txq_trans_update(txq);
4281 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4283 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4286 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4288 return netdev_class_create_file_ns(class_attr, NULL);
4291 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4293 netdev_class_remove_file_ns(class_attr, NULL);
4296 extern const struct kobj_ns_type_operations net_ns_type_operations;
4298 const char *netdev_drivername(const struct net_device *dev);
4300 void linkwatch_run_queue(void);
4302 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4303 netdev_features_t f2)
4305 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4306 if (f1 & NETIF_F_HW_CSUM)
4307 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4309 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4315 static inline netdev_features_t netdev_get_wanted_features(
4316 struct net_device *dev)
4318 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4320 netdev_features_t netdev_increment_features(netdev_features_t all,
4321 netdev_features_t one, netdev_features_t mask);
4323 /* Allow TSO being used on stacked device :
4324 * Performing the GSO segmentation before last device
4325 * is a performance improvement.
4327 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4328 netdev_features_t mask)
4330 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4333 int __netdev_update_features(struct net_device *dev);
4334 void netdev_update_features(struct net_device *dev);
4335 void netdev_change_features(struct net_device *dev);
4337 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4338 struct net_device *dev);
4340 netdev_features_t passthru_features_check(struct sk_buff *skb,
4341 struct net_device *dev,
4342 netdev_features_t features);
4343 netdev_features_t netif_skb_features(struct sk_buff *skb);
4345 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4347 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4349 /* check flags correspondence */
4350 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4351 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4352 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4353 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4354 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4355 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4356 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4357 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4358 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4359 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4360 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4361 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4362 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4363 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4364 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4365 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4366 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4367 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4369 return (features & feature) == feature;
4372 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4374 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4375 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4378 static inline bool netif_needs_gso(struct sk_buff *skb,
4379 netdev_features_t features)
4381 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4382 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4383 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4386 static inline void netif_set_gso_max_size(struct net_device *dev,
4389 dev->gso_max_size = size;
4392 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4393 int pulled_hlen, u16 mac_offset,
4396 skb->protocol = protocol;
4397 skb->encapsulation = 1;
4398 skb_push(skb, pulled_hlen);
4399 skb_reset_transport_header(skb);
4400 skb->mac_header = mac_offset;
4401 skb->network_header = skb->mac_header + mac_len;
4402 skb->mac_len = mac_len;
4405 static inline bool netif_is_macsec(const struct net_device *dev)
4407 return dev->priv_flags & IFF_MACSEC;
4410 static inline bool netif_is_macvlan(const struct net_device *dev)
4412 return dev->priv_flags & IFF_MACVLAN;
4415 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4417 return dev->priv_flags & IFF_MACVLAN_PORT;
4420 static inline bool netif_is_bond_master(const struct net_device *dev)
4422 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4425 static inline bool netif_is_bond_slave(const struct net_device *dev)
4427 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4430 static inline bool netif_supports_nofcs(struct net_device *dev)
4432 return dev->priv_flags & IFF_SUPP_NOFCS;
4435 static inline bool netif_is_l3_master(const struct net_device *dev)
4437 return dev->priv_flags & IFF_L3MDEV_MASTER;
4440 static inline bool netif_is_l3_slave(const struct net_device *dev)
4442 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4445 static inline bool netif_is_bridge_master(const struct net_device *dev)
4447 return dev->priv_flags & IFF_EBRIDGE;
4450 static inline bool netif_is_bridge_port(const struct net_device *dev)
4452 return dev->priv_flags & IFF_BRIDGE_PORT;
4455 static inline bool netif_is_ovs_master(const struct net_device *dev)
4457 return dev->priv_flags & IFF_OPENVSWITCH;
4460 static inline bool netif_is_ovs_port(const struct net_device *dev)
4462 return dev->priv_flags & IFF_OVS_DATAPATH;
4465 static inline bool netif_is_team_master(const struct net_device *dev)
4467 return dev->priv_flags & IFF_TEAM;
4470 static inline bool netif_is_team_port(const struct net_device *dev)
4472 return dev->priv_flags & IFF_TEAM_PORT;
4475 static inline bool netif_is_lag_master(const struct net_device *dev)
4477 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4480 static inline bool netif_is_lag_port(const struct net_device *dev)
4482 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4485 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4487 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4490 static inline bool netif_is_failover(const struct net_device *dev)
4492 return dev->priv_flags & IFF_FAILOVER;
4495 static inline bool netif_is_failover_slave(const struct net_device *dev)
4497 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4500 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4501 static inline void netif_keep_dst(struct net_device *dev)
4503 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4506 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4507 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4509 /* TODO: reserve and use an additional IFF bit, if we get more users */
4510 return dev->priv_flags & IFF_MACSEC;
4513 extern struct pernet_operations __net_initdata loopback_net_ops;
4515 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4517 /* netdev_printk helpers, similar to dev_printk */
4519 static inline const char *netdev_name(const struct net_device *dev)
4521 if (!dev->name[0] || strchr(dev->name, '%'))
4522 return "(unnamed net_device)";
4526 static inline bool netdev_unregistering(const struct net_device *dev)
4528 return dev->reg_state == NETREG_UNREGISTERING;
4531 static inline const char *netdev_reg_state(const struct net_device *dev)
4533 switch (dev->reg_state) {
4534 case NETREG_UNINITIALIZED: return " (uninitialized)";
4535 case NETREG_REGISTERED: return "";
4536 case NETREG_UNREGISTERING: return " (unregistering)";
4537 case NETREG_UNREGISTERED: return " (unregistered)";
4538 case NETREG_RELEASED: return " (released)";
4539 case NETREG_DUMMY: return " (dummy)";
4542 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4543 return " (unknown)";
4547 void netdev_printk(const char *level, const struct net_device *dev,
4548 const char *format, ...);
4550 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4552 void netdev_alert(const struct net_device *dev, const char *format, ...);
4554 void netdev_crit(const struct net_device *dev, const char *format, ...);
4556 void netdev_err(const struct net_device *dev, const char *format, ...);
4558 void netdev_warn(const struct net_device *dev, const char *format, ...);
4560 void netdev_notice(const struct net_device *dev, const char *format, ...);
4562 void netdev_info(const struct net_device *dev, const char *format, ...);
4564 #define netdev_level_once(level, dev, fmt, ...) \
4566 static bool __print_once __read_mostly; \
4568 if (!__print_once) { \
4569 __print_once = true; \
4570 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4574 #define netdev_emerg_once(dev, fmt, ...) \
4575 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4576 #define netdev_alert_once(dev, fmt, ...) \
4577 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4578 #define netdev_crit_once(dev, fmt, ...) \
4579 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4580 #define netdev_err_once(dev, fmt, ...) \
4581 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4582 #define netdev_warn_once(dev, fmt, ...) \
4583 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4584 #define netdev_notice_once(dev, fmt, ...) \
4585 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4586 #define netdev_info_once(dev, fmt, ...) \
4587 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4589 #define MODULE_ALIAS_NETDEV(device) \
4590 MODULE_ALIAS("netdev-" device)
4592 #if defined(CONFIG_DYNAMIC_DEBUG)
4593 #define netdev_dbg(__dev, format, args...) \
4595 dynamic_netdev_dbg(__dev, format, ##args); \
4597 #elif defined(DEBUG)
4598 #define netdev_dbg(__dev, format, args...) \
4599 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4601 #define netdev_dbg(__dev, format, args...) \
4604 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4608 #if defined(VERBOSE_DEBUG)
4609 #define netdev_vdbg netdev_dbg
4612 #define netdev_vdbg(dev, format, args...) \
4615 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4621 * netdev_WARN() acts like dev_printk(), but with the key difference
4622 * of using a WARN/WARN_ON to get the message out, including the
4623 * file/line information and a backtrace.
4625 #define netdev_WARN(dev, format, args...) \
4626 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4627 netdev_reg_state(dev), ##args)
4629 #define netdev_WARN_ONCE(dev, format, args...) \
4630 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4631 netdev_reg_state(dev), ##args)
4633 /* netif printk helpers, similar to netdev_printk */
4635 #define netif_printk(priv, type, level, dev, fmt, args...) \
4637 if (netif_msg_##type(priv)) \
4638 netdev_printk(level, (dev), fmt, ##args); \
4641 #define netif_level(level, priv, type, dev, fmt, args...) \
4643 if (netif_msg_##type(priv)) \
4644 netdev_##level(dev, fmt, ##args); \
4647 #define netif_emerg(priv, type, dev, fmt, args...) \
4648 netif_level(emerg, priv, type, dev, fmt, ##args)
4649 #define netif_alert(priv, type, dev, fmt, args...) \
4650 netif_level(alert, priv, type, dev, fmt, ##args)
4651 #define netif_crit(priv, type, dev, fmt, args...) \
4652 netif_level(crit, priv, type, dev, fmt, ##args)
4653 #define netif_err(priv, type, dev, fmt, args...) \
4654 netif_level(err, priv, type, dev, fmt, ##args)
4655 #define netif_warn(priv, type, dev, fmt, args...) \
4656 netif_level(warn, priv, type, dev, fmt, ##args)
4657 #define netif_notice(priv, type, dev, fmt, args...) \
4658 netif_level(notice, priv, type, dev, fmt, ##args)
4659 #define netif_info(priv, type, dev, fmt, args...) \
4660 netif_level(info, priv, type, dev, fmt, ##args)
4662 #if defined(CONFIG_DYNAMIC_DEBUG)
4663 #define netif_dbg(priv, type, netdev, format, args...) \
4665 if (netif_msg_##type(priv)) \
4666 dynamic_netdev_dbg(netdev, format, ##args); \
4668 #elif defined(DEBUG)
4669 #define netif_dbg(priv, type, dev, format, args...) \
4670 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4672 #define netif_dbg(priv, type, dev, format, args...) \
4675 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4680 /* if @cond then downgrade to debug, else print at @level */
4681 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4684 netif_dbg(priv, type, netdev, fmt, ##args); \
4686 netif_ ## level(priv, type, netdev, fmt, ##args); \
4689 #if defined(VERBOSE_DEBUG)
4690 #define netif_vdbg netif_dbg
4692 #define netif_vdbg(priv, type, dev, format, args...) \
4695 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4701 * The list of packet types we will receive (as opposed to discard)
4702 * and the routines to invoke.
4704 * Why 16. Because with 16 the only overlap we get on a hash of the
4705 * low nibble of the protocol value is RARP/SNAP/X.25.
4719 #define PTYPE_HASH_SIZE (16)
4720 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4722 #endif /* _LINUX_NETDEVICE_H */