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/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 #include <linux/hashtable.h>
62 /* 802.15.4 specific */
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
77 * Transmit return codes: transmit return codes originate from three different
80 * - qdisc return codes
81 * - driver transmit return codes
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
90 * others are propagated to higher layers.
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS 0x00
95 #define NET_XMIT_DROP 0x01 /* skb dropped */
96 #define NET_XMIT_CN 0x02 /* congestion notification */
97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK 0xf0
109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
113 typedef enum netdev_tx netdev_tx_t;
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 static inline bool dev_xmit_complete(int rc)
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 if (likely(rc < NET_XMIT_MASK))
134 * Compute the worst-case header length according to the protocols
138 #if defined(CONFIG_HYPERV_NET)
139 # define LL_MAX_HEADER 128
140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
141 # if defined(CONFIG_MAC80211_MESH)
142 # define LL_MAX_HEADER 128
144 # define LL_MAX_HEADER 96
147 # define LL_MAX_HEADER 32
150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
152 #define MAX_HEADER LL_MAX_HEADER
154 #define MAX_HEADER (LL_MAX_HEADER + 48)
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
162 struct net_device_stats {
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
172 unsigned long collisions;
173 unsigned long rx_length_errors;
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
189 #include <linux/cache.h>
190 #include <linux/skbuff.h>
193 #include <linux/static_key.h>
194 extern struct static_key rps_needed;
195 extern struct static_key rfs_needed;
202 struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
215 struct rcu_head rcu_head;
218 struct netdev_hw_addr_list {
219 struct list_head list;
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
274 bool (*validate)(const char *ll_header, unsigned int len);
277 /* These flag bits are private to the generic network queueing
278 * layer; they may not be explicitly referenced by any other
282 enum netdev_state_t {
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
292 * This structure holds boot-time configured netdevice settings. They
293 * are then used in the device probing.
295 struct netdev_boot_setup {
299 #define NETDEV_BOOT_SETUP_MAX 8
301 int __init netdev_boot_setup(char *str);
304 * Structure for NAPI scheduling similar to tasklet but with weighting
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-CPU poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
313 struct list_head poll_list;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
322 struct net_device *dev;
323 struct sk_buff *gro_list;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
337 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
341 NAPIF_STATE_SCHED = (1UL << NAPI_STATE_SCHED),
342 NAPIF_STATE_DISABLE = (1UL << NAPI_STATE_DISABLE),
343 NAPIF_STATE_NPSVC = (1UL << NAPI_STATE_NPSVC),
344 NAPIF_STATE_HASHED = (1UL << NAPI_STATE_HASHED),
345 NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL),
346 NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL),
356 typedef enum gro_result gro_result_t;
359 * enum rx_handler_result - Possible return values for rx_handlers.
360 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
362 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
363 * case skb->dev was changed by rx_handler.
364 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
365 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
367 * rx_handlers are functions called from inside __netif_receive_skb(), to do
368 * special processing of the skb, prior to delivery to protocol handlers.
370 * Currently, a net_device can only have a single rx_handler registered. Trying
371 * to register a second rx_handler will return -EBUSY.
373 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
374 * To unregister a rx_handler on a net_device, use
375 * netdev_rx_handler_unregister().
377 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
380 * If the rx_handler consumed the skb in some way, it should return
381 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
382 * the skb to be delivered in some other way.
384 * If the rx_handler changed skb->dev, to divert the skb to another
385 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
386 * new device will be called if it exists.
388 * If the rx_handler decides the skb should be ignored, it should return
389 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
390 * are registered on exact device (ptype->dev == skb->dev).
392 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
393 * delivered, it should return RX_HANDLER_PASS.
395 * A device without a registered rx_handler will behave as if rx_handler
396 * returned RX_HANDLER_PASS.
399 enum rx_handler_result {
405 typedef enum rx_handler_result rx_handler_result_t;
406 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
408 void __napi_schedule(struct napi_struct *n);
409 void __napi_schedule_irqoff(struct napi_struct *n);
411 static inline bool napi_disable_pending(struct napi_struct *n)
413 return test_bit(NAPI_STATE_DISABLE, &n->state);
417 * napi_schedule_prep - check if NAPI can be scheduled
420 * Test if NAPI routine is already running, and if not mark
421 * it as running. This is used as a condition variable to
422 * insure only one NAPI poll instance runs. We also make
423 * sure there is no pending NAPI disable.
425 static inline bool napi_schedule_prep(struct napi_struct *n)
427 return !napi_disable_pending(n) &&
428 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
432 * napi_schedule - schedule NAPI poll
435 * Schedule NAPI poll routine to be called if it is not already
438 static inline void napi_schedule(struct napi_struct *n)
440 if (napi_schedule_prep(n))
445 * napi_schedule_irqoff - schedule NAPI poll
448 * Variant of napi_schedule(), assuming hard irqs are masked.
450 static inline void napi_schedule_irqoff(struct napi_struct *n)
452 if (napi_schedule_prep(n))
453 __napi_schedule_irqoff(n);
456 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
457 static inline bool napi_reschedule(struct napi_struct *napi)
459 if (napi_schedule_prep(napi)) {
460 __napi_schedule(napi);
466 bool __napi_complete(struct napi_struct *n);
467 bool napi_complete_done(struct napi_struct *n, int work_done);
469 * napi_complete - NAPI processing complete
472 * Mark NAPI processing as complete.
473 * Consider using napi_complete_done() instead.
474 * Return false if device should avoid rearming interrupts.
476 static inline bool napi_complete(struct napi_struct *n)
478 return napi_complete_done(n, 0);
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: NAPI context
485 * Warning: caller must observe RCU grace period
486 * before freeing memory containing @napi, if
487 * this function returns true.
488 * Note: core networking stack automatically calls it
489 * from netif_napi_del().
490 * Drivers might want to call this helper to combine all
491 * the needed RCU grace periods into a single one.
493 bool napi_hash_del(struct napi_struct *napi);
496 * napi_disable - prevent NAPI from scheduling
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
502 void napi_disable(struct napi_struct *n);
505 * napi_enable - enable NAPI scheduling
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
511 static inline void napi_enable(struct napi_struct *n)
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
514 smp_mb__before_atomic();
515 clear_bit(NAPI_STATE_SCHED, &n->state);
516 clear_bit(NAPI_STATE_NPSVC, &n->state);
520 * napi_synchronize - wait until NAPI is not running
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
527 static inline void napi_synchronize(const struct napi_struct *n)
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
536 enum netdev_queue_state_t {
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
539 __QUEUE_STATE_FROZEN,
542 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
546 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
562 struct netdev_queue {
566 struct net_device *dev;
567 struct Qdisc __rcu *qdisc;
568 struct Qdisc *qdisc_sleeping;
572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
575 unsigned long tx_maxrate;
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
580 unsigned long trans_timeout;
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 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
631 struct rps_dev_flow {
634 unsigned int last_qtail;
636 #define RPS_NO_FILTER 0xffff
639 * The rps_dev_flow_table structure contains a table of flow mappings.
641 struct rps_dev_flow_table {
644 struct rps_dev_flow flows[0];
646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
647 ((_num) * sizeof(struct rps_dev_flow)))
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
654 * rps_cpu_mask is used to partition the space, depending on number of
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
657 * meaning we use 32-6=26 bits for the hash.
659 struct rps_sock_flow_table {
662 u32 ents[0] ____cacheline_aligned_in_smp;
664 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
666 #define RPS_NO_CPU 0xffff
668 extern u32 rps_cpu_mask;
669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
678 /* We only give a hint, preemption can change CPU under us */
679 val |= raw_smp_processor_id();
681 if (table->ents[index] != val)
682 table->ents[index] = val;
686 #ifdef CONFIG_RFS_ACCEL
687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
690 #endif /* CONFIG_RPS */
692 /* This structure contains an instance of an RX queue. */
693 struct netdev_rx_queue {
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
699 struct net_device *dev;
700 } ____cacheline_aligned_in_smp;
703 * RX queue sysfs structures and functions.
705 struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
720 unsigned int alloc_len;
724 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
725 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
731 struct xps_dev_maps {
733 struct xps_map __rcu *cpu_map[0];
735 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
737 #endif /* CONFIG_XPS */
739 #define TC_MAX_QUEUE 16
740 #define TC_BITMASK 15
741 /* HW offloaded queuing disciplines txq count and offset maps */
742 struct netdev_tc_txq {
747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
752 struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
760 char model_description[256];
764 #define MAX_PHYS_ITEM_ID_LEN 32
766 /* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
769 struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
771 unsigned char id_len;
774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
784 /* These structures hold the attributes of qdisc and classifiers
785 * that are being passed to the netdevice through the setup_tc op.
795 struct tc_cls_u32_offload;
797 struct tc_to_netdev {
801 struct tc_cls_u32_offload *cls_u32;
802 struct tc_cls_flower_offload *cls_flower;
803 struct tc_cls_matchall_offload *cls_mall;
804 struct tc_cls_bpf_offload *cls_bpf;
809 /* These structures hold the attributes of xdp state that are being passed
810 * to the netdevice through the xdp op.
812 enum xdp_netdev_command {
813 /* Set or clear a bpf program used in the earliest stages of packet
814 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
815 * is responsible for calling bpf_prog_put on any old progs that are
816 * stored. In case of error, the callee need not release the new prog
817 * reference, but on success it takes ownership and must bpf_prog_put
818 * when it is no longer used.
821 /* Check if a bpf program is set on the device. The callee should
822 * return true if a program is currently attached and running.
828 enum xdp_netdev_command command;
831 struct bpf_prog *prog;
838 * This structure defines the management hooks for network devices.
839 * The following hooks can be defined; unless noted otherwise, they are
840 * optional and can be filled with a null pointer.
842 * int (*ndo_init)(struct net_device *dev);
843 * This function is called once when a network device is registered.
844 * The network device can use this for any late stage initialization
845 * or semantic validation. It can fail with an error code which will
846 * be propagated back to register_netdev.
848 * void (*ndo_uninit)(struct net_device *dev);
849 * This function is called when device is unregistered or when registration
850 * fails. It is not called if init fails.
852 * int (*ndo_open)(struct net_device *dev);
853 * This function is called when a network device transitions to the up
856 * int (*ndo_stop)(struct net_device *dev);
857 * This function is called when a network device transitions to the down
860 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
861 * struct net_device *dev);
862 * Called when a packet needs to be transmitted.
863 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
864 * the queue before that can happen; it's for obsolete devices and weird
865 * corner cases, but the stack really does a non-trivial amount
866 * of useless work if you return NETDEV_TX_BUSY.
867 * Required; cannot be NULL.
869 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
870 * netdev_features_t features);
871 * Adjusts the requested feature flags according to device-specific
872 * constraints, and returns the resulting flags. Must not modify
875 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
876 * void *accel_priv, select_queue_fallback_t fallback);
877 * Called to decide which queue to use when device supports multiple
880 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
881 * This function is called to allow device receiver to make
882 * changes to configuration when multicast or promiscuous is enabled.
884 * void (*ndo_set_rx_mode)(struct net_device *dev);
885 * This function is called device changes address list filtering.
886 * If driver handles unicast address filtering, it should set
887 * IFF_UNICAST_FLT in its priv_flags.
889 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
890 * This function is called when the Media Access Control address
891 * needs to be changed. If this interface is not defined, the
892 * MAC address can not be changed.
894 * int (*ndo_validate_addr)(struct net_device *dev);
895 * Test if Media Access Control address is valid for the device.
897 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
898 * Called when a user requests an ioctl which can't be handled by
899 * the generic interface code. If not defined ioctls return
900 * not supported error code.
902 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
903 * Used to set network devices bus interface parameters. This interface
904 * is retained for legacy reasons; new devices should use the bus
905 * interface (PCI) for low level management.
907 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
908 * Called when a user wants to change the Maximum Transfer Unit
909 * of a device. If not defined, any request to change MTU will
910 * will return an error.
912 * void (*ndo_tx_timeout)(struct net_device *dev);
913 * Callback used when the transmitter has not made any progress
914 * for dev->watchdog ticks.
916 * void (*ndo_get_stats64)(struct net_device *dev,
917 * struct rtnl_link_stats64 *storage);
918 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
919 * Called when a user wants to get the network device usage
920 * statistics. Drivers must do one of the following:
921 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
922 * rtnl_link_stats64 structure passed by the caller.
923 * 2. Define @ndo_get_stats to update a net_device_stats structure
924 * (which should normally be dev->stats) and return a pointer to
925 * it. The structure may be changed asynchronously only if each
926 * field is written atomically.
927 * 3. Update dev->stats asynchronously and atomically, and define
930 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
931 * Return true if this device supports offload stats of this attr_id.
933 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
935 * Get statistics for offload operations by attr_id. Write it into the
938 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
939 * If device supports VLAN filtering this function is called when a
940 * VLAN id is registered.
942 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
943 * If device supports VLAN filtering this function is called when a
944 * VLAN id is unregistered.
946 * void (*ndo_poll_controller)(struct net_device *dev);
948 * SR-IOV management functions.
949 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
950 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
951 * u8 qos, __be16 proto);
952 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
954 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
955 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
956 * int (*ndo_get_vf_config)(struct net_device *dev,
957 * int vf, struct ifla_vf_info *ivf);
958 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
959 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
960 * struct nlattr *port[]);
962 * Enable or disable the VF ability to query its RSS Redirection Table and
963 * Hash Key. This is needed since on some devices VF share this information
964 * with PF and querying it may introduce a theoretical security risk.
965 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
966 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
967 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
968 * __be16 protocol, struct tc_to_netdev *tc);
969 * Called to setup any 'tc' scheduler, classifier or action on @dev.
970 * This is always called from the stack with the rtnl lock held and netif
971 * tx queues stopped. This allows the netdevice to perform queue
974 * Fiber Channel over Ethernet (FCoE) offload functions.
975 * int (*ndo_fcoe_enable)(struct net_device *dev);
976 * Called when the FCoE protocol stack wants to start using LLD for FCoE
977 * so the underlying device can perform whatever needed configuration or
978 * initialization to support acceleration of FCoE traffic.
980 * int (*ndo_fcoe_disable)(struct net_device *dev);
981 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
982 * so the underlying device can perform whatever needed clean-ups to
983 * stop supporting acceleration of FCoE traffic.
985 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
986 * struct scatterlist *sgl, unsigned int sgc);
987 * Called when the FCoE Initiator wants to initialize an I/O that
988 * is a possible candidate for Direct Data Placement (DDP). The LLD can
989 * perform necessary setup and returns 1 to indicate the device is set up
990 * successfully to perform DDP on this I/O, otherwise this returns 0.
992 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
993 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
994 * indicated by the FC exchange id 'xid', so the underlying device can
995 * clean up and reuse resources for later DDP requests.
997 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
998 * struct scatterlist *sgl, unsigned int sgc);
999 * Called when the FCoE Target wants to initialize an I/O that
1000 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1001 * perform necessary setup and returns 1 to indicate the device is set up
1002 * successfully to perform DDP on this I/O, otherwise this returns 0.
1004 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1005 * struct netdev_fcoe_hbainfo *hbainfo);
1006 * Called when the FCoE Protocol stack wants information on the underlying
1007 * device. This information is utilized by the FCoE protocol stack to
1008 * register attributes with Fiber Channel management service as per the
1009 * FC-GS Fabric Device Management Information(FDMI) specification.
1011 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1012 * Called when the underlying device wants to override default World Wide
1013 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1014 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1015 * protocol stack to use.
1018 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1019 * u16 rxq_index, u32 flow_id);
1020 * Set hardware filter for RFS. rxq_index is the target queue index;
1021 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1022 * Return the filter ID on success, or a negative error code.
1024 * Slave management functions (for bridge, bonding, etc).
1025 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1026 * Called to make another netdev an underling.
1028 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1029 * Called to release previously enslaved netdev.
1031 * Feature/offload setting functions.
1032 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1033 * Called to update device configuration to new features. Passed
1034 * feature set might be less than what was returned by ndo_fix_features()).
1035 * Must return >0 or -errno if it changed dev->features itself.
1037 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1038 * struct net_device *dev,
1039 * const unsigned char *addr, u16 vid, u16 flags)
1040 * Adds an FDB entry to dev for addr.
1041 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1042 * struct net_device *dev,
1043 * const unsigned char *addr, u16 vid)
1044 * Deletes the FDB entry from dev coresponding to addr.
1045 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1046 * struct net_device *dev, struct net_device *filter_dev,
1048 * Used to add FDB entries to dump requests. Implementers should add
1049 * entries to skb and update idx with the number of entries.
1051 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1053 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1054 * struct net_device *dev, u32 filter_mask,
1056 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1059 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1060 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1061 * which do not represent real hardware may define this to allow their
1062 * userspace components to manage their virtual carrier state. Devices
1063 * that determine carrier state from physical hardware properties (eg
1064 * network cables) or protocol-dependent mechanisms (eg
1065 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1067 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1068 * struct netdev_phys_item_id *ppid);
1069 * Called to get ID of physical port of this device. If driver does
1070 * not implement this, it is assumed that the hw is not able to have
1071 * multiple net devices on single physical port.
1073 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1074 * struct udp_tunnel_info *ti);
1075 * Called by UDP tunnel to notify a driver about the UDP port and socket
1076 * address family that a UDP tunnel is listnening to. It is called only
1077 * when a new port starts listening. The operation is protected by the
1080 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1081 * struct udp_tunnel_info *ti);
1082 * Called by UDP tunnel to notify the driver about a UDP port and socket
1083 * address family that the UDP tunnel is not listening to anymore. The
1084 * operation is protected by the RTNL.
1086 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1087 * struct net_device *dev)
1088 * Called by upper layer devices to accelerate switching or other
1089 * station functionality into hardware. 'pdev is the lowerdev
1090 * to use for the offload and 'dev' is the net device that will
1091 * back the offload. Returns a pointer to the private structure
1092 * the upper layer will maintain.
1093 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1094 * Called by upper layer device to delete the station created
1095 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1096 * the station and priv is the structure returned by the add
1098 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1099 * struct net_device *dev,
1101 * Callback to use for xmit over the accelerated station. This
1102 * is used in place of ndo_start_xmit on accelerated net
1104 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1105 * struct net_device *dev
1106 * netdev_features_t features);
1107 * Called by core transmit path to determine if device is capable of
1108 * performing offload operations on a given packet. This is to give
1109 * the device an opportunity to implement any restrictions that cannot
1110 * be otherwise expressed by feature flags. The check is called with
1111 * the set of features that the stack has calculated and it returns
1112 * those the driver believes to be appropriate.
1113 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1114 * int queue_index, u32 maxrate);
1115 * Called when a user wants to set a max-rate limitation of specific
1117 * int (*ndo_get_iflink)(const struct net_device *dev);
1118 * Called to get the iflink value of this device.
1119 * void (*ndo_change_proto_down)(struct net_device *dev,
1121 * This function is used to pass protocol port error state information
1122 * to the switch driver. The switch driver can react to the proto_down
1123 * by doing a phys down on the associated switch port.
1124 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1125 * This function is used to get egress tunnel information for given skb.
1126 * This is useful for retrieving outer tunnel header parameters while
1128 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1129 * This function is used to specify the headroom that the skb must
1130 * consider when allocation skb during packet reception. Setting
1131 * appropriate rx headroom value allows avoiding skb head copy on
1132 * forward. Setting a negative value resets the rx headroom to the
1134 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1135 * This function is used to set or query state related to XDP on the
1136 * netdevice. See definition of enum xdp_netdev_command for details.
1139 struct net_device_ops {
1140 int (*ndo_init)(struct net_device *dev);
1141 void (*ndo_uninit)(struct net_device *dev);
1142 int (*ndo_open)(struct net_device *dev);
1143 int (*ndo_stop)(struct net_device *dev);
1144 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1145 struct net_device *dev);
1146 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1147 struct net_device *dev,
1148 netdev_features_t features);
1149 u16 (*ndo_select_queue)(struct net_device *dev,
1150 struct sk_buff *skb,
1152 select_queue_fallback_t fallback);
1153 void (*ndo_change_rx_flags)(struct net_device *dev,
1155 void (*ndo_set_rx_mode)(struct net_device *dev);
1156 int (*ndo_set_mac_address)(struct net_device *dev,
1158 int (*ndo_validate_addr)(struct net_device *dev);
1159 int (*ndo_do_ioctl)(struct net_device *dev,
1160 struct ifreq *ifr, int cmd);
1161 int (*ndo_set_config)(struct net_device *dev,
1163 int (*ndo_change_mtu)(struct net_device *dev,
1165 int (*ndo_neigh_setup)(struct net_device *dev,
1166 struct neigh_parms *);
1167 void (*ndo_tx_timeout) (struct net_device *dev);
1169 void (*ndo_get_stats64)(struct net_device *dev,
1170 struct rtnl_link_stats64 *storage);
1171 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1172 int (*ndo_get_offload_stats)(int attr_id,
1173 const struct net_device *dev,
1175 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1177 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1178 __be16 proto, u16 vid);
1179 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1180 __be16 proto, u16 vid);
1181 #ifdef CONFIG_NET_POLL_CONTROLLER
1182 void (*ndo_poll_controller)(struct net_device *dev);
1183 int (*ndo_netpoll_setup)(struct net_device *dev,
1184 struct netpoll_info *info);
1185 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1187 #ifdef CONFIG_NET_RX_BUSY_POLL
1188 int (*ndo_busy_poll)(struct napi_struct *dev);
1190 int (*ndo_set_vf_mac)(struct net_device *dev,
1191 int queue, u8 *mac);
1192 int (*ndo_set_vf_vlan)(struct net_device *dev,
1193 int queue, u16 vlan,
1194 u8 qos, __be16 proto);
1195 int (*ndo_set_vf_rate)(struct net_device *dev,
1196 int vf, int min_tx_rate,
1198 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1199 int vf, bool setting);
1200 int (*ndo_set_vf_trust)(struct net_device *dev,
1201 int vf, bool setting);
1202 int (*ndo_get_vf_config)(struct net_device *dev,
1204 struct ifla_vf_info *ivf);
1205 int (*ndo_set_vf_link_state)(struct net_device *dev,
1206 int vf, int link_state);
1207 int (*ndo_get_vf_stats)(struct net_device *dev,
1209 struct ifla_vf_stats
1211 int (*ndo_set_vf_port)(struct net_device *dev,
1213 struct nlattr *port[]);
1214 int (*ndo_get_vf_port)(struct net_device *dev,
1215 int vf, struct sk_buff *skb);
1216 int (*ndo_set_vf_guid)(struct net_device *dev,
1219 int (*ndo_set_vf_rss_query_en)(
1220 struct net_device *dev,
1221 int vf, bool setting);
1222 int (*ndo_setup_tc)(struct net_device *dev,
1225 struct tc_to_netdev *tc);
1226 #if IS_ENABLED(CONFIG_FCOE)
1227 int (*ndo_fcoe_enable)(struct net_device *dev);
1228 int (*ndo_fcoe_disable)(struct net_device *dev);
1229 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1231 struct scatterlist *sgl,
1233 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1235 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1237 struct scatterlist *sgl,
1239 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1240 struct netdev_fcoe_hbainfo *hbainfo);
1243 #if IS_ENABLED(CONFIG_LIBFCOE)
1244 #define NETDEV_FCOE_WWNN 0
1245 #define NETDEV_FCOE_WWPN 1
1246 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1247 u64 *wwn, int type);
1250 #ifdef CONFIG_RFS_ACCEL
1251 int (*ndo_rx_flow_steer)(struct net_device *dev,
1252 const struct sk_buff *skb,
1256 int (*ndo_add_slave)(struct net_device *dev,
1257 struct net_device *slave_dev);
1258 int (*ndo_del_slave)(struct net_device *dev,
1259 struct net_device *slave_dev);
1260 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1261 netdev_features_t features);
1262 int (*ndo_set_features)(struct net_device *dev,
1263 netdev_features_t features);
1264 int (*ndo_neigh_construct)(struct net_device *dev,
1265 struct neighbour *n);
1266 void (*ndo_neigh_destroy)(struct net_device *dev,
1267 struct neighbour *n);
1269 int (*ndo_fdb_add)(struct ndmsg *ndm,
1270 struct nlattr *tb[],
1271 struct net_device *dev,
1272 const unsigned char *addr,
1275 int (*ndo_fdb_del)(struct ndmsg *ndm,
1276 struct nlattr *tb[],
1277 struct net_device *dev,
1278 const unsigned char *addr,
1280 int (*ndo_fdb_dump)(struct sk_buff *skb,
1281 struct netlink_callback *cb,
1282 struct net_device *dev,
1283 struct net_device *filter_dev,
1286 int (*ndo_bridge_setlink)(struct net_device *dev,
1287 struct nlmsghdr *nlh,
1289 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1291 struct net_device *dev,
1294 int (*ndo_bridge_dellink)(struct net_device *dev,
1295 struct nlmsghdr *nlh,
1297 int (*ndo_change_carrier)(struct net_device *dev,
1299 int (*ndo_get_phys_port_id)(struct net_device *dev,
1300 struct netdev_phys_item_id *ppid);
1301 int (*ndo_get_phys_port_name)(struct net_device *dev,
1302 char *name, size_t len);
1303 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1304 struct udp_tunnel_info *ti);
1305 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1306 struct udp_tunnel_info *ti);
1307 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1308 struct net_device *dev);
1309 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1312 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1313 struct net_device *dev,
1315 int (*ndo_get_lock_subclass)(struct net_device *dev);
1316 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1319 int (*ndo_get_iflink)(const struct net_device *dev);
1320 int (*ndo_change_proto_down)(struct net_device *dev,
1322 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1323 struct sk_buff *skb);
1324 void (*ndo_set_rx_headroom)(struct net_device *dev,
1325 int needed_headroom);
1326 int (*ndo_xdp)(struct net_device *dev,
1327 struct netdev_xdp *xdp);
1331 * enum net_device_priv_flags - &struct net_device priv_flags
1333 * These are the &struct net_device, they are only set internally
1334 * by drivers and used in the kernel. These flags are invisible to
1335 * userspace; this means that the order of these flags can change
1336 * during any kernel release.
1338 * You should have a pretty good reason to be extending these flags.
1340 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1341 * @IFF_EBRIDGE: Ethernet bridging device
1342 * @IFF_BONDING: bonding master or slave
1343 * @IFF_ISATAP: ISATAP interface (RFC4214)
1344 * @IFF_WAN_HDLC: WAN HDLC device
1345 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1347 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1348 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1349 * @IFF_MACVLAN_PORT: device used as macvlan port
1350 * @IFF_BRIDGE_PORT: device used as bridge port
1351 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1352 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1353 * @IFF_UNICAST_FLT: Supports unicast filtering
1354 * @IFF_TEAM_PORT: device used as team port
1355 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1356 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1357 * change when it's running
1358 * @IFF_MACVLAN: Macvlan device
1359 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1360 * underlying stacked devices
1361 * @IFF_IPVLAN_MASTER: IPvlan master device
1362 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1363 * @IFF_L3MDEV_MASTER: device is an L3 master device
1364 * @IFF_NO_QUEUE: device can run without qdisc attached
1365 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1366 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1367 * @IFF_TEAM: device is a team device
1368 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1369 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1370 * entity (i.e. the master device for bridged veth)
1371 * @IFF_MACSEC: device is a MACsec device
1373 enum netdev_priv_flags {
1374 IFF_802_1Q_VLAN = 1<<0,
1378 IFF_WAN_HDLC = 1<<4,
1379 IFF_XMIT_DST_RELEASE = 1<<5,
1380 IFF_DONT_BRIDGE = 1<<6,
1381 IFF_DISABLE_NETPOLL = 1<<7,
1382 IFF_MACVLAN_PORT = 1<<8,
1383 IFF_BRIDGE_PORT = 1<<9,
1384 IFF_OVS_DATAPATH = 1<<10,
1385 IFF_TX_SKB_SHARING = 1<<11,
1386 IFF_UNICAST_FLT = 1<<12,
1387 IFF_TEAM_PORT = 1<<13,
1388 IFF_SUPP_NOFCS = 1<<14,
1389 IFF_LIVE_ADDR_CHANGE = 1<<15,
1390 IFF_MACVLAN = 1<<16,
1391 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1392 IFF_IPVLAN_MASTER = 1<<18,
1393 IFF_IPVLAN_SLAVE = 1<<19,
1394 IFF_L3MDEV_MASTER = 1<<20,
1395 IFF_NO_QUEUE = 1<<21,
1396 IFF_OPENVSWITCH = 1<<22,
1397 IFF_L3MDEV_SLAVE = 1<<23,
1399 IFF_RXFH_CONFIGURED = 1<<25,
1400 IFF_PHONY_HEADROOM = 1<<26,
1404 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1405 #define IFF_EBRIDGE IFF_EBRIDGE
1406 #define IFF_BONDING IFF_BONDING
1407 #define IFF_ISATAP IFF_ISATAP
1408 #define IFF_WAN_HDLC IFF_WAN_HDLC
1409 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1410 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1411 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1412 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1413 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1414 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1415 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1416 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1417 #define IFF_TEAM_PORT IFF_TEAM_PORT
1418 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1419 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1420 #define IFF_MACVLAN IFF_MACVLAN
1421 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1422 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1423 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1424 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1425 #define IFF_NO_QUEUE IFF_NO_QUEUE
1426 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1427 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1428 #define IFF_TEAM IFF_TEAM
1429 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1430 #define IFF_MACSEC IFF_MACSEC
1433 * struct net_device - The DEVICE structure.
1434 * Actually, this whole structure is a big mistake. It mixes I/O
1435 * data with strictly "high-level" data, and it has to know about
1436 * almost every data structure used in the INET module.
1438 * @name: This is the first field of the "visible" part of this structure
1439 * (i.e. as seen by users in the "Space.c" file). It is the name
1442 * @name_hlist: Device name hash chain, please keep it close to name[]
1443 * @ifalias: SNMP alias
1444 * @mem_end: Shared memory end
1445 * @mem_start: Shared memory start
1446 * @base_addr: Device I/O address
1447 * @irq: Device IRQ number
1449 * @carrier_changes: Stats to monitor carrier on<->off transitions
1451 * @state: Generic network queuing layer state, see netdev_state_t
1452 * @dev_list: The global list of network devices
1453 * @napi_list: List entry used for polling NAPI devices
1454 * @unreg_list: List entry when we are unregistering the
1455 * device; see the function unregister_netdev
1456 * @close_list: List entry used when we are closing the device
1457 * @ptype_all: Device-specific packet handlers for all protocols
1458 * @ptype_specific: Device-specific, protocol-specific packet handlers
1460 * @adj_list: Directly linked devices, like slaves for bonding
1461 * @features: Currently active device features
1462 * @hw_features: User-changeable features
1464 * @wanted_features: User-requested features
1465 * @vlan_features: Mask of features inheritable by VLAN devices
1467 * @hw_enc_features: Mask of features inherited by encapsulating devices
1468 * This field indicates what encapsulation
1469 * offloads the hardware is capable of doing,
1470 * and drivers will need to set them appropriately.
1472 * @mpls_features: Mask of features inheritable by MPLS
1474 * @ifindex: interface index
1475 * @group: The group the device belongs to
1477 * @stats: Statistics struct, which was left as a legacy, use
1478 * rtnl_link_stats64 instead
1480 * @rx_dropped: Dropped packets by core network,
1481 * do not use this in drivers
1482 * @tx_dropped: Dropped packets by core network,
1483 * do not use this in drivers
1484 * @rx_nohandler: nohandler dropped packets by core network on
1485 * inactive devices, do not use this in drivers
1487 * @wireless_handlers: List of functions to handle Wireless Extensions,
1489 * see <net/iw_handler.h> for details.
1490 * @wireless_data: Instance data managed by the core of wireless extensions
1492 * @netdev_ops: Includes several pointers to callbacks,
1493 * if one wants to override the ndo_*() functions
1494 * @ethtool_ops: Management operations
1495 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1496 * discovery handling. Necessary for e.g. 6LoWPAN.
1497 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1498 * of Layer 2 headers.
1500 * @flags: Interface flags (a la BSD)
1501 * @priv_flags: Like 'flags' but invisible to userspace,
1502 * see if.h for the definitions
1503 * @gflags: Global flags ( kept as legacy )
1504 * @padded: How much padding added by alloc_netdev()
1505 * @operstate: RFC2863 operstate
1506 * @link_mode: Mapping policy to operstate
1507 * @if_port: Selectable AUI, TP, ...
1509 * @mtu: Interface MTU value
1510 * @min_mtu: Interface Minimum MTU value
1511 * @max_mtu: Interface Maximum MTU value
1512 * @type: Interface hardware type
1513 * @hard_header_len: Maximum hardware header length.
1515 * @needed_headroom: Extra headroom the hardware may need, but not in all
1516 * cases can this be guaranteed
1517 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1518 * cases can this be guaranteed. Some cases also use
1519 * LL_MAX_HEADER instead to allocate the skb
1521 * interface address info:
1523 * @perm_addr: Permanent hw address
1524 * @addr_assign_type: Hw address assignment type
1525 * @addr_len: Hardware address length
1526 * @neigh_priv_len: Used in neigh_alloc()
1527 * @dev_id: Used to differentiate devices that share
1528 * the same link layer address
1529 * @dev_port: Used to differentiate devices that share
1531 * @addr_list_lock: XXX: need comments on this one
1532 * @uc_promisc: Counter that indicates promiscuous mode
1533 * has been enabled due to the need to listen to
1534 * additional unicast addresses in a device that
1535 * does not implement ndo_set_rx_mode()
1536 * @uc: unicast mac addresses
1537 * @mc: multicast mac addresses
1538 * @dev_addrs: list of device hw addresses
1539 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1540 * @promiscuity: Number of times the NIC is told to work in
1541 * promiscuous mode; if it becomes 0 the NIC will
1542 * exit promiscuous mode
1543 * @allmulti: Counter, enables or disables allmulticast mode
1545 * @vlan_info: VLAN info
1546 * @dsa_ptr: dsa specific data
1547 * @tipc_ptr: TIPC specific data
1548 * @atalk_ptr: AppleTalk link
1549 * @ip_ptr: IPv4 specific data
1550 * @dn_ptr: DECnet specific data
1551 * @ip6_ptr: IPv6 specific data
1552 * @ax25_ptr: AX.25 specific data
1553 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1555 * @dev_addr: Hw address (before bcast,
1556 * because most packets are unicast)
1558 * @_rx: Array of RX queues
1559 * @num_rx_queues: Number of RX queues
1560 * allocated at register_netdev() time
1561 * @real_num_rx_queues: Number of RX queues currently active in device
1563 * @rx_handler: handler for received packets
1564 * @rx_handler_data: XXX: need comments on this one
1565 * @ingress_queue: XXX: need comments on this one
1566 * @broadcast: hw bcast address
1568 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1569 * indexed by RX queue number. Assigned by driver.
1570 * This must only be set if the ndo_rx_flow_steer
1571 * operation is defined
1572 * @index_hlist: Device index hash chain
1574 * @_tx: Array of TX queues
1575 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1576 * @real_num_tx_queues: Number of TX queues currently active in device
1577 * @qdisc: Root qdisc from userspace point of view
1578 * @tx_queue_len: Max frames per queue allowed
1579 * @tx_global_lock: XXX: need comments on this one
1581 * @xps_maps: XXX: need comments on this one
1583 * @watchdog_timeo: Represents the timeout that is used by
1584 * the watchdog (see dev_watchdog())
1585 * @watchdog_timer: List of timers
1587 * @pcpu_refcnt: Number of references to this device
1588 * @todo_list: Delayed register/unregister
1589 * @link_watch_list: XXX: need comments on this one
1591 * @reg_state: Register/unregister state machine
1592 * @dismantle: Device is going to be freed
1593 * @rtnl_link_state: This enum represents the phases of creating
1596 * @destructor: Called from unregister,
1597 * can be used to call free_netdev
1598 * @npinfo: XXX: need comments on this one
1599 * @nd_net: Network namespace this network device is inside
1601 * @ml_priv: Mid-layer private
1602 * @lstats: Loopback statistics
1603 * @tstats: Tunnel statistics
1604 * @dstats: Dummy statistics
1605 * @vstats: Virtual ethernet statistics
1610 * @dev: Class/net/name entry
1611 * @sysfs_groups: Space for optional device, statistics and wireless
1614 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1615 * @rtnl_link_ops: Rtnl_link_ops
1617 * @gso_max_size: Maximum size of generic segmentation offload
1618 * @gso_max_segs: Maximum number of segments that can be passed to the
1621 * @dcbnl_ops: Data Center Bridging netlink ops
1622 * @num_tc: Number of traffic classes in the net device
1623 * @tc_to_txq: XXX: need comments on this one
1624 * @prio_tc_map: XXX: need comments on this one
1626 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1628 * @priomap: XXX: need comments on this one
1629 * @phydev: Physical device may attach itself
1630 * for hardware timestamping
1632 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1633 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1635 * @proto_down: protocol port state information can be sent to the
1636 * switch driver and used to set the phys state of the
1639 * FIXME: cleanup struct net_device such that network protocol info
1644 char name[IFNAMSIZ];
1645 struct hlist_node name_hlist;
1648 * I/O specific fields
1649 * FIXME: Merge these and struct ifmap into one
1651 unsigned long mem_end;
1652 unsigned long mem_start;
1653 unsigned long base_addr;
1656 atomic_t carrier_changes;
1659 * Some hardware also needs these fields (state,dev_list,
1660 * napi_list,unreg_list,close_list) but they are not
1661 * part of the usual set specified in Space.c.
1664 unsigned long state;
1666 struct list_head dev_list;
1667 struct list_head napi_list;
1668 struct list_head unreg_list;
1669 struct list_head close_list;
1670 struct list_head ptype_all;
1671 struct list_head ptype_specific;
1674 struct list_head upper;
1675 struct list_head lower;
1678 netdev_features_t features;
1679 netdev_features_t hw_features;
1680 netdev_features_t wanted_features;
1681 netdev_features_t vlan_features;
1682 netdev_features_t hw_enc_features;
1683 netdev_features_t mpls_features;
1684 netdev_features_t gso_partial_features;
1689 struct net_device_stats stats;
1691 atomic_long_t rx_dropped;
1692 atomic_long_t tx_dropped;
1693 atomic_long_t rx_nohandler;
1695 #ifdef CONFIG_WIRELESS_EXT
1696 const struct iw_handler_def *wireless_handlers;
1697 struct iw_public_data *wireless_data;
1699 const struct net_device_ops *netdev_ops;
1700 const struct ethtool_ops *ethtool_ops;
1701 #ifdef CONFIG_NET_SWITCHDEV
1702 const struct switchdev_ops *switchdev_ops;
1704 #ifdef CONFIG_NET_L3_MASTER_DEV
1705 const struct l3mdev_ops *l3mdev_ops;
1707 #if IS_ENABLED(CONFIG_IPV6)
1708 const struct ndisc_ops *ndisc_ops;
1711 const struct header_ops *header_ops;
1714 unsigned int priv_flags;
1716 unsigned short gflags;
1717 unsigned short padded;
1719 unsigned char operstate;
1720 unsigned char link_mode;
1722 unsigned char if_port;
1726 unsigned int min_mtu;
1727 unsigned int max_mtu;
1728 unsigned short type;
1729 unsigned short hard_header_len;
1731 unsigned short needed_headroom;
1732 unsigned short needed_tailroom;
1734 /* Interface address info. */
1735 unsigned char perm_addr[MAX_ADDR_LEN];
1736 unsigned char addr_assign_type;
1737 unsigned char addr_len;
1738 unsigned short neigh_priv_len;
1739 unsigned short dev_id;
1740 unsigned short dev_port;
1741 spinlock_t addr_list_lock;
1742 unsigned char name_assign_type;
1744 struct netdev_hw_addr_list uc;
1745 struct netdev_hw_addr_list mc;
1746 struct netdev_hw_addr_list dev_addrs;
1749 struct kset *queues_kset;
1751 unsigned int promiscuity;
1752 unsigned int allmulti;
1755 /* Protocol-specific pointers */
1757 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1758 struct vlan_info __rcu *vlan_info;
1760 #if IS_ENABLED(CONFIG_NET_DSA)
1761 struct dsa_switch_tree *dsa_ptr;
1763 #if IS_ENABLED(CONFIG_TIPC)
1764 struct tipc_bearer __rcu *tipc_ptr;
1767 struct in_device __rcu *ip_ptr;
1768 struct dn_dev __rcu *dn_ptr;
1769 struct inet6_dev __rcu *ip6_ptr;
1771 struct wireless_dev *ieee80211_ptr;
1772 struct wpan_dev *ieee802154_ptr;
1773 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1774 struct mpls_dev __rcu *mpls_ptr;
1778 * Cache lines mostly used on receive path (including eth_type_trans())
1780 /* Interface address info used in eth_type_trans() */
1781 unsigned char *dev_addr;
1784 struct netdev_rx_queue *_rx;
1786 unsigned int num_rx_queues;
1787 unsigned int real_num_rx_queues;
1790 unsigned long gro_flush_timeout;
1791 rx_handler_func_t __rcu *rx_handler;
1792 void __rcu *rx_handler_data;
1794 #ifdef CONFIG_NET_CLS_ACT
1795 struct tcf_proto __rcu *ingress_cl_list;
1797 struct netdev_queue __rcu *ingress_queue;
1798 #ifdef CONFIG_NETFILTER_INGRESS
1799 struct nf_hook_entry __rcu *nf_hooks_ingress;
1802 unsigned char broadcast[MAX_ADDR_LEN];
1803 #ifdef CONFIG_RFS_ACCEL
1804 struct cpu_rmap *rx_cpu_rmap;
1806 struct hlist_node index_hlist;
1809 * Cache lines mostly used on transmit path
1811 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1812 unsigned int num_tx_queues;
1813 unsigned int real_num_tx_queues;
1814 struct Qdisc *qdisc;
1815 #ifdef CONFIG_NET_SCHED
1816 DECLARE_HASHTABLE (qdisc_hash, 4);
1818 unsigned long tx_queue_len;
1819 spinlock_t tx_global_lock;
1823 struct xps_dev_maps __rcu *xps_maps;
1825 #ifdef CONFIG_NET_CLS_ACT
1826 struct tcf_proto __rcu *egress_cl_list;
1829 /* These may be needed for future network-power-down code. */
1830 struct timer_list watchdog_timer;
1832 int __percpu *pcpu_refcnt;
1833 struct list_head todo_list;
1835 struct list_head link_watch_list;
1837 enum { NETREG_UNINITIALIZED=0,
1838 NETREG_REGISTERED, /* completed register_netdevice */
1839 NETREG_UNREGISTERING, /* called unregister_netdevice */
1840 NETREG_UNREGISTERED, /* completed unregister todo */
1841 NETREG_RELEASED, /* called free_netdev */
1842 NETREG_DUMMY, /* dummy device for NAPI poll */
1848 RTNL_LINK_INITIALIZED,
1849 RTNL_LINK_INITIALIZING,
1850 } rtnl_link_state:16;
1852 void (*destructor)(struct net_device *dev);
1854 #ifdef CONFIG_NETPOLL
1855 struct netpoll_info __rcu *npinfo;
1858 possible_net_t nd_net;
1860 /* mid-layer private */
1863 struct pcpu_lstats __percpu *lstats;
1864 struct pcpu_sw_netstats __percpu *tstats;
1865 struct pcpu_dstats __percpu *dstats;
1866 struct pcpu_vstats __percpu *vstats;
1869 struct garp_port __rcu *garp_port;
1870 struct mrp_port __rcu *mrp_port;
1873 const struct attribute_group *sysfs_groups[4];
1874 const struct attribute_group *sysfs_rx_queue_group;
1876 const struct rtnl_link_ops *rtnl_link_ops;
1878 /* for setting kernel sock attribute on TCP connection setup */
1879 #define GSO_MAX_SIZE 65536
1880 unsigned int gso_max_size;
1881 #define GSO_MAX_SEGS 65535
1885 const struct dcbnl_rtnl_ops *dcbnl_ops;
1888 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1889 u8 prio_tc_map[TC_BITMASK + 1];
1891 #if IS_ENABLED(CONFIG_FCOE)
1892 unsigned int fcoe_ddp_xid;
1894 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1895 struct netprio_map __rcu *priomap;
1897 struct phy_device *phydev;
1898 struct lock_class_key *qdisc_tx_busylock;
1899 struct lock_class_key *qdisc_running_key;
1902 #define to_net_dev(d) container_of(d, struct net_device, dev)
1904 #define NETDEV_ALIGN 32
1907 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1909 return dev->prio_tc_map[prio & TC_BITMASK];
1913 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1915 if (tc >= dev->num_tc)
1918 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1922 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1923 void netdev_reset_tc(struct net_device *dev);
1924 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1925 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1928 int netdev_get_num_tc(struct net_device *dev)
1934 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1937 return &dev->_tx[index];
1940 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1941 const struct sk_buff *skb)
1943 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1946 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1947 void (*f)(struct net_device *,
1948 struct netdev_queue *,
1954 for (i = 0; i < dev->num_tx_queues; i++)
1955 f(dev, &dev->_tx[i], arg);
1958 #define netdev_lockdep_set_classes(dev) \
1960 static struct lock_class_key qdisc_tx_busylock_key; \
1961 static struct lock_class_key qdisc_running_key; \
1962 static struct lock_class_key qdisc_xmit_lock_key; \
1963 static struct lock_class_key dev_addr_list_lock_key; \
1966 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1967 (dev)->qdisc_running_key = &qdisc_running_key; \
1968 lockdep_set_class(&(dev)->addr_list_lock, \
1969 &dev_addr_list_lock_key); \
1970 for (i = 0; i < (dev)->num_tx_queues; i++) \
1971 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1972 &qdisc_xmit_lock_key); \
1975 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1976 struct sk_buff *skb,
1979 /* returns the headroom that the master device needs to take in account
1980 * when forwarding to this dev
1982 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1984 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1987 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1989 if (dev->netdev_ops->ndo_set_rx_headroom)
1990 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1993 /* set the device rx headroom to the dev's default */
1994 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1996 netdev_set_rx_headroom(dev, -1);
2000 * Net namespace inlines
2003 struct net *dev_net(const struct net_device *dev)
2005 return read_pnet(&dev->nd_net);
2009 void dev_net_set(struct net_device *dev, struct net *net)
2011 write_pnet(&dev->nd_net, net);
2014 static inline bool netdev_uses_dsa(struct net_device *dev)
2016 #if IS_ENABLED(CONFIG_NET_DSA)
2017 if (dev->dsa_ptr != NULL)
2018 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2024 * netdev_priv - access network device private data
2025 * @dev: network device
2027 * Get network device private data
2029 static inline void *netdev_priv(const struct net_device *dev)
2031 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2034 /* Set the sysfs physical device reference for the network logical device
2035 * if set prior to registration will cause a symlink during initialization.
2037 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2039 /* Set the sysfs device type for the network logical device to allow
2040 * fine-grained identification of different network device types. For
2041 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2043 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2045 /* Default NAPI poll() weight
2046 * Device drivers are strongly advised to not use bigger value
2048 #define NAPI_POLL_WEIGHT 64
2051 * netif_napi_add - initialize a NAPI context
2052 * @dev: network device
2053 * @napi: NAPI context
2054 * @poll: polling function
2055 * @weight: default weight
2057 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2058 * *any* of the other NAPI-related functions.
2060 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2061 int (*poll)(struct napi_struct *, int), int weight);
2064 * netif_tx_napi_add - initialize a NAPI context
2065 * @dev: network device
2066 * @napi: NAPI context
2067 * @poll: polling function
2068 * @weight: default weight
2070 * This variant of netif_napi_add() should be used from drivers using NAPI
2071 * to exclusively poll a TX queue.
2072 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2074 static inline void netif_tx_napi_add(struct net_device *dev,
2075 struct napi_struct *napi,
2076 int (*poll)(struct napi_struct *, int),
2079 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2080 netif_napi_add(dev, napi, poll, weight);
2084 * netif_napi_del - remove a NAPI context
2085 * @napi: NAPI context
2087 * netif_napi_del() removes a NAPI context from the network device NAPI list
2089 void netif_napi_del(struct napi_struct *napi);
2091 struct napi_gro_cb {
2092 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2095 /* Length of frag0. */
2096 unsigned int frag0_len;
2098 /* This indicates where we are processing relative to skb->data. */
2101 /* This is non-zero if the packet cannot be merged with the new skb. */
2104 /* Save the IP ID here and check when we get to the transport layer */
2107 /* Number of segments aggregated. */
2110 /* Start offset for remote checksum offload */
2111 u16 gro_remcsum_start;
2113 /* jiffies when first packet was created/queued */
2116 /* Used in ipv6_gro_receive() and foo-over-udp */
2119 /* This is non-zero if the packet may be of the same flow. */
2122 /* Used in tunnel GRO receive */
2125 /* GRO checksum is valid */
2128 /* Number of checksums via CHECKSUM_UNNECESSARY */
2133 #define NAPI_GRO_FREE 1
2134 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2136 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2139 /* Used in GRE, set in fou/gue_gro_receive */
2142 /* Used to determine if flush_id can be ignored */
2145 /* Number of gro_receive callbacks this packet already went through */
2146 u8 recursion_counter:4;
2150 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2153 /* used in skb_gro_receive() slow path */
2154 struct sk_buff *last;
2157 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2159 #define GRO_RECURSION_LIMIT 15
2160 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2162 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2165 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2166 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2167 struct sk_buff **head,
2168 struct sk_buff *skb)
2170 if (unlikely(gro_recursion_inc_test(skb))) {
2171 NAPI_GRO_CB(skb)->flush |= 1;
2175 return cb(head, skb);
2178 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2180 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2182 struct sk_buff **head,
2183 struct sk_buff *skb)
2185 if (unlikely(gro_recursion_inc_test(skb))) {
2186 NAPI_GRO_CB(skb)->flush |= 1;
2190 return cb(sk, head, skb);
2193 struct packet_type {
2194 __be16 type; /* This is really htons(ether_type). */
2195 struct net_device *dev; /* NULL is wildcarded here */
2196 int (*func) (struct sk_buff *,
2197 struct net_device *,
2198 struct packet_type *,
2199 struct net_device *);
2200 bool (*id_match)(struct packet_type *ptype,
2202 void *af_packet_priv;
2203 struct list_head list;
2206 struct offload_callbacks {
2207 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2208 netdev_features_t features);
2209 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2210 struct sk_buff *skb);
2211 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2214 struct packet_offload {
2215 __be16 type; /* This is really htons(ether_type). */
2217 struct offload_callbacks callbacks;
2218 struct list_head list;
2221 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2222 struct pcpu_sw_netstats {
2227 struct u64_stats_sync syncp;
2230 #define __netdev_alloc_pcpu_stats(type, gfp) \
2232 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2235 for_each_possible_cpu(__cpu) { \
2236 typeof(type) *stat; \
2237 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2238 u64_stats_init(&stat->syncp); \
2244 #define netdev_alloc_pcpu_stats(type) \
2245 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2247 enum netdev_lag_tx_type {
2248 NETDEV_LAG_TX_TYPE_UNKNOWN,
2249 NETDEV_LAG_TX_TYPE_RANDOM,
2250 NETDEV_LAG_TX_TYPE_BROADCAST,
2251 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2252 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2253 NETDEV_LAG_TX_TYPE_HASH,
2256 struct netdev_lag_upper_info {
2257 enum netdev_lag_tx_type tx_type;
2260 struct netdev_lag_lower_state_info {
2265 #include <linux/notifier.h>
2267 /* netdevice notifier chain. Please remember to update the rtnetlink
2268 * notification exclusion list in rtnetlink_event() when adding new
2271 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2272 #define NETDEV_DOWN 0x0002
2273 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2274 detected a hardware crash and restarted
2275 - we can use this eg to kick tcp sessions
2277 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2278 #define NETDEV_REGISTER 0x0005
2279 #define NETDEV_UNREGISTER 0x0006
2280 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2281 #define NETDEV_CHANGEADDR 0x0008
2282 #define NETDEV_GOING_DOWN 0x0009
2283 #define NETDEV_CHANGENAME 0x000A
2284 #define NETDEV_FEAT_CHANGE 0x000B
2285 #define NETDEV_BONDING_FAILOVER 0x000C
2286 #define NETDEV_PRE_UP 0x000D
2287 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2288 #define NETDEV_POST_TYPE_CHANGE 0x000F
2289 #define NETDEV_POST_INIT 0x0010
2290 #define NETDEV_UNREGISTER_FINAL 0x0011
2291 #define NETDEV_RELEASE 0x0012
2292 #define NETDEV_NOTIFY_PEERS 0x0013
2293 #define NETDEV_JOIN 0x0014
2294 #define NETDEV_CHANGEUPPER 0x0015
2295 #define NETDEV_RESEND_IGMP 0x0016
2296 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2297 #define NETDEV_CHANGEINFODATA 0x0018
2298 #define NETDEV_BONDING_INFO 0x0019
2299 #define NETDEV_PRECHANGEUPPER 0x001A
2300 #define NETDEV_CHANGELOWERSTATE 0x001B
2301 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2302 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2304 int register_netdevice_notifier(struct notifier_block *nb);
2305 int unregister_netdevice_notifier(struct notifier_block *nb);
2307 struct netdev_notifier_info {
2308 struct net_device *dev;
2311 struct netdev_notifier_change_info {
2312 struct netdev_notifier_info info; /* must be first */
2313 unsigned int flags_changed;
2316 struct netdev_notifier_changeupper_info {
2317 struct netdev_notifier_info info; /* must be first */
2318 struct net_device *upper_dev; /* new upper dev */
2319 bool master; /* is upper dev master */
2320 bool linking; /* is the notification for link or unlink */
2321 void *upper_info; /* upper dev info */
2324 struct netdev_notifier_changelowerstate_info {
2325 struct netdev_notifier_info info; /* must be first */
2326 void *lower_state_info; /* is lower dev state */
2329 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2330 struct net_device *dev)
2335 static inline struct net_device *
2336 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2341 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2344 extern rwlock_t dev_base_lock; /* Device list lock */
2346 #define for_each_netdev(net, d) \
2347 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2348 #define for_each_netdev_reverse(net, d) \
2349 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2350 #define for_each_netdev_rcu(net, d) \
2351 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2352 #define for_each_netdev_safe(net, d, n) \
2353 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2354 #define for_each_netdev_continue(net, d) \
2355 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2356 #define for_each_netdev_continue_rcu(net, d) \
2357 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2358 #define for_each_netdev_in_bond_rcu(bond, slave) \
2359 for_each_netdev_rcu(&init_net, slave) \
2360 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2361 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2363 static inline struct net_device *next_net_device(struct net_device *dev)
2365 struct list_head *lh;
2369 lh = dev->dev_list.next;
2370 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2373 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2375 struct list_head *lh;
2379 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2380 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2383 static inline struct net_device *first_net_device(struct net *net)
2385 return list_empty(&net->dev_base_head) ? NULL :
2386 net_device_entry(net->dev_base_head.next);
2389 static inline struct net_device *first_net_device_rcu(struct net *net)
2391 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2393 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2396 int netdev_boot_setup_check(struct net_device *dev);
2397 unsigned long netdev_boot_base(const char *prefix, int unit);
2398 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2399 const char *hwaddr);
2400 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2401 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2402 void dev_add_pack(struct packet_type *pt);
2403 void dev_remove_pack(struct packet_type *pt);
2404 void __dev_remove_pack(struct packet_type *pt);
2405 void dev_add_offload(struct packet_offload *po);
2406 void dev_remove_offload(struct packet_offload *po);
2408 int dev_get_iflink(const struct net_device *dev);
2409 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2410 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2411 unsigned short mask);
2412 struct net_device *dev_get_by_name(struct net *net, const char *name);
2413 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2414 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2415 int dev_alloc_name(struct net_device *dev, const char *name);
2416 int dev_open(struct net_device *dev);
2417 int dev_close(struct net_device *dev);
2418 int dev_close_many(struct list_head *head, bool unlink);
2419 void dev_disable_lro(struct net_device *dev);
2420 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2421 int dev_queue_xmit(struct sk_buff *skb);
2422 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2423 int register_netdevice(struct net_device *dev);
2424 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2425 void unregister_netdevice_many(struct list_head *head);
2426 static inline void unregister_netdevice(struct net_device *dev)
2428 unregister_netdevice_queue(dev, NULL);
2431 int netdev_refcnt_read(const struct net_device *dev);
2432 void free_netdev(struct net_device *dev);
2433 void netdev_freemem(struct net_device *dev);
2434 void synchronize_net(void);
2435 int init_dummy_netdev(struct net_device *dev);
2437 DECLARE_PER_CPU(int, xmit_recursion);
2438 #define XMIT_RECURSION_LIMIT 10
2440 static inline int dev_recursion_level(void)
2442 return this_cpu_read(xmit_recursion);
2445 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2446 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2447 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2448 int netdev_get_name(struct net *net, char *name, int ifindex);
2449 int dev_restart(struct net_device *dev);
2450 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2452 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2454 return NAPI_GRO_CB(skb)->data_offset;
2457 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2459 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2462 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2464 NAPI_GRO_CB(skb)->data_offset += len;
2467 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2468 unsigned int offset)
2470 return NAPI_GRO_CB(skb)->frag0 + offset;
2473 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2475 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2478 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2480 NAPI_GRO_CB(skb)->frag0 = NULL;
2481 NAPI_GRO_CB(skb)->frag0_len = 0;
2484 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2485 unsigned int offset)
2487 if (!pskb_may_pull(skb, hlen))
2490 skb_gro_frag0_invalidate(skb);
2491 return skb->data + offset;
2494 static inline void *skb_gro_network_header(struct sk_buff *skb)
2496 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2497 skb_network_offset(skb);
2500 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2501 const void *start, unsigned int len)
2503 if (NAPI_GRO_CB(skb)->csum_valid)
2504 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2505 csum_partial(start, len, 0));
2508 /* GRO checksum functions. These are logical equivalents of the normal
2509 * checksum functions (in skbuff.h) except that they operate on the GRO
2510 * offsets and fields in sk_buff.
2513 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2515 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2517 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2520 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2524 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2525 skb_checksum_start_offset(skb) <
2526 skb_gro_offset(skb)) &&
2527 !skb_at_gro_remcsum_start(skb) &&
2528 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2529 (!zero_okay || check));
2532 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2535 if (NAPI_GRO_CB(skb)->csum_valid &&
2536 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2539 NAPI_GRO_CB(skb)->csum = psum;
2541 return __skb_gro_checksum_complete(skb);
2544 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2546 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2547 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2548 NAPI_GRO_CB(skb)->csum_cnt--;
2550 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2551 * verified a new top level checksum or an encapsulated one
2552 * during GRO. This saves work if we fallback to normal path.
2554 __skb_incr_checksum_unnecessary(skb);
2558 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2561 __sum16 __ret = 0; \
2562 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2563 __ret = __skb_gro_checksum_validate_complete(skb, \
2564 compute_pseudo(skb, proto)); \
2566 __skb_mark_checksum_bad(skb); \
2568 skb_gro_incr_csum_unnecessary(skb); \
2572 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2573 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2575 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2577 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2579 #define skb_gro_checksum_simple_validate(skb) \
2580 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2582 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2584 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2585 !NAPI_GRO_CB(skb)->csum_valid);
2588 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2589 __sum16 check, __wsum pseudo)
2591 NAPI_GRO_CB(skb)->csum = ~pseudo;
2592 NAPI_GRO_CB(skb)->csum_valid = 1;
2595 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2597 if (__skb_gro_checksum_convert_check(skb)) \
2598 __skb_gro_checksum_convert(skb, check, \
2599 compute_pseudo(skb, proto)); \
2602 struct gro_remcsum {
2607 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2613 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2614 unsigned int off, size_t hdrlen,
2615 int start, int offset,
2616 struct gro_remcsum *grc,
2620 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2622 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2625 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2629 ptr = skb_gro_header_fast(skb, off);
2630 if (skb_gro_header_hard(skb, off + plen)) {
2631 ptr = skb_gro_header_slow(skb, off + plen, off);
2636 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2639 /* Adjust skb->csum since we changed the packet */
2640 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2642 grc->offset = off + hdrlen + offset;
2648 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2649 struct gro_remcsum *grc)
2652 size_t plen = grc->offset + sizeof(u16);
2657 ptr = skb_gro_header_fast(skb, grc->offset);
2658 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2659 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2664 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2667 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2668 unsigned short type,
2669 const void *daddr, const void *saddr,
2672 if (!dev->header_ops || !dev->header_ops->create)
2675 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2678 static inline int dev_parse_header(const struct sk_buff *skb,
2679 unsigned char *haddr)
2681 const struct net_device *dev = skb->dev;
2683 if (!dev->header_ops || !dev->header_ops->parse)
2685 return dev->header_ops->parse(skb, haddr);
2688 /* ll_header must have at least hard_header_len allocated */
2689 static inline bool dev_validate_header(const struct net_device *dev,
2690 char *ll_header, int len)
2692 if (likely(len >= dev->hard_header_len))
2695 if (capable(CAP_SYS_RAWIO)) {
2696 memset(ll_header + len, 0, dev->hard_header_len - len);
2700 if (dev->header_ops && dev->header_ops->validate)
2701 return dev->header_ops->validate(ll_header, len);
2706 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2707 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2708 static inline int unregister_gifconf(unsigned int family)
2710 return register_gifconf(family, NULL);
2713 #ifdef CONFIG_NET_FLOW_LIMIT
2714 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2715 struct sd_flow_limit {
2717 unsigned int num_buckets;
2718 unsigned int history_head;
2719 u16 history[FLOW_LIMIT_HISTORY];
2723 extern int netdev_flow_limit_table_len;
2724 #endif /* CONFIG_NET_FLOW_LIMIT */
2727 * Incoming packets are placed on per-CPU queues
2729 struct softnet_data {
2730 struct list_head poll_list;
2731 struct sk_buff_head process_queue;
2734 unsigned int processed;
2735 unsigned int time_squeeze;
2736 unsigned int received_rps;
2738 struct softnet_data *rps_ipi_list;
2740 #ifdef CONFIG_NET_FLOW_LIMIT
2741 struct sd_flow_limit __rcu *flow_limit;
2743 struct Qdisc *output_queue;
2744 struct Qdisc **output_queue_tailp;
2745 struct sk_buff *completion_queue;
2748 /* input_queue_head should be written by cpu owning this struct,
2749 * and only read by other cpus. Worth using a cache line.
2751 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2753 /* Elements below can be accessed between CPUs for RPS/RFS */
2754 struct call_single_data csd ____cacheline_aligned_in_smp;
2755 struct softnet_data *rps_ipi_next;
2757 unsigned int input_queue_tail;
2759 unsigned int dropped;
2760 struct sk_buff_head input_pkt_queue;
2761 struct napi_struct backlog;
2765 static inline void input_queue_head_incr(struct softnet_data *sd)
2768 sd->input_queue_head++;
2772 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2773 unsigned int *qtail)
2776 *qtail = ++sd->input_queue_tail;
2780 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2782 void __netif_schedule(struct Qdisc *q);
2783 void netif_schedule_queue(struct netdev_queue *txq);
2785 static inline void netif_tx_schedule_all(struct net_device *dev)
2789 for (i = 0; i < dev->num_tx_queues; i++)
2790 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2793 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2795 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2799 * netif_start_queue - allow transmit
2800 * @dev: network device
2802 * Allow upper layers to call the device hard_start_xmit routine.
2804 static inline void netif_start_queue(struct net_device *dev)
2806 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2809 static inline void netif_tx_start_all_queues(struct net_device *dev)
2813 for (i = 0; i < dev->num_tx_queues; i++) {
2814 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2815 netif_tx_start_queue(txq);
2819 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2822 * netif_wake_queue - restart transmit
2823 * @dev: network device
2825 * Allow upper layers to call the device hard_start_xmit routine.
2826 * Used for flow control when transmit resources are available.
2828 static inline void netif_wake_queue(struct net_device *dev)
2830 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2833 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2837 for (i = 0; i < dev->num_tx_queues; i++) {
2838 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2839 netif_tx_wake_queue(txq);
2843 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2845 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2849 * netif_stop_queue - stop transmitted packets
2850 * @dev: network device
2852 * Stop upper layers calling the device hard_start_xmit routine.
2853 * Used for flow control when transmit resources are unavailable.
2855 static inline void netif_stop_queue(struct net_device *dev)
2857 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2860 void netif_tx_stop_all_queues(struct net_device *dev);
2862 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2864 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2868 * netif_queue_stopped - test if transmit queue is flowblocked
2869 * @dev: network device
2871 * Test if transmit queue on device is currently unable to send.
2873 static inline bool netif_queue_stopped(const struct net_device *dev)
2875 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2878 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2880 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2884 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2886 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2890 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2892 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2896 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2897 * @dev_queue: pointer to transmit queue
2899 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2900 * to give appropriate hint to the CPU.
2902 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2905 prefetchw(&dev_queue->dql.num_queued);
2910 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2911 * @dev_queue: pointer to transmit queue
2913 * BQL enabled drivers might use this helper in their TX completion path,
2914 * to give appropriate hint to the CPU.
2916 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2919 prefetchw(&dev_queue->dql.limit);
2923 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2927 dql_queued(&dev_queue->dql, bytes);
2929 if (likely(dql_avail(&dev_queue->dql) >= 0))
2932 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2935 * The XOFF flag must be set before checking the dql_avail below,
2936 * because in netdev_tx_completed_queue we update the dql_completed
2937 * before checking the XOFF flag.
2941 /* check again in case another CPU has just made room avail */
2942 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2943 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2948 * netdev_sent_queue - report the number of bytes queued to hardware
2949 * @dev: network device
2950 * @bytes: number of bytes queued to the hardware device queue
2952 * Report the number of bytes queued for sending/completion to the network
2953 * device hardware queue. @bytes should be a good approximation and should
2954 * exactly match netdev_completed_queue() @bytes
2956 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2958 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2961 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2962 unsigned int pkts, unsigned int bytes)
2965 if (unlikely(!bytes))
2968 dql_completed(&dev_queue->dql, bytes);
2971 * Without the memory barrier there is a small possiblity that
2972 * netdev_tx_sent_queue will miss the update and cause the queue to
2973 * be stopped forever
2977 if (dql_avail(&dev_queue->dql) < 0)
2980 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2981 netif_schedule_queue(dev_queue);
2986 * netdev_completed_queue - report bytes and packets completed by device
2987 * @dev: network device
2988 * @pkts: actual number of packets sent over the medium
2989 * @bytes: actual number of bytes sent over the medium
2991 * Report the number of bytes and packets transmitted by the network device
2992 * hardware queue over the physical medium, @bytes must exactly match the
2993 * @bytes amount passed to netdev_sent_queue()
2995 static inline void netdev_completed_queue(struct net_device *dev,
2996 unsigned int pkts, unsigned int bytes)
2998 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3001 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3004 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3010 * netdev_reset_queue - reset the packets and bytes count of a network device
3011 * @dev_queue: network device
3013 * Reset the bytes and packet count of a network device and clear the
3014 * software flow control OFF bit for this network device
3016 static inline void netdev_reset_queue(struct net_device *dev_queue)
3018 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3022 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3023 * @dev: network device
3024 * @queue_index: given tx queue index
3026 * Returns 0 if given tx queue index >= number of device tx queues,
3027 * otherwise returns the originally passed tx queue index.
3029 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3031 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3032 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3033 dev->name, queue_index,
3034 dev->real_num_tx_queues);
3042 * netif_running - test if up
3043 * @dev: network device
3045 * Test if the device has been brought up.
3047 static inline bool netif_running(const struct net_device *dev)
3049 return test_bit(__LINK_STATE_START, &dev->state);
3053 * Routines to manage the subqueues on a device. We only need start,
3054 * stop, and a check if it's stopped. All other device management is
3055 * done at the overall netdevice level.
3056 * Also test the device if we're multiqueue.
3060 * netif_start_subqueue - allow sending packets on subqueue
3061 * @dev: network device
3062 * @queue_index: sub queue index
3064 * Start individual transmit queue of a device with multiple transmit queues.
3066 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3068 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3070 netif_tx_start_queue(txq);
3074 * netif_stop_subqueue - stop sending packets on subqueue
3075 * @dev: network device
3076 * @queue_index: sub queue index
3078 * Stop individual transmit queue of a device with multiple transmit queues.
3080 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3082 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3083 netif_tx_stop_queue(txq);
3087 * netif_subqueue_stopped - test status of subqueue
3088 * @dev: network device
3089 * @queue_index: sub queue index
3091 * Check individual transmit queue of a device with multiple transmit queues.
3093 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3096 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3098 return netif_tx_queue_stopped(txq);
3101 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3102 struct sk_buff *skb)
3104 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3108 * netif_wake_subqueue - allow sending packets on subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3112 * Resume individual transmit queue of a device with multiple transmit queues.
3114 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3116 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3118 netif_tx_wake_queue(txq);
3122 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3125 static inline int netif_set_xps_queue(struct net_device *dev,
3126 const struct cpumask *mask,
3133 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3134 unsigned int num_tx_queues);
3137 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3138 * as a distribution range limit for the returned value.
3140 static inline u16 skb_tx_hash(const struct net_device *dev,
3141 struct sk_buff *skb)
3143 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3147 * netif_is_multiqueue - test if device has multiple transmit queues
3148 * @dev: network device
3150 * Check if device has multiple transmit queues
3152 static inline bool netif_is_multiqueue(const struct net_device *dev)
3154 return dev->num_tx_queues > 1;
3157 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3160 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3162 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3170 static inline unsigned int get_netdev_rx_queue_index(
3171 struct netdev_rx_queue *queue)
3173 struct net_device *dev = queue->dev;
3174 int index = queue - dev->_rx;
3176 BUG_ON(index >= dev->num_rx_queues);
3181 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3182 int netif_get_num_default_rss_queues(void);
3184 enum skb_free_reason {
3185 SKB_REASON_CONSUMED,
3189 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3190 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3193 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3194 * interrupt context or with hardware interrupts being disabled.
3195 * (in_irq() || irqs_disabled())
3197 * We provide four helpers that can be used in following contexts :
3199 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3200 * replacing kfree_skb(skb)
3202 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3203 * Typically used in place of consume_skb(skb) in TX completion path
3205 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3206 * replacing kfree_skb(skb)
3208 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3209 * and consumed a packet. Used in place of consume_skb(skb)
3211 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3213 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3216 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3218 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3221 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3223 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3226 static inline void dev_consume_skb_any(struct sk_buff *skb)
3228 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3231 int netif_rx(struct sk_buff *skb);
3232 int netif_rx_ni(struct sk_buff *skb);
3233 int netif_receive_skb(struct sk_buff *skb);
3234 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3235 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3236 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3237 gro_result_t napi_gro_frags(struct napi_struct *napi);
3238 struct packet_offload *gro_find_receive_by_type(__be16 type);
3239 struct packet_offload *gro_find_complete_by_type(__be16 type);
3241 static inline void napi_free_frags(struct napi_struct *napi)
3243 kfree_skb(napi->skb);
3247 bool netdev_is_rx_handler_busy(struct net_device *dev);
3248 int netdev_rx_handler_register(struct net_device *dev,
3249 rx_handler_func_t *rx_handler,
3250 void *rx_handler_data);
3251 void netdev_rx_handler_unregister(struct net_device *dev);
3253 bool dev_valid_name(const char *name);
3254 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3255 int dev_ethtool(struct net *net, struct ifreq *);
3256 unsigned int dev_get_flags(const struct net_device *);
3257 int __dev_change_flags(struct net_device *, unsigned int flags);
3258 int dev_change_flags(struct net_device *, unsigned int);
3259 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3260 unsigned int gchanges);
3261 int dev_change_name(struct net_device *, const char *);
3262 int dev_set_alias(struct net_device *, const char *, size_t);
3263 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3264 int dev_set_mtu(struct net_device *, int);
3265 void dev_set_group(struct net_device *, int);
3266 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3267 int dev_change_carrier(struct net_device *, bool new_carrier);
3268 int dev_get_phys_port_id(struct net_device *dev,
3269 struct netdev_phys_item_id *ppid);
3270 int dev_get_phys_port_name(struct net_device *dev,
3271 char *name, size_t len);
3272 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3273 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
3274 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3275 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3276 struct netdev_queue *txq, int *ret);
3277 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3278 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3279 bool is_skb_forwardable(const struct net_device *dev,
3280 const struct sk_buff *skb);
3282 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3283 struct sk_buff *skb)
3285 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3286 unlikely(!is_skb_forwardable(dev, skb))) {
3287 atomic_long_inc(&dev->rx_dropped);
3292 skb_scrub_packet(skb, true);
3297 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3299 extern int netdev_budget;
3301 /* Called by rtnetlink.c:rtnl_unlock() */
3302 void netdev_run_todo(void);
3305 * dev_put - release reference to device
3306 * @dev: network device
3308 * Release reference to device to allow it to be freed.
3310 static inline void dev_put(struct net_device *dev)
3312 this_cpu_dec(*dev->pcpu_refcnt);
3316 * dev_hold - get reference to device
3317 * @dev: network device
3319 * Hold reference to device to keep it from being freed.
3321 static inline void dev_hold(struct net_device *dev)
3323 this_cpu_inc(*dev->pcpu_refcnt);
3326 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3327 * and _off may be called from IRQ context, but it is caller
3328 * who is responsible for serialization of these calls.
3330 * The name carrier is inappropriate, these functions should really be
3331 * called netif_lowerlayer_*() because they represent the state of any
3332 * kind of lower layer not just hardware media.
3335 void linkwatch_init_dev(struct net_device *dev);
3336 void linkwatch_fire_event(struct net_device *dev);
3337 void linkwatch_forget_dev(struct net_device *dev);
3340 * netif_carrier_ok - test if carrier present
3341 * @dev: network device
3343 * Check if carrier is present on device
3345 static inline bool netif_carrier_ok(const struct net_device *dev)
3347 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3350 unsigned long dev_trans_start(struct net_device *dev);
3352 void __netdev_watchdog_up(struct net_device *dev);
3354 void netif_carrier_on(struct net_device *dev);
3356 void netif_carrier_off(struct net_device *dev);
3359 * netif_dormant_on - mark device as dormant.
3360 * @dev: network device
3362 * Mark device as dormant (as per RFC2863).
3364 * The dormant state indicates that the relevant interface is not
3365 * actually in a condition to pass packets (i.e., it is not 'up') but is
3366 * in a "pending" state, waiting for some external event. For "on-
3367 * demand" interfaces, this new state identifies the situation where the
3368 * interface is waiting for events to place it in the up state.
3370 static inline void netif_dormant_on(struct net_device *dev)
3372 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3373 linkwatch_fire_event(dev);
3377 * netif_dormant_off - set device as not dormant.
3378 * @dev: network device
3380 * Device is not in dormant state.
3382 static inline void netif_dormant_off(struct net_device *dev)
3384 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3385 linkwatch_fire_event(dev);
3389 * netif_dormant - test if carrier present
3390 * @dev: network device
3392 * Check if carrier is present on device
3394 static inline bool netif_dormant(const struct net_device *dev)
3396 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3401 * netif_oper_up - test if device is operational
3402 * @dev: network device
3404 * Check if carrier is operational
3406 static inline bool netif_oper_up(const struct net_device *dev)
3408 return (dev->operstate == IF_OPER_UP ||
3409 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3413 * netif_device_present - is device available or removed
3414 * @dev: network device
3416 * Check if device has not been removed from system.
3418 static inline bool netif_device_present(struct net_device *dev)
3420 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3423 void netif_device_detach(struct net_device *dev);
3425 void netif_device_attach(struct net_device *dev);
3428 * Network interface message level settings
3432 NETIF_MSG_DRV = 0x0001,
3433 NETIF_MSG_PROBE = 0x0002,
3434 NETIF_MSG_LINK = 0x0004,
3435 NETIF_MSG_TIMER = 0x0008,
3436 NETIF_MSG_IFDOWN = 0x0010,
3437 NETIF_MSG_IFUP = 0x0020,
3438 NETIF_MSG_RX_ERR = 0x0040,
3439 NETIF_MSG_TX_ERR = 0x0080,
3440 NETIF_MSG_TX_QUEUED = 0x0100,
3441 NETIF_MSG_INTR = 0x0200,
3442 NETIF_MSG_TX_DONE = 0x0400,
3443 NETIF_MSG_RX_STATUS = 0x0800,
3444 NETIF_MSG_PKTDATA = 0x1000,
3445 NETIF_MSG_HW = 0x2000,
3446 NETIF_MSG_WOL = 0x4000,
3449 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3450 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3451 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3452 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3453 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3454 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3455 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3456 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3457 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3458 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3459 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3460 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3461 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3462 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3463 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3465 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3468 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3469 return default_msg_enable_bits;
3470 if (debug_value == 0) /* no output */
3472 /* set low N bits */
3473 return (1 << debug_value) - 1;
3476 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3478 spin_lock(&txq->_xmit_lock);
3479 txq->xmit_lock_owner = cpu;
3482 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3484 __acquire(&txq->_xmit_lock);
3488 static inline void __netif_tx_release(struct netdev_queue *txq)
3490 __release(&txq->_xmit_lock);
3493 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3495 spin_lock_bh(&txq->_xmit_lock);
3496 txq->xmit_lock_owner = smp_processor_id();
3499 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3501 bool ok = spin_trylock(&txq->_xmit_lock);
3503 txq->xmit_lock_owner = smp_processor_id();
3507 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3509 txq->xmit_lock_owner = -1;
3510 spin_unlock(&txq->_xmit_lock);
3513 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3515 txq->xmit_lock_owner = -1;
3516 spin_unlock_bh(&txq->_xmit_lock);
3519 static inline void txq_trans_update(struct netdev_queue *txq)
3521 if (txq->xmit_lock_owner != -1)
3522 txq->trans_start = jiffies;
3525 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3526 static inline void netif_trans_update(struct net_device *dev)
3528 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3530 if (txq->trans_start != jiffies)
3531 txq->trans_start = jiffies;
3535 * netif_tx_lock - grab network device transmit lock
3536 * @dev: network device
3538 * Get network device transmit lock
3540 static inline void netif_tx_lock(struct net_device *dev)
3545 spin_lock(&dev->tx_global_lock);
3546 cpu = smp_processor_id();
3547 for (i = 0; i < dev->num_tx_queues; i++) {
3548 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3550 /* We are the only thread of execution doing a
3551 * freeze, but we have to grab the _xmit_lock in
3552 * order to synchronize with threads which are in
3553 * the ->hard_start_xmit() handler and already
3554 * checked the frozen bit.
3556 __netif_tx_lock(txq, cpu);
3557 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3558 __netif_tx_unlock(txq);
3562 static inline void netif_tx_lock_bh(struct net_device *dev)
3568 static inline void netif_tx_unlock(struct net_device *dev)
3572 for (i = 0; i < dev->num_tx_queues; i++) {
3573 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3575 /* No need to grab the _xmit_lock here. If the
3576 * queue is not stopped for another reason, we
3579 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 netif_schedule_queue(txq);
3582 spin_unlock(&dev->tx_global_lock);
3585 static inline void netif_tx_unlock_bh(struct net_device *dev)
3587 netif_tx_unlock(dev);
3591 #define HARD_TX_LOCK(dev, txq, cpu) { \
3592 if ((dev->features & NETIF_F_LLTX) == 0) { \
3593 __netif_tx_lock(txq, cpu); \
3595 __netif_tx_acquire(txq); \
3599 #define HARD_TX_TRYLOCK(dev, txq) \
3600 (((dev->features & NETIF_F_LLTX) == 0) ? \
3601 __netif_tx_trylock(txq) : \
3602 __netif_tx_acquire(txq))
3604 #define HARD_TX_UNLOCK(dev, txq) { \
3605 if ((dev->features & NETIF_F_LLTX) == 0) { \
3606 __netif_tx_unlock(txq); \
3608 __netif_tx_release(txq); \
3612 static inline void netif_tx_disable(struct net_device *dev)
3618 cpu = smp_processor_id();
3619 for (i = 0; i < dev->num_tx_queues; i++) {
3620 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3622 __netif_tx_lock(txq, cpu);
3623 netif_tx_stop_queue(txq);
3624 __netif_tx_unlock(txq);
3629 static inline void netif_addr_lock(struct net_device *dev)
3631 spin_lock(&dev->addr_list_lock);
3634 static inline void netif_addr_lock_nested(struct net_device *dev)
3636 int subclass = SINGLE_DEPTH_NESTING;
3638 if (dev->netdev_ops->ndo_get_lock_subclass)
3639 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3641 spin_lock_nested(&dev->addr_list_lock, subclass);
3644 static inline void netif_addr_lock_bh(struct net_device *dev)
3646 spin_lock_bh(&dev->addr_list_lock);
3649 static inline void netif_addr_unlock(struct net_device *dev)
3651 spin_unlock(&dev->addr_list_lock);
3654 static inline void netif_addr_unlock_bh(struct net_device *dev)
3656 spin_unlock_bh(&dev->addr_list_lock);
3660 * dev_addrs walker. Should be used only for read access. Call with
3661 * rcu_read_lock held.
3663 #define for_each_dev_addr(dev, ha) \
3664 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3666 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3668 void ether_setup(struct net_device *dev);
3670 /* Support for loadable net-drivers */
3671 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3672 unsigned char name_assign_type,
3673 void (*setup)(struct net_device *),
3674 unsigned int txqs, unsigned int rxqs);
3675 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3676 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3678 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3679 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3682 int register_netdev(struct net_device *dev);
3683 void unregister_netdev(struct net_device *dev);
3685 /* General hardware address lists handling functions */
3686 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3687 struct netdev_hw_addr_list *from_list, int addr_len);
3688 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3689 struct netdev_hw_addr_list *from_list, int addr_len);
3690 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3691 struct net_device *dev,
3692 int (*sync)(struct net_device *, const unsigned char *),
3693 int (*unsync)(struct net_device *,
3694 const unsigned char *));
3695 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3696 struct net_device *dev,
3697 int (*unsync)(struct net_device *,
3698 const unsigned char *));
3699 void __hw_addr_init(struct netdev_hw_addr_list *list);
3701 /* Functions used for device addresses handling */
3702 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3703 unsigned char addr_type);
3704 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3705 unsigned char addr_type);
3706 void dev_addr_flush(struct net_device *dev);
3707 int dev_addr_init(struct net_device *dev);
3709 /* Functions used for unicast addresses handling */
3710 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3711 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3712 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3713 int dev_uc_sync(struct net_device *to, struct net_device *from);
3714 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3715 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3716 void dev_uc_flush(struct net_device *dev);
3717 void dev_uc_init(struct net_device *dev);
3720 * __dev_uc_sync - Synchonize device's unicast list
3721 * @dev: device to sync
3722 * @sync: function to call if address should be added
3723 * @unsync: function to call if address should be removed
3725 * Add newly added addresses to the interface, and release
3726 * addresses that have been deleted.
3728 static inline int __dev_uc_sync(struct net_device *dev,
3729 int (*sync)(struct net_device *,
3730 const unsigned char *),
3731 int (*unsync)(struct net_device *,
3732 const unsigned char *))
3734 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3738 * __dev_uc_unsync - Remove synchronized addresses from device
3739 * @dev: device to sync
3740 * @unsync: function to call if address should be removed
3742 * Remove all addresses that were added to the device by dev_uc_sync().
3744 static inline void __dev_uc_unsync(struct net_device *dev,
3745 int (*unsync)(struct net_device *,
3746 const unsigned char *))
3748 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3751 /* Functions used for multicast addresses handling */
3752 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3753 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3754 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3755 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3756 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3757 int dev_mc_sync(struct net_device *to, struct net_device *from);
3758 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3759 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3760 void dev_mc_flush(struct net_device *dev);
3761 void dev_mc_init(struct net_device *dev);
3764 * __dev_mc_sync - Synchonize device's multicast list
3765 * @dev: device to sync
3766 * @sync: function to call if address should be added
3767 * @unsync: function to call if address should be removed
3769 * Add newly added addresses to the interface, and release
3770 * addresses that have been deleted.
3772 static inline int __dev_mc_sync(struct net_device *dev,
3773 int (*sync)(struct net_device *,
3774 const unsigned char *),
3775 int (*unsync)(struct net_device *,
3776 const unsigned char *))
3778 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3782 * __dev_mc_unsync - Remove synchronized addresses from device
3783 * @dev: device to sync
3784 * @unsync: function to call if address should be removed
3786 * Remove all addresses that were added to the device by dev_mc_sync().
3788 static inline void __dev_mc_unsync(struct net_device *dev,
3789 int (*unsync)(struct net_device *,
3790 const unsigned char *))
3792 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3795 /* Functions used for secondary unicast and multicast support */
3796 void dev_set_rx_mode(struct net_device *dev);
3797 void __dev_set_rx_mode(struct net_device *dev);
3798 int dev_set_promiscuity(struct net_device *dev, int inc);
3799 int dev_set_allmulti(struct net_device *dev, int inc);
3800 void netdev_state_change(struct net_device *dev);
3801 void netdev_notify_peers(struct net_device *dev);
3802 void netdev_features_change(struct net_device *dev);
3803 /* Load a device via the kmod */
3804 void dev_load(struct net *net, const char *name);
3805 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3806 struct rtnl_link_stats64 *storage);
3807 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3808 const struct net_device_stats *netdev_stats);
3810 extern int netdev_max_backlog;
3811 extern int netdev_tstamp_prequeue;
3812 extern int weight_p;
3813 extern int dev_weight_rx_bias;
3814 extern int dev_weight_tx_bias;
3815 extern int dev_rx_weight;
3816 extern int dev_tx_weight;
3818 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3819 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3820 struct list_head **iter);
3821 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3822 struct list_head **iter);
3824 /* iterate through upper list, must be called under RCU read lock */
3825 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3826 for (iter = &(dev)->adj_list.upper, \
3827 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3829 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3831 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3832 int (*fn)(struct net_device *upper_dev,
3836 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3837 struct net_device *upper_dev);
3839 void *netdev_lower_get_next_private(struct net_device *dev,
3840 struct list_head **iter);
3841 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3842 struct list_head **iter);
3844 #define netdev_for_each_lower_private(dev, priv, iter) \
3845 for (iter = (dev)->adj_list.lower.next, \
3846 priv = netdev_lower_get_next_private(dev, &(iter)); \
3848 priv = netdev_lower_get_next_private(dev, &(iter)))
3850 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3851 for (iter = &(dev)->adj_list.lower, \
3852 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3854 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3856 void *netdev_lower_get_next(struct net_device *dev,
3857 struct list_head **iter);
3859 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3860 for (iter = (dev)->adj_list.lower.next, \
3861 ldev = netdev_lower_get_next(dev, &(iter)); \
3863 ldev = netdev_lower_get_next(dev, &(iter)))
3865 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3866 struct list_head **iter);
3867 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3868 struct list_head **iter);
3870 int netdev_walk_all_lower_dev(struct net_device *dev,
3871 int (*fn)(struct net_device *lower_dev,
3874 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3875 int (*fn)(struct net_device *lower_dev,
3879 void *netdev_adjacent_get_private(struct list_head *adj_list);
3880 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3881 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3882 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3883 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3884 int netdev_master_upper_dev_link(struct net_device *dev,
3885 struct net_device *upper_dev,
3886 void *upper_priv, void *upper_info);
3887 void netdev_upper_dev_unlink(struct net_device *dev,
3888 struct net_device *upper_dev);
3889 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3890 void *netdev_lower_dev_get_private(struct net_device *dev,
3891 struct net_device *lower_dev);
3892 void netdev_lower_state_changed(struct net_device *lower_dev,
3893 void *lower_state_info);
3894 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3895 struct neighbour *n);
3896 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3897 struct neighbour *n);
3899 /* RSS keys are 40 or 52 bytes long */
3900 #define NETDEV_RSS_KEY_LEN 52
3901 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3902 void netdev_rss_key_fill(void *buffer, size_t len);
3904 int dev_get_nest_level(struct net_device *dev);
3905 int skb_checksum_help(struct sk_buff *skb);
3906 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3907 netdev_features_t features, bool tx_path);
3908 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3909 netdev_features_t features);
3911 struct netdev_bonding_info {
3916 struct netdev_notifier_bonding_info {
3917 struct netdev_notifier_info info; /* must be first */
3918 struct netdev_bonding_info bonding_info;
3921 void netdev_bonding_info_change(struct net_device *dev,
3922 struct netdev_bonding_info *bonding_info);
3925 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3927 return __skb_gso_segment(skb, features, true);
3929 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3931 static inline bool can_checksum_protocol(netdev_features_t features,
3934 if (protocol == htons(ETH_P_FCOE))
3935 return !!(features & NETIF_F_FCOE_CRC);
3937 /* Assume this is an IP checksum (not SCTP CRC) */
3939 if (features & NETIF_F_HW_CSUM) {
3940 /* Can checksum everything */
3945 case htons(ETH_P_IP):
3946 return !!(features & NETIF_F_IP_CSUM);
3947 case htons(ETH_P_IPV6):
3948 return !!(features & NETIF_F_IPV6_CSUM);
3955 void netdev_rx_csum_fault(struct net_device *dev);
3957 static inline void netdev_rx_csum_fault(struct net_device *dev)
3961 /* rx skb timestamps */
3962 void net_enable_timestamp(void);
3963 void net_disable_timestamp(void);
3965 #ifdef CONFIG_PROC_FS
3966 int __init dev_proc_init(void);
3968 #define dev_proc_init() 0
3971 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3972 struct sk_buff *skb, struct net_device *dev,
3975 skb->xmit_more = more ? 1 : 0;
3976 return ops->ndo_start_xmit(skb, dev);
3979 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3980 struct netdev_queue *txq, bool more)
3982 const struct net_device_ops *ops = dev->netdev_ops;
3985 rc = __netdev_start_xmit(ops, skb, dev, more);
3986 if (rc == NETDEV_TX_OK)
3987 txq_trans_update(txq);
3992 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3994 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3997 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3999 return netdev_class_create_file_ns(class_attr, NULL);
4002 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4004 netdev_class_remove_file_ns(class_attr, NULL);
4007 extern struct kobj_ns_type_operations net_ns_type_operations;
4009 const char *netdev_drivername(const struct net_device *dev);
4011 void linkwatch_run_queue(void);
4013 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4014 netdev_features_t f2)
4016 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4017 if (f1 & NETIF_F_HW_CSUM)
4018 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4020 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4026 static inline netdev_features_t netdev_get_wanted_features(
4027 struct net_device *dev)
4029 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4031 netdev_features_t netdev_increment_features(netdev_features_t all,
4032 netdev_features_t one, netdev_features_t mask);
4034 /* Allow TSO being used on stacked device :
4035 * Performing the GSO segmentation before last device
4036 * is a performance improvement.
4038 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4039 netdev_features_t mask)
4041 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4044 int __netdev_update_features(struct net_device *dev);
4045 void netdev_update_features(struct net_device *dev);
4046 void netdev_change_features(struct net_device *dev);
4048 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4049 struct net_device *dev);
4051 netdev_features_t passthru_features_check(struct sk_buff *skb,
4052 struct net_device *dev,
4053 netdev_features_t features);
4054 netdev_features_t netif_skb_features(struct sk_buff *skb);
4056 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4058 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4060 /* check flags correspondence */
4061 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4062 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4063 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4064 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4065 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4066 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4067 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4068 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4069 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4070 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4071 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4072 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4073 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4074 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4075 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4076 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4078 return (features & feature) == feature;
4081 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4083 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4084 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4087 static inline bool netif_needs_gso(struct sk_buff *skb,
4088 netdev_features_t features)
4090 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4091 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4092 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4095 static inline void netif_set_gso_max_size(struct net_device *dev,
4098 dev->gso_max_size = size;
4101 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4102 int pulled_hlen, u16 mac_offset,
4105 skb->protocol = protocol;
4106 skb->encapsulation = 1;
4107 skb_push(skb, pulled_hlen);
4108 skb_reset_transport_header(skb);
4109 skb->mac_header = mac_offset;
4110 skb->network_header = skb->mac_header + mac_len;
4111 skb->mac_len = mac_len;
4114 static inline bool netif_is_macsec(const struct net_device *dev)
4116 return dev->priv_flags & IFF_MACSEC;
4119 static inline bool netif_is_macvlan(const struct net_device *dev)
4121 return dev->priv_flags & IFF_MACVLAN;
4124 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4126 return dev->priv_flags & IFF_MACVLAN_PORT;
4129 static inline bool netif_is_ipvlan(const struct net_device *dev)
4131 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4134 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4136 return dev->priv_flags & IFF_IPVLAN_MASTER;
4139 static inline bool netif_is_bond_master(const struct net_device *dev)
4141 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4144 static inline bool netif_is_bond_slave(const struct net_device *dev)
4146 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4149 static inline bool netif_supports_nofcs(struct net_device *dev)
4151 return dev->priv_flags & IFF_SUPP_NOFCS;
4154 static inline bool netif_is_l3_master(const struct net_device *dev)
4156 return dev->priv_flags & IFF_L3MDEV_MASTER;
4159 static inline bool netif_is_l3_slave(const struct net_device *dev)
4161 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4164 static inline bool netif_is_bridge_master(const struct net_device *dev)
4166 return dev->priv_flags & IFF_EBRIDGE;
4169 static inline bool netif_is_bridge_port(const struct net_device *dev)
4171 return dev->priv_flags & IFF_BRIDGE_PORT;
4174 static inline bool netif_is_ovs_master(const struct net_device *dev)
4176 return dev->priv_flags & IFF_OPENVSWITCH;
4179 static inline bool netif_is_team_master(const struct net_device *dev)
4181 return dev->priv_flags & IFF_TEAM;
4184 static inline bool netif_is_team_port(const struct net_device *dev)
4186 return dev->priv_flags & IFF_TEAM_PORT;
4189 static inline bool netif_is_lag_master(const struct net_device *dev)
4191 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4194 static inline bool netif_is_lag_port(const struct net_device *dev)
4196 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4199 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4201 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4204 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4205 static inline void netif_keep_dst(struct net_device *dev)
4207 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4210 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4211 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4213 /* TODO: reserve and use an additional IFF bit, if we get more users */
4214 return dev->priv_flags & IFF_MACSEC;
4217 extern struct pernet_operations __net_initdata loopback_net_ops;
4219 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4221 /* netdev_printk helpers, similar to dev_printk */
4223 static inline const char *netdev_name(const struct net_device *dev)
4225 if (!dev->name[0] || strchr(dev->name, '%'))
4226 return "(unnamed net_device)";
4230 static inline const char *netdev_reg_state(const struct net_device *dev)
4232 switch (dev->reg_state) {
4233 case NETREG_UNINITIALIZED: return " (uninitialized)";
4234 case NETREG_REGISTERED: return "";
4235 case NETREG_UNREGISTERING: return " (unregistering)";
4236 case NETREG_UNREGISTERED: return " (unregistered)";
4237 case NETREG_RELEASED: return " (released)";
4238 case NETREG_DUMMY: return " (dummy)";
4241 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4242 return " (unknown)";
4246 void netdev_printk(const char *level, const struct net_device *dev,
4247 const char *format, ...);
4249 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4251 void netdev_alert(const struct net_device *dev, const char *format, ...);
4253 void netdev_crit(const struct net_device *dev, const char *format, ...);
4255 void netdev_err(const struct net_device *dev, const char *format, ...);
4257 void netdev_warn(const struct net_device *dev, const char *format, ...);
4259 void netdev_notice(const struct net_device *dev, const char *format, ...);
4261 void netdev_info(const struct net_device *dev, const char *format, ...);
4263 #define MODULE_ALIAS_NETDEV(device) \
4264 MODULE_ALIAS("netdev-" device)
4266 #if defined(CONFIG_DYNAMIC_DEBUG)
4267 #define netdev_dbg(__dev, format, args...) \
4269 dynamic_netdev_dbg(__dev, format, ##args); \
4271 #elif defined(DEBUG)
4272 #define netdev_dbg(__dev, format, args...) \
4273 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4275 #define netdev_dbg(__dev, format, args...) \
4278 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4282 #if defined(VERBOSE_DEBUG)
4283 #define netdev_vdbg netdev_dbg
4286 #define netdev_vdbg(dev, format, args...) \
4289 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4295 * netdev_WARN() acts like dev_printk(), but with the key difference
4296 * of using a WARN/WARN_ON to get the message out, including the
4297 * file/line information and a backtrace.
4299 #define netdev_WARN(dev, format, args...) \
4300 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4301 netdev_reg_state(dev), ##args)
4303 /* netif printk helpers, similar to netdev_printk */
4305 #define netif_printk(priv, type, level, dev, fmt, args...) \
4307 if (netif_msg_##type(priv)) \
4308 netdev_printk(level, (dev), fmt, ##args); \
4311 #define netif_level(level, priv, type, dev, fmt, args...) \
4313 if (netif_msg_##type(priv)) \
4314 netdev_##level(dev, fmt, ##args); \
4317 #define netif_emerg(priv, type, dev, fmt, args...) \
4318 netif_level(emerg, priv, type, dev, fmt, ##args)
4319 #define netif_alert(priv, type, dev, fmt, args...) \
4320 netif_level(alert, priv, type, dev, fmt, ##args)
4321 #define netif_crit(priv, type, dev, fmt, args...) \
4322 netif_level(crit, priv, type, dev, fmt, ##args)
4323 #define netif_err(priv, type, dev, fmt, args...) \
4324 netif_level(err, priv, type, dev, fmt, ##args)
4325 #define netif_warn(priv, type, dev, fmt, args...) \
4326 netif_level(warn, priv, type, dev, fmt, ##args)
4327 #define netif_notice(priv, type, dev, fmt, args...) \
4328 netif_level(notice, priv, type, dev, fmt, ##args)
4329 #define netif_info(priv, type, dev, fmt, args...) \
4330 netif_level(info, priv, type, dev, fmt, ##args)
4332 #if defined(CONFIG_DYNAMIC_DEBUG)
4333 #define netif_dbg(priv, type, netdev, format, args...) \
4335 if (netif_msg_##type(priv)) \
4336 dynamic_netdev_dbg(netdev, format, ##args); \
4338 #elif defined(DEBUG)
4339 #define netif_dbg(priv, type, dev, format, args...) \
4340 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4342 #define netif_dbg(priv, type, dev, format, args...) \
4345 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4350 #if defined(VERBOSE_DEBUG)
4351 #define netif_vdbg netif_dbg
4353 #define netif_vdbg(priv, type, dev, format, args...) \
4356 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4362 * The list of packet types we will receive (as opposed to discard)
4363 * and the routines to invoke.
4365 * Why 16. Because with 16 the only overlap we get on a hash of the
4366 * low nibble of the protocol value is RARP/SNAP/X.25.
4368 * NOTE: That is no longer true with the addition of VLAN tags. Not
4369 * sure which should go first, but I bet it won't make much
4370 * difference if we are running VLANs. The good news is that
4371 * this protocol won't be in the list unless compiled in, so
4372 * the average user (w/out VLANs) will not be adversely affected.
4388 #define PTYPE_HASH_SIZE (16)
4389 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4391 #endif /* _LINUX_NETDEVICE_H */