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;
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
214 struct rcu_head rcu_head;
217 struct netdev_hw_addr_list {
218 struct list_head list;
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
242 /* cached hardware header; allow for machine alignment needs. */
243 #define HH_DATA_MOD 16
244 #define HH_DATA_OFF(__len) \
245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
246 #define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
259 #define LL_RESERVED_SPACE(dev) \
260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
267 const void *saddr, unsigned int len);
268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
273 bool (*validate)(const char *ll_header, unsigned int len);
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
281 enum netdev_state_t {
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
294 struct netdev_boot_setup {
298 #define NETDEV_BOOT_SETUP_MAX 8
300 int __init netdev_boot_setup(char *str);
303 * Structure for NAPI scheduling similar to tasklet but with weighting
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
309 * to the per-CPU poll_list, and whoever clears that bit
310 * can remove from the list right before clearing the bit.
312 struct list_head poll_list;
316 unsigned int gro_count;
317 int (*poll)(struct napi_struct *, int);
318 #ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
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 void __napi_complete(struct napi_struct *n);
467 void 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.
475 static inline void napi_complete(struct napi_struct *n)
477 return napi_complete_done(n, 0);
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: NAPI context
484 * Warning: caller must observe RCU grace period
485 * before freeing memory containing @napi, if
486 * this function returns true.
487 * Note: core networking stack automatically calls it
488 * from netif_napi_del().
489 * Drivers might want to call this helper to combine all
490 * the needed RCU grace periods into a single one.
492 bool napi_hash_del(struct napi_struct *napi);
495 * napi_disable - prevent NAPI from scheduling
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
501 void napi_disable(struct napi_struct *n);
504 * napi_enable - enable NAPI scheduling
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
510 static inline void napi_enable(struct napi_struct *n)
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
513 smp_mb__before_atomic();
514 clear_bit(NAPI_STATE_SCHED, &n->state);
515 clear_bit(NAPI_STATE_NPSVC, &n->state);
519 * napi_synchronize - wait until NAPI is not running
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
526 static inline void napi_synchronize(const struct napi_struct *n)
528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
535 enum netdev_queue_state_t {
536 __QUEUE_STATE_DRV_XOFF,
537 __QUEUE_STATE_STACK_XOFF,
538 __QUEUE_STATE_FROZEN,
541 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
542 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
543 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
546 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
552 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
553 * netif_tx_* functions below are used to manipulate this flag. The
554 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
555 * queue independently. The netif_xmit_*stopped functions below are called
556 * to check if the queue has been stopped by the driver or stack (either
557 * of the XOFF bits are set in the state). Drivers should not need to call
558 * netif_xmit*stopped functions, they should only be using netif_tx_*.
561 struct netdev_queue {
565 struct net_device *dev;
566 struct Qdisc __rcu *qdisc;
567 struct Qdisc *qdisc_sleeping;
571 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
574 unsigned long tx_maxrate;
576 * Number of TX timeouts for this queue
577 * (/sys/class/net/DEV/Q/trans_timeout)
579 unsigned long trans_timeout;
583 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
586 * Time (in jiffies) of last Tx
588 unsigned long trans_start;
595 } ____cacheline_aligned_in_smp;
597 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
623 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
630 struct rps_dev_flow {
633 unsigned int last_qtail;
635 #define RPS_NO_FILTER 0xffff
638 * The rps_dev_flow_table structure contains a table of flow mappings.
640 struct rps_dev_flow_table {
643 struct rps_dev_flow flows[0];
645 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
646 ((_num) * sizeof(struct rps_dev_flow)))
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
651 * Each entry is a 32bit value. Upper part is the high-order bits
652 * of flow hash, lower part is CPU number.
653 * rps_cpu_mask is used to partition the space, depending on number of
654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
656 * meaning we use 32-6=26 bits for the hash.
658 struct rps_sock_flow_table {
661 u32 ents[0] ____cacheline_aligned_in_smp;
663 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
665 #define RPS_NO_CPU 0xffff
667 extern u32 rps_cpu_mask;
668 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
677 /* We only give a hint, preemption can change CPU under us */
678 val |= raw_smp_processor_id();
680 if (table->ents[index] != val)
681 table->ents[index] = val;
685 #ifdef CONFIG_RFS_ACCEL
686 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
689 #endif /* CONFIG_RPS */
691 /* This structure contains an instance of an RX queue. */
692 struct netdev_rx_queue {
694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
698 struct net_device *dev;
699 } ____cacheline_aligned_in_smp;
702 * RX queue sysfs structures and functions.
704 struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
719 unsigned int alloc_len;
723 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
724 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 struct xps_dev_maps {
732 struct xps_map __rcu *cpu_map[0];
734 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
736 #endif /* CONFIG_XPS */
738 #define TC_MAX_QUEUE 16
739 #define TC_BITMASK 15
740 /* HW offloaded queuing disciplines txq count and offset maps */
741 struct netdev_tc_txq {
746 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
751 struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
759 char model_description[256];
763 #define MAX_PHYS_ITEM_ID_LEN 32
765 /* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
768 struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
770 unsigned char id_len;
773 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
780 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
783 /* These structures hold the attributes of qdisc and classifiers
784 * that are being passed to the netdevice through the setup_tc op.
794 struct tc_cls_u32_offload;
796 struct tc_to_netdev {
800 struct tc_cls_u32_offload *cls_u32;
801 struct tc_cls_flower_offload *cls_flower;
802 struct tc_cls_matchall_offload *cls_mall;
803 struct tc_cls_bpf_offload *cls_bpf;
807 /* These structures hold the attributes of xdp state that are being passed
808 * to the netdevice through the xdp op.
810 enum xdp_netdev_command {
811 /* Set or clear a bpf program used in the earliest stages of packet
812 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
813 * is responsible for calling bpf_prog_put on any old progs that are
814 * stored. In case of error, the callee need not release the new prog
815 * reference, but on success it takes ownership and must bpf_prog_put
816 * when it is no longer used.
819 /* Check if a bpf program is set on the device. The callee should
820 * return true if a program is currently attached and running.
826 enum xdp_netdev_command command;
829 struct bpf_prog *prog;
836 * This structure defines the management hooks for network devices.
837 * The following hooks can be defined; unless noted otherwise, they are
838 * optional and can be filled with a null pointer.
840 * int (*ndo_init)(struct net_device *dev);
841 * This function is called once when a network device is registered.
842 * The network device can use this for any late stage initialization
843 * or semantic validation. It can fail with an error code which will
844 * be propagated back to register_netdev.
846 * void (*ndo_uninit)(struct net_device *dev);
847 * This function is called when device is unregistered or when registration
848 * fails. It is not called if init fails.
850 * int (*ndo_open)(struct net_device *dev);
851 * This function is called when a network device transitions to the up
854 * int (*ndo_stop)(struct net_device *dev);
855 * This function is called when a network device transitions to the down
858 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
859 * struct net_device *dev);
860 * Called when a packet needs to be transmitted.
861 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
862 * the queue before that can happen; it's for obsolete devices and weird
863 * corner cases, but the stack really does a non-trivial amount
864 * of useless work if you return NETDEV_TX_BUSY.
865 * Required; cannot be NULL.
867 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
868 * netdev_features_t features);
869 * Adjusts the requested feature flags according to device-specific
870 * constraints, and returns the resulting flags. Must not modify
873 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
874 * void *accel_priv, select_queue_fallback_t fallback);
875 * Called to decide which queue to use when device supports multiple
878 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
879 * This function is called to allow device receiver to make
880 * changes to configuration when multicast or promiscuous is enabled.
882 * void (*ndo_set_rx_mode)(struct net_device *dev);
883 * This function is called device changes address list filtering.
884 * If driver handles unicast address filtering, it should set
885 * IFF_UNICAST_FLT in its priv_flags.
887 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
888 * This function is called when the Media Access Control address
889 * needs to be changed. If this interface is not defined, the
890 * MAC address can not be changed.
892 * int (*ndo_validate_addr)(struct net_device *dev);
893 * Test if Media Access Control address is valid for the device.
895 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
896 * Called when a user requests an ioctl which can't be handled by
897 * the generic interface code. If not defined ioctls return
898 * not supported error code.
900 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
901 * Used to set network devices bus interface parameters. This interface
902 * is retained for legacy reasons; new devices should use the bus
903 * interface (PCI) for low level management.
905 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
906 * Called when a user wants to change the Maximum Transfer Unit
907 * of a device. If not defined, any request to change MTU will
908 * will return an error.
910 * void (*ndo_tx_timeout)(struct net_device *dev);
911 * Callback used when the transmitter has not made any progress
912 * for dev->watchdog ticks.
914 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
915 * struct rtnl_link_stats64 *storage);
916 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
917 * Called when a user wants to get the network device usage
918 * statistics. Drivers must do one of the following:
919 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
920 * rtnl_link_stats64 structure passed by the caller.
921 * 2. Define @ndo_get_stats to update a net_device_stats structure
922 * (which should normally be dev->stats) and return a pointer to
923 * it. The structure may be changed asynchronously only if each
924 * field is written atomically.
925 * 3. Update dev->stats asynchronously and atomically, and define
928 * bool (*ndo_has_offload_stats)(int attr_id)
929 * Return true if this device supports offload stats of this attr_id.
931 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
933 * Get statistics for offload operations by attr_id. Write it into the
936 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
937 * If device supports VLAN filtering this function is called when a
938 * VLAN id is registered.
940 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
941 * If device supports VLAN filtering this function is called when a
942 * VLAN id is unregistered.
944 * void (*ndo_poll_controller)(struct net_device *dev);
946 * SR-IOV management functions.
947 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
948 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
949 * u8 qos, __be16 proto);
950 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
952 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
953 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
954 * int (*ndo_get_vf_config)(struct net_device *dev,
955 * int vf, struct ifla_vf_info *ivf);
956 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
957 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
958 * struct nlattr *port[]);
960 * Enable or disable the VF ability to query its RSS Redirection Table and
961 * Hash Key. This is needed since on some devices VF share this information
962 * with PF and querying it may introduce a theoretical security risk.
963 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
964 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
965 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
966 * Called to setup 'tc' number of traffic classes in the net device. This
967 * is always called from the stack with the rtnl lock held and netif tx
968 * queues stopped. This allows the netdevice to perform queue management
971 * Fiber Channel over Ethernet (FCoE) offload functions.
972 * int (*ndo_fcoe_enable)(struct net_device *dev);
973 * Called when the FCoE protocol stack wants to start using LLD for FCoE
974 * so the underlying device can perform whatever needed configuration or
975 * initialization to support acceleration of FCoE traffic.
977 * int (*ndo_fcoe_disable)(struct net_device *dev);
978 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
979 * so the underlying device can perform whatever needed clean-ups to
980 * stop supporting acceleration of FCoE traffic.
982 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
983 * struct scatterlist *sgl, unsigned int sgc);
984 * Called when the FCoE Initiator wants to initialize an I/O that
985 * is a possible candidate for Direct Data Placement (DDP). The LLD can
986 * perform necessary setup and returns 1 to indicate the device is set up
987 * successfully to perform DDP on this I/O, otherwise this returns 0.
989 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
990 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
991 * indicated by the FC exchange id 'xid', so the underlying device can
992 * clean up and reuse resources for later DDP requests.
994 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
995 * struct scatterlist *sgl, unsigned int sgc);
996 * Called when the FCoE Target wants to initialize an I/O that
997 * is a possible candidate for Direct Data Placement (DDP). The LLD can
998 * perform necessary setup and returns 1 to indicate the device is set up
999 * successfully to perform DDP on this I/O, otherwise this returns 0.
1001 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1002 * struct netdev_fcoe_hbainfo *hbainfo);
1003 * Called when the FCoE Protocol stack wants information on the underlying
1004 * device. This information is utilized by the FCoE protocol stack to
1005 * register attributes with Fiber Channel management service as per the
1006 * FC-GS Fabric Device Management Information(FDMI) specification.
1008 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1009 * Called when the underlying device wants to override default World Wide
1010 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1011 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1012 * protocol stack to use.
1015 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1016 * u16 rxq_index, u32 flow_id);
1017 * Set hardware filter for RFS. rxq_index is the target queue index;
1018 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1019 * Return the filter ID on success, or a negative error code.
1021 * Slave management functions (for bridge, bonding, etc).
1022 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1023 * Called to make another netdev an underling.
1025 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1026 * Called to release previously enslaved netdev.
1028 * Feature/offload setting functions.
1029 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1030 * Called to update device configuration to new features. Passed
1031 * feature set might be less than what was returned by ndo_fix_features()).
1032 * Must return >0 or -errno if it changed dev->features itself.
1034 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1035 * struct net_device *dev,
1036 * const unsigned char *addr, u16 vid, u16 flags)
1037 * Adds an FDB entry to dev for addr.
1038 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1039 * struct net_device *dev,
1040 * const unsigned char *addr, u16 vid)
1041 * Deletes the FDB entry from dev coresponding to addr.
1042 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1043 * struct net_device *dev, struct net_device *filter_dev,
1045 * Used to add FDB entries to dump requests. Implementers should add
1046 * entries to skb and update idx with the number of entries.
1048 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1050 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1051 * struct net_device *dev, u32 filter_mask,
1053 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1056 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1057 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1058 * which do not represent real hardware may define this to allow their
1059 * userspace components to manage their virtual carrier state. Devices
1060 * that determine carrier state from physical hardware properties (eg
1061 * network cables) or protocol-dependent mechanisms (eg
1062 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1064 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1065 * struct netdev_phys_item_id *ppid);
1066 * Called to get ID of physical port of this device. If driver does
1067 * not implement this, it is assumed that the hw is not able to have
1068 * multiple net devices on single physical port.
1070 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1071 * struct udp_tunnel_info *ti);
1072 * Called by UDP tunnel to notify a driver about the UDP port and socket
1073 * address family that a UDP tunnel is listnening to. It is called only
1074 * when a new port starts listening. The operation is protected by the
1077 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1078 * struct udp_tunnel_info *ti);
1079 * Called by UDP tunnel to notify the driver about a UDP port and socket
1080 * address family that the UDP tunnel is not listening to anymore. The
1081 * operation is protected by the RTNL.
1083 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1084 * struct net_device *dev)
1085 * Called by upper layer devices to accelerate switching or other
1086 * station functionality into hardware. 'pdev is the lowerdev
1087 * to use for the offload and 'dev' is the net device that will
1088 * back the offload. Returns a pointer to the private structure
1089 * the upper layer will maintain.
1090 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1091 * Called by upper layer device to delete the station created
1092 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1093 * the station and priv is the structure returned by the add
1095 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1096 * struct net_device *dev,
1098 * Callback to use for xmit over the accelerated station. This
1099 * is used in place of ndo_start_xmit on accelerated net
1101 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1102 * struct net_device *dev
1103 * netdev_features_t features);
1104 * Called by core transmit path to determine if device is capable of
1105 * performing offload operations on a given packet. This is to give
1106 * the device an opportunity to implement any restrictions that cannot
1107 * be otherwise expressed by feature flags. The check is called with
1108 * the set of features that the stack has calculated and it returns
1109 * those the driver believes to be appropriate.
1110 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1111 * int queue_index, u32 maxrate);
1112 * Called when a user wants to set a max-rate limitation of specific
1114 * int (*ndo_get_iflink)(const struct net_device *dev);
1115 * Called to get the iflink value of this device.
1116 * void (*ndo_change_proto_down)(struct net_device *dev,
1118 * This function is used to pass protocol port error state information
1119 * to the switch driver. The switch driver can react to the proto_down
1120 * by doing a phys down on the associated switch port.
1121 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1122 * This function is used to get egress tunnel information for given skb.
1123 * This is useful for retrieving outer tunnel header parameters while
1125 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1126 * This function is used to specify the headroom that the skb must
1127 * consider when allocation skb during packet reception. Setting
1128 * appropriate rx headroom value allows avoiding skb head copy on
1129 * forward. Setting a negative value resets the rx headroom to the
1131 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1132 * This function is used to set or query state related to XDP on the
1133 * netdevice. See definition of enum xdp_netdev_command for details.
1136 struct net_device_ops {
1137 int (*ndo_init)(struct net_device *dev);
1138 void (*ndo_uninit)(struct net_device *dev);
1139 int (*ndo_open)(struct net_device *dev);
1140 int (*ndo_stop)(struct net_device *dev);
1141 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1142 struct net_device *dev);
1143 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1144 struct net_device *dev,
1145 netdev_features_t features);
1146 u16 (*ndo_select_queue)(struct net_device *dev,
1147 struct sk_buff *skb,
1149 select_queue_fallback_t fallback);
1150 void (*ndo_change_rx_flags)(struct net_device *dev,
1152 void (*ndo_set_rx_mode)(struct net_device *dev);
1153 int (*ndo_set_mac_address)(struct net_device *dev,
1155 int (*ndo_validate_addr)(struct net_device *dev);
1156 int (*ndo_do_ioctl)(struct net_device *dev,
1157 struct ifreq *ifr, int cmd);
1158 int (*ndo_set_config)(struct net_device *dev,
1160 int (*ndo_change_mtu)(struct net_device *dev,
1162 int (*ndo_neigh_setup)(struct net_device *dev,
1163 struct neigh_parms *);
1164 void (*ndo_tx_timeout) (struct net_device *dev);
1166 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1167 struct rtnl_link_stats64 *storage);
1168 bool (*ndo_has_offload_stats)(int attr_id);
1169 int (*ndo_get_offload_stats)(int attr_id,
1170 const struct net_device *dev,
1172 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1175 __be16 proto, u16 vid);
1176 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1177 __be16 proto, u16 vid);
1178 #ifdef CONFIG_NET_POLL_CONTROLLER
1179 void (*ndo_poll_controller)(struct net_device *dev);
1180 int (*ndo_netpoll_setup)(struct net_device *dev,
1181 struct netpoll_info *info);
1182 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1184 #ifdef CONFIG_NET_RX_BUSY_POLL
1185 int (*ndo_busy_poll)(struct napi_struct *dev);
1187 int (*ndo_set_vf_mac)(struct net_device *dev,
1188 int queue, u8 *mac);
1189 int (*ndo_set_vf_vlan)(struct net_device *dev,
1190 int queue, u16 vlan,
1191 u8 qos, __be16 proto);
1192 int (*ndo_set_vf_rate)(struct net_device *dev,
1193 int vf, int min_tx_rate,
1195 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1196 int vf, bool setting);
1197 int (*ndo_set_vf_trust)(struct net_device *dev,
1198 int vf, bool setting);
1199 int (*ndo_get_vf_config)(struct net_device *dev,
1201 struct ifla_vf_info *ivf);
1202 int (*ndo_set_vf_link_state)(struct net_device *dev,
1203 int vf, int link_state);
1204 int (*ndo_get_vf_stats)(struct net_device *dev,
1206 struct ifla_vf_stats
1208 int (*ndo_set_vf_port)(struct net_device *dev,
1210 struct nlattr *port[]);
1211 int (*ndo_get_vf_port)(struct net_device *dev,
1212 int vf, struct sk_buff *skb);
1213 int (*ndo_set_vf_guid)(struct net_device *dev,
1216 int (*ndo_set_vf_rss_query_en)(
1217 struct net_device *dev,
1218 int vf, bool setting);
1219 int (*ndo_setup_tc)(struct net_device *dev,
1222 struct tc_to_netdev *tc);
1223 #if IS_ENABLED(CONFIG_FCOE)
1224 int (*ndo_fcoe_enable)(struct net_device *dev);
1225 int (*ndo_fcoe_disable)(struct net_device *dev);
1226 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1228 struct scatterlist *sgl,
1230 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1232 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1234 struct scatterlist *sgl,
1236 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1237 struct netdev_fcoe_hbainfo *hbainfo);
1240 #if IS_ENABLED(CONFIG_LIBFCOE)
1241 #define NETDEV_FCOE_WWNN 0
1242 #define NETDEV_FCOE_WWPN 1
1243 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1244 u64 *wwn, int type);
1247 #ifdef CONFIG_RFS_ACCEL
1248 int (*ndo_rx_flow_steer)(struct net_device *dev,
1249 const struct sk_buff *skb,
1253 int (*ndo_add_slave)(struct net_device *dev,
1254 struct net_device *slave_dev);
1255 int (*ndo_del_slave)(struct net_device *dev,
1256 struct net_device *slave_dev);
1257 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1258 netdev_features_t features);
1259 int (*ndo_set_features)(struct net_device *dev,
1260 netdev_features_t features);
1261 int (*ndo_neigh_construct)(struct net_device *dev,
1262 struct neighbour *n);
1263 void (*ndo_neigh_destroy)(struct net_device *dev,
1264 struct neighbour *n);
1266 int (*ndo_fdb_add)(struct ndmsg *ndm,
1267 struct nlattr *tb[],
1268 struct net_device *dev,
1269 const unsigned char *addr,
1272 int (*ndo_fdb_del)(struct ndmsg *ndm,
1273 struct nlattr *tb[],
1274 struct net_device *dev,
1275 const unsigned char *addr,
1277 int (*ndo_fdb_dump)(struct sk_buff *skb,
1278 struct netlink_callback *cb,
1279 struct net_device *dev,
1280 struct net_device *filter_dev,
1283 int (*ndo_bridge_setlink)(struct net_device *dev,
1284 struct nlmsghdr *nlh,
1286 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1288 struct net_device *dev,
1291 int (*ndo_bridge_dellink)(struct net_device *dev,
1292 struct nlmsghdr *nlh,
1294 int (*ndo_change_carrier)(struct net_device *dev,
1296 int (*ndo_get_phys_port_id)(struct net_device *dev,
1297 struct netdev_phys_item_id *ppid);
1298 int (*ndo_get_phys_port_name)(struct net_device *dev,
1299 char *name, size_t len);
1300 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1301 struct udp_tunnel_info *ti);
1302 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1303 struct udp_tunnel_info *ti);
1304 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1305 struct net_device *dev);
1306 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1309 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1310 struct net_device *dev,
1312 int (*ndo_get_lock_subclass)(struct net_device *dev);
1313 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1316 int (*ndo_get_iflink)(const struct net_device *dev);
1317 int (*ndo_change_proto_down)(struct net_device *dev,
1319 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1320 struct sk_buff *skb);
1321 void (*ndo_set_rx_headroom)(struct net_device *dev,
1322 int needed_headroom);
1323 int (*ndo_xdp)(struct net_device *dev,
1324 struct netdev_xdp *xdp);
1328 * enum net_device_priv_flags - &struct net_device priv_flags
1330 * These are the &struct net_device, they are only set internally
1331 * by drivers and used in the kernel. These flags are invisible to
1332 * userspace; this means that the order of these flags can change
1333 * during any kernel release.
1335 * You should have a pretty good reason to be extending these flags.
1337 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1338 * @IFF_EBRIDGE: Ethernet bridging device
1339 * @IFF_BONDING: bonding master or slave
1340 * @IFF_ISATAP: ISATAP interface (RFC4214)
1341 * @IFF_WAN_HDLC: WAN HDLC device
1342 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1344 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1345 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1346 * @IFF_MACVLAN_PORT: device used as macvlan port
1347 * @IFF_BRIDGE_PORT: device used as bridge port
1348 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1349 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1350 * @IFF_UNICAST_FLT: Supports unicast filtering
1351 * @IFF_TEAM_PORT: device used as team port
1352 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1353 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1354 * change when it's running
1355 * @IFF_MACVLAN: Macvlan device
1356 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1357 * underlying stacked devices
1358 * @IFF_IPVLAN_MASTER: IPvlan master device
1359 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1360 * @IFF_L3MDEV_MASTER: device is an L3 master device
1361 * @IFF_NO_QUEUE: device can run without qdisc attached
1362 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1363 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1364 * @IFF_TEAM: device is a team device
1365 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1366 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1367 * entity (i.e. the master device for bridged veth)
1368 * @IFF_MACSEC: device is a MACsec device
1370 enum netdev_priv_flags {
1371 IFF_802_1Q_VLAN = 1<<0,
1375 IFF_WAN_HDLC = 1<<4,
1376 IFF_XMIT_DST_RELEASE = 1<<5,
1377 IFF_DONT_BRIDGE = 1<<6,
1378 IFF_DISABLE_NETPOLL = 1<<7,
1379 IFF_MACVLAN_PORT = 1<<8,
1380 IFF_BRIDGE_PORT = 1<<9,
1381 IFF_OVS_DATAPATH = 1<<10,
1382 IFF_TX_SKB_SHARING = 1<<11,
1383 IFF_UNICAST_FLT = 1<<12,
1384 IFF_TEAM_PORT = 1<<13,
1385 IFF_SUPP_NOFCS = 1<<14,
1386 IFF_LIVE_ADDR_CHANGE = 1<<15,
1387 IFF_MACVLAN = 1<<16,
1388 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1389 IFF_IPVLAN_MASTER = 1<<18,
1390 IFF_IPVLAN_SLAVE = 1<<19,
1391 IFF_L3MDEV_MASTER = 1<<20,
1392 IFF_NO_QUEUE = 1<<21,
1393 IFF_OPENVSWITCH = 1<<22,
1394 IFF_L3MDEV_SLAVE = 1<<23,
1396 IFF_RXFH_CONFIGURED = 1<<25,
1397 IFF_PHONY_HEADROOM = 1<<26,
1401 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1402 #define IFF_EBRIDGE IFF_EBRIDGE
1403 #define IFF_BONDING IFF_BONDING
1404 #define IFF_ISATAP IFF_ISATAP
1405 #define IFF_WAN_HDLC IFF_WAN_HDLC
1406 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1407 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1408 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1409 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1410 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1411 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1412 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1413 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1414 #define IFF_TEAM_PORT IFF_TEAM_PORT
1415 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1416 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1417 #define IFF_MACVLAN IFF_MACVLAN
1418 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1419 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1420 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1421 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1422 #define IFF_NO_QUEUE IFF_NO_QUEUE
1423 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1424 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1425 #define IFF_TEAM IFF_TEAM
1426 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1427 #define IFF_MACSEC IFF_MACSEC
1430 * struct net_device - The DEVICE structure.
1431 * Actually, this whole structure is a big mistake. It mixes I/O
1432 * data with strictly "high-level" data, and it has to know about
1433 * almost every data structure used in the INET module.
1435 * @name: This is the first field of the "visible" part of this structure
1436 * (i.e. as seen by users in the "Space.c" file). It is the name
1439 * @name_hlist: Device name hash chain, please keep it close to name[]
1440 * @ifalias: SNMP alias
1441 * @mem_end: Shared memory end
1442 * @mem_start: Shared memory start
1443 * @base_addr: Device I/O address
1444 * @irq: Device IRQ number
1446 * @carrier_changes: Stats to monitor carrier on<->off transitions
1448 * @state: Generic network queuing layer state, see netdev_state_t
1449 * @dev_list: The global list of network devices
1450 * @napi_list: List entry used for polling NAPI devices
1451 * @unreg_list: List entry when we are unregistering the
1452 * device; see the function unregister_netdev
1453 * @close_list: List entry used when we are closing the device
1454 * @ptype_all: Device-specific packet handlers for all protocols
1455 * @ptype_specific: Device-specific, protocol-specific packet handlers
1457 * @adj_list: Directly linked devices, like slaves for bonding
1458 * @features: Currently active device features
1459 * @hw_features: User-changeable features
1461 * @wanted_features: User-requested features
1462 * @vlan_features: Mask of features inheritable by VLAN devices
1464 * @hw_enc_features: Mask of features inherited by encapsulating devices
1465 * This field indicates what encapsulation
1466 * offloads the hardware is capable of doing,
1467 * and drivers will need to set them appropriately.
1469 * @mpls_features: Mask of features inheritable by MPLS
1471 * @ifindex: interface index
1472 * @group: The group the device belongs to
1474 * @stats: Statistics struct, which was left as a legacy, use
1475 * rtnl_link_stats64 instead
1477 * @rx_dropped: Dropped packets by core network,
1478 * do not use this in drivers
1479 * @tx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
1481 * @rx_nohandler: nohandler dropped packets by core network on
1482 * inactive devices, do not use this in drivers
1484 * @wireless_handlers: List of functions to handle Wireless Extensions,
1486 * see <net/iw_handler.h> for details.
1487 * @wireless_data: Instance data managed by the core of wireless extensions
1489 * @netdev_ops: Includes several pointers to callbacks,
1490 * if one wants to override the ndo_*() functions
1491 * @ethtool_ops: Management operations
1492 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1493 * discovery handling. Necessary for e.g. 6LoWPAN.
1494 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1495 * of Layer 2 headers.
1497 * @flags: Interface flags (a la BSD)
1498 * @priv_flags: Like 'flags' but invisible to userspace,
1499 * see if.h for the definitions
1500 * @gflags: Global flags ( kept as legacy )
1501 * @padded: How much padding added by alloc_netdev()
1502 * @operstate: RFC2863 operstate
1503 * @link_mode: Mapping policy to operstate
1504 * @if_port: Selectable AUI, TP, ...
1506 * @mtu: Interface MTU value
1507 * @min_mtu: Interface Minimum MTU value
1508 * @max_mtu: Interface Maximum MTU value
1509 * @type: Interface hardware type
1510 * @hard_header_len: Maximum hardware header length.
1512 * @needed_headroom: Extra headroom the hardware may need, but not in all
1513 * cases can this be guaranteed
1514 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1515 * cases can this be guaranteed. Some cases also use
1516 * LL_MAX_HEADER instead to allocate the skb
1518 * interface address info:
1520 * @perm_addr: Permanent hw address
1521 * @addr_assign_type: Hw address assignment type
1522 * @addr_len: Hardware address length
1523 * @neigh_priv_len: Used in neigh_alloc()
1524 * @dev_id: Used to differentiate devices that share
1525 * the same link layer address
1526 * @dev_port: Used to differentiate devices that share
1528 * @addr_list_lock: XXX: need comments on this one
1529 * @uc_promisc: Counter that indicates promiscuous mode
1530 * has been enabled due to the need to listen to
1531 * additional unicast addresses in a device that
1532 * does not implement ndo_set_rx_mode()
1533 * @uc: unicast mac addresses
1534 * @mc: multicast mac addresses
1535 * @dev_addrs: list of device hw addresses
1536 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1537 * @promiscuity: Number of times the NIC is told to work in
1538 * promiscuous mode; if it becomes 0 the NIC will
1539 * exit promiscuous mode
1540 * @allmulti: Counter, enables or disables allmulticast mode
1542 * @vlan_info: VLAN info
1543 * @dsa_ptr: dsa specific data
1544 * @tipc_ptr: TIPC specific data
1545 * @atalk_ptr: AppleTalk link
1546 * @ip_ptr: IPv4 specific data
1547 * @dn_ptr: DECnet specific data
1548 * @ip6_ptr: IPv6 specific data
1549 * @ax25_ptr: AX.25 specific data
1550 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1552 * @last_rx: Time of last Rx
1553 * @dev_addr: Hw address (before bcast,
1554 * because most packets are unicast)
1556 * @_rx: Array of RX queues
1557 * @num_rx_queues: Number of RX queues
1558 * allocated at register_netdev() time
1559 * @real_num_rx_queues: Number of RX queues currently active in device
1561 * @rx_handler: handler for received packets
1562 * @rx_handler_data: XXX: need comments on this one
1563 * @ingress_queue: XXX: need comments on this one
1564 * @broadcast: hw bcast address
1566 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1567 * indexed by RX queue number. Assigned by driver.
1568 * This must only be set if the ndo_rx_flow_steer
1569 * operation is defined
1570 * @index_hlist: Device index hash chain
1572 * @_tx: Array of TX queues
1573 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1574 * @real_num_tx_queues: Number of TX queues currently active in device
1575 * @qdisc: Root qdisc from userspace point of view
1576 * @tx_queue_len: Max frames per queue allowed
1577 * @tx_global_lock: XXX: need comments on this one
1579 * @xps_maps: XXX: need comments on this one
1581 * @watchdog_timeo: Represents the timeout that is used by
1582 * the watchdog (see dev_watchdog())
1583 * @watchdog_timer: List of timers
1585 * @pcpu_refcnt: Number of references to this device
1586 * @todo_list: Delayed register/unregister
1587 * @link_watch_list: XXX: need comments on this one
1589 * @reg_state: Register/unregister state machine
1590 * @dismantle: Device is going to be freed
1591 * @rtnl_link_state: This enum represents the phases of creating
1594 * @destructor: Called from unregister,
1595 * can be used to call free_netdev
1596 * @npinfo: XXX: need comments on this one
1597 * @nd_net: Network namespace this network device is inside
1599 * @ml_priv: Mid-layer private
1600 * @lstats: Loopback statistics
1601 * @tstats: Tunnel statistics
1602 * @dstats: Dummy statistics
1603 * @vstats: Virtual ethernet statistics
1608 * @dev: Class/net/name entry
1609 * @sysfs_groups: Space for optional device, statistics and wireless
1612 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1613 * @rtnl_link_ops: Rtnl_link_ops
1615 * @gso_max_size: Maximum size of generic segmentation offload
1616 * @gso_max_segs: Maximum number of segments that can be passed to the
1619 * @dcbnl_ops: Data Center Bridging netlink ops
1620 * @num_tc: Number of traffic classes in the net device
1621 * @tc_to_txq: XXX: need comments on this one
1622 * @prio_tc_map XXX: need comments on this one
1624 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1626 * @priomap: XXX: need comments on this one
1627 * @phydev: Physical device may attach itself
1628 * for hardware timestamping
1630 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1631 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1633 * @proto_down: protocol port state information can be sent to the
1634 * switch driver and used to set the phys state of the
1637 * FIXME: cleanup struct net_device such that network protocol info
1642 char name[IFNAMSIZ];
1643 struct hlist_node name_hlist;
1646 * I/O specific fields
1647 * FIXME: Merge these and struct ifmap into one
1649 unsigned long mem_end;
1650 unsigned long mem_start;
1651 unsigned long base_addr;
1654 atomic_t carrier_changes;
1657 * Some hardware also needs these fields (state,dev_list,
1658 * napi_list,unreg_list,close_list) but they are not
1659 * part of the usual set specified in Space.c.
1662 unsigned long state;
1664 struct list_head dev_list;
1665 struct list_head napi_list;
1666 struct list_head unreg_list;
1667 struct list_head close_list;
1668 struct list_head ptype_all;
1669 struct list_head ptype_specific;
1672 struct list_head upper;
1673 struct list_head lower;
1676 netdev_features_t features;
1677 netdev_features_t hw_features;
1678 netdev_features_t wanted_features;
1679 netdev_features_t vlan_features;
1680 netdev_features_t hw_enc_features;
1681 netdev_features_t mpls_features;
1682 netdev_features_t gso_partial_features;
1687 struct net_device_stats stats;
1689 atomic_long_t rx_dropped;
1690 atomic_long_t tx_dropped;
1691 atomic_long_t rx_nohandler;
1693 #ifdef CONFIG_WIRELESS_EXT
1694 const struct iw_handler_def *wireless_handlers;
1695 struct iw_public_data *wireless_data;
1697 const struct net_device_ops *netdev_ops;
1698 const struct ethtool_ops *ethtool_ops;
1699 #ifdef CONFIG_NET_SWITCHDEV
1700 const struct switchdev_ops *switchdev_ops;
1702 #ifdef CONFIG_NET_L3_MASTER_DEV
1703 const struct l3mdev_ops *l3mdev_ops;
1705 #if IS_ENABLED(CONFIG_IPV6)
1706 const struct ndisc_ops *ndisc_ops;
1709 const struct header_ops *header_ops;
1712 unsigned int priv_flags;
1714 unsigned short gflags;
1715 unsigned short padded;
1717 unsigned char operstate;
1718 unsigned char link_mode;
1720 unsigned char if_port;
1724 unsigned int min_mtu;
1725 unsigned int max_mtu;
1726 unsigned short type;
1727 unsigned short hard_header_len;
1729 unsigned short needed_headroom;
1730 unsigned short needed_tailroom;
1732 /* Interface address info. */
1733 unsigned char perm_addr[MAX_ADDR_LEN];
1734 unsigned char addr_assign_type;
1735 unsigned char addr_len;
1736 unsigned short neigh_priv_len;
1737 unsigned short dev_id;
1738 unsigned short dev_port;
1739 spinlock_t addr_list_lock;
1740 unsigned char name_assign_type;
1742 struct netdev_hw_addr_list uc;
1743 struct netdev_hw_addr_list mc;
1744 struct netdev_hw_addr_list dev_addrs;
1747 struct kset *queues_kset;
1749 unsigned int promiscuity;
1750 unsigned int allmulti;
1753 /* Protocol-specific pointers */
1755 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1756 struct vlan_info __rcu *vlan_info;
1758 #if IS_ENABLED(CONFIG_NET_DSA)
1759 struct dsa_switch_tree *dsa_ptr;
1761 #if IS_ENABLED(CONFIG_TIPC)
1762 struct tipc_bearer __rcu *tipc_ptr;
1765 struct in_device __rcu *ip_ptr;
1766 struct dn_dev __rcu *dn_ptr;
1767 struct inet6_dev __rcu *ip6_ptr;
1769 struct wireless_dev *ieee80211_ptr;
1770 struct wpan_dev *ieee802154_ptr;
1771 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1772 struct mpls_dev __rcu *mpls_ptr;
1776 * Cache lines mostly used on receive path (including eth_type_trans())
1778 unsigned long last_rx;
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_header_slow(struct sk_buff *skb, unsigned int hlen,
2479 unsigned int offset)
2481 if (!pskb_may_pull(skb, hlen))
2484 NAPI_GRO_CB(skb)->frag0 = NULL;
2485 NAPI_GRO_CB(skb)->frag0_len = 0;
2486 return skb->data + offset;
2489 static inline void *skb_gro_network_header(struct sk_buff *skb)
2491 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2492 skb_network_offset(skb);
2495 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2496 const void *start, unsigned int len)
2498 if (NAPI_GRO_CB(skb)->csum_valid)
2499 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2500 csum_partial(start, len, 0));
2503 /* GRO checksum functions. These are logical equivalents of the normal
2504 * checksum functions (in skbuff.h) except that they operate on the GRO
2505 * offsets and fields in sk_buff.
2508 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2510 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2512 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2515 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2519 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2520 skb_checksum_start_offset(skb) <
2521 skb_gro_offset(skb)) &&
2522 !skb_at_gro_remcsum_start(skb) &&
2523 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2524 (!zero_okay || check));
2527 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2530 if (NAPI_GRO_CB(skb)->csum_valid &&
2531 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2534 NAPI_GRO_CB(skb)->csum = psum;
2536 return __skb_gro_checksum_complete(skb);
2539 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2541 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2542 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2543 NAPI_GRO_CB(skb)->csum_cnt--;
2545 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2546 * verified a new top level checksum or an encapsulated one
2547 * during GRO. This saves work if we fallback to normal path.
2549 __skb_incr_checksum_unnecessary(skb);
2553 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2556 __sum16 __ret = 0; \
2557 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2558 __ret = __skb_gro_checksum_validate_complete(skb, \
2559 compute_pseudo(skb, proto)); \
2561 __skb_mark_checksum_bad(skb); \
2563 skb_gro_incr_csum_unnecessary(skb); \
2567 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2568 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2570 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2572 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2574 #define skb_gro_checksum_simple_validate(skb) \
2575 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2577 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2579 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2580 !NAPI_GRO_CB(skb)->csum_valid);
2583 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2584 __sum16 check, __wsum pseudo)
2586 NAPI_GRO_CB(skb)->csum = ~pseudo;
2587 NAPI_GRO_CB(skb)->csum_valid = 1;
2590 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2592 if (__skb_gro_checksum_convert_check(skb)) \
2593 __skb_gro_checksum_convert(skb, check, \
2594 compute_pseudo(skb, proto)); \
2597 struct gro_remcsum {
2602 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2608 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2609 unsigned int off, size_t hdrlen,
2610 int start, int offset,
2611 struct gro_remcsum *grc,
2615 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2617 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2620 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2624 ptr = skb_gro_header_fast(skb, off);
2625 if (skb_gro_header_hard(skb, off + plen)) {
2626 ptr = skb_gro_header_slow(skb, off + plen, off);
2631 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2634 /* Adjust skb->csum since we changed the packet */
2635 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2637 grc->offset = off + hdrlen + offset;
2643 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2644 struct gro_remcsum *grc)
2647 size_t plen = grc->offset + sizeof(u16);
2652 ptr = skb_gro_header_fast(skb, grc->offset);
2653 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2654 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2659 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2662 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2663 unsigned short type,
2664 const void *daddr, const void *saddr,
2667 if (!dev->header_ops || !dev->header_ops->create)
2670 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2673 static inline int dev_parse_header(const struct sk_buff *skb,
2674 unsigned char *haddr)
2676 const struct net_device *dev = skb->dev;
2678 if (!dev->header_ops || !dev->header_ops->parse)
2680 return dev->header_ops->parse(skb, haddr);
2683 /* ll_header must have at least hard_header_len allocated */
2684 static inline bool dev_validate_header(const struct net_device *dev,
2685 char *ll_header, int len)
2687 if (likely(len >= dev->hard_header_len))
2690 if (capable(CAP_SYS_RAWIO)) {
2691 memset(ll_header + len, 0, dev->hard_header_len - len);
2695 if (dev->header_ops && dev->header_ops->validate)
2696 return dev->header_ops->validate(ll_header, len);
2701 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2702 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2703 static inline int unregister_gifconf(unsigned int family)
2705 return register_gifconf(family, NULL);
2708 #ifdef CONFIG_NET_FLOW_LIMIT
2709 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2710 struct sd_flow_limit {
2712 unsigned int num_buckets;
2713 unsigned int history_head;
2714 u16 history[FLOW_LIMIT_HISTORY];
2718 extern int netdev_flow_limit_table_len;
2719 #endif /* CONFIG_NET_FLOW_LIMIT */
2722 * Incoming packets are placed on per-CPU queues
2724 struct softnet_data {
2725 struct list_head poll_list;
2726 struct sk_buff_head process_queue;
2729 unsigned int processed;
2730 unsigned int time_squeeze;
2731 unsigned int received_rps;
2733 struct softnet_data *rps_ipi_list;
2735 #ifdef CONFIG_NET_FLOW_LIMIT
2736 struct sd_flow_limit __rcu *flow_limit;
2738 struct Qdisc *output_queue;
2739 struct Qdisc **output_queue_tailp;
2740 struct sk_buff *completion_queue;
2743 /* input_queue_head should be written by cpu owning this struct,
2744 * and only read by other cpus. Worth using a cache line.
2746 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2748 /* Elements below can be accessed between CPUs for RPS/RFS */
2749 struct call_single_data csd ____cacheline_aligned_in_smp;
2750 struct softnet_data *rps_ipi_next;
2752 unsigned int input_queue_tail;
2754 unsigned int dropped;
2755 struct sk_buff_head input_pkt_queue;
2756 struct napi_struct backlog;
2760 static inline void input_queue_head_incr(struct softnet_data *sd)
2763 sd->input_queue_head++;
2767 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2768 unsigned int *qtail)
2771 *qtail = ++sd->input_queue_tail;
2775 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2777 void __netif_schedule(struct Qdisc *q);
2778 void netif_schedule_queue(struct netdev_queue *txq);
2780 static inline void netif_tx_schedule_all(struct net_device *dev)
2784 for (i = 0; i < dev->num_tx_queues; i++)
2785 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2788 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2790 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2794 * netif_start_queue - allow transmit
2795 * @dev: network device
2797 * Allow upper layers to call the device hard_start_xmit routine.
2799 static inline void netif_start_queue(struct net_device *dev)
2801 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2804 static inline void netif_tx_start_all_queues(struct net_device *dev)
2808 for (i = 0; i < dev->num_tx_queues; i++) {
2809 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2810 netif_tx_start_queue(txq);
2814 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2817 * netif_wake_queue - restart transmit
2818 * @dev: network device
2820 * Allow upper layers to call the device hard_start_xmit routine.
2821 * Used for flow control when transmit resources are available.
2823 static inline void netif_wake_queue(struct net_device *dev)
2825 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2828 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2832 for (i = 0; i < dev->num_tx_queues; i++) {
2833 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2834 netif_tx_wake_queue(txq);
2838 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2840 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2844 * netif_stop_queue - stop transmitted packets
2845 * @dev: network device
2847 * Stop upper layers calling the device hard_start_xmit routine.
2848 * Used for flow control when transmit resources are unavailable.
2850 static inline void netif_stop_queue(struct net_device *dev)
2852 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2855 void netif_tx_stop_all_queues(struct net_device *dev);
2857 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2859 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2863 * netif_queue_stopped - test if transmit queue is flowblocked
2864 * @dev: network device
2866 * Test if transmit queue on device is currently unable to send.
2868 static inline bool netif_queue_stopped(const struct net_device *dev)
2870 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2873 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2875 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2879 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2881 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2885 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2887 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2891 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2892 * @dev_queue: pointer to transmit queue
2894 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2895 * to give appropriate hint to the CPU.
2897 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2900 prefetchw(&dev_queue->dql.num_queued);
2905 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2906 * @dev_queue: pointer to transmit queue
2908 * BQL enabled drivers might use this helper in their TX completion path,
2909 * to give appropriate hint to the CPU.
2911 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2914 prefetchw(&dev_queue->dql.limit);
2918 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2922 dql_queued(&dev_queue->dql, bytes);
2924 if (likely(dql_avail(&dev_queue->dql) >= 0))
2927 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2930 * The XOFF flag must be set before checking the dql_avail below,
2931 * because in netdev_tx_completed_queue we update the dql_completed
2932 * before checking the XOFF flag.
2936 /* check again in case another CPU has just made room avail */
2937 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2938 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2943 * netdev_sent_queue - report the number of bytes queued to hardware
2944 * @dev: network device
2945 * @bytes: number of bytes queued to the hardware device queue
2947 * Report the number of bytes queued for sending/completion to the network
2948 * device hardware queue. @bytes should be a good approximation and should
2949 * exactly match netdev_completed_queue() @bytes
2951 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2953 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2956 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2957 unsigned int pkts, unsigned int bytes)
2960 if (unlikely(!bytes))
2963 dql_completed(&dev_queue->dql, bytes);
2966 * Without the memory barrier there is a small possiblity that
2967 * netdev_tx_sent_queue will miss the update and cause the queue to
2968 * be stopped forever
2972 if (dql_avail(&dev_queue->dql) < 0)
2975 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2976 netif_schedule_queue(dev_queue);
2981 * netdev_completed_queue - report bytes and packets completed by device
2982 * @dev: network device
2983 * @pkts: actual number of packets sent over the medium
2984 * @bytes: actual number of bytes sent over the medium
2986 * Report the number of bytes and packets transmitted by the network device
2987 * hardware queue over the physical medium, @bytes must exactly match the
2988 * @bytes amount passed to netdev_sent_queue()
2990 static inline void netdev_completed_queue(struct net_device *dev,
2991 unsigned int pkts, unsigned int bytes)
2993 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2996 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2999 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3005 * netdev_reset_queue - reset the packets and bytes count of a network device
3006 * @dev_queue: network device
3008 * Reset the bytes and packet count of a network device and clear the
3009 * software flow control OFF bit for this network device
3011 static inline void netdev_reset_queue(struct net_device *dev_queue)
3013 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3017 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3018 * @dev: network device
3019 * @queue_index: given tx queue index
3021 * Returns 0 if given tx queue index >= number of device tx queues,
3022 * otherwise returns the originally passed tx queue index.
3024 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3026 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3027 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3028 dev->name, queue_index,
3029 dev->real_num_tx_queues);
3037 * netif_running - test if up
3038 * @dev: network device
3040 * Test if the device has been brought up.
3042 static inline bool netif_running(const struct net_device *dev)
3044 return test_bit(__LINK_STATE_START, &dev->state);
3048 * Routines to manage the subqueues on a device. We only need start,
3049 * stop, and a check if it's stopped. All other device management is
3050 * done at the overall netdevice level.
3051 * Also test the device if we're multiqueue.
3055 * netif_start_subqueue - allow sending packets on subqueue
3056 * @dev: network device
3057 * @queue_index: sub queue index
3059 * Start individual transmit queue of a device with multiple transmit queues.
3061 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3063 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3065 netif_tx_start_queue(txq);
3069 * netif_stop_subqueue - stop sending packets on subqueue
3070 * @dev: network device
3071 * @queue_index: sub queue index
3073 * Stop individual transmit queue of a device with multiple transmit queues.
3075 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3077 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3078 netif_tx_stop_queue(txq);
3082 * netif_subqueue_stopped - test status of subqueue
3083 * @dev: network device
3084 * @queue_index: sub queue index
3086 * Check individual transmit queue of a device with multiple transmit queues.
3088 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3091 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3093 return netif_tx_queue_stopped(txq);
3096 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3097 struct sk_buff *skb)
3099 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3102 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3105 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3108 static inline int netif_set_xps_queue(struct net_device *dev,
3109 const struct cpumask *mask,
3116 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3117 unsigned int num_tx_queues);
3120 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3121 * as a distribution range limit for the returned value.
3123 static inline u16 skb_tx_hash(const struct net_device *dev,
3124 struct sk_buff *skb)
3126 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3130 * netif_is_multiqueue - test if device has multiple transmit queues
3131 * @dev: network device
3133 * Check if device has multiple transmit queues
3135 static inline bool netif_is_multiqueue(const struct net_device *dev)
3137 return dev->num_tx_queues > 1;
3140 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3143 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3145 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3153 static inline unsigned int get_netdev_rx_queue_index(
3154 struct netdev_rx_queue *queue)
3156 struct net_device *dev = queue->dev;
3157 int index = queue - dev->_rx;
3159 BUG_ON(index >= dev->num_rx_queues);
3164 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3165 int netif_get_num_default_rss_queues(void);
3167 enum skb_free_reason {
3168 SKB_REASON_CONSUMED,
3172 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3173 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3176 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3177 * interrupt context or with hardware interrupts being disabled.
3178 * (in_irq() || irqs_disabled())
3180 * We provide four helpers that can be used in following contexts :
3182 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3183 * replacing kfree_skb(skb)
3185 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3186 * Typically used in place of consume_skb(skb) in TX completion path
3188 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3189 * replacing kfree_skb(skb)
3191 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3192 * and consumed a packet. Used in place of consume_skb(skb)
3194 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3196 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3199 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3201 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3204 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3206 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3209 static inline void dev_consume_skb_any(struct sk_buff *skb)
3211 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3214 int netif_rx(struct sk_buff *skb);
3215 int netif_rx_ni(struct sk_buff *skb);
3216 int netif_receive_skb(struct sk_buff *skb);
3217 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3218 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3219 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3220 gro_result_t napi_gro_frags(struct napi_struct *napi);
3221 struct packet_offload *gro_find_receive_by_type(__be16 type);
3222 struct packet_offload *gro_find_complete_by_type(__be16 type);
3224 static inline void napi_free_frags(struct napi_struct *napi)
3226 kfree_skb(napi->skb);
3230 bool netdev_is_rx_handler_busy(struct net_device *dev);
3231 int netdev_rx_handler_register(struct net_device *dev,
3232 rx_handler_func_t *rx_handler,
3233 void *rx_handler_data);
3234 void netdev_rx_handler_unregister(struct net_device *dev);
3236 bool dev_valid_name(const char *name);
3237 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3238 int dev_ethtool(struct net *net, struct ifreq *);
3239 unsigned int dev_get_flags(const struct net_device *);
3240 int __dev_change_flags(struct net_device *, unsigned int flags);
3241 int dev_change_flags(struct net_device *, unsigned int);
3242 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3243 unsigned int gchanges);
3244 int dev_change_name(struct net_device *, const char *);
3245 int dev_set_alias(struct net_device *, const char *, size_t);
3246 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3247 int dev_set_mtu(struct net_device *, int);
3248 void dev_set_group(struct net_device *, int);
3249 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3250 int dev_change_carrier(struct net_device *, bool new_carrier);
3251 int dev_get_phys_port_id(struct net_device *dev,
3252 struct netdev_phys_item_id *ppid);
3253 int dev_get_phys_port_name(struct net_device *dev,
3254 char *name, size_t len);
3255 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3256 int dev_change_xdp_fd(struct net_device *dev, int fd);
3257 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3258 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3259 struct netdev_queue *txq, int *ret);
3260 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3261 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3262 bool is_skb_forwardable(const struct net_device *dev,
3263 const struct sk_buff *skb);
3265 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3266 struct sk_buff *skb)
3268 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3269 unlikely(!is_skb_forwardable(dev, skb))) {
3270 atomic_long_inc(&dev->rx_dropped);
3275 skb_scrub_packet(skb, true);
3280 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3282 extern int netdev_budget;
3284 /* Called by rtnetlink.c:rtnl_unlock() */
3285 void netdev_run_todo(void);
3288 * dev_put - release reference to device
3289 * @dev: network device
3291 * Release reference to device to allow it to be freed.
3293 static inline void dev_put(struct net_device *dev)
3295 this_cpu_dec(*dev->pcpu_refcnt);
3299 * dev_hold - get reference to device
3300 * @dev: network device
3302 * Hold reference to device to keep it from being freed.
3304 static inline void dev_hold(struct net_device *dev)
3306 this_cpu_inc(*dev->pcpu_refcnt);
3309 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3310 * and _off may be called from IRQ context, but it is caller
3311 * who is responsible for serialization of these calls.
3313 * The name carrier is inappropriate, these functions should really be
3314 * called netif_lowerlayer_*() because they represent the state of any
3315 * kind of lower layer not just hardware media.
3318 void linkwatch_init_dev(struct net_device *dev);
3319 void linkwatch_fire_event(struct net_device *dev);
3320 void linkwatch_forget_dev(struct net_device *dev);
3323 * netif_carrier_ok - test if carrier present
3324 * @dev: network device
3326 * Check if carrier is present on device
3328 static inline bool netif_carrier_ok(const struct net_device *dev)
3330 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3333 unsigned long dev_trans_start(struct net_device *dev);
3335 void __netdev_watchdog_up(struct net_device *dev);
3337 void netif_carrier_on(struct net_device *dev);
3339 void netif_carrier_off(struct net_device *dev);
3342 * netif_dormant_on - mark device as dormant.
3343 * @dev: network device
3345 * Mark device as dormant (as per RFC2863).
3347 * The dormant state indicates that the relevant interface is not
3348 * actually in a condition to pass packets (i.e., it is not 'up') but is
3349 * in a "pending" state, waiting for some external event. For "on-
3350 * demand" interfaces, this new state identifies the situation where the
3351 * interface is waiting for events to place it in the up state.
3353 static inline void netif_dormant_on(struct net_device *dev)
3355 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3356 linkwatch_fire_event(dev);
3360 * netif_dormant_off - set device as not dormant.
3361 * @dev: network device
3363 * Device is not in dormant state.
3365 static inline void netif_dormant_off(struct net_device *dev)
3367 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3368 linkwatch_fire_event(dev);
3372 * netif_dormant - test if carrier present
3373 * @dev: network device
3375 * Check if carrier is present on device
3377 static inline bool netif_dormant(const struct net_device *dev)
3379 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3384 * netif_oper_up - test if device is operational
3385 * @dev: network device
3387 * Check if carrier is operational
3389 static inline bool netif_oper_up(const struct net_device *dev)
3391 return (dev->operstate == IF_OPER_UP ||
3392 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3396 * netif_device_present - is device available or removed
3397 * @dev: network device
3399 * Check if device has not been removed from system.
3401 static inline bool netif_device_present(struct net_device *dev)
3403 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3406 void netif_device_detach(struct net_device *dev);
3408 void netif_device_attach(struct net_device *dev);
3411 * Network interface message level settings
3415 NETIF_MSG_DRV = 0x0001,
3416 NETIF_MSG_PROBE = 0x0002,
3417 NETIF_MSG_LINK = 0x0004,
3418 NETIF_MSG_TIMER = 0x0008,
3419 NETIF_MSG_IFDOWN = 0x0010,
3420 NETIF_MSG_IFUP = 0x0020,
3421 NETIF_MSG_RX_ERR = 0x0040,
3422 NETIF_MSG_TX_ERR = 0x0080,
3423 NETIF_MSG_TX_QUEUED = 0x0100,
3424 NETIF_MSG_INTR = 0x0200,
3425 NETIF_MSG_TX_DONE = 0x0400,
3426 NETIF_MSG_RX_STATUS = 0x0800,
3427 NETIF_MSG_PKTDATA = 0x1000,
3428 NETIF_MSG_HW = 0x2000,
3429 NETIF_MSG_WOL = 0x4000,
3432 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3433 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3434 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3435 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3436 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3437 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3438 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3439 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3440 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3441 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3442 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3443 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3444 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3445 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3446 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3448 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3451 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3452 return default_msg_enable_bits;
3453 if (debug_value == 0) /* no output */
3455 /* set low N bits */
3456 return (1 << debug_value) - 1;
3459 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3461 spin_lock(&txq->_xmit_lock);
3462 txq->xmit_lock_owner = cpu;
3465 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3467 spin_lock_bh(&txq->_xmit_lock);
3468 txq->xmit_lock_owner = smp_processor_id();
3471 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3473 bool ok = spin_trylock(&txq->_xmit_lock);
3475 txq->xmit_lock_owner = smp_processor_id();
3479 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3481 txq->xmit_lock_owner = -1;
3482 spin_unlock(&txq->_xmit_lock);
3485 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3487 txq->xmit_lock_owner = -1;
3488 spin_unlock_bh(&txq->_xmit_lock);
3491 static inline void txq_trans_update(struct netdev_queue *txq)
3493 if (txq->xmit_lock_owner != -1)
3494 txq->trans_start = jiffies;
3497 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3498 static inline void netif_trans_update(struct net_device *dev)
3500 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3502 if (txq->trans_start != jiffies)
3503 txq->trans_start = jiffies;
3507 * netif_tx_lock - grab network device transmit lock
3508 * @dev: network device
3510 * Get network device transmit lock
3512 static inline void netif_tx_lock(struct net_device *dev)
3517 spin_lock(&dev->tx_global_lock);
3518 cpu = smp_processor_id();
3519 for (i = 0; i < dev->num_tx_queues; i++) {
3520 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3522 /* We are the only thread of execution doing a
3523 * freeze, but we have to grab the _xmit_lock in
3524 * order to synchronize with threads which are in
3525 * the ->hard_start_xmit() handler and already
3526 * checked the frozen bit.
3528 __netif_tx_lock(txq, cpu);
3529 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3530 __netif_tx_unlock(txq);
3534 static inline void netif_tx_lock_bh(struct net_device *dev)
3540 static inline void netif_tx_unlock(struct net_device *dev)
3544 for (i = 0; i < dev->num_tx_queues; i++) {
3545 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3547 /* No need to grab the _xmit_lock here. If the
3548 * queue is not stopped for another reason, we
3551 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3552 netif_schedule_queue(txq);
3554 spin_unlock(&dev->tx_global_lock);
3557 static inline void netif_tx_unlock_bh(struct net_device *dev)
3559 netif_tx_unlock(dev);
3563 #define HARD_TX_LOCK(dev, txq, cpu) { \
3564 if ((dev->features & NETIF_F_LLTX) == 0) { \
3565 __netif_tx_lock(txq, cpu); \
3569 #define HARD_TX_TRYLOCK(dev, txq) \
3570 (((dev->features & NETIF_F_LLTX) == 0) ? \
3571 __netif_tx_trylock(txq) : \
3574 #define HARD_TX_UNLOCK(dev, txq) { \
3575 if ((dev->features & NETIF_F_LLTX) == 0) { \
3576 __netif_tx_unlock(txq); \
3580 static inline void netif_tx_disable(struct net_device *dev)
3586 cpu = smp_processor_id();
3587 for (i = 0; i < dev->num_tx_queues; i++) {
3588 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3590 __netif_tx_lock(txq, cpu);
3591 netif_tx_stop_queue(txq);
3592 __netif_tx_unlock(txq);
3597 static inline void netif_addr_lock(struct net_device *dev)
3599 spin_lock(&dev->addr_list_lock);
3602 static inline void netif_addr_lock_nested(struct net_device *dev)
3604 int subclass = SINGLE_DEPTH_NESTING;
3606 if (dev->netdev_ops->ndo_get_lock_subclass)
3607 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3609 spin_lock_nested(&dev->addr_list_lock, subclass);
3612 static inline void netif_addr_lock_bh(struct net_device *dev)
3614 spin_lock_bh(&dev->addr_list_lock);
3617 static inline void netif_addr_unlock(struct net_device *dev)
3619 spin_unlock(&dev->addr_list_lock);
3622 static inline void netif_addr_unlock_bh(struct net_device *dev)
3624 spin_unlock_bh(&dev->addr_list_lock);
3628 * dev_addrs walker. Should be used only for read access. Call with
3629 * rcu_read_lock held.
3631 #define for_each_dev_addr(dev, ha) \
3632 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3634 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3636 void ether_setup(struct net_device *dev);
3638 /* Support for loadable net-drivers */
3639 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3640 unsigned char name_assign_type,
3641 void (*setup)(struct net_device *),
3642 unsigned int txqs, unsigned int rxqs);
3643 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3644 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3646 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3647 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3650 int register_netdev(struct net_device *dev);
3651 void unregister_netdev(struct net_device *dev);
3653 /* General hardware address lists handling functions */
3654 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3655 struct netdev_hw_addr_list *from_list, int addr_len);
3656 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3657 struct netdev_hw_addr_list *from_list, int addr_len);
3658 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3659 struct net_device *dev,
3660 int (*sync)(struct net_device *, const unsigned char *),
3661 int (*unsync)(struct net_device *,
3662 const unsigned char *));
3663 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3664 struct net_device *dev,
3665 int (*unsync)(struct net_device *,
3666 const unsigned char *));
3667 void __hw_addr_init(struct netdev_hw_addr_list *list);
3669 /* Functions used for device addresses handling */
3670 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3671 unsigned char addr_type);
3672 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3673 unsigned char addr_type);
3674 void dev_addr_flush(struct net_device *dev);
3675 int dev_addr_init(struct net_device *dev);
3677 /* Functions used for unicast addresses handling */
3678 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3679 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3680 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3681 int dev_uc_sync(struct net_device *to, struct net_device *from);
3682 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3683 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3684 void dev_uc_flush(struct net_device *dev);
3685 void dev_uc_init(struct net_device *dev);
3688 * __dev_uc_sync - Synchonize device's unicast list
3689 * @dev: device to sync
3690 * @sync: function to call if address should be added
3691 * @unsync: function to call if address should be removed
3693 * Add newly added addresses to the interface, and release
3694 * addresses that have been deleted.
3696 static inline int __dev_uc_sync(struct net_device *dev,
3697 int (*sync)(struct net_device *,
3698 const unsigned char *),
3699 int (*unsync)(struct net_device *,
3700 const unsigned char *))
3702 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3706 * __dev_uc_unsync - Remove synchronized addresses from device
3707 * @dev: device to sync
3708 * @unsync: function to call if address should be removed
3710 * Remove all addresses that were added to the device by dev_uc_sync().
3712 static inline void __dev_uc_unsync(struct net_device *dev,
3713 int (*unsync)(struct net_device *,
3714 const unsigned char *))
3716 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3719 /* Functions used for multicast addresses handling */
3720 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3721 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3722 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3723 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3724 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3725 int dev_mc_sync(struct net_device *to, struct net_device *from);
3726 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3727 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3728 void dev_mc_flush(struct net_device *dev);
3729 void dev_mc_init(struct net_device *dev);
3732 * __dev_mc_sync - Synchonize device's multicast list
3733 * @dev: device to sync
3734 * @sync: function to call if address should be added
3735 * @unsync: function to call if address should be removed
3737 * Add newly added addresses to the interface, and release
3738 * addresses that have been deleted.
3740 static inline int __dev_mc_sync(struct net_device *dev,
3741 int (*sync)(struct net_device *,
3742 const unsigned char *),
3743 int (*unsync)(struct net_device *,
3744 const unsigned char *))
3746 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3750 * __dev_mc_unsync - Remove synchronized addresses from device
3751 * @dev: device to sync
3752 * @unsync: function to call if address should be removed
3754 * Remove all addresses that were added to the device by dev_mc_sync().
3756 static inline void __dev_mc_unsync(struct net_device *dev,
3757 int (*unsync)(struct net_device *,
3758 const unsigned char *))
3760 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3763 /* Functions used for secondary unicast and multicast support */
3764 void dev_set_rx_mode(struct net_device *dev);
3765 void __dev_set_rx_mode(struct net_device *dev);
3766 int dev_set_promiscuity(struct net_device *dev, int inc);
3767 int dev_set_allmulti(struct net_device *dev, int inc);
3768 void netdev_state_change(struct net_device *dev);
3769 void netdev_notify_peers(struct net_device *dev);
3770 void netdev_features_change(struct net_device *dev);
3771 /* Load a device via the kmod */
3772 void dev_load(struct net *net, const char *name);
3773 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3774 struct rtnl_link_stats64 *storage);
3775 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3776 const struct net_device_stats *netdev_stats);
3778 extern int netdev_max_backlog;
3779 extern int netdev_tstamp_prequeue;
3780 extern int weight_p;
3782 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3783 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3784 struct list_head **iter);
3785 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3786 struct list_head **iter);
3788 /* iterate through upper list, must be called under RCU read lock */
3789 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3790 for (iter = &(dev)->adj_list.upper, \
3791 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3793 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3795 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3796 int (*fn)(struct net_device *upper_dev,
3800 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3801 struct net_device *upper_dev);
3803 void *netdev_lower_get_next_private(struct net_device *dev,
3804 struct list_head **iter);
3805 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3806 struct list_head **iter);
3808 #define netdev_for_each_lower_private(dev, priv, iter) \
3809 for (iter = (dev)->adj_list.lower.next, \
3810 priv = netdev_lower_get_next_private(dev, &(iter)); \
3812 priv = netdev_lower_get_next_private(dev, &(iter)))
3814 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3815 for (iter = &(dev)->adj_list.lower, \
3816 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3818 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3820 void *netdev_lower_get_next(struct net_device *dev,
3821 struct list_head **iter);
3823 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3824 for (iter = (dev)->adj_list.lower.next, \
3825 ldev = netdev_lower_get_next(dev, &(iter)); \
3827 ldev = netdev_lower_get_next(dev, &(iter)))
3829 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3830 struct list_head **iter);
3831 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3832 struct list_head **iter);
3834 int netdev_walk_all_lower_dev(struct net_device *dev,
3835 int (*fn)(struct net_device *lower_dev,
3838 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3839 int (*fn)(struct net_device *lower_dev,
3843 void *netdev_adjacent_get_private(struct list_head *adj_list);
3844 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3845 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3846 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3847 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3848 int netdev_master_upper_dev_link(struct net_device *dev,
3849 struct net_device *upper_dev,
3850 void *upper_priv, void *upper_info);
3851 void netdev_upper_dev_unlink(struct net_device *dev,
3852 struct net_device *upper_dev);
3853 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3854 void *netdev_lower_dev_get_private(struct net_device *dev,
3855 struct net_device *lower_dev);
3856 void netdev_lower_state_changed(struct net_device *lower_dev,
3857 void *lower_state_info);
3858 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3859 struct neighbour *n);
3860 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3861 struct neighbour *n);
3863 /* RSS keys are 40 or 52 bytes long */
3864 #define NETDEV_RSS_KEY_LEN 52
3865 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3866 void netdev_rss_key_fill(void *buffer, size_t len);
3868 int dev_get_nest_level(struct net_device *dev);
3869 int skb_checksum_help(struct sk_buff *skb);
3870 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3871 netdev_features_t features, bool tx_path);
3872 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3873 netdev_features_t features);
3875 struct netdev_bonding_info {
3880 struct netdev_notifier_bonding_info {
3881 struct netdev_notifier_info info; /* must be first */
3882 struct netdev_bonding_info bonding_info;
3885 void netdev_bonding_info_change(struct net_device *dev,
3886 struct netdev_bonding_info *bonding_info);
3889 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3891 return __skb_gso_segment(skb, features, true);
3893 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3895 static inline bool can_checksum_protocol(netdev_features_t features,
3898 if (protocol == htons(ETH_P_FCOE))
3899 return !!(features & NETIF_F_FCOE_CRC);
3901 /* Assume this is an IP checksum (not SCTP CRC) */
3903 if (features & NETIF_F_HW_CSUM) {
3904 /* Can checksum everything */
3909 case htons(ETH_P_IP):
3910 return !!(features & NETIF_F_IP_CSUM);
3911 case htons(ETH_P_IPV6):
3912 return !!(features & NETIF_F_IPV6_CSUM);
3919 void netdev_rx_csum_fault(struct net_device *dev);
3921 static inline void netdev_rx_csum_fault(struct net_device *dev)
3925 /* rx skb timestamps */
3926 void net_enable_timestamp(void);
3927 void net_disable_timestamp(void);
3929 #ifdef CONFIG_PROC_FS
3930 int __init dev_proc_init(void);
3932 #define dev_proc_init() 0
3935 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3936 struct sk_buff *skb, struct net_device *dev,
3939 skb->xmit_more = more ? 1 : 0;
3940 return ops->ndo_start_xmit(skb, dev);
3943 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3944 struct netdev_queue *txq, bool more)
3946 const struct net_device_ops *ops = dev->netdev_ops;
3949 rc = __netdev_start_xmit(ops, skb, dev, more);
3950 if (rc == NETDEV_TX_OK)
3951 txq_trans_update(txq);
3956 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3958 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3961 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3963 return netdev_class_create_file_ns(class_attr, NULL);
3966 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3968 netdev_class_remove_file_ns(class_attr, NULL);
3971 extern struct kobj_ns_type_operations net_ns_type_operations;
3973 const char *netdev_drivername(const struct net_device *dev);
3975 void linkwatch_run_queue(void);
3977 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3978 netdev_features_t f2)
3980 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3981 if (f1 & NETIF_F_HW_CSUM)
3982 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3984 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3990 static inline netdev_features_t netdev_get_wanted_features(
3991 struct net_device *dev)
3993 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3995 netdev_features_t netdev_increment_features(netdev_features_t all,
3996 netdev_features_t one, netdev_features_t mask);
3998 /* Allow TSO being used on stacked device :
3999 * Performing the GSO segmentation before last device
4000 * is a performance improvement.
4002 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4003 netdev_features_t mask)
4005 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4008 int __netdev_update_features(struct net_device *dev);
4009 void netdev_update_features(struct net_device *dev);
4010 void netdev_change_features(struct net_device *dev);
4012 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4013 struct net_device *dev);
4015 netdev_features_t passthru_features_check(struct sk_buff *skb,
4016 struct net_device *dev,
4017 netdev_features_t features);
4018 netdev_features_t netif_skb_features(struct sk_buff *skb);
4020 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4022 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4024 /* check flags correspondence */
4025 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4026 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4027 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4028 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4029 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4030 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4031 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4032 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4033 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4034 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4035 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4036 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4037 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4038 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4039 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4040 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4042 return (features & feature) == feature;
4045 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4047 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4048 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4051 static inline bool netif_needs_gso(struct sk_buff *skb,
4052 netdev_features_t features)
4054 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4055 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4056 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4059 static inline void netif_set_gso_max_size(struct net_device *dev,
4062 dev->gso_max_size = size;
4065 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4066 int pulled_hlen, u16 mac_offset,
4069 skb->protocol = protocol;
4070 skb->encapsulation = 1;
4071 skb_push(skb, pulled_hlen);
4072 skb_reset_transport_header(skb);
4073 skb->mac_header = mac_offset;
4074 skb->network_header = skb->mac_header + mac_len;
4075 skb->mac_len = mac_len;
4078 static inline bool netif_is_macsec(const struct net_device *dev)
4080 return dev->priv_flags & IFF_MACSEC;
4083 static inline bool netif_is_macvlan(const struct net_device *dev)
4085 return dev->priv_flags & IFF_MACVLAN;
4088 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4090 return dev->priv_flags & IFF_MACVLAN_PORT;
4093 static inline bool netif_is_ipvlan(const struct net_device *dev)
4095 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4098 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4100 return dev->priv_flags & IFF_IPVLAN_MASTER;
4103 static inline bool netif_is_bond_master(const struct net_device *dev)
4105 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4108 static inline bool netif_is_bond_slave(const struct net_device *dev)
4110 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4113 static inline bool netif_supports_nofcs(struct net_device *dev)
4115 return dev->priv_flags & IFF_SUPP_NOFCS;
4118 static inline bool netif_is_l3_master(const struct net_device *dev)
4120 return dev->priv_flags & IFF_L3MDEV_MASTER;
4123 static inline bool netif_is_l3_slave(const struct net_device *dev)
4125 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4128 static inline bool netif_is_bridge_master(const struct net_device *dev)
4130 return dev->priv_flags & IFF_EBRIDGE;
4133 static inline bool netif_is_bridge_port(const struct net_device *dev)
4135 return dev->priv_flags & IFF_BRIDGE_PORT;
4138 static inline bool netif_is_ovs_master(const struct net_device *dev)
4140 return dev->priv_flags & IFF_OPENVSWITCH;
4143 static inline bool netif_is_team_master(const struct net_device *dev)
4145 return dev->priv_flags & IFF_TEAM;
4148 static inline bool netif_is_team_port(const struct net_device *dev)
4150 return dev->priv_flags & IFF_TEAM_PORT;
4153 static inline bool netif_is_lag_master(const struct net_device *dev)
4155 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4158 static inline bool netif_is_lag_port(const struct net_device *dev)
4160 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4163 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4165 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4168 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4169 static inline void netif_keep_dst(struct net_device *dev)
4171 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4174 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4175 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4177 /* TODO: reserve and use an additional IFF bit, if we get more users */
4178 return dev->priv_flags & IFF_MACSEC;
4181 extern struct pernet_operations __net_initdata loopback_net_ops;
4183 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4185 /* netdev_printk helpers, similar to dev_printk */
4187 static inline const char *netdev_name(const struct net_device *dev)
4189 if (!dev->name[0] || strchr(dev->name, '%'))
4190 return "(unnamed net_device)";
4194 static inline const char *netdev_reg_state(const struct net_device *dev)
4196 switch (dev->reg_state) {
4197 case NETREG_UNINITIALIZED: return " (uninitialized)";
4198 case NETREG_REGISTERED: return "";
4199 case NETREG_UNREGISTERING: return " (unregistering)";
4200 case NETREG_UNREGISTERED: return " (unregistered)";
4201 case NETREG_RELEASED: return " (released)";
4202 case NETREG_DUMMY: return " (dummy)";
4205 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4206 return " (unknown)";
4210 void netdev_printk(const char *level, const struct net_device *dev,
4211 const char *format, ...);
4213 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4215 void netdev_alert(const struct net_device *dev, const char *format, ...);
4217 void netdev_crit(const struct net_device *dev, const char *format, ...);
4219 void netdev_err(const struct net_device *dev, const char *format, ...);
4221 void netdev_warn(const struct net_device *dev, const char *format, ...);
4223 void netdev_notice(const struct net_device *dev, const char *format, ...);
4225 void netdev_info(const struct net_device *dev, const char *format, ...);
4227 #define MODULE_ALIAS_NETDEV(device) \
4228 MODULE_ALIAS("netdev-" device)
4230 #if defined(CONFIG_DYNAMIC_DEBUG)
4231 #define netdev_dbg(__dev, format, args...) \
4233 dynamic_netdev_dbg(__dev, format, ##args); \
4235 #elif defined(DEBUG)
4236 #define netdev_dbg(__dev, format, args...) \
4237 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4239 #define netdev_dbg(__dev, format, args...) \
4242 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4246 #if defined(VERBOSE_DEBUG)
4247 #define netdev_vdbg netdev_dbg
4250 #define netdev_vdbg(dev, format, args...) \
4253 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4259 * netdev_WARN() acts like dev_printk(), but with the key difference
4260 * of using a WARN/WARN_ON to get the message out, including the
4261 * file/line information and a backtrace.
4263 #define netdev_WARN(dev, format, args...) \
4264 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4265 netdev_reg_state(dev), ##args)
4267 /* netif printk helpers, similar to netdev_printk */
4269 #define netif_printk(priv, type, level, dev, fmt, args...) \
4271 if (netif_msg_##type(priv)) \
4272 netdev_printk(level, (dev), fmt, ##args); \
4275 #define netif_level(level, priv, type, dev, fmt, args...) \
4277 if (netif_msg_##type(priv)) \
4278 netdev_##level(dev, fmt, ##args); \
4281 #define netif_emerg(priv, type, dev, fmt, args...) \
4282 netif_level(emerg, priv, type, dev, fmt, ##args)
4283 #define netif_alert(priv, type, dev, fmt, args...) \
4284 netif_level(alert, priv, type, dev, fmt, ##args)
4285 #define netif_crit(priv, type, dev, fmt, args...) \
4286 netif_level(crit, priv, type, dev, fmt, ##args)
4287 #define netif_err(priv, type, dev, fmt, args...) \
4288 netif_level(err, priv, type, dev, fmt, ##args)
4289 #define netif_warn(priv, type, dev, fmt, args...) \
4290 netif_level(warn, priv, type, dev, fmt, ##args)
4291 #define netif_notice(priv, type, dev, fmt, args...) \
4292 netif_level(notice, priv, type, dev, fmt, ##args)
4293 #define netif_info(priv, type, dev, fmt, args...) \
4294 netif_level(info, priv, type, dev, fmt, ##args)
4296 #if defined(CONFIG_DYNAMIC_DEBUG)
4297 #define netif_dbg(priv, type, netdev, format, args...) \
4299 if (netif_msg_##type(priv)) \
4300 dynamic_netdev_dbg(netdev, format, ##args); \
4302 #elif defined(DEBUG)
4303 #define netif_dbg(priv, type, dev, format, args...) \
4304 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4306 #define netif_dbg(priv, type, dev, format, args...) \
4309 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4314 #if defined(VERBOSE_DEBUG)
4315 #define netif_vdbg netif_dbg
4317 #define netif_vdbg(priv, type, dev, format, args...) \
4320 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4326 * The list of packet types we will receive (as opposed to discard)
4327 * and the routines to invoke.
4329 * Why 16. Because with 16 the only overlap we get on a hash of the
4330 * low nibble of the protocol value is RARP/SNAP/X.25.
4332 * NOTE: That is no longer true with the addition of VLAN tags. Not
4333 * sure which should go first, but I bet it won't make much
4334 * difference if we are running VLANs. The good news is that
4335 * this protocol won't be in the list unless compiled in, so
4336 * the average user (w/out VLANs) will not be adversely affected.
4352 #define PTYPE_HASH_SIZE (16)
4353 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4355 #endif /* _LINUX_NETDEVICE_H */