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>
45 #include <net/dcbnl.h>
47 #include <net/netprio_cgroup.h>
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
59 struct dsa_switch_tree;
63 /* 802.15.4 specific */
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
70 void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
78 * Transmit return codes: transmit return codes originate from three different
81 * - qdisc return codes
82 * - driver transmit return codes
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91 * others are propagated to higher layers.
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK 0xf0
110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 typedef enum netdev_tx netdev_tx_t;
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 static inline bool dev_xmit_complete(int rc)
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 if (likely(rc < NET_XMIT_MASK))
135 * Compute the worst-case header length according to the protocols
139 #if defined(CONFIG_HYPERV_NET)
140 # define LL_MAX_HEADER 128
141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
145 # define LL_MAX_HEADER 96
148 # define LL_MAX_HEADER 32
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 extern struct static_key rfs_needed;
203 struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
207 #define NETDEV_HW_ADDR_T_LAN 1
208 #define NETDEV_HW_ADDR_T_SAN 2
209 #define NETDEV_HW_ADDR_T_SLAVE 3
210 #define NETDEV_HW_ADDR_T_UNICAST 4
211 #define NETDEV_HW_ADDR_T_MULTICAST 5
216 struct rcu_head rcu_head;
219 struct netdev_hw_addr_list {
220 struct list_head list;
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
274 bool (*validate)(const char *ll_header, unsigned int len);
277 /* These flag bits are private to the generic network queueing
278 * layer; they may not be explicitly referenced by any other
282 enum netdev_state_t {
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
292 * This structure holds boot-time configured netdevice settings. They
293 * are then used in the device probing.
295 struct netdev_boot_setup {
299 #define NETDEV_BOOT_SETUP_MAX 8
301 int __init netdev_boot_setup(char *str);
304 * Structure for NAPI scheduling similar to tasklet but with weighting
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-CPU poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
313 struct list_head poll_list;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
322 struct net_device *dev;
323 struct sk_buff *gro_list;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_MISSED, /* reschedule a napi */
334 NAPI_STATE_DISABLE, /* Disable pending */
335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
359 typedef enum gro_result gro_result_t;
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
383 * If the rx_handler consumed the skb in some way, it should return
384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
385 * the skb to be delivered in some other way.
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
391 * If the rx_handler decides the skb should be ignored, it should return
392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
393 * are registered on exact device (ptype->dev == skb->dev).
395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
396 * delivered, it should return RX_HANDLER_PASS.
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
402 enum rx_handler_result {
408 typedef enum rx_handler_result rx_handler_result_t;
409 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
411 void __napi_schedule(struct napi_struct *n);
412 void __napi_schedule_irqoff(struct napi_struct *n);
414 static inline bool napi_disable_pending(struct napi_struct *n)
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
419 bool napi_schedule_prep(struct napi_struct *n);
422 * napi_schedule - schedule NAPI poll
425 * Schedule NAPI poll routine to be called if it is not already
428 static inline void napi_schedule(struct napi_struct *n)
430 if (napi_schedule_prep(n))
435 * napi_schedule_irqoff - schedule NAPI poll
438 * Variant of napi_schedule(), assuming hard irqs are masked.
440 static inline void napi_schedule_irqoff(struct napi_struct *n)
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
447 static inline bool napi_reschedule(struct napi_struct *napi)
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
456 bool napi_complete_done(struct napi_struct *n, int work_done);
458 * napi_complete - NAPI processing complete
461 * Mark NAPI processing as complete.
462 * Consider using napi_complete_done() instead.
463 * Return false if device should avoid rearming interrupts.
465 static inline bool napi_complete(struct napi_struct *n)
467 return napi_complete_done(n, 0);
471 * napi_hash_del - remove a NAPI from global table
472 * @napi: NAPI context
474 * Warning: caller must observe RCU grace period
475 * before freeing memory containing @napi, if
476 * this function returns true.
477 * Note: core networking stack automatically calls it
478 * from netif_napi_del().
479 * Drivers might want to call this helper to combine all
480 * the needed RCU grace periods into a single one.
482 bool napi_hash_del(struct napi_struct *napi);
485 * napi_disable - prevent NAPI from scheduling
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
491 void napi_disable(struct napi_struct *n);
494 * napi_enable - enable NAPI scheduling
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
500 static inline void napi_enable(struct napi_struct *n)
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
503 smp_mb__before_atomic();
504 clear_bit(NAPI_STATE_SCHED, &n->state);
505 clear_bit(NAPI_STATE_NPSVC, &n->state);
509 * napi_synchronize - wait until NAPI is not running
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct *n)
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
525 enum netdev_queue_state_t {
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
528 __QUEUE_STATE_FROZEN,
531 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
551 struct netdev_queue {
555 struct net_device *dev;
556 struct Qdisc __rcu *qdisc;
557 struct Qdisc *qdisc_sleeping;
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
564 unsigned long tx_maxrate;
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
569 unsigned long trans_timeout;
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
576 * Time (in jiffies) of last Tx
578 unsigned long trans_start;
585 } ____cacheline_aligned_in_smp;
587 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
613 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
620 struct rps_dev_flow {
623 unsigned int last_qtail;
625 #define RPS_NO_FILTER 0xffff
628 * The rps_dev_flow_table structure contains a table of flow mappings.
630 struct rps_dev_flow_table {
633 struct rps_dev_flow flows[0];
635 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
636 ((_num) * sizeof(struct rps_dev_flow)))
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
643 * rps_cpu_mask is used to partition the space, depending on number of
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
646 * meaning we use 32-6=26 bits for the hash.
648 struct rps_sock_flow_table {
651 u32 ents[0] ____cacheline_aligned_in_smp;
653 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
655 #define RPS_NO_CPU 0xffff
657 extern u32 rps_cpu_mask;
658 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
667 /* We only give a hint, preemption can change CPU under us */
668 val |= raw_smp_processor_id();
670 if (table->ents[index] != val)
671 table->ents[index] = val;
675 #ifdef CONFIG_RFS_ACCEL
676 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
679 #endif /* CONFIG_RPS */
681 /* This structure contains an instance of an RX queue. */
682 struct netdev_rx_queue {
684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
688 struct net_device *dev;
689 } ____cacheline_aligned_in_smp;
692 * RX queue sysfs structures and functions.
694 struct rx_queue_attribute {
695 struct attribute attr;
696 ssize_t (*show)(struct netdev_rx_queue *queue,
697 struct rx_queue_attribute *attr, char *buf);
698 ssize_t (*store)(struct netdev_rx_queue *queue,
699 struct rx_queue_attribute *attr, const char *buf, size_t len);
704 * This structure holds an XPS map which can be of variable length. The
705 * map is an array of queues.
709 unsigned int alloc_len;
713 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
714 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715 - sizeof(struct xps_map)) / sizeof(u16))
718 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 struct xps_dev_maps {
722 struct xps_map __rcu *cpu_map[0];
724 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
726 #endif /* CONFIG_XPS */
728 #define TC_MAX_QUEUE 16
729 #define TC_BITMASK 15
730 /* HW offloaded queuing disciplines txq count and offset maps */
731 struct netdev_tc_txq {
736 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738 * This structure is to hold information about the device
739 * configured to run FCoE protocol stack.
741 struct netdev_fcoe_hbainfo {
742 char manufacturer[64];
743 char serial_number[64];
744 char hardware_version[64];
745 char driver_version[64];
746 char optionrom_version[64];
747 char firmware_version[64];
749 char model_description[256];
753 #define MAX_PHYS_ITEM_ID_LEN 32
755 /* This structure holds a unique identifier to identify some
756 * physical item (port for example) used by a netdevice.
758 struct netdev_phys_item_id {
759 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
760 unsigned char id_len;
763 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
764 struct netdev_phys_item_id *b)
766 return a->id_len == b->id_len &&
767 memcmp(a->id, b->id, a->id_len) == 0;
770 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
773 /* These structures hold the attributes of qdisc and classifiers
774 * that are being passed to the netdevice through the setup_tc op.
784 struct tc_cls_u32_offload;
786 struct tc_to_netdev {
789 struct tc_cls_u32_offload *cls_u32;
790 struct tc_cls_flower_offload *cls_flower;
791 struct tc_cls_matchall_offload *cls_mall;
792 struct tc_cls_bpf_offload *cls_bpf;
793 struct tc_mqprio_qopt *mqprio;
798 /* These structures hold the attributes of xdp state that are being passed
799 * to the netdevice through the xdp op.
801 enum xdp_netdev_command {
802 /* Set or clear a bpf program used in the earliest stages of packet
803 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
804 * is responsible for calling bpf_prog_put on any old progs that are
805 * stored. In case of error, the callee need not release the new prog
806 * reference, but on success it takes ownership and must bpf_prog_put
807 * when it is no longer used.
810 /* Check if a bpf program is set on the device. The callee should
811 * return true if a program is currently attached and running.
817 enum xdp_netdev_command command;
820 struct bpf_prog *prog;
826 #ifdef CONFIG_XFRM_OFFLOAD
828 int (*xdo_dev_state_add) (struct xfrm_state *x);
829 void (*xdo_dev_state_delete) (struct xfrm_state *x);
830 void (*xdo_dev_state_free) (struct xfrm_state *x);
831 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
832 struct xfrm_state *x);
837 * This structure defines the management hooks for network devices.
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
841 * int (*ndo_init)(struct net_device *dev);
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
851 * int (*ndo_open)(struct net_device *dev);
852 * This function is called when a network device transitions to the up
855 * int (*ndo_stop)(struct net_device *dev);
856 * This function is called when a network device transitions to the down
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
861 * Called when a packet needs to be transmitted.
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
866 * Required; cannot be NULL.
868 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
869 * struct net_device *dev
870 * netdev_features_t features);
871 * Called by core transmit path to determine if device is capable of
872 * performing offload operations on a given packet. This is to give
873 * the device an opportunity to implement any restrictions that cannot
874 * be otherwise expressed by feature flags. The check is called with
875 * the set of features that the stack has calculated and it returns
876 * those the driver believes to be appropriate.
878 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
879 * void *accel_priv, select_queue_fallback_t fallback);
880 * Called to decide which queue to use when device supports multiple
883 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
884 * This function is called to allow device receiver to make
885 * changes to configuration when multicast or promiscuous is enabled.
887 * void (*ndo_set_rx_mode)(struct net_device *dev);
888 * This function is called device changes address list filtering.
889 * If driver handles unicast address filtering, it should set
890 * IFF_UNICAST_FLT in its priv_flags.
892 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
893 * This function is called when the Media Access Control address
894 * needs to be changed. If this interface is not defined, the
895 * MAC address can not be changed.
897 * int (*ndo_validate_addr)(struct net_device *dev);
898 * Test if Media Access Control address is valid for the device.
900 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
901 * Called when a user requests an ioctl which can't be handled by
902 * the generic interface code. If not defined ioctls return
903 * not supported error code.
905 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
906 * Used to set network devices bus interface parameters. This interface
907 * is retained for legacy reasons; new devices should use the bus
908 * interface (PCI) for low level management.
910 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
911 * Called when a user wants to change the Maximum Transfer Unit
912 * of a device. If not defined, any request to change MTU will
913 * will return an error.
915 * void (*ndo_tx_timeout)(struct net_device *dev);
916 * Callback used when the transmitter has not made any progress
917 * for dev->watchdog ticks.
919 * void (*ndo_get_stats64)(struct net_device *dev,
920 * struct rtnl_link_stats64 *storage);
921 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
922 * Called when a user wants to get the network device usage
923 * statistics. Drivers must do one of the following:
924 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
925 * rtnl_link_stats64 structure passed by the caller.
926 * 2. Define @ndo_get_stats to update a net_device_stats structure
927 * (which should normally be dev->stats) and return a pointer to
928 * it. The structure may be changed asynchronously only if each
929 * field is written atomically.
930 * 3. Update dev->stats asynchronously and atomically, and define
933 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
934 * Return true if this device supports offload stats of this attr_id.
936 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
938 * Get statistics for offload operations by attr_id. Write it into the
941 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
942 * If device supports VLAN filtering this function is called when a
943 * VLAN id is registered.
945 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
946 * If device supports VLAN filtering this function is called when a
947 * VLAN id is unregistered.
949 * void (*ndo_poll_controller)(struct net_device *dev);
951 * SR-IOV management functions.
952 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
953 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
954 * u8 qos, __be16 proto);
955 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
957 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
958 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
959 * int (*ndo_get_vf_config)(struct net_device *dev,
960 * int vf, struct ifla_vf_info *ivf);
961 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
962 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
963 * struct nlattr *port[]);
965 * Enable or disable the VF ability to query its RSS Redirection Table and
966 * Hash Key. This is needed since on some devices VF share this information
967 * with PF and querying it may introduce a theoretical security risk.
968 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
969 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
970 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
971 * __be16 protocol, struct tc_to_netdev *tc);
972 * Called to setup any 'tc' scheduler, classifier or action on @dev.
973 * This is always called from the stack with the rtnl lock held and netif
974 * tx queues stopped. This allows the netdevice to perform queue
977 * Fiber Channel over Ethernet (FCoE) offload functions.
978 * int (*ndo_fcoe_enable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to start using LLD for FCoE
980 * so the underlying device can perform whatever needed configuration or
981 * initialization to support acceleration of FCoE traffic.
983 * int (*ndo_fcoe_disable)(struct net_device *dev);
984 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
985 * so the underlying device can perform whatever needed clean-ups to
986 * stop supporting acceleration of FCoE traffic.
988 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
989 * struct scatterlist *sgl, unsigned int sgc);
990 * Called when the FCoE Initiator wants to initialize an I/O that
991 * is a possible candidate for Direct Data Placement (DDP). The LLD can
992 * perform necessary setup and returns 1 to indicate the device is set up
993 * successfully to perform DDP on this I/O, otherwise this returns 0.
995 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
996 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
997 * indicated by the FC exchange id 'xid', so the underlying device can
998 * clean up and reuse resources for later DDP requests.
1000 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1001 * struct scatterlist *sgl, unsigned int sgc);
1002 * Called when the FCoE Target wants to initialize an I/O that
1003 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1004 * perform necessary setup and returns 1 to indicate the device is set up
1005 * successfully to perform DDP on this I/O, otherwise this returns 0.
1007 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1008 * struct netdev_fcoe_hbainfo *hbainfo);
1009 * Called when the FCoE Protocol stack wants information on the underlying
1010 * device. This information is utilized by the FCoE protocol stack to
1011 * register attributes with Fiber Channel management service as per the
1012 * FC-GS Fabric Device Management Information(FDMI) specification.
1014 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1015 * Called when the underlying device wants to override default World Wide
1016 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1017 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1018 * protocol stack to use.
1021 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1022 * u16 rxq_index, u32 flow_id);
1023 * Set hardware filter for RFS. rxq_index is the target queue index;
1024 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1025 * Return the filter ID on success, or a negative error code.
1027 * Slave management functions (for bridge, bonding, etc).
1028 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1029 * Called to make another netdev an underling.
1031 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1032 * Called to release previously enslaved netdev.
1034 * Feature/offload setting functions.
1035 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1036 * netdev_features_t features);
1037 * Adjusts the requested feature flags according to device-specific
1038 * constraints, and returns the resulting flags. Must not modify
1041 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1042 * Called to update device configuration to new features. Passed
1043 * feature set might be less than what was returned by ndo_fix_features()).
1044 * Must return >0 or -errno if it changed dev->features itself.
1046 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1047 * struct net_device *dev,
1048 * const unsigned char *addr, u16 vid, u16 flags)
1049 * Adds an FDB entry to dev for addr.
1050 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1051 * struct net_device *dev,
1052 * const unsigned char *addr, u16 vid)
1053 * Deletes the FDB entry from dev coresponding to addr.
1054 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1055 * struct net_device *dev, struct net_device *filter_dev,
1057 * Used to add FDB entries to dump requests. Implementers should add
1058 * entries to skb and update idx with the number of entries.
1060 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1062 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1063 * struct net_device *dev, u32 filter_mask,
1065 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1068 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1069 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1070 * which do not represent real hardware may define this to allow their
1071 * userspace components to manage their virtual carrier state. Devices
1072 * that determine carrier state from physical hardware properties (eg
1073 * network cables) or protocol-dependent mechanisms (eg
1074 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1076 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1077 * struct netdev_phys_item_id *ppid);
1078 * Called to get ID of physical port of this device. If driver does
1079 * not implement this, it is assumed that the hw is not able to have
1080 * multiple net devices on single physical port.
1082 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1083 * struct udp_tunnel_info *ti);
1084 * Called by UDP tunnel to notify a driver about the UDP port and socket
1085 * address family that a UDP tunnel is listnening to. It is called only
1086 * when a new port starts listening. The operation is protected by the
1089 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1090 * struct udp_tunnel_info *ti);
1091 * Called by UDP tunnel to notify the driver about a UDP port and socket
1092 * address family that the UDP tunnel is not listening to anymore. The
1093 * operation is protected by the RTNL.
1095 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1096 * struct net_device *dev)
1097 * Called by upper layer devices to accelerate switching or other
1098 * station functionality into hardware. 'pdev is the lowerdev
1099 * to use for the offload and 'dev' is the net device that will
1100 * back the offload. Returns a pointer to the private structure
1101 * the upper layer will maintain.
1102 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1103 * Called by upper layer device to delete the station created
1104 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1105 * the station and priv is the structure returned by the add
1107 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1108 * struct net_device *dev,
1110 * Callback to use for xmit over the accelerated station. This
1111 * is used in place of ndo_start_xmit on accelerated net
1113 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1114 * int queue_index, u32 maxrate);
1115 * Called when a user wants to set a max-rate limitation of specific
1117 * int (*ndo_get_iflink)(const struct net_device *dev);
1118 * Called to get the iflink value of this device.
1119 * void (*ndo_change_proto_down)(struct net_device *dev,
1121 * This function is used to pass protocol port error state information
1122 * to the switch driver. The switch driver can react to the proto_down
1123 * by doing a phys down on the associated switch port.
1124 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1125 * This function is used to get egress tunnel information for given skb.
1126 * This is useful for retrieving outer tunnel header parameters while
1128 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1129 * This function is used to specify the headroom that the skb must
1130 * consider when allocation skb during packet reception. Setting
1131 * appropriate rx headroom value allows avoiding skb head copy on
1132 * forward. Setting a negative value resets the rx headroom to the
1134 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1135 * This function is used to set or query state related to XDP on the
1136 * netdevice. See definition of enum xdp_netdev_command for details.
1139 struct net_device_ops {
1140 int (*ndo_init)(struct net_device *dev);
1141 void (*ndo_uninit)(struct net_device *dev);
1142 int (*ndo_open)(struct net_device *dev);
1143 int (*ndo_stop)(struct net_device *dev);
1144 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1145 struct net_device *dev);
1146 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1147 struct net_device *dev,
1148 netdev_features_t features);
1149 u16 (*ndo_select_queue)(struct net_device *dev,
1150 struct sk_buff *skb,
1152 select_queue_fallback_t fallback);
1153 void (*ndo_change_rx_flags)(struct net_device *dev,
1155 void (*ndo_set_rx_mode)(struct net_device *dev);
1156 int (*ndo_set_mac_address)(struct net_device *dev,
1158 int (*ndo_validate_addr)(struct net_device *dev);
1159 int (*ndo_do_ioctl)(struct net_device *dev,
1160 struct ifreq *ifr, int cmd);
1161 int (*ndo_set_config)(struct net_device *dev,
1163 int (*ndo_change_mtu)(struct net_device *dev,
1165 int (*ndo_neigh_setup)(struct net_device *dev,
1166 struct neigh_parms *);
1167 void (*ndo_tx_timeout) (struct net_device *dev);
1169 void (*ndo_get_stats64)(struct net_device *dev,
1170 struct rtnl_link_stats64 *storage);
1171 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1172 int (*ndo_get_offload_stats)(int attr_id,
1173 const struct net_device *dev,
1175 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1177 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1178 __be16 proto, u16 vid);
1179 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1180 __be16 proto, u16 vid);
1181 #ifdef CONFIG_NET_POLL_CONTROLLER
1182 void (*ndo_poll_controller)(struct net_device *dev);
1183 int (*ndo_netpoll_setup)(struct net_device *dev,
1184 struct netpoll_info *info);
1185 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1187 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.
1511 * @min_header_len: Minimum hardware header length
1513 * @needed_headroom: Extra headroom the hardware may need, but not in all
1514 * cases can this be guaranteed
1515 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1516 * cases can this be guaranteed. Some cases also use
1517 * LL_MAX_HEADER instead to allocate the skb
1519 * interface address info:
1521 * @perm_addr: Permanent hw address
1522 * @addr_assign_type: Hw address assignment type
1523 * @addr_len: Hardware address length
1524 * @neigh_priv_len: Used in neigh_alloc()
1525 * @dev_id: Used to differentiate devices that share
1526 * the same link layer address
1527 * @dev_port: Used to differentiate devices that share
1529 * @addr_list_lock: XXX: need comments on this one
1530 * @uc_promisc: Counter that indicates promiscuous mode
1531 * has been enabled due to the need to listen to
1532 * additional unicast addresses in a device that
1533 * does not implement ndo_set_rx_mode()
1534 * @uc: unicast mac addresses
1535 * @mc: multicast mac addresses
1536 * @dev_addrs: list of device hw addresses
1537 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1538 * @promiscuity: Number of times the NIC is told to work in
1539 * promiscuous mode; if it becomes 0 the NIC will
1540 * exit promiscuous mode
1541 * @allmulti: Counter, enables or disables allmulticast mode
1543 * @vlan_info: VLAN info
1544 * @dsa_ptr: dsa specific data
1545 * @tipc_ptr: TIPC specific data
1546 * @atalk_ptr: AppleTalk link
1547 * @ip_ptr: IPv4 specific data
1548 * @dn_ptr: DECnet specific data
1549 * @ip6_ptr: IPv6 specific data
1550 * @ax25_ptr: AX.25 specific data
1551 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
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;
1710 const struct xfrmdev_ops *xfrmdev_ops;
1713 const struct header_ops *header_ops;
1716 unsigned int priv_flags;
1718 unsigned short gflags;
1719 unsigned short padded;
1721 unsigned char operstate;
1722 unsigned char link_mode;
1724 unsigned char if_port;
1728 unsigned int min_mtu;
1729 unsigned int max_mtu;
1730 unsigned short type;
1731 unsigned short hard_header_len;
1732 unsigned char min_header_len;
1734 unsigned short needed_headroom;
1735 unsigned short needed_tailroom;
1737 /* Interface address info. */
1738 unsigned char perm_addr[MAX_ADDR_LEN];
1739 unsigned char addr_assign_type;
1740 unsigned char addr_len;
1741 unsigned short neigh_priv_len;
1742 unsigned short dev_id;
1743 unsigned short dev_port;
1744 spinlock_t addr_list_lock;
1745 unsigned char name_assign_type;
1747 struct netdev_hw_addr_list uc;
1748 struct netdev_hw_addr_list mc;
1749 struct netdev_hw_addr_list dev_addrs;
1752 struct kset *queues_kset;
1754 unsigned int promiscuity;
1755 unsigned int allmulti;
1758 /* Protocol-specific pointers */
1760 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1761 struct vlan_info __rcu *vlan_info;
1763 #if IS_ENABLED(CONFIG_NET_DSA)
1764 struct dsa_switch_tree *dsa_ptr;
1766 #if IS_ENABLED(CONFIG_TIPC)
1767 struct tipc_bearer __rcu *tipc_ptr;
1770 struct in_device __rcu *ip_ptr;
1771 struct dn_dev __rcu *dn_ptr;
1772 struct inet6_dev __rcu *ip6_ptr;
1774 struct wireless_dev *ieee80211_ptr;
1775 struct wpan_dev *ieee802154_ptr;
1776 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1777 struct mpls_dev __rcu *mpls_ptr;
1781 * Cache lines mostly used on receive path (including eth_type_trans())
1783 /* Interface address info used in eth_type_trans() */
1784 unsigned char *dev_addr;
1787 struct netdev_rx_queue *_rx;
1789 unsigned int num_rx_queues;
1790 unsigned int real_num_rx_queues;
1793 struct bpf_prog __rcu *xdp_prog;
1794 unsigned long gro_flush_timeout;
1795 rx_handler_func_t __rcu *rx_handler;
1796 void __rcu *rx_handler_data;
1798 #ifdef CONFIG_NET_CLS_ACT
1799 struct tcf_proto __rcu *ingress_cl_list;
1801 struct netdev_queue __rcu *ingress_queue;
1802 #ifdef CONFIG_NETFILTER_INGRESS
1803 struct nf_hook_entry __rcu *nf_hooks_ingress;
1806 unsigned char broadcast[MAX_ADDR_LEN];
1807 #ifdef CONFIG_RFS_ACCEL
1808 struct cpu_rmap *rx_cpu_rmap;
1810 struct hlist_node index_hlist;
1813 * Cache lines mostly used on transmit path
1815 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1816 unsigned int num_tx_queues;
1817 unsigned int real_num_tx_queues;
1818 struct Qdisc *qdisc;
1819 #ifdef CONFIG_NET_SCHED
1820 DECLARE_HASHTABLE (qdisc_hash, 4);
1822 unsigned long tx_queue_len;
1823 spinlock_t tx_global_lock;
1827 struct xps_dev_maps __rcu *xps_maps;
1829 #ifdef CONFIG_NET_CLS_ACT
1830 struct tcf_proto __rcu *egress_cl_list;
1833 /* These may be needed for future network-power-down code. */
1834 struct timer_list watchdog_timer;
1836 int __percpu *pcpu_refcnt;
1837 struct list_head todo_list;
1839 struct list_head link_watch_list;
1841 enum { NETREG_UNINITIALIZED=0,
1842 NETREG_REGISTERED, /* completed register_netdevice */
1843 NETREG_UNREGISTERING, /* called unregister_netdevice */
1844 NETREG_UNREGISTERED, /* completed unregister todo */
1845 NETREG_RELEASED, /* called free_netdev */
1846 NETREG_DUMMY, /* dummy device for NAPI poll */
1852 RTNL_LINK_INITIALIZED,
1853 RTNL_LINK_INITIALIZING,
1854 } rtnl_link_state:16;
1856 void (*destructor)(struct net_device *dev);
1858 #ifdef CONFIG_NETPOLL
1859 struct netpoll_info __rcu *npinfo;
1862 possible_net_t nd_net;
1864 /* mid-layer private */
1867 struct pcpu_lstats __percpu *lstats;
1868 struct pcpu_sw_netstats __percpu *tstats;
1869 struct pcpu_dstats __percpu *dstats;
1870 struct pcpu_vstats __percpu *vstats;
1873 #if IS_ENABLED(CONFIG_GARP)
1874 struct garp_port __rcu *garp_port;
1876 #if IS_ENABLED(CONFIG_MRP)
1877 struct mrp_port __rcu *mrp_port;
1881 const struct attribute_group *sysfs_groups[4];
1882 const struct attribute_group *sysfs_rx_queue_group;
1884 const struct rtnl_link_ops *rtnl_link_ops;
1886 /* for setting kernel sock attribute on TCP connection setup */
1887 #define GSO_MAX_SIZE 65536
1888 unsigned int gso_max_size;
1889 #define GSO_MAX_SEGS 65535
1893 const struct dcbnl_rtnl_ops *dcbnl_ops;
1896 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1897 u8 prio_tc_map[TC_BITMASK + 1];
1899 #if IS_ENABLED(CONFIG_FCOE)
1900 unsigned int fcoe_ddp_xid;
1902 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1903 struct netprio_map __rcu *priomap;
1905 struct phy_device *phydev;
1906 struct lock_class_key *qdisc_tx_busylock;
1907 struct lock_class_key *qdisc_running_key;
1910 #define to_net_dev(d) container_of(d, struct net_device, dev)
1912 static inline bool netif_elide_gro(const struct net_device *dev)
1914 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1919 #define NETDEV_ALIGN 32
1922 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1924 return dev->prio_tc_map[prio & TC_BITMASK];
1928 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1930 if (tc >= dev->num_tc)
1933 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1937 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1938 void netdev_reset_tc(struct net_device *dev);
1939 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1940 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1943 int netdev_get_num_tc(struct net_device *dev)
1949 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1952 return &dev->_tx[index];
1955 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1956 const struct sk_buff *skb)
1958 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1961 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1962 void (*f)(struct net_device *,
1963 struct netdev_queue *,
1969 for (i = 0; i < dev->num_tx_queues; i++)
1970 f(dev, &dev->_tx[i], arg);
1973 #define netdev_lockdep_set_classes(dev) \
1975 static struct lock_class_key qdisc_tx_busylock_key; \
1976 static struct lock_class_key qdisc_running_key; \
1977 static struct lock_class_key qdisc_xmit_lock_key; \
1978 static struct lock_class_key dev_addr_list_lock_key; \
1981 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1982 (dev)->qdisc_running_key = &qdisc_running_key; \
1983 lockdep_set_class(&(dev)->addr_list_lock, \
1984 &dev_addr_list_lock_key); \
1985 for (i = 0; i < (dev)->num_tx_queues; i++) \
1986 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1987 &qdisc_xmit_lock_key); \
1990 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1991 struct sk_buff *skb,
1994 /* returns the headroom that the master device needs to take in account
1995 * when forwarding to this dev
1997 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1999 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2002 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2004 if (dev->netdev_ops->ndo_set_rx_headroom)
2005 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2008 /* set the device rx headroom to the dev's default */
2009 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2011 netdev_set_rx_headroom(dev, -1);
2015 * Net namespace inlines
2018 struct net *dev_net(const struct net_device *dev)
2020 return read_pnet(&dev->nd_net);
2024 void dev_net_set(struct net_device *dev, struct net *net)
2026 write_pnet(&dev->nd_net, net);
2030 * netdev_priv - access network device private data
2031 * @dev: network device
2033 * Get network device private data
2035 static inline void *netdev_priv(const struct net_device *dev)
2037 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2040 /* Set the sysfs physical device reference for the network logical device
2041 * if set prior to registration will cause a symlink during initialization.
2043 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2045 /* Set the sysfs device type for the network logical device to allow
2046 * fine-grained identification of different network device types. For
2047 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2049 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2051 /* Default NAPI poll() weight
2052 * Device drivers are strongly advised to not use bigger value
2054 #define NAPI_POLL_WEIGHT 64
2057 * netif_napi_add - initialize a NAPI context
2058 * @dev: network device
2059 * @napi: NAPI context
2060 * @poll: polling function
2061 * @weight: default weight
2063 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2064 * *any* of the other NAPI-related functions.
2066 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2067 int (*poll)(struct napi_struct *, int), int weight);
2070 * netif_tx_napi_add - initialize a NAPI context
2071 * @dev: network device
2072 * @napi: NAPI context
2073 * @poll: polling function
2074 * @weight: default weight
2076 * This variant of netif_napi_add() should be used from drivers using NAPI
2077 * to exclusively poll a TX queue.
2078 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2080 static inline void netif_tx_napi_add(struct net_device *dev,
2081 struct napi_struct *napi,
2082 int (*poll)(struct napi_struct *, int),
2085 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2086 netif_napi_add(dev, napi, poll, weight);
2090 * netif_napi_del - remove a NAPI context
2091 * @napi: NAPI context
2093 * netif_napi_del() removes a NAPI context from the network device NAPI list
2095 void netif_napi_del(struct napi_struct *napi);
2097 struct napi_gro_cb {
2098 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2101 /* Length of frag0. */
2102 unsigned int frag0_len;
2104 /* This indicates where we are processing relative to skb->data. */
2107 /* This is non-zero if the packet cannot be merged with the new skb. */
2110 /* Save the IP ID here and check when we get to the transport layer */
2113 /* Number of segments aggregated. */
2116 /* Start offset for remote checksum offload */
2117 u16 gro_remcsum_start;
2119 /* jiffies when first packet was created/queued */
2122 /* Used in ipv6_gro_receive() and foo-over-udp */
2125 /* This is non-zero if the packet may be of the same flow. */
2128 /* Used in tunnel GRO receive */
2131 /* GRO checksum is valid */
2134 /* Number of checksums via CHECKSUM_UNNECESSARY */
2139 #define NAPI_GRO_FREE 1
2140 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2142 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2145 /* Used in GRE, set in fou/gue_gro_receive */
2148 /* Used to determine if flush_id can be ignored */
2151 /* Number of gro_receive callbacks this packet already went through */
2152 u8 recursion_counter:4;
2156 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2159 /* used in skb_gro_receive() slow path */
2160 struct sk_buff *last;
2163 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2165 #define GRO_RECURSION_LIMIT 15
2166 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2168 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2171 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2172 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2173 struct sk_buff **head,
2174 struct sk_buff *skb)
2176 if (unlikely(gro_recursion_inc_test(skb))) {
2177 NAPI_GRO_CB(skb)->flush |= 1;
2181 return cb(head, skb);
2184 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2186 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2188 struct sk_buff **head,
2189 struct sk_buff *skb)
2191 if (unlikely(gro_recursion_inc_test(skb))) {
2192 NAPI_GRO_CB(skb)->flush |= 1;
2196 return cb(sk, head, skb);
2199 struct packet_type {
2200 __be16 type; /* This is really htons(ether_type). */
2201 struct net_device *dev; /* NULL is wildcarded here */
2202 int (*func) (struct sk_buff *,
2203 struct net_device *,
2204 struct packet_type *,
2205 struct net_device *);
2206 bool (*id_match)(struct packet_type *ptype,
2208 void *af_packet_priv;
2209 struct list_head list;
2212 struct offload_callbacks {
2213 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2214 netdev_features_t features);
2215 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2216 struct sk_buff *skb);
2217 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2220 struct packet_offload {
2221 __be16 type; /* This is really htons(ether_type). */
2223 struct offload_callbacks callbacks;
2224 struct list_head list;
2227 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2228 struct pcpu_sw_netstats {
2233 struct u64_stats_sync syncp;
2236 #define __netdev_alloc_pcpu_stats(type, gfp) \
2238 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2241 for_each_possible_cpu(__cpu) { \
2242 typeof(type) *stat; \
2243 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2244 u64_stats_init(&stat->syncp); \
2250 #define netdev_alloc_pcpu_stats(type) \
2251 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2253 enum netdev_lag_tx_type {
2254 NETDEV_LAG_TX_TYPE_UNKNOWN,
2255 NETDEV_LAG_TX_TYPE_RANDOM,
2256 NETDEV_LAG_TX_TYPE_BROADCAST,
2257 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2258 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2259 NETDEV_LAG_TX_TYPE_HASH,
2262 struct netdev_lag_upper_info {
2263 enum netdev_lag_tx_type tx_type;
2266 struct netdev_lag_lower_state_info {
2271 #include <linux/notifier.h>
2273 /* netdevice notifier chain. Please remember to update the rtnetlink
2274 * notification exclusion list in rtnetlink_event() when adding new
2277 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2278 #define NETDEV_DOWN 0x0002
2279 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2280 detected a hardware crash and restarted
2281 - we can use this eg to kick tcp sessions
2283 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2284 #define NETDEV_REGISTER 0x0005
2285 #define NETDEV_UNREGISTER 0x0006
2286 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2287 #define NETDEV_CHANGEADDR 0x0008
2288 #define NETDEV_GOING_DOWN 0x0009
2289 #define NETDEV_CHANGENAME 0x000A
2290 #define NETDEV_FEAT_CHANGE 0x000B
2291 #define NETDEV_BONDING_FAILOVER 0x000C
2292 #define NETDEV_PRE_UP 0x000D
2293 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2294 #define NETDEV_POST_TYPE_CHANGE 0x000F
2295 #define NETDEV_POST_INIT 0x0010
2296 #define NETDEV_UNREGISTER_FINAL 0x0011
2297 #define NETDEV_RELEASE 0x0012
2298 #define NETDEV_NOTIFY_PEERS 0x0013
2299 #define NETDEV_JOIN 0x0014
2300 #define NETDEV_CHANGEUPPER 0x0015
2301 #define NETDEV_RESEND_IGMP 0x0016
2302 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2303 #define NETDEV_CHANGEINFODATA 0x0018
2304 #define NETDEV_BONDING_INFO 0x0019
2305 #define NETDEV_PRECHANGEUPPER 0x001A
2306 #define NETDEV_CHANGELOWERSTATE 0x001B
2307 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2308 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2310 int register_netdevice_notifier(struct notifier_block *nb);
2311 int unregister_netdevice_notifier(struct notifier_block *nb);
2313 struct netdev_notifier_info {
2314 struct net_device *dev;
2317 struct netdev_notifier_change_info {
2318 struct netdev_notifier_info info; /* must be first */
2319 unsigned int flags_changed;
2322 struct netdev_notifier_changeupper_info {
2323 struct netdev_notifier_info info; /* must be first */
2324 struct net_device *upper_dev; /* new upper dev */
2325 bool master; /* is upper dev master */
2326 bool linking; /* is the notification for link or unlink */
2327 void *upper_info; /* upper dev info */
2330 struct netdev_notifier_changelowerstate_info {
2331 struct netdev_notifier_info info; /* must be first */
2332 void *lower_state_info; /* is lower dev state */
2335 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2336 struct net_device *dev)
2341 static inline struct net_device *
2342 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2347 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2350 extern rwlock_t dev_base_lock; /* Device list lock */
2352 #define for_each_netdev(net, d) \
2353 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2354 #define for_each_netdev_reverse(net, d) \
2355 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2356 #define for_each_netdev_rcu(net, d) \
2357 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2358 #define for_each_netdev_safe(net, d, n) \
2359 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2360 #define for_each_netdev_continue(net, d) \
2361 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2362 #define for_each_netdev_continue_rcu(net, d) \
2363 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2364 #define for_each_netdev_in_bond_rcu(bond, slave) \
2365 for_each_netdev_rcu(&init_net, slave) \
2366 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2367 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2369 static inline struct net_device *next_net_device(struct net_device *dev)
2371 struct list_head *lh;
2375 lh = dev->dev_list.next;
2376 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2379 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2381 struct list_head *lh;
2385 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2386 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2389 static inline struct net_device *first_net_device(struct net *net)
2391 return list_empty(&net->dev_base_head) ? NULL :
2392 net_device_entry(net->dev_base_head.next);
2395 static inline struct net_device *first_net_device_rcu(struct net *net)
2397 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2399 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2402 int netdev_boot_setup_check(struct net_device *dev);
2403 unsigned long netdev_boot_base(const char *prefix, int unit);
2404 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2405 const char *hwaddr);
2406 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2407 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2408 void dev_add_pack(struct packet_type *pt);
2409 void dev_remove_pack(struct packet_type *pt);
2410 void __dev_remove_pack(struct packet_type *pt);
2411 void dev_add_offload(struct packet_offload *po);
2412 void dev_remove_offload(struct packet_offload *po);
2414 int dev_get_iflink(const struct net_device *dev);
2415 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2416 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2417 unsigned short mask);
2418 struct net_device *dev_get_by_name(struct net *net, const char *name);
2419 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2420 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2421 int dev_alloc_name(struct net_device *dev, const char *name);
2422 int dev_open(struct net_device *dev);
2423 int dev_close(struct net_device *dev);
2424 int dev_close_many(struct list_head *head, bool unlink);
2425 void dev_disable_lro(struct net_device *dev);
2426 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2427 int dev_queue_xmit(struct sk_buff *skb);
2428 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2429 int register_netdevice(struct net_device *dev);
2430 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2431 void unregister_netdevice_many(struct list_head *head);
2432 static inline void unregister_netdevice(struct net_device *dev)
2434 unregister_netdevice_queue(dev, NULL);
2437 int netdev_refcnt_read(const struct net_device *dev);
2438 void free_netdev(struct net_device *dev);
2439 void netdev_freemem(struct net_device *dev);
2440 void synchronize_net(void);
2441 int init_dummy_netdev(struct net_device *dev);
2443 DECLARE_PER_CPU(int, xmit_recursion);
2444 #define XMIT_RECURSION_LIMIT 10
2446 static inline int dev_recursion_level(void)
2448 return this_cpu_read(xmit_recursion);
2451 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2452 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2453 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2454 int netdev_get_name(struct net *net, char *name, int ifindex);
2455 int dev_restart(struct net_device *dev);
2456 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2458 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2460 return NAPI_GRO_CB(skb)->data_offset;
2463 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2465 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2468 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2470 NAPI_GRO_CB(skb)->data_offset += len;
2473 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2474 unsigned int offset)
2476 return NAPI_GRO_CB(skb)->frag0 + offset;
2479 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2481 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2484 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2486 NAPI_GRO_CB(skb)->frag0 = NULL;
2487 NAPI_GRO_CB(skb)->frag0_len = 0;
2490 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2491 unsigned int offset)
2493 if (!pskb_may_pull(skb, hlen))
2496 skb_gro_frag0_invalidate(skb);
2497 return skb->data + offset;
2500 static inline void *skb_gro_network_header(struct sk_buff *skb)
2502 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2503 skb_network_offset(skb);
2506 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2507 const void *start, unsigned int len)
2509 if (NAPI_GRO_CB(skb)->csum_valid)
2510 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2511 csum_partial(start, len, 0));
2514 /* GRO checksum functions. These are logical equivalents of the normal
2515 * checksum functions (in skbuff.h) except that they operate on the GRO
2516 * offsets and fields in sk_buff.
2519 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2521 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2523 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2526 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2530 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2531 skb_checksum_start_offset(skb) <
2532 skb_gro_offset(skb)) &&
2533 !skb_at_gro_remcsum_start(skb) &&
2534 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2535 (!zero_okay || check));
2538 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2541 if (NAPI_GRO_CB(skb)->csum_valid &&
2542 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2545 NAPI_GRO_CB(skb)->csum = psum;
2547 return __skb_gro_checksum_complete(skb);
2550 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2552 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2553 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2554 NAPI_GRO_CB(skb)->csum_cnt--;
2556 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2557 * verified a new top level checksum or an encapsulated one
2558 * during GRO. This saves work if we fallback to normal path.
2560 __skb_incr_checksum_unnecessary(skb);
2564 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2567 __sum16 __ret = 0; \
2568 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2569 __ret = __skb_gro_checksum_validate_complete(skb, \
2570 compute_pseudo(skb, proto)); \
2572 __skb_mark_checksum_bad(skb); \
2574 skb_gro_incr_csum_unnecessary(skb); \
2578 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2579 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2581 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2583 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2585 #define skb_gro_checksum_simple_validate(skb) \
2586 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2588 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2590 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2591 !NAPI_GRO_CB(skb)->csum_valid);
2594 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2595 __sum16 check, __wsum pseudo)
2597 NAPI_GRO_CB(skb)->csum = ~pseudo;
2598 NAPI_GRO_CB(skb)->csum_valid = 1;
2601 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2603 if (__skb_gro_checksum_convert_check(skb)) \
2604 __skb_gro_checksum_convert(skb, check, \
2605 compute_pseudo(skb, proto)); \
2608 struct gro_remcsum {
2613 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2619 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2620 unsigned int off, size_t hdrlen,
2621 int start, int offset,
2622 struct gro_remcsum *grc,
2626 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2628 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2631 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2635 ptr = skb_gro_header_fast(skb, off);
2636 if (skb_gro_header_hard(skb, off + plen)) {
2637 ptr = skb_gro_header_slow(skb, off + plen, off);
2642 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2645 /* Adjust skb->csum since we changed the packet */
2646 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2648 grc->offset = off + hdrlen + offset;
2654 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2655 struct gro_remcsum *grc)
2658 size_t plen = grc->offset + sizeof(u16);
2663 ptr = skb_gro_header_fast(skb, grc->offset);
2664 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2665 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2670 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2673 #ifdef CONFIG_XFRM_OFFLOAD
2674 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2676 if (PTR_ERR(pp) != -EINPROGRESS)
2677 NAPI_GRO_CB(skb)->flush |= flush;
2680 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2682 NAPI_GRO_CB(skb)->flush |= flush;
2686 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2687 unsigned short type,
2688 const void *daddr, const void *saddr,
2691 if (!dev->header_ops || !dev->header_ops->create)
2694 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2697 static inline int dev_parse_header(const struct sk_buff *skb,
2698 unsigned char *haddr)
2700 const struct net_device *dev = skb->dev;
2702 if (!dev->header_ops || !dev->header_ops->parse)
2704 return dev->header_ops->parse(skb, haddr);
2707 /* ll_header must have at least hard_header_len allocated */
2708 static inline bool dev_validate_header(const struct net_device *dev,
2709 char *ll_header, int len)
2711 if (likely(len >= dev->hard_header_len))
2713 if (len < dev->min_header_len)
2716 if (capable(CAP_SYS_RAWIO)) {
2717 memset(ll_header + len, 0, dev->hard_header_len - len);
2721 if (dev->header_ops && dev->header_ops->validate)
2722 return dev->header_ops->validate(ll_header, len);
2727 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2728 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2729 static inline int unregister_gifconf(unsigned int family)
2731 return register_gifconf(family, NULL);
2734 #ifdef CONFIG_NET_FLOW_LIMIT
2735 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2736 struct sd_flow_limit {
2738 unsigned int num_buckets;
2739 unsigned int history_head;
2740 u16 history[FLOW_LIMIT_HISTORY];
2744 extern int netdev_flow_limit_table_len;
2745 #endif /* CONFIG_NET_FLOW_LIMIT */
2748 * Incoming packets are placed on per-CPU queues
2750 struct softnet_data {
2751 struct list_head poll_list;
2752 struct sk_buff_head process_queue;
2755 unsigned int processed;
2756 unsigned int time_squeeze;
2757 unsigned int received_rps;
2759 struct softnet_data *rps_ipi_list;
2761 #ifdef CONFIG_NET_FLOW_LIMIT
2762 struct sd_flow_limit __rcu *flow_limit;
2764 struct Qdisc *output_queue;
2765 struct Qdisc **output_queue_tailp;
2766 struct sk_buff *completion_queue;
2769 /* input_queue_head should be written by cpu owning this struct,
2770 * and only read by other cpus. Worth using a cache line.
2772 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2774 /* Elements below can be accessed between CPUs for RPS/RFS */
2775 struct call_single_data csd ____cacheline_aligned_in_smp;
2776 struct softnet_data *rps_ipi_next;
2778 unsigned int input_queue_tail;
2780 unsigned int dropped;
2781 struct sk_buff_head input_pkt_queue;
2782 struct napi_struct backlog;
2786 static inline void input_queue_head_incr(struct softnet_data *sd)
2789 sd->input_queue_head++;
2793 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2794 unsigned int *qtail)
2797 *qtail = ++sd->input_queue_tail;
2801 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2803 void __netif_schedule(struct Qdisc *q);
2804 void netif_schedule_queue(struct netdev_queue *txq);
2806 static inline void netif_tx_schedule_all(struct net_device *dev)
2810 for (i = 0; i < dev->num_tx_queues; i++)
2811 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2814 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2816 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2820 * netif_start_queue - allow transmit
2821 * @dev: network device
2823 * Allow upper layers to call the device hard_start_xmit routine.
2825 static inline void netif_start_queue(struct net_device *dev)
2827 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2830 static inline void netif_tx_start_all_queues(struct net_device *dev)
2834 for (i = 0; i < dev->num_tx_queues; i++) {
2835 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2836 netif_tx_start_queue(txq);
2840 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2843 * netif_wake_queue - restart transmit
2844 * @dev: network device
2846 * Allow upper layers to call the device hard_start_xmit routine.
2847 * Used for flow control when transmit resources are available.
2849 static inline void netif_wake_queue(struct net_device *dev)
2851 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2854 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2858 for (i = 0; i < dev->num_tx_queues; i++) {
2859 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2860 netif_tx_wake_queue(txq);
2864 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2866 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2870 * netif_stop_queue - stop transmitted packets
2871 * @dev: network device
2873 * Stop upper layers calling the device hard_start_xmit routine.
2874 * Used for flow control when transmit resources are unavailable.
2876 static inline void netif_stop_queue(struct net_device *dev)
2878 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2881 void netif_tx_stop_all_queues(struct net_device *dev);
2883 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2885 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2889 * netif_queue_stopped - test if transmit queue is flowblocked
2890 * @dev: network device
2892 * Test if transmit queue on device is currently unable to send.
2894 static inline bool netif_queue_stopped(const struct net_device *dev)
2896 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2899 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2901 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2905 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2907 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2911 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2913 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2917 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2918 * @dev_queue: pointer to transmit queue
2920 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2921 * to give appropriate hint to the CPU.
2923 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2926 prefetchw(&dev_queue->dql.num_queued);
2931 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2932 * @dev_queue: pointer to transmit queue
2934 * BQL enabled drivers might use this helper in their TX completion path,
2935 * to give appropriate hint to the CPU.
2937 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2940 prefetchw(&dev_queue->dql.limit);
2944 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2948 dql_queued(&dev_queue->dql, bytes);
2950 if (likely(dql_avail(&dev_queue->dql) >= 0))
2953 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2956 * The XOFF flag must be set before checking the dql_avail below,
2957 * because in netdev_tx_completed_queue we update the dql_completed
2958 * before checking the XOFF flag.
2962 /* check again in case another CPU has just made room avail */
2963 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2964 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2969 * netdev_sent_queue - report the number of bytes queued to hardware
2970 * @dev: network device
2971 * @bytes: number of bytes queued to the hardware device queue
2973 * Report the number of bytes queued for sending/completion to the network
2974 * device hardware queue. @bytes should be a good approximation and should
2975 * exactly match netdev_completed_queue() @bytes
2977 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2979 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2982 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2983 unsigned int pkts, unsigned int bytes)
2986 if (unlikely(!bytes))
2989 dql_completed(&dev_queue->dql, bytes);
2992 * Without the memory barrier there is a small possiblity that
2993 * netdev_tx_sent_queue will miss the update and cause the queue to
2994 * be stopped forever
2998 if (dql_avail(&dev_queue->dql) < 0)
3001 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3002 netif_schedule_queue(dev_queue);
3007 * netdev_completed_queue - report bytes and packets completed by device
3008 * @dev: network device
3009 * @pkts: actual number of packets sent over the medium
3010 * @bytes: actual number of bytes sent over the medium
3012 * Report the number of bytes and packets transmitted by the network device
3013 * hardware queue over the physical medium, @bytes must exactly match the
3014 * @bytes amount passed to netdev_sent_queue()
3016 static inline void netdev_completed_queue(struct net_device *dev,
3017 unsigned int pkts, unsigned int bytes)
3019 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3022 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3025 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3031 * netdev_reset_queue - reset the packets and bytes count of a network device
3032 * @dev_queue: network device
3034 * Reset the bytes and packet count of a network device and clear the
3035 * software flow control OFF bit for this network device
3037 static inline void netdev_reset_queue(struct net_device *dev_queue)
3039 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3043 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3044 * @dev: network device
3045 * @queue_index: given tx queue index
3047 * Returns 0 if given tx queue index >= number of device tx queues,
3048 * otherwise returns the originally passed tx queue index.
3050 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3052 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3053 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3054 dev->name, queue_index,
3055 dev->real_num_tx_queues);
3063 * netif_running - test if up
3064 * @dev: network device
3066 * Test if the device has been brought up.
3068 static inline bool netif_running(const struct net_device *dev)
3070 return test_bit(__LINK_STATE_START, &dev->state);
3074 * Routines to manage the subqueues on a device. We only need start,
3075 * stop, and a check if it's stopped. All other device management is
3076 * done at the overall netdevice level.
3077 * Also test the device if we're multiqueue.
3081 * netif_start_subqueue - allow sending packets on subqueue
3082 * @dev: network device
3083 * @queue_index: sub queue index
3085 * Start individual transmit queue of a device with multiple transmit queues.
3087 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3089 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3091 netif_tx_start_queue(txq);
3095 * netif_stop_subqueue - stop sending packets on subqueue
3096 * @dev: network device
3097 * @queue_index: sub queue index
3099 * Stop individual transmit queue of a device with multiple transmit queues.
3101 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3103 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3104 netif_tx_stop_queue(txq);
3108 * netif_subqueue_stopped - test status of subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3112 * Check individual transmit queue of a device with multiple transmit queues.
3114 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3117 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3119 return netif_tx_queue_stopped(txq);
3122 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3123 struct sk_buff *skb)
3125 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3129 * netif_wake_subqueue - allow sending packets on subqueue
3130 * @dev: network device
3131 * @queue_index: sub queue index
3133 * Resume individual transmit queue of a device with multiple transmit queues.
3135 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3137 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3139 netif_tx_wake_queue(txq);
3143 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3146 static inline int netif_set_xps_queue(struct net_device *dev,
3147 const struct cpumask *mask,
3154 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3155 unsigned int num_tx_queues);
3158 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3159 * as a distribution range limit for the returned value.
3161 static inline u16 skb_tx_hash(const struct net_device *dev,
3162 struct sk_buff *skb)
3164 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3168 * netif_is_multiqueue - test if device has multiple transmit queues
3169 * @dev: network device
3171 * Check if device has multiple transmit queues
3173 static inline bool netif_is_multiqueue(const struct net_device *dev)
3175 return dev->num_tx_queues > 1;
3178 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3181 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3183 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3191 static inline unsigned int get_netdev_rx_queue_index(
3192 struct netdev_rx_queue *queue)
3194 struct net_device *dev = queue->dev;
3195 int index = queue - dev->_rx;
3197 BUG_ON(index >= dev->num_rx_queues);
3202 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3203 int netif_get_num_default_rss_queues(void);
3205 enum skb_free_reason {
3206 SKB_REASON_CONSUMED,
3210 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3211 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3214 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3215 * interrupt context or with hardware interrupts being disabled.
3216 * (in_irq() || irqs_disabled())
3218 * We provide four helpers that can be used in following contexts :
3220 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3221 * replacing kfree_skb(skb)
3223 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3224 * Typically used in place of consume_skb(skb) in TX completion path
3226 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3227 * replacing kfree_skb(skb)
3229 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3230 * and consumed a packet. Used in place of consume_skb(skb)
3232 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3234 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3237 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3239 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3242 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3244 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3247 static inline void dev_consume_skb_any(struct sk_buff *skb)
3249 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3252 int netif_rx(struct sk_buff *skb);
3253 int netif_rx_ni(struct sk_buff *skb);
3254 int netif_receive_skb(struct sk_buff *skb);
3255 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3256 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3257 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3258 gro_result_t napi_gro_frags(struct napi_struct *napi);
3259 struct packet_offload *gro_find_receive_by_type(__be16 type);
3260 struct packet_offload *gro_find_complete_by_type(__be16 type);
3262 static inline void napi_free_frags(struct napi_struct *napi)
3264 kfree_skb(napi->skb);
3268 bool netdev_is_rx_handler_busy(struct net_device *dev);
3269 int netdev_rx_handler_register(struct net_device *dev,
3270 rx_handler_func_t *rx_handler,
3271 void *rx_handler_data);
3272 void netdev_rx_handler_unregister(struct net_device *dev);
3274 bool dev_valid_name(const char *name);
3275 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3276 int dev_ethtool(struct net *net, struct ifreq *);
3277 unsigned int dev_get_flags(const struct net_device *);
3278 int __dev_change_flags(struct net_device *, unsigned int flags);
3279 int dev_change_flags(struct net_device *, unsigned int);
3280 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3281 unsigned int gchanges);
3282 int dev_change_name(struct net_device *, const char *);
3283 int dev_set_alias(struct net_device *, const char *, size_t);
3284 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3285 int dev_set_mtu(struct net_device *, int);
3286 void dev_set_group(struct net_device *, int);
3287 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3288 int dev_change_carrier(struct net_device *, bool new_carrier);
3289 int dev_get_phys_port_id(struct net_device *dev,
3290 struct netdev_phys_item_id *ppid);
3291 int dev_get_phys_port_name(struct net_device *dev,
3292 char *name, size_t len);
3293 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3294 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
3295 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3296 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3297 struct netdev_queue *txq, int *ret);
3298 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3299 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3300 bool is_skb_forwardable(const struct net_device *dev,
3301 const struct sk_buff *skb);
3303 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3304 struct sk_buff *skb)
3306 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3307 unlikely(!is_skb_forwardable(dev, skb))) {
3308 atomic_long_inc(&dev->rx_dropped);
3313 skb_scrub_packet(skb, true);
3318 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3320 extern int netdev_budget;
3321 extern unsigned int netdev_budget_usecs;
3323 /* Called by rtnetlink.c:rtnl_unlock() */
3324 void netdev_run_todo(void);
3327 * dev_put - release reference to device
3328 * @dev: network device
3330 * Release reference to device to allow it to be freed.
3332 static inline void dev_put(struct net_device *dev)
3334 this_cpu_dec(*dev->pcpu_refcnt);
3338 * dev_hold - get reference to device
3339 * @dev: network device
3341 * Hold reference to device to keep it from being freed.
3343 static inline void dev_hold(struct net_device *dev)
3345 this_cpu_inc(*dev->pcpu_refcnt);
3348 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3349 * and _off may be called from IRQ context, but it is caller
3350 * who is responsible for serialization of these calls.
3352 * The name carrier is inappropriate, these functions should really be
3353 * called netif_lowerlayer_*() because they represent the state of any
3354 * kind of lower layer not just hardware media.
3357 void linkwatch_init_dev(struct net_device *dev);
3358 void linkwatch_fire_event(struct net_device *dev);
3359 void linkwatch_forget_dev(struct net_device *dev);
3362 * netif_carrier_ok - test if carrier present
3363 * @dev: network device
3365 * Check if carrier is present on device
3367 static inline bool netif_carrier_ok(const struct net_device *dev)
3369 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3372 unsigned long dev_trans_start(struct net_device *dev);
3374 void __netdev_watchdog_up(struct net_device *dev);
3376 void netif_carrier_on(struct net_device *dev);
3378 void netif_carrier_off(struct net_device *dev);
3381 * netif_dormant_on - mark device as dormant.
3382 * @dev: network device
3384 * Mark device as dormant (as per RFC2863).
3386 * The dormant state indicates that the relevant interface is not
3387 * actually in a condition to pass packets (i.e., it is not 'up') but is
3388 * in a "pending" state, waiting for some external event. For "on-
3389 * demand" interfaces, this new state identifies the situation where the
3390 * interface is waiting for events to place it in the up state.
3392 static inline void netif_dormant_on(struct net_device *dev)
3394 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3395 linkwatch_fire_event(dev);
3399 * netif_dormant_off - set device as not dormant.
3400 * @dev: network device
3402 * Device is not in dormant state.
3404 static inline void netif_dormant_off(struct net_device *dev)
3406 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3407 linkwatch_fire_event(dev);
3411 * netif_dormant - test if carrier present
3412 * @dev: network device
3414 * Check if carrier is present on device
3416 static inline bool netif_dormant(const struct net_device *dev)
3418 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3423 * netif_oper_up - test if device is operational
3424 * @dev: network device
3426 * Check if carrier is operational
3428 static inline bool netif_oper_up(const struct net_device *dev)
3430 return (dev->operstate == IF_OPER_UP ||
3431 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3435 * netif_device_present - is device available or removed
3436 * @dev: network device
3438 * Check if device has not been removed from system.
3440 static inline bool netif_device_present(struct net_device *dev)
3442 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3445 void netif_device_detach(struct net_device *dev);
3447 void netif_device_attach(struct net_device *dev);
3450 * Network interface message level settings
3454 NETIF_MSG_DRV = 0x0001,
3455 NETIF_MSG_PROBE = 0x0002,
3456 NETIF_MSG_LINK = 0x0004,
3457 NETIF_MSG_TIMER = 0x0008,
3458 NETIF_MSG_IFDOWN = 0x0010,
3459 NETIF_MSG_IFUP = 0x0020,
3460 NETIF_MSG_RX_ERR = 0x0040,
3461 NETIF_MSG_TX_ERR = 0x0080,
3462 NETIF_MSG_TX_QUEUED = 0x0100,
3463 NETIF_MSG_INTR = 0x0200,
3464 NETIF_MSG_TX_DONE = 0x0400,
3465 NETIF_MSG_RX_STATUS = 0x0800,
3466 NETIF_MSG_PKTDATA = 0x1000,
3467 NETIF_MSG_HW = 0x2000,
3468 NETIF_MSG_WOL = 0x4000,
3471 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3472 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3473 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3474 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3475 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3476 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3477 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3478 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3479 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3480 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3481 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3482 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3483 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3484 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3485 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3487 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3490 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3491 return default_msg_enable_bits;
3492 if (debug_value == 0) /* no output */
3494 /* set low N bits */
3495 return (1 << debug_value) - 1;
3498 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3500 spin_lock(&txq->_xmit_lock);
3501 txq->xmit_lock_owner = cpu;
3504 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3506 __acquire(&txq->_xmit_lock);
3510 static inline void __netif_tx_release(struct netdev_queue *txq)
3512 __release(&txq->_xmit_lock);
3515 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3517 spin_lock_bh(&txq->_xmit_lock);
3518 txq->xmit_lock_owner = smp_processor_id();
3521 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3523 bool ok = spin_trylock(&txq->_xmit_lock);
3525 txq->xmit_lock_owner = smp_processor_id();
3529 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3531 txq->xmit_lock_owner = -1;
3532 spin_unlock(&txq->_xmit_lock);
3535 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3537 txq->xmit_lock_owner = -1;
3538 spin_unlock_bh(&txq->_xmit_lock);
3541 static inline void txq_trans_update(struct netdev_queue *txq)
3543 if (txq->xmit_lock_owner != -1)
3544 txq->trans_start = jiffies;
3547 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3548 static inline void netif_trans_update(struct net_device *dev)
3550 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3552 if (txq->trans_start != jiffies)
3553 txq->trans_start = jiffies;
3557 * netif_tx_lock - grab network device transmit lock
3558 * @dev: network device
3560 * Get network device transmit lock
3562 static inline void netif_tx_lock(struct net_device *dev)
3567 spin_lock(&dev->tx_global_lock);
3568 cpu = smp_processor_id();
3569 for (i = 0; i < dev->num_tx_queues; i++) {
3570 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3572 /* We are the only thread of execution doing a
3573 * freeze, but we have to grab the _xmit_lock in
3574 * order to synchronize with threads which are in
3575 * the ->hard_start_xmit() handler and already
3576 * checked the frozen bit.
3578 __netif_tx_lock(txq, cpu);
3579 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 __netif_tx_unlock(txq);
3584 static inline void netif_tx_lock_bh(struct net_device *dev)
3590 static inline void netif_tx_unlock(struct net_device *dev)
3594 for (i = 0; i < dev->num_tx_queues; i++) {
3595 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3597 /* No need to grab the _xmit_lock here. If the
3598 * queue is not stopped for another reason, we
3601 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3602 netif_schedule_queue(txq);
3604 spin_unlock(&dev->tx_global_lock);
3607 static inline void netif_tx_unlock_bh(struct net_device *dev)
3609 netif_tx_unlock(dev);
3613 #define HARD_TX_LOCK(dev, txq, cpu) { \
3614 if ((dev->features & NETIF_F_LLTX) == 0) { \
3615 __netif_tx_lock(txq, cpu); \
3617 __netif_tx_acquire(txq); \
3621 #define HARD_TX_TRYLOCK(dev, txq) \
3622 (((dev->features & NETIF_F_LLTX) == 0) ? \
3623 __netif_tx_trylock(txq) : \
3624 __netif_tx_acquire(txq))
3626 #define HARD_TX_UNLOCK(dev, txq) { \
3627 if ((dev->features & NETIF_F_LLTX) == 0) { \
3628 __netif_tx_unlock(txq); \
3630 __netif_tx_release(txq); \
3634 static inline void netif_tx_disable(struct net_device *dev)
3640 cpu = smp_processor_id();
3641 for (i = 0; i < dev->num_tx_queues; i++) {
3642 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3644 __netif_tx_lock(txq, cpu);
3645 netif_tx_stop_queue(txq);
3646 __netif_tx_unlock(txq);
3651 static inline void netif_addr_lock(struct net_device *dev)
3653 spin_lock(&dev->addr_list_lock);
3656 static inline void netif_addr_lock_nested(struct net_device *dev)
3658 int subclass = SINGLE_DEPTH_NESTING;
3660 if (dev->netdev_ops->ndo_get_lock_subclass)
3661 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3663 spin_lock_nested(&dev->addr_list_lock, subclass);
3666 static inline void netif_addr_lock_bh(struct net_device *dev)
3668 spin_lock_bh(&dev->addr_list_lock);
3671 static inline void netif_addr_unlock(struct net_device *dev)
3673 spin_unlock(&dev->addr_list_lock);
3676 static inline void netif_addr_unlock_bh(struct net_device *dev)
3678 spin_unlock_bh(&dev->addr_list_lock);
3682 * dev_addrs walker. Should be used only for read access. Call with
3683 * rcu_read_lock held.
3685 #define for_each_dev_addr(dev, ha) \
3686 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3688 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3690 void ether_setup(struct net_device *dev);
3692 /* Support for loadable net-drivers */
3693 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3694 unsigned char name_assign_type,
3695 void (*setup)(struct net_device *),
3696 unsigned int txqs, unsigned int rxqs);
3697 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3698 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3700 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3701 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3704 int register_netdev(struct net_device *dev);
3705 void unregister_netdev(struct net_device *dev);
3707 /* General hardware address lists handling functions */
3708 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3709 struct netdev_hw_addr_list *from_list, int addr_len);
3710 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3711 struct netdev_hw_addr_list *from_list, int addr_len);
3712 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3713 struct net_device *dev,
3714 int (*sync)(struct net_device *, const unsigned char *),
3715 int (*unsync)(struct net_device *,
3716 const unsigned char *));
3717 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3718 struct net_device *dev,
3719 int (*unsync)(struct net_device *,
3720 const unsigned char *));
3721 void __hw_addr_init(struct netdev_hw_addr_list *list);
3723 /* Functions used for device addresses handling */
3724 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3725 unsigned char addr_type);
3726 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3727 unsigned char addr_type);
3728 void dev_addr_flush(struct net_device *dev);
3729 int dev_addr_init(struct net_device *dev);
3731 /* Functions used for unicast addresses handling */
3732 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3733 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3734 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3735 int dev_uc_sync(struct net_device *to, struct net_device *from);
3736 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3737 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3738 void dev_uc_flush(struct net_device *dev);
3739 void dev_uc_init(struct net_device *dev);
3742 * __dev_uc_sync - Synchonize device's unicast list
3743 * @dev: device to sync
3744 * @sync: function to call if address should be added
3745 * @unsync: function to call if address should be removed
3747 * Add newly added addresses to the interface, and release
3748 * addresses that have been deleted.
3750 static inline int __dev_uc_sync(struct net_device *dev,
3751 int (*sync)(struct net_device *,
3752 const unsigned char *),
3753 int (*unsync)(struct net_device *,
3754 const unsigned char *))
3756 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3760 * __dev_uc_unsync - Remove synchronized addresses from device
3761 * @dev: device to sync
3762 * @unsync: function to call if address should be removed
3764 * Remove all addresses that were added to the device by dev_uc_sync().
3766 static inline void __dev_uc_unsync(struct net_device *dev,
3767 int (*unsync)(struct net_device *,
3768 const unsigned char *))
3770 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3773 /* Functions used for multicast addresses handling */
3774 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3775 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3776 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3777 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3778 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3779 int dev_mc_sync(struct net_device *to, struct net_device *from);
3780 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3781 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3782 void dev_mc_flush(struct net_device *dev);
3783 void dev_mc_init(struct net_device *dev);
3786 * __dev_mc_sync - Synchonize device's multicast list
3787 * @dev: device to sync
3788 * @sync: function to call if address should be added
3789 * @unsync: function to call if address should be removed
3791 * Add newly added addresses to the interface, and release
3792 * addresses that have been deleted.
3794 static inline int __dev_mc_sync(struct net_device *dev,
3795 int (*sync)(struct net_device *,
3796 const unsigned char *),
3797 int (*unsync)(struct net_device *,
3798 const unsigned char *))
3800 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3804 * __dev_mc_unsync - Remove synchronized addresses from device
3805 * @dev: device to sync
3806 * @unsync: function to call if address should be removed
3808 * Remove all addresses that were added to the device by dev_mc_sync().
3810 static inline void __dev_mc_unsync(struct net_device *dev,
3811 int (*unsync)(struct net_device *,
3812 const unsigned char *))
3814 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3817 /* Functions used for secondary unicast and multicast support */
3818 void dev_set_rx_mode(struct net_device *dev);
3819 void __dev_set_rx_mode(struct net_device *dev);
3820 int dev_set_promiscuity(struct net_device *dev, int inc);
3821 int dev_set_allmulti(struct net_device *dev, int inc);
3822 void netdev_state_change(struct net_device *dev);
3823 void netdev_notify_peers(struct net_device *dev);
3824 void netdev_features_change(struct net_device *dev);
3825 /* Load a device via the kmod */
3826 void dev_load(struct net *net, const char *name);
3827 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3828 struct rtnl_link_stats64 *storage);
3829 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3830 const struct net_device_stats *netdev_stats);
3832 extern int netdev_max_backlog;
3833 extern int netdev_tstamp_prequeue;
3834 extern int weight_p;
3835 extern int dev_weight_rx_bias;
3836 extern int dev_weight_tx_bias;
3837 extern int dev_rx_weight;
3838 extern int dev_tx_weight;
3840 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3841 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3842 struct list_head **iter);
3843 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3844 struct list_head **iter);
3846 /* iterate through upper list, must be called under RCU read lock */
3847 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3848 for (iter = &(dev)->adj_list.upper, \
3849 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3851 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3853 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3854 int (*fn)(struct net_device *upper_dev,
3858 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3859 struct net_device *upper_dev);
3861 void *netdev_lower_get_next_private(struct net_device *dev,
3862 struct list_head **iter);
3863 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3864 struct list_head **iter);
3866 #define netdev_for_each_lower_private(dev, priv, iter) \
3867 for (iter = (dev)->adj_list.lower.next, \
3868 priv = netdev_lower_get_next_private(dev, &(iter)); \
3870 priv = netdev_lower_get_next_private(dev, &(iter)))
3872 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3873 for (iter = &(dev)->adj_list.lower, \
3874 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3876 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3878 void *netdev_lower_get_next(struct net_device *dev,
3879 struct list_head **iter);
3881 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3882 for (iter = (dev)->adj_list.lower.next, \
3883 ldev = netdev_lower_get_next(dev, &(iter)); \
3885 ldev = netdev_lower_get_next(dev, &(iter)))
3887 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3888 struct list_head **iter);
3889 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3890 struct list_head **iter);
3892 int netdev_walk_all_lower_dev(struct net_device *dev,
3893 int (*fn)(struct net_device *lower_dev,
3896 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3897 int (*fn)(struct net_device *lower_dev,
3901 void *netdev_adjacent_get_private(struct list_head *adj_list);
3902 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3903 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3904 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3905 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3906 int netdev_master_upper_dev_link(struct net_device *dev,
3907 struct net_device *upper_dev,
3908 void *upper_priv, void *upper_info);
3909 void netdev_upper_dev_unlink(struct net_device *dev,
3910 struct net_device *upper_dev);
3911 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3912 void *netdev_lower_dev_get_private(struct net_device *dev,
3913 struct net_device *lower_dev);
3914 void netdev_lower_state_changed(struct net_device *lower_dev,
3915 void *lower_state_info);
3917 /* RSS keys are 40 or 52 bytes long */
3918 #define NETDEV_RSS_KEY_LEN 52
3919 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3920 void netdev_rss_key_fill(void *buffer, size_t len);
3922 int dev_get_nest_level(struct net_device *dev);
3923 int skb_checksum_help(struct sk_buff *skb);
3924 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3925 netdev_features_t features, bool tx_path);
3926 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3927 netdev_features_t features);
3929 struct netdev_bonding_info {
3934 struct netdev_notifier_bonding_info {
3935 struct netdev_notifier_info info; /* must be first */
3936 struct netdev_bonding_info bonding_info;
3939 void netdev_bonding_info_change(struct net_device *dev,
3940 struct netdev_bonding_info *bonding_info);
3943 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3945 return __skb_gso_segment(skb, features, true);
3947 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3949 static inline bool can_checksum_protocol(netdev_features_t features,
3952 if (protocol == htons(ETH_P_FCOE))
3953 return !!(features & NETIF_F_FCOE_CRC);
3955 /* Assume this is an IP checksum (not SCTP CRC) */
3957 if (features & NETIF_F_HW_CSUM) {
3958 /* Can checksum everything */
3963 case htons(ETH_P_IP):
3964 return !!(features & NETIF_F_IP_CSUM);
3965 case htons(ETH_P_IPV6):
3966 return !!(features & NETIF_F_IPV6_CSUM);
3973 void netdev_rx_csum_fault(struct net_device *dev);
3975 static inline void netdev_rx_csum_fault(struct net_device *dev)
3979 /* rx skb timestamps */
3980 void net_enable_timestamp(void);
3981 void net_disable_timestamp(void);
3983 #ifdef CONFIG_PROC_FS
3984 int __init dev_proc_init(void);
3986 #define dev_proc_init() 0
3989 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3990 struct sk_buff *skb, struct net_device *dev,
3993 skb->xmit_more = more ? 1 : 0;
3994 return ops->ndo_start_xmit(skb, dev);
3997 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3998 struct netdev_queue *txq, bool more)
4000 const struct net_device_ops *ops = dev->netdev_ops;
4003 rc = __netdev_start_xmit(ops, skb, dev, more);
4004 if (rc == NETDEV_TX_OK)
4005 txq_trans_update(txq);
4010 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4012 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4015 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4017 return netdev_class_create_file_ns(class_attr, NULL);
4020 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4022 netdev_class_remove_file_ns(class_attr, NULL);
4025 extern struct kobj_ns_type_operations net_ns_type_operations;
4027 const char *netdev_drivername(const struct net_device *dev);
4029 void linkwatch_run_queue(void);
4031 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4032 netdev_features_t f2)
4034 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4035 if (f1 & NETIF_F_HW_CSUM)
4036 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4038 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4044 static inline netdev_features_t netdev_get_wanted_features(
4045 struct net_device *dev)
4047 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4049 netdev_features_t netdev_increment_features(netdev_features_t all,
4050 netdev_features_t one, netdev_features_t mask);
4052 /* Allow TSO being used on stacked device :
4053 * Performing the GSO segmentation before last device
4054 * is a performance improvement.
4056 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4057 netdev_features_t mask)
4059 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4062 int __netdev_update_features(struct net_device *dev);
4063 void netdev_update_features(struct net_device *dev);
4064 void netdev_change_features(struct net_device *dev);
4066 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4067 struct net_device *dev);
4069 netdev_features_t passthru_features_check(struct sk_buff *skb,
4070 struct net_device *dev,
4071 netdev_features_t features);
4072 netdev_features_t netif_skb_features(struct sk_buff *skb);
4074 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4076 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4078 /* check flags correspondence */
4079 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4080 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4081 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4082 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4083 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4084 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4085 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4086 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4087 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4088 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4089 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4090 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4091 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4092 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4093 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4094 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4095 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4097 return (features & feature) == feature;
4100 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4102 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4103 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4106 static inline bool netif_needs_gso(struct sk_buff *skb,
4107 netdev_features_t features)
4109 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4110 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4111 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4114 static inline void netif_set_gso_max_size(struct net_device *dev,
4117 dev->gso_max_size = size;
4120 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4121 int pulled_hlen, u16 mac_offset,
4124 skb->protocol = protocol;
4125 skb->encapsulation = 1;
4126 skb_push(skb, pulled_hlen);
4127 skb_reset_transport_header(skb);
4128 skb->mac_header = mac_offset;
4129 skb->network_header = skb->mac_header + mac_len;
4130 skb->mac_len = mac_len;
4133 static inline bool netif_is_macsec(const struct net_device *dev)
4135 return dev->priv_flags & IFF_MACSEC;
4138 static inline bool netif_is_macvlan(const struct net_device *dev)
4140 return dev->priv_flags & IFF_MACVLAN;
4143 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4145 return dev->priv_flags & IFF_MACVLAN_PORT;
4148 static inline bool netif_is_ipvlan(const struct net_device *dev)
4150 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4153 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4155 return dev->priv_flags & IFF_IPVLAN_MASTER;
4158 static inline bool netif_is_bond_master(const struct net_device *dev)
4160 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4163 static inline bool netif_is_bond_slave(const struct net_device *dev)
4165 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4168 static inline bool netif_supports_nofcs(struct net_device *dev)
4170 return dev->priv_flags & IFF_SUPP_NOFCS;
4173 static inline bool netif_is_l3_master(const struct net_device *dev)
4175 return dev->priv_flags & IFF_L3MDEV_MASTER;
4178 static inline bool netif_is_l3_slave(const struct net_device *dev)
4180 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4183 static inline bool netif_is_bridge_master(const struct net_device *dev)
4185 return dev->priv_flags & IFF_EBRIDGE;
4188 static inline bool netif_is_bridge_port(const struct net_device *dev)
4190 return dev->priv_flags & IFF_BRIDGE_PORT;
4193 static inline bool netif_is_ovs_master(const struct net_device *dev)
4195 return dev->priv_flags & IFF_OPENVSWITCH;
4198 static inline bool netif_is_ovs_port(const struct net_device *dev)
4200 return dev->priv_flags & IFF_OVS_DATAPATH;
4203 static inline bool netif_is_team_master(const struct net_device *dev)
4205 return dev->priv_flags & IFF_TEAM;
4208 static inline bool netif_is_team_port(const struct net_device *dev)
4210 return dev->priv_flags & IFF_TEAM_PORT;
4213 static inline bool netif_is_lag_master(const struct net_device *dev)
4215 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4218 static inline bool netif_is_lag_port(const struct net_device *dev)
4220 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4223 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4225 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4228 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4229 static inline void netif_keep_dst(struct net_device *dev)
4231 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4234 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4235 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4237 /* TODO: reserve and use an additional IFF bit, if we get more users */
4238 return dev->priv_flags & IFF_MACSEC;
4241 extern struct pernet_operations __net_initdata loopback_net_ops;
4243 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4245 /* netdev_printk helpers, similar to dev_printk */
4247 static inline const char *netdev_name(const struct net_device *dev)
4249 if (!dev->name[0] || strchr(dev->name, '%'))
4250 return "(unnamed net_device)";
4254 static inline const char *netdev_reg_state(const struct net_device *dev)
4256 switch (dev->reg_state) {
4257 case NETREG_UNINITIALIZED: return " (uninitialized)";
4258 case NETREG_REGISTERED: return "";
4259 case NETREG_UNREGISTERING: return " (unregistering)";
4260 case NETREG_UNREGISTERED: return " (unregistered)";
4261 case NETREG_RELEASED: return " (released)";
4262 case NETREG_DUMMY: return " (dummy)";
4265 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4266 return " (unknown)";
4270 void netdev_printk(const char *level, const struct net_device *dev,
4271 const char *format, ...);
4273 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4275 void netdev_alert(const struct net_device *dev, const char *format, ...);
4277 void netdev_crit(const struct net_device *dev, const char *format, ...);
4279 void netdev_err(const struct net_device *dev, const char *format, ...);
4281 void netdev_warn(const struct net_device *dev, const char *format, ...);
4283 void netdev_notice(const struct net_device *dev, const char *format, ...);
4285 void netdev_info(const struct net_device *dev, const char *format, ...);
4287 #define MODULE_ALIAS_NETDEV(device) \
4288 MODULE_ALIAS("netdev-" device)
4290 #if defined(CONFIG_DYNAMIC_DEBUG)
4291 #define netdev_dbg(__dev, format, args...) \
4293 dynamic_netdev_dbg(__dev, format, ##args); \
4295 #elif defined(DEBUG)
4296 #define netdev_dbg(__dev, format, args...) \
4297 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4299 #define netdev_dbg(__dev, format, args...) \
4302 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4306 #if defined(VERBOSE_DEBUG)
4307 #define netdev_vdbg netdev_dbg
4310 #define netdev_vdbg(dev, format, args...) \
4313 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4319 * netdev_WARN() acts like dev_printk(), but with the key difference
4320 * of using a WARN/WARN_ON to get the message out, including the
4321 * file/line information and a backtrace.
4323 #define netdev_WARN(dev, format, args...) \
4324 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4325 netdev_reg_state(dev), ##args)
4327 /* netif printk helpers, similar to netdev_printk */
4329 #define netif_printk(priv, type, level, dev, fmt, args...) \
4331 if (netif_msg_##type(priv)) \
4332 netdev_printk(level, (dev), fmt, ##args); \
4335 #define netif_level(level, priv, type, dev, fmt, args...) \
4337 if (netif_msg_##type(priv)) \
4338 netdev_##level(dev, fmt, ##args); \
4341 #define netif_emerg(priv, type, dev, fmt, args...) \
4342 netif_level(emerg, priv, type, dev, fmt, ##args)
4343 #define netif_alert(priv, type, dev, fmt, args...) \
4344 netif_level(alert, priv, type, dev, fmt, ##args)
4345 #define netif_crit(priv, type, dev, fmt, args...) \
4346 netif_level(crit, priv, type, dev, fmt, ##args)
4347 #define netif_err(priv, type, dev, fmt, args...) \
4348 netif_level(err, priv, type, dev, fmt, ##args)
4349 #define netif_warn(priv, type, dev, fmt, args...) \
4350 netif_level(warn, priv, type, dev, fmt, ##args)
4351 #define netif_notice(priv, type, dev, fmt, args...) \
4352 netif_level(notice, priv, type, dev, fmt, ##args)
4353 #define netif_info(priv, type, dev, fmt, args...) \
4354 netif_level(info, priv, type, dev, fmt, ##args)
4356 #if defined(CONFIG_DYNAMIC_DEBUG)
4357 #define netif_dbg(priv, type, netdev, format, args...) \
4359 if (netif_msg_##type(priv)) \
4360 dynamic_netdev_dbg(netdev, format, ##args); \
4362 #elif defined(DEBUG)
4363 #define netif_dbg(priv, type, dev, format, args...) \
4364 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4366 #define netif_dbg(priv, type, dev, format, args...) \
4369 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4374 /* if @cond then downgrade to debug, else print at @level */
4375 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4378 netif_dbg(priv, type, netdev, fmt, ##args); \
4380 netif_ ## level(priv, type, netdev, fmt, ##args); \
4383 #if defined(VERBOSE_DEBUG)
4384 #define netif_vdbg netif_dbg
4386 #define netif_vdbg(priv, type, dev, format, args...) \
4389 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4395 * The list of packet types we will receive (as opposed to discard)
4396 * and the routines to invoke.
4398 * Why 16. Because with 16 the only overlap we get on a hash of the
4399 * low nibble of the protocol value is RARP/SNAP/X.25.
4401 * NOTE: That is no longer true with the addition of VLAN tags. Not
4402 * sure which should go first, but I bet it won't make much
4403 * difference if we are running VLANs. The good news is that
4404 * this protocol won't be in the list unless compiled in, so
4405 * the average user (w/out VLANs) will not be adversely affected.
4421 #define PTYPE_HASH_SIZE (16)
4422 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4424 #endif /* _LINUX_NETDEVICE_H */