2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
44 #include <net/dcbnl.h>
46 #include <net/netprio_cgroup.h>
48 #include <linux/netdev_features.h>
49 #include <linux/neighbour.h>
50 #include <uapi/linux/netdevice.h>
51 #include <uapi/linux/if_bonding.h>
52 #include <uapi/linux/pkt_cls.h>
53 #include <linux/hashtable.h>
58 struct dsa_switch_tree;
62 /* 802.15.4 specific */
65 /* UDP Tunnel offloads */
66 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 + READ_ONCE((dev)->needed_headroom)) \
262 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
264 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
265 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
270 const void *saddr, unsigned int len);
271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
276 bool (*validate)(const char *ll_header, unsigned int len);
277 __be16 (*parse_protocol)(const struct sk_buff *skb);
280 /* These flag bits are private to the generic network queueing
281 * layer; they may not be explicitly referenced by any other
285 enum netdev_state_t {
287 __LINK_STATE_PRESENT,
288 __LINK_STATE_NOCARRIER,
289 __LINK_STATE_LINKWATCH_PENDING,
290 __LINK_STATE_DORMANT,
295 * This structure holds boot-time configured netdevice settings. They
296 * are then used in the device probing.
298 struct netdev_boot_setup {
302 #define NETDEV_BOOT_SETUP_MAX 8
304 int __init netdev_boot_setup(char *str);
307 * Structure for NAPI scheduling similar to tasklet but with weighting
310 /* The poll_list must only be managed by the entity which
311 * changes the state of the NAPI_STATE_SCHED bit. This means
312 * whoever atomically sets that bit can add this napi_struct
313 * to the per-CPU poll_list, and whoever clears that bit
314 * can remove from the list right before clearing the bit.
316 struct list_head poll_list;
320 unsigned int gro_count;
321 int (*poll)(struct napi_struct *, int);
322 #ifdef CONFIG_NETPOLL
325 struct net_device *dev;
326 struct sk_buff *gro_list;
328 struct hrtimer timer;
329 struct list_head dev_list;
330 struct hlist_node napi_hash_node;
331 unsigned int napi_id;
335 NAPI_STATE_SCHED, /* Poll is scheduled */
336 NAPI_STATE_MISSED, /* reschedule a napi */
337 NAPI_STATE_DISABLE, /* Disable pending */
338 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
339 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
340 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
341 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
345 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
346 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
347 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
348 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
349 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
350 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
351 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
362 typedef enum gro_result gro_result_t;
365 * enum rx_handler_result - Possible return values for rx_handlers.
366 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
368 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
369 * case skb->dev was changed by rx_handler.
370 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
371 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
373 * rx_handlers are functions called from inside __netif_receive_skb(), to do
374 * special processing of the skb, prior to delivery to protocol handlers.
376 * Currently, a net_device can only have a single rx_handler registered. Trying
377 * to register a second rx_handler will return -EBUSY.
379 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
380 * To unregister a rx_handler on a net_device, use
381 * netdev_rx_handler_unregister().
383 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
386 * If the rx_handler consumed the skb in some way, it should return
387 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
388 * the skb to be delivered in some other way.
390 * If the rx_handler changed skb->dev, to divert the skb to another
391 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
392 * new device will be called if it exists.
394 * If the rx_handler decides the skb should be ignored, it should return
395 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
396 * are registered on exact device (ptype->dev == skb->dev).
398 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
399 * delivered, it should return RX_HANDLER_PASS.
401 * A device without a registered rx_handler will behave as if rx_handler
402 * returned RX_HANDLER_PASS.
405 enum rx_handler_result {
411 typedef enum rx_handler_result rx_handler_result_t;
412 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
414 void __napi_schedule(struct napi_struct *n);
415 void __napi_schedule_irqoff(struct napi_struct *n);
417 static inline bool napi_disable_pending(struct napi_struct *n)
419 return test_bit(NAPI_STATE_DISABLE, &n->state);
422 bool napi_schedule_prep(struct napi_struct *n);
425 * napi_schedule - schedule NAPI poll
428 * Schedule NAPI poll routine to be called if it is not already
431 static inline void napi_schedule(struct napi_struct *n)
433 if (napi_schedule_prep(n))
438 * napi_schedule_irqoff - schedule NAPI poll
441 * Variant of napi_schedule(), assuming hard irqs are masked.
443 static inline void napi_schedule_irqoff(struct napi_struct *n)
445 if (napi_schedule_prep(n))
446 __napi_schedule_irqoff(n);
449 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
450 static inline bool napi_reschedule(struct napi_struct *napi)
452 if (napi_schedule_prep(napi)) {
453 __napi_schedule(napi);
459 bool napi_complete_done(struct napi_struct *n, int work_done);
461 * napi_complete - NAPI processing complete
464 * Mark NAPI processing as complete.
465 * Consider using napi_complete_done() instead.
466 * Return false if device should avoid rearming interrupts.
468 static inline bool napi_complete(struct napi_struct *n)
470 return napi_complete_done(n, 0);
474 * napi_hash_del - remove a NAPI from global table
475 * @napi: NAPI context
477 * Warning: caller must observe RCU grace period
478 * before freeing memory containing @napi, if
479 * this function returns true.
480 * Note: core networking stack automatically calls it
481 * from netif_napi_del().
482 * Drivers might want to call this helper to combine all
483 * the needed RCU grace periods into a single one.
485 bool napi_hash_del(struct napi_struct *napi);
488 * napi_disable - prevent NAPI from scheduling
491 * Stop NAPI from being scheduled on this context.
492 * Waits till any outstanding processing completes.
494 void napi_disable(struct napi_struct *n);
497 * napi_enable - enable NAPI scheduling
500 * Resume NAPI from being scheduled on this context.
501 * Must be paired with napi_disable.
503 static inline void napi_enable(struct napi_struct *n)
505 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
506 smp_mb__before_atomic();
507 clear_bit(NAPI_STATE_SCHED, &n->state);
508 clear_bit(NAPI_STATE_NPSVC, &n->state);
512 * napi_synchronize - wait until NAPI is not running
515 * Wait until NAPI is done being scheduled on this context.
516 * Waits till any outstanding processing completes but
517 * does not disable future activations.
519 static inline void napi_synchronize(const struct napi_struct *n)
521 if (IS_ENABLED(CONFIG_SMP))
522 while (test_bit(NAPI_STATE_SCHED, &n->state))
528 enum netdev_queue_state_t {
529 __QUEUE_STATE_DRV_XOFF,
530 __QUEUE_STATE_STACK_XOFF,
531 __QUEUE_STATE_FROZEN,
534 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
535 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
538 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
539 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
541 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
545 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
546 * netif_tx_* functions below are used to manipulate this flag. The
547 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
548 * queue independently. The netif_xmit_*stopped functions below are called
549 * to check if the queue has been stopped by the driver or stack (either
550 * of the XOFF bits are set in the state). Drivers should not need to call
551 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 struct netdev_queue {
558 struct net_device *dev;
559 struct Qdisc __rcu *qdisc;
560 struct Qdisc *qdisc_sleeping;
564 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 unsigned long tx_maxrate;
569 * Number of TX timeouts for this queue
570 * (/sys/class/net/DEV/Q/trans_timeout)
572 unsigned long trans_timeout;
576 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
579 * Time (in jiffies) of last Tx
581 unsigned long trans_start;
588 } ____cacheline_aligned_in_smp;
590 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
592 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 * This structure holds an RPS map which can be of variable length. The
609 * map is an array of CPUs.
616 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
619 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
620 * tail pointer for that CPU's input queue at the time of last enqueue, and
621 * a hardware filter index.
623 struct rps_dev_flow {
626 unsigned int last_qtail;
628 #define RPS_NO_FILTER 0xffff
631 * The rps_dev_flow_table structure contains a table of flow mappings.
633 struct rps_dev_flow_table {
636 struct rps_dev_flow flows[0];
638 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
639 ((_num) * sizeof(struct rps_dev_flow)))
642 * The rps_sock_flow_table contains mappings of flows to the last CPU
643 * on which they were processed by the application (set in recvmsg).
644 * Each entry is a 32bit value. Upper part is the high-order bits
645 * of flow hash, lower part is CPU number.
646 * rps_cpu_mask is used to partition the space, depending on number of
647 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
648 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
649 * meaning we use 32-6=26 bits for the hash.
651 struct rps_sock_flow_table {
654 u32 ents[0] ____cacheline_aligned_in_smp;
656 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
658 #define RPS_NO_CPU 0xffff
660 extern u32 rps_cpu_mask;
661 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
663 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 unsigned int index = hash & table->mask;
668 u32 val = hash & ~rps_cpu_mask;
670 /* We only give a hint, preemption can change CPU under us */
671 val |= raw_smp_processor_id();
673 /* The following WRITE_ONCE() is paired with the READ_ONCE()
674 * here, and another one in get_rps_cpu().
676 if (READ_ONCE(table->ents[index]) != val)
677 WRITE_ONCE(table->ents[index], val);
681 #ifdef CONFIG_RFS_ACCEL
682 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
685 #endif /* CONFIG_RPS */
687 /* This structure contains an instance of an RX queue. */
688 struct netdev_rx_queue {
690 struct rps_map __rcu *rps_map;
691 struct rps_dev_flow_table __rcu *rps_flow_table;
694 struct net_device *dev;
695 } ____cacheline_aligned_in_smp;
698 * RX queue sysfs structures and functions.
700 struct rx_queue_attribute {
701 struct attribute attr;
702 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
703 ssize_t (*store)(struct netdev_rx_queue *queue,
704 const char *buf, size_t len);
709 * This structure holds an XPS map which can be of variable length. The
710 * map is an array of queues.
714 unsigned int alloc_len;
718 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
719 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
720 - sizeof(struct xps_map)) / sizeof(u16))
723 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 struct xps_dev_maps {
727 struct xps_map __rcu *cpu_map[0];
729 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
730 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
731 #endif /* CONFIG_XPS */
733 #define TC_MAX_QUEUE 16
734 #define TC_BITMASK 15
735 /* HW offloaded queuing disciplines txq count and offset maps */
736 struct netdev_tc_txq {
741 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743 * This structure is to hold information about the device
744 * configured to run FCoE protocol stack.
746 struct netdev_fcoe_hbainfo {
747 char manufacturer[64];
748 char serial_number[64];
749 char hardware_version[64];
750 char driver_version[64];
751 char optionrom_version[64];
752 char firmware_version[64];
754 char model_description[256];
758 #define MAX_PHYS_ITEM_ID_LEN 32
760 /* This structure holds a unique identifier to identify some
761 * physical item (port for example) used by a netdevice.
763 struct netdev_phys_item_id {
764 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
765 unsigned char id_len;
768 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
769 struct netdev_phys_item_id *b)
771 return a->id_len == b->id_len &&
772 memcmp(a->id, b->id, a->id_len) == 0;
775 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
776 struct sk_buff *skb);
782 TC_SETUP_CLSMATCHALL,
786 /* These structures hold the attributes of xdp state that are being passed
787 * to the netdevice through the xdp op.
789 enum xdp_netdev_command {
790 /* Set or clear a bpf program used in the earliest stages of packet
791 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
792 * is responsible for calling bpf_prog_put on any old progs that are
793 * stored. In case of error, the callee need not release the new prog
794 * reference, but on success it takes ownership and must bpf_prog_put
795 * when it is no longer used.
799 /* Check if a bpf program is set on the device. The callee should
800 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
801 * is equivalent to XDP_ATTACHED_DRV.
806 struct netlink_ext_ack;
809 enum xdp_netdev_command command;
814 struct bpf_prog *prog;
815 struct netlink_ext_ack *extack;
825 #ifdef CONFIG_XFRM_OFFLOAD
827 int (*xdo_dev_state_add) (struct xfrm_state *x);
828 void (*xdo_dev_state_delete) (struct xfrm_state *x);
829 void (*xdo_dev_state_free) (struct xfrm_state *x);
830 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
831 struct xfrm_state *x);
836 * This structure defines the management hooks for network devices.
837 * The following hooks can be defined; unless noted otherwise, they are
838 * optional and can be filled with a null pointer.
840 * int (*ndo_init)(struct net_device *dev);
841 * This function is called once when a network device is registered.
842 * The network device can use this for any late stage initialization
843 * or semantic validation. It can fail with an error code which will
844 * be propagated back to register_netdev.
846 * void (*ndo_uninit)(struct net_device *dev);
847 * This function is called when device is unregistered or when registration
848 * fails. It is not called if init fails.
850 * int (*ndo_open)(struct net_device *dev);
851 * This function is called when a network device transitions to the up
854 * int (*ndo_stop)(struct net_device *dev);
855 * This function is called when a network device transitions to the down
858 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
859 * struct net_device *dev);
860 * Called when a packet needs to be transmitted.
861 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
862 * the queue before that can happen; it's for obsolete devices and weird
863 * corner cases, but the stack really does a non-trivial amount
864 * of useless work if you return NETDEV_TX_BUSY.
865 * Required; cannot be NULL.
867 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
868 * struct net_device *dev
869 * netdev_features_t features);
870 * Called by core transmit path to determine if device is capable of
871 * performing offload operations on a given packet. This is to give
872 * the device an opportunity to implement any restrictions that cannot
873 * be otherwise expressed by feature flags. The check is called with
874 * the set of features that the stack has calculated and it returns
875 * those the driver believes to be appropriate.
877 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
878 * void *accel_priv, select_queue_fallback_t fallback);
879 * Called to decide which queue to use when device supports multiple
882 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
883 * This function is called to allow device receiver to make
884 * changes to configuration when multicast or promiscuous is enabled.
886 * void (*ndo_set_rx_mode)(struct net_device *dev);
887 * This function is called device changes address list filtering.
888 * If driver handles unicast address filtering, it should set
889 * IFF_UNICAST_FLT in its priv_flags.
891 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
892 * This function is called when the Media Access Control address
893 * needs to be changed. If this interface is not defined, the
894 * MAC address can not be changed.
896 * int (*ndo_validate_addr)(struct net_device *dev);
897 * Test if Media Access Control address is valid for the device.
899 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
900 * Called when a user requests an ioctl which can't be handled by
901 * the generic interface code. If not defined ioctls return
902 * not supported error code.
904 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
905 * Used to set network devices bus interface parameters. This interface
906 * is retained for legacy reasons; new devices should use the bus
907 * interface (PCI) for low level management.
909 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
910 * Called when a user wants to change the Maximum Transfer Unit
913 * void (*ndo_tx_timeout)(struct net_device *dev);
914 * Callback used when the transmitter has not made any progress
915 * for dev->watchdog ticks.
917 * void (*ndo_get_stats64)(struct net_device *dev,
918 * struct rtnl_link_stats64 *storage);
919 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
920 * Called when a user wants to get the network device usage
921 * statistics. Drivers must do one of the following:
922 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
923 * rtnl_link_stats64 structure passed by the caller.
924 * 2. Define @ndo_get_stats to update a net_device_stats structure
925 * (which should normally be dev->stats) and return a pointer to
926 * it. The structure may be changed asynchronously only if each
927 * field is written atomically.
928 * 3. Update dev->stats asynchronously and atomically, and define
931 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
932 * Return true if this device supports offload stats of this attr_id.
934 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
936 * Get statistics for offload operations by attr_id. Write it into the
939 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
940 * If device supports VLAN filtering this function is called when a
941 * VLAN id is registered.
943 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
944 * If device supports VLAN filtering this function is called when a
945 * VLAN id is unregistered.
947 * void (*ndo_poll_controller)(struct net_device *dev);
949 * SR-IOV management functions.
950 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
951 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
952 * u8 qos, __be16 proto);
953 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
955 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
956 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
957 * int (*ndo_get_vf_config)(struct net_device *dev,
958 * int vf, struct ifla_vf_info *ivf);
959 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
960 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
961 * struct nlattr *port[]);
963 * Enable or disable the VF ability to query its RSS Redirection Table and
964 * Hash Key. This is needed since on some devices VF share this information
965 * with PF and querying it may introduce a theoretical security risk.
966 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
967 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
968 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
970 * Called to setup any 'tc' scheduler, classifier or action on @dev.
971 * This is always called from the stack with the rtnl lock held and netif
972 * tx queues stopped. This allows the netdevice to perform queue
975 * Fiber Channel over Ethernet (FCoE) offload functions.
976 * int (*ndo_fcoe_enable)(struct net_device *dev);
977 * Called when the FCoE protocol stack wants to start using LLD for FCoE
978 * so the underlying device can perform whatever needed configuration or
979 * initialization to support acceleration of FCoE traffic.
981 * int (*ndo_fcoe_disable)(struct net_device *dev);
982 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
983 * so the underlying device can perform whatever needed clean-ups to
984 * stop supporting acceleration of FCoE traffic.
986 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
987 * struct scatterlist *sgl, unsigned int sgc);
988 * Called when the FCoE Initiator wants to initialize an I/O that
989 * is a possible candidate for Direct Data Placement (DDP). The LLD can
990 * perform necessary setup and returns 1 to indicate the device is set up
991 * successfully to perform DDP on this I/O, otherwise this returns 0.
993 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
994 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
995 * indicated by the FC exchange id 'xid', so the underlying device can
996 * clean up and reuse resources for later DDP requests.
998 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
999 * struct scatterlist *sgl, unsigned int sgc);
1000 * Called when the FCoE Target wants to initialize an I/O that
1001 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1002 * perform necessary setup and returns 1 to indicate the device is set up
1003 * successfully to perform DDP on this I/O, otherwise this returns 0.
1005 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1006 * struct netdev_fcoe_hbainfo *hbainfo);
1007 * Called when the FCoE Protocol stack wants information on the underlying
1008 * device. This information is utilized by the FCoE protocol stack to
1009 * register attributes with Fiber Channel management service as per the
1010 * FC-GS Fabric Device Management Information(FDMI) specification.
1012 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1013 * Called when the underlying device wants to override default World Wide
1014 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1015 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1016 * protocol stack to use.
1019 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1020 * u16 rxq_index, u32 flow_id);
1021 * Set hardware filter for RFS. rxq_index is the target queue index;
1022 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1023 * Return the filter ID on success, or a negative error code.
1025 * Slave management functions (for bridge, bonding, etc).
1026 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1027 * Called to make another netdev an underling.
1029 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1030 * Called to release previously enslaved netdev.
1032 * Feature/offload setting functions.
1033 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1034 * netdev_features_t features);
1035 * Adjusts the requested feature flags according to device-specific
1036 * constraints, and returns the resulting flags. Must not modify
1039 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1040 * Called to update device configuration to new features. Passed
1041 * feature set might be less than what was returned by ndo_fix_features()).
1042 * Must return >0 or -errno if it changed dev->features itself.
1044 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1045 * struct net_device *dev,
1046 * const unsigned char *addr, u16 vid, u16 flags)
1047 * Adds an FDB entry to dev for addr.
1048 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1049 * struct net_device *dev,
1050 * const unsigned char *addr, u16 vid)
1051 * Deletes the FDB entry from dev coresponding to addr.
1052 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1053 * struct net_device *dev, struct net_device *filter_dev,
1055 * Used to add FDB entries to dump requests. Implementers should add
1056 * entries to skb and update idx with the number of entries.
1058 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1060 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1061 * struct net_device *dev, u32 filter_mask,
1063 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1066 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1067 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1068 * which do not represent real hardware may define this to allow their
1069 * userspace components to manage their virtual carrier state. Devices
1070 * that determine carrier state from physical hardware properties (eg
1071 * network cables) or protocol-dependent mechanisms (eg
1072 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1074 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1075 * struct netdev_phys_item_id *ppid);
1076 * Called to get ID of physical port of this device. If driver does
1077 * not implement this, it is assumed that the hw is not able to have
1078 * multiple net devices on single physical port.
1080 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1081 * struct udp_tunnel_info *ti);
1082 * Called by UDP tunnel to notify a driver about the UDP port and socket
1083 * address family that a UDP tunnel is listnening to. It is called only
1084 * when a new port starts listening. The operation is protected by the
1087 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1088 * struct udp_tunnel_info *ti);
1089 * Called by UDP tunnel to notify the driver about a UDP port and socket
1090 * address family that the UDP tunnel is not listening to anymore. The
1091 * operation is protected by the RTNL.
1093 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1094 * struct net_device *dev)
1095 * Called by upper layer devices to accelerate switching or other
1096 * station functionality into hardware. 'pdev is the lowerdev
1097 * to use for the offload and 'dev' is the net device that will
1098 * back the offload. Returns a pointer to the private structure
1099 * the upper layer will maintain.
1100 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1101 * Called by upper layer device to delete the station created
1102 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1103 * the station and priv is the structure returned by the add
1105 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1106 * int queue_index, u32 maxrate);
1107 * Called when a user wants to set a max-rate limitation of specific
1109 * int (*ndo_get_iflink)(const struct net_device *dev);
1110 * Called to get the iflink value of this device.
1111 * void (*ndo_change_proto_down)(struct net_device *dev,
1113 * This function is used to pass protocol port error state information
1114 * to the switch driver. The switch driver can react to the proto_down
1115 * by doing a phys down on the associated switch port.
1116 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1117 * This function is used to get egress tunnel information for given skb.
1118 * This is useful for retrieving outer tunnel header parameters while
1120 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1121 * This function is used to specify the headroom that the skb must
1122 * consider when allocation skb during packet reception. Setting
1123 * appropriate rx headroom value allows avoiding skb head copy on
1124 * forward. Setting a negative value resets the rx headroom to the
1126 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1127 * This function is used to set or query state related to XDP on the
1128 * netdevice. See definition of enum xdp_netdev_command for details.
1129 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1130 * This function is used to submit a XDP packet for transmit on a
1132 * void (*ndo_xdp_flush)(struct net_device *dev);
1133 * This function is used to inform the driver to flush a particular
1134 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1136 struct net_device_ops {
1137 int (*ndo_init)(struct net_device *dev);
1138 void (*ndo_uninit)(struct net_device *dev);
1139 int (*ndo_open)(struct net_device *dev);
1140 int (*ndo_stop)(struct net_device *dev);
1141 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1142 struct net_device *dev);
1143 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1144 struct net_device *dev,
1145 netdev_features_t features);
1146 u16 (*ndo_select_queue)(struct net_device *dev,
1147 struct sk_buff *skb,
1149 select_queue_fallback_t fallback);
1150 void (*ndo_change_rx_flags)(struct net_device *dev,
1152 void (*ndo_set_rx_mode)(struct net_device *dev);
1153 int (*ndo_set_mac_address)(struct net_device *dev,
1155 int (*ndo_validate_addr)(struct net_device *dev);
1156 int (*ndo_do_ioctl)(struct net_device *dev,
1157 struct ifreq *ifr, int cmd);
1158 int (*ndo_set_config)(struct net_device *dev,
1160 int (*ndo_change_mtu)(struct net_device *dev,
1162 int (*ndo_neigh_setup)(struct net_device *dev,
1163 struct neigh_parms *);
1164 void (*ndo_tx_timeout) (struct net_device *dev);
1166 void (*ndo_get_stats64)(struct net_device *dev,
1167 struct rtnl_link_stats64 *storage);
1168 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1169 int (*ndo_get_offload_stats)(int attr_id,
1170 const struct net_device *dev,
1172 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1175 __be16 proto, u16 vid);
1176 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1177 __be16 proto, u16 vid);
1178 #ifdef CONFIG_NET_POLL_CONTROLLER
1179 void (*ndo_poll_controller)(struct net_device *dev);
1180 int (*ndo_netpoll_setup)(struct net_device *dev,
1181 struct netpoll_info *info);
1182 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1184 int (*ndo_set_vf_mac)(struct net_device *dev,
1185 int queue, u8 *mac);
1186 int (*ndo_set_vf_vlan)(struct net_device *dev,
1187 int queue, u16 vlan,
1188 u8 qos, __be16 proto);
1189 int (*ndo_set_vf_rate)(struct net_device *dev,
1190 int vf, int min_tx_rate,
1192 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1193 int vf, bool setting);
1194 int (*ndo_set_vf_trust)(struct net_device *dev,
1195 int vf, bool setting);
1196 int (*ndo_get_vf_config)(struct net_device *dev,
1198 struct ifla_vf_info *ivf);
1199 int (*ndo_set_vf_link_state)(struct net_device *dev,
1200 int vf, int link_state);
1201 int (*ndo_get_vf_stats)(struct net_device *dev,
1203 struct ifla_vf_stats
1205 int (*ndo_set_vf_port)(struct net_device *dev,
1207 struct nlattr *port[]);
1208 int (*ndo_get_vf_port)(struct net_device *dev,
1209 int vf, struct sk_buff *skb);
1210 int (*ndo_set_vf_guid)(struct net_device *dev,
1213 int (*ndo_set_vf_rss_query_en)(
1214 struct net_device *dev,
1215 int vf, bool setting);
1216 int (*ndo_setup_tc)(struct net_device *dev,
1217 enum tc_setup_type type,
1219 #if IS_ENABLED(CONFIG_FCOE)
1220 int (*ndo_fcoe_enable)(struct net_device *dev);
1221 int (*ndo_fcoe_disable)(struct net_device *dev);
1222 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1224 struct scatterlist *sgl,
1226 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1228 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1230 struct scatterlist *sgl,
1232 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1233 struct netdev_fcoe_hbainfo *hbainfo);
1236 #if IS_ENABLED(CONFIG_LIBFCOE)
1237 #define NETDEV_FCOE_WWNN 0
1238 #define NETDEV_FCOE_WWPN 1
1239 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1240 u64 *wwn, int type);
1243 #ifdef CONFIG_RFS_ACCEL
1244 int (*ndo_rx_flow_steer)(struct net_device *dev,
1245 const struct sk_buff *skb,
1249 int (*ndo_add_slave)(struct net_device *dev,
1250 struct net_device *slave_dev);
1251 int (*ndo_del_slave)(struct net_device *dev,
1252 struct net_device *slave_dev);
1253 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1254 netdev_features_t features);
1255 int (*ndo_set_features)(struct net_device *dev,
1256 netdev_features_t features);
1257 int (*ndo_neigh_construct)(struct net_device *dev,
1258 struct neighbour *n);
1259 void (*ndo_neigh_destroy)(struct net_device *dev,
1260 struct neighbour *n);
1262 int (*ndo_fdb_add)(struct ndmsg *ndm,
1263 struct nlattr *tb[],
1264 struct net_device *dev,
1265 const unsigned char *addr,
1268 int (*ndo_fdb_del)(struct ndmsg *ndm,
1269 struct nlattr *tb[],
1270 struct net_device *dev,
1271 const unsigned char *addr,
1273 int (*ndo_fdb_dump)(struct sk_buff *skb,
1274 struct netlink_callback *cb,
1275 struct net_device *dev,
1276 struct net_device *filter_dev,
1279 int (*ndo_bridge_setlink)(struct net_device *dev,
1280 struct nlmsghdr *nlh,
1282 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1284 struct net_device *dev,
1287 int (*ndo_bridge_dellink)(struct net_device *dev,
1288 struct nlmsghdr *nlh,
1290 int (*ndo_change_carrier)(struct net_device *dev,
1292 int (*ndo_get_phys_port_id)(struct net_device *dev,
1293 struct netdev_phys_item_id *ppid);
1294 int (*ndo_get_phys_port_name)(struct net_device *dev,
1295 char *name, size_t len);
1296 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1297 struct udp_tunnel_info *ti);
1298 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1299 struct udp_tunnel_info *ti);
1300 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1301 struct net_device *dev);
1302 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1305 int (*ndo_get_lock_subclass)(struct net_device *dev);
1306 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1309 int (*ndo_get_iflink)(const struct net_device *dev);
1310 int (*ndo_change_proto_down)(struct net_device *dev,
1312 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1313 struct sk_buff *skb);
1314 void (*ndo_set_rx_headroom)(struct net_device *dev,
1315 int needed_headroom);
1316 int (*ndo_xdp)(struct net_device *dev,
1317 struct netdev_xdp *xdp);
1318 int (*ndo_xdp_xmit)(struct net_device *dev,
1319 struct xdp_buff *xdp);
1320 void (*ndo_xdp_flush)(struct net_device *dev);
1324 * enum net_device_priv_flags - &struct net_device priv_flags
1326 * These are the &struct net_device, they are only set internally
1327 * by drivers and used in the kernel. These flags are invisible to
1328 * userspace; this means that the order of these flags can change
1329 * during any kernel release.
1331 * You should have a pretty good reason to be extending these flags.
1333 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1334 * @IFF_EBRIDGE: Ethernet bridging device
1335 * @IFF_BONDING: bonding master or slave
1336 * @IFF_ISATAP: ISATAP interface (RFC4214)
1337 * @IFF_WAN_HDLC: WAN HDLC device
1338 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1340 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1341 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1342 * @IFF_MACVLAN_PORT: device used as macvlan port
1343 * @IFF_BRIDGE_PORT: device used as bridge port
1344 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1345 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1346 * @IFF_UNICAST_FLT: Supports unicast filtering
1347 * @IFF_TEAM_PORT: device used as team port
1348 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1349 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1350 * change when it's running
1351 * @IFF_MACVLAN: Macvlan device
1352 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1353 * underlying stacked devices
1354 * @IFF_IPVLAN_MASTER: IPvlan master device
1355 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1356 * @IFF_L3MDEV_MASTER: device is an L3 master device
1357 * @IFF_NO_QUEUE: device can run without qdisc attached
1358 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1359 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1360 * @IFF_TEAM: device is a team device
1361 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1362 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1363 * entity (i.e. the master device for bridged veth)
1364 * @IFF_MACSEC: device is a MACsec device
1365 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1367 enum netdev_priv_flags {
1368 IFF_802_1Q_VLAN = 1<<0,
1372 IFF_WAN_HDLC = 1<<4,
1373 IFF_XMIT_DST_RELEASE = 1<<5,
1374 IFF_DONT_BRIDGE = 1<<6,
1375 IFF_DISABLE_NETPOLL = 1<<7,
1376 IFF_MACVLAN_PORT = 1<<8,
1377 IFF_BRIDGE_PORT = 1<<9,
1378 IFF_OVS_DATAPATH = 1<<10,
1379 IFF_TX_SKB_SHARING = 1<<11,
1380 IFF_UNICAST_FLT = 1<<12,
1381 IFF_TEAM_PORT = 1<<13,
1382 IFF_SUPP_NOFCS = 1<<14,
1383 IFF_LIVE_ADDR_CHANGE = 1<<15,
1384 IFF_MACVLAN = 1<<16,
1385 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1386 IFF_IPVLAN_MASTER = 1<<18,
1387 IFF_IPVLAN_SLAVE = 1<<19,
1388 IFF_L3MDEV_MASTER = 1<<20,
1389 IFF_NO_QUEUE = 1<<21,
1390 IFF_OPENVSWITCH = 1<<22,
1391 IFF_L3MDEV_SLAVE = 1<<23,
1393 IFF_RXFH_CONFIGURED = 1<<25,
1394 IFF_PHONY_HEADROOM = 1<<26,
1396 IFF_L3MDEV_RX_HANDLER = 1<<28,
1399 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1400 #define IFF_EBRIDGE IFF_EBRIDGE
1401 #define IFF_BONDING IFF_BONDING
1402 #define IFF_ISATAP IFF_ISATAP
1403 #define IFF_WAN_HDLC IFF_WAN_HDLC
1404 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1405 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1406 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1407 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1408 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1409 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1410 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1411 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1412 #define IFF_TEAM_PORT IFF_TEAM_PORT
1413 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1414 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1415 #define IFF_MACVLAN IFF_MACVLAN
1416 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1417 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1418 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1419 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1420 #define IFF_NO_QUEUE IFF_NO_QUEUE
1421 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1422 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1423 #define IFF_TEAM IFF_TEAM
1424 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1425 #define IFF_MACSEC IFF_MACSEC
1426 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1429 * 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 * @ip6_ptr: IPv6 specific data
1549 * @ax25_ptr: AX.25 specific data
1550 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1552 * @dev_addr: Hw address (before bcast,
1553 * because most packets are unicast)
1555 * @_rx: Array of RX queues
1556 * @num_rx_queues: Number of RX queues
1557 * allocated at register_netdev() time
1558 * @real_num_rx_queues: Number of RX queues currently active in device
1560 * @rx_handler: handler for received packets
1561 * @rx_handler_data: XXX: need comments on this one
1562 * @ingress_queue: XXX: need comments on this one
1563 * @broadcast: hw bcast address
1565 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1566 * indexed by RX queue number. Assigned by driver.
1567 * This must only be set if the ndo_rx_flow_steer
1568 * operation is defined
1569 * @index_hlist: Device index hash chain
1571 * @_tx: Array of TX queues
1572 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1573 * @real_num_tx_queues: Number of TX queues currently active in device
1574 * @qdisc: Root qdisc from userspace point of view
1575 * @tx_queue_len: Max frames per queue allowed
1576 * @tx_global_lock: XXX: need comments on this one
1578 * @xps_maps: XXX: need comments on this one
1580 * @watchdog_timeo: Represents the timeout that is used by
1581 * the watchdog (see dev_watchdog())
1582 * @watchdog_timer: List of timers
1584 * @pcpu_refcnt: Number of references to this device
1585 * @todo_list: Delayed register/unregister
1586 * @link_watch_list: XXX: need comments on this one
1588 * @reg_state: Register/unregister state machine
1589 * @dismantle: Device is going to be freed
1590 * @rtnl_link_state: This enum represents the phases of creating
1593 * @needs_free_netdev: Should unregister perform free_netdev?
1594 * @priv_destructor: Called from unregister
1595 * @npinfo: XXX: need comments on this one
1596 * @nd_net: Network namespace this network device is inside
1598 * @ml_priv: Mid-layer private
1599 * @lstats: Loopback statistics
1600 * @tstats: Tunnel statistics
1601 * @dstats: Dummy statistics
1602 * @vstats: Virtual ethernet statistics
1607 * @dev: Class/net/name entry
1608 * @sysfs_groups: Space for optional device, statistics and wireless
1611 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1612 * @rtnl_link_ops: Rtnl_link_ops
1614 * @gso_max_size: Maximum size of generic segmentation offload
1615 * @gso_max_segs: Maximum number of segments that can be passed to the
1618 * @dcbnl_ops: Data Center Bridging netlink ops
1619 * @num_tc: Number of traffic classes in the net device
1620 * @tc_to_txq: XXX: need comments on this one
1621 * @prio_tc_map: XXX: need comments on this one
1623 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1625 * @priomap: XXX: need comments on this one
1626 * @phydev: Physical device may attach itself
1627 * for hardware timestamping
1629 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1630 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1632 * @proto_down: protocol port state information can be sent to the
1633 * switch driver and used to set the phys state of the
1636 * FIXME: cleanup struct net_device such that network protocol info
1641 char name[IFNAMSIZ];
1642 struct hlist_node name_hlist;
1645 * I/O specific fields
1646 * FIXME: Merge these and struct ifmap into one
1648 unsigned long mem_end;
1649 unsigned long mem_start;
1650 unsigned long base_addr;
1653 atomic_t carrier_changes;
1656 * Some hardware also needs these fields (state,dev_list,
1657 * napi_list,unreg_list,close_list) but they are not
1658 * part of the usual set specified in Space.c.
1661 unsigned long state;
1663 struct list_head dev_list;
1664 struct list_head napi_list;
1665 struct list_head unreg_list;
1666 struct list_head close_list;
1667 struct list_head ptype_all;
1668 struct list_head ptype_specific;
1671 struct list_head upper;
1672 struct list_head lower;
1675 netdev_features_t features;
1676 netdev_features_t hw_features;
1677 netdev_features_t wanted_features;
1678 netdev_features_t vlan_features;
1679 netdev_features_t hw_enc_features;
1680 netdev_features_t mpls_features;
1681 netdev_features_t gso_partial_features;
1686 struct net_device_stats stats;
1688 atomic_long_t rx_dropped;
1689 atomic_long_t tx_dropped;
1690 atomic_long_t rx_nohandler;
1692 #ifdef CONFIG_WIRELESS_EXT
1693 const struct iw_handler_def *wireless_handlers;
1694 struct iw_public_data *wireless_data;
1696 const struct net_device_ops *netdev_ops;
1697 const struct ethtool_ops *ethtool_ops;
1698 #ifdef CONFIG_NET_SWITCHDEV
1699 const struct switchdev_ops *switchdev_ops;
1701 #ifdef CONFIG_NET_L3_MASTER_DEV
1702 const struct l3mdev_ops *l3mdev_ops;
1704 #if IS_ENABLED(CONFIG_IPV6)
1705 const struct ndisc_ops *ndisc_ops;
1709 const struct xfrmdev_ops *xfrmdev_ops;
1712 const struct header_ops *header_ops;
1715 unsigned int priv_flags;
1717 unsigned short gflags;
1718 unsigned short padded;
1720 unsigned char operstate;
1721 unsigned char link_mode;
1723 unsigned char if_port;
1726 /* Note : dev->mtu is often read without holding a lock.
1727 * Writers usually hold RTNL.
1728 * It is recommended to use READ_ONCE() to annotate the reads,
1729 * and to use WRITE_ONCE() to annotate the writes.
1732 unsigned int min_mtu;
1733 unsigned int max_mtu;
1734 unsigned short type;
1735 unsigned short hard_header_len;
1736 unsigned char min_header_len;
1738 unsigned short needed_headroom;
1739 unsigned short needed_tailroom;
1741 /* Interface address info. */
1742 unsigned char perm_addr[MAX_ADDR_LEN];
1743 unsigned char addr_assign_type;
1744 unsigned char addr_len;
1745 unsigned short neigh_priv_len;
1746 unsigned short dev_id;
1747 unsigned short dev_port;
1748 spinlock_t addr_list_lock;
1749 unsigned char name_assign_type;
1751 struct netdev_hw_addr_list uc;
1752 struct netdev_hw_addr_list mc;
1753 struct netdev_hw_addr_list dev_addrs;
1756 struct kset *queues_kset;
1758 unsigned int promiscuity;
1759 unsigned int allmulti;
1762 /* Protocol-specific pointers */
1764 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1765 struct vlan_info __rcu *vlan_info;
1767 #if IS_ENABLED(CONFIG_NET_DSA)
1768 struct dsa_switch_tree *dsa_ptr;
1770 #if IS_ENABLED(CONFIG_TIPC)
1771 struct tipc_bearer __rcu *tipc_ptr;
1774 struct in_device __rcu *ip_ptr;
1775 struct inet6_dev __rcu *ip6_ptr;
1777 struct wireless_dev *ieee80211_ptr;
1778 struct wpan_dev *ieee802154_ptr;
1779 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1780 struct mpls_dev __rcu *mpls_ptr;
1784 * Cache lines mostly used on receive path (including eth_type_trans())
1786 /* Interface address info used in eth_type_trans() */
1787 unsigned char *dev_addr;
1790 struct netdev_rx_queue *_rx;
1792 unsigned int num_rx_queues;
1793 unsigned int real_num_rx_queues;
1796 struct bpf_prog __rcu *xdp_prog;
1797 unsigned long gro_flush_timeout;
1798 rx_handler_func_t __rcu *rx_handler;
1799 void __rcu *rx_handler_data;
1801 #ifdef CONFIG_NET_CLS_ACT
1802 struct tcf_proto __rcu *ingress_cl_list;
1804 struct netdev_queue __rcu *ingress_queue;
1805 #ifdef CONFIG_NETFILTER_INGRESS
1806 struct nf_hook_entries __rcu *nf_hooks_ingress;
1809 unsigned char broadcast[MAX_ADDR_LEN];
1810 #ifdef CONFIG_RFS_ACCEL
1811 struct cpu_rmap *rx_cpu_rmap;
1813 struct hlist_node index_hlist;
1816 * Cache lines mostly used on transmit path
1818 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1819 unsigned int num_tx_queues;
1820 unsigned int real_num_tx_queues;
1821 struct Qdisc *qdisc;
1822 #ifdef CONFIG_NET_SCHED
1823 DECLARE_HASHTABLE (qdisc_hash, 4);
1825 unsigned int tx_queue_len;
1826 spinlock_t tx_global_lock;
1830 struct xps_dev_maps __rcu *xps_maps;
1832 #ifdef CONFIG_NET_CLS_ACT
1833 struct tcf_proto __rcu *egress_cl_list;
1836 /* These may be needed for future network-power-down code. */
1837 struct timer_list watchdog_timer;
1839 int __percpu *pcpu_refcnt;
1840 struct list_head todo_list;
1842 struct list_head link_watch_list;
1844 enum { NETREG_UNINITIALIZED=0,
1845 NETREG_REGISTERED, /* completed register_netdevice */
1846 NETREG_UNREGISTERING, /* called unregister_netdevice */
1847 NETREG_UNREGISTERED, /* completed unregister todo */
1848 NETREG_RELEASED, /* called free_netdev */
1849 NETREG_DUMMY, /* dummy device for NAPI poll */
1855 RTNL_LINK_INITIALIZED,
1856 RTNL_LINK_INITIALIZING,
1857 } rtnl_link_state:16;
1859 bool needs_free_netdev;
1860 void (*priv_destructor)(struct net_device *dev);
1862 #ifdef CONFIG_NETPOLL
1863 struct netpoll_info __rcu *npinfo;
1866 possible_net_t nd_net;
1868 /* mid-layer private */
1871 struct pcpu_lstats __percpu *lstats;
1872 struct pcpu_sw_netstats __percpu *tstats;
1873 struct pcpu_dstats __percpu *dstats;
1874 struct pcpu_vstats __percpu *vstats;
1877 #if IS_ENABLED(CONFIG_GARP)
1878 struct garp_port __rcu *garp_port;
1880 #if IS_ENABLED(CONFIG_MRP)
1881 struct mrp_port __rcu *mrp_port;
1885 const struct attribute_group *sysfs_groups[4];
1886 const struct attribute_group *sysfs_rx_queue_group;
1888 const struct rtnl_link_ops *rtnl_link_ops;
1890 /* for setting kernel sock attribute on TCP connection setup */
1891 #define GSO_MAX_SIZE 65536
1892 unsigned int gso_max_size;
1893 #define GSO_MAX_SEGS 65535
1897 const struct dcbnl_rtnl_ops *dcbnl_ops;
1900 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1901 u8 prio_tc_map[TC_BITMASK + 1];
1903 #if IS_ENABLED(CONFIG_FCOE)
1904 unsigned int fcoe_ddp_xid;
1906 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1907 struct netprio_map __rcu *priomap;
1909 struct phy_device *phydev;
1910 struct lock_class_key *qdisc_tx_busylock;
1911 struct lock_class_key *qdisc_running_key;
1914 #define to_net_dev(d) container_of(d, struct net_device, dev)
1916 static inline bool netif_elide_gro(const struct net_device *dev)
1918 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1923 #define NETDEV_ALIGN 32
1926 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1928 return dev->prio_tc_map[prio & TC_BITMASK];
1932 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1934 if (tc >= dev->num_tc)
1937 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1941 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1942 void netdev_reset_tc(struct net_device *dev);
1943 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1944 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1947 int netdev_get_num_tc(struct net_device *dev)
1953 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1956 return &dev->_tx[index];
1959 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1960 const struct sk_buff *skb)
1962 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1965 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1966 void (*f)(struct net_device *,
1967 struct netdev_queue *,
1973 for (i = 0; i < dev->num_tx_queues; i++)
1974 f(dev, &dev->_tx[i], arg);
1977 #define netdev_lockdep_set_classes(dev) \
1979 static struct lock_class_key qdisc_tx_busylock_key; \
1980 static struct lock_class_key qdisc_running_key; \
1981 static struct lock_class_key qdisc_xmit_lock_key; \
1982 static struct lock_class_key dev_addr_list_lock_key; \
1985 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1986 (dev)->qdisc_running_key = &qdisc_running_key; \
1987 lockdep_set_class(&(dev)->addr_list_lock, \
1988 &dev_addr_list_lock_key); \
1989 for (i = 0; i < (dev)->num_tx_queues; i++) \
1990 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1991 &qdisc_xmit_lock_key); \
1994 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1995 struct sk_buff *skb,
1998 /* returns the headroom that the master device needs to take in account
1999 * when forwarding to this dev
2001 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2003 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2006 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2008 if (dev->netdev_ops->ndo_set_rx_headroom)
2009 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2012 /* set the device rx headroom to the dev's default */
2013 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2015 netdev_set_rx_headroom(dev, -1);
2019 * Net namespace inlines
2022 struct net *dev_net(const struct net_device *dev)
2024 return read_pnet(&dev->nd_net);
2028 void dev_net_set(struct net_device *dev, struct net *net)
2030 write_pnet(&dev->nd_net, net);
2034 * netdev_priv - access network device private data
2035 * @dev: network device
2037 * Get network device private data
2039 static inline void *netdev_priv(const struct net_device *dev)
2041 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2044 /* Set the sysfs physical device reference for the network logical device
2045 * if set prior to registration will cause a symlink during initialization.
2047 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2049 /* Set the sysfs device type for the network logical device to allow
2050 * fine-grained identification of different network device types. For
2051 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2053 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2055 /* Default NAPI poll() weight
2056 * Device drivers are strongly advised to not use bigger value
2058 #define NAPI_POLL_WEIGHT 64
2061 * netif_napi_add - initialize a NAPI context
2062 * @dev: network device
2063 * @napi: NAPI context
2064 * @poll: polling function
2065 * @weight: default weight
2067 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2068 * *any* of the other NAPI-related functions.
2070 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2071 int (*poll)(struct napi_struct *, int), int weight);
2074 * netif_tx_napi_add - initialize a NAPI context
2075 * @dev: network device
2076 * @napi: NAPI context
2077 * @poll: polling function
2078 * @weight: default weight
2080 * This variant of netif_napi_add() should be used from drivers using NAPI
2081 * to exclusively poll a TX queue.
2082 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2084 static inline void netif_tx_napi_add(struct net_device *dev,
2085 struct napi_struct *napi,
2086 int (*poll)(struct napi_struct *, int),
2089 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2090 netif_napi_add(dev, napi, poll, weight);
2094 * netif_napi_del - remove a NAPI context
2095 * @napi: NAPI context
2097 * netif_napi_del() removes a NAPI context from the network device NAPI list
2099 void netif_napi_del(struct napi_struct *napi);
2101 struct napi_gro_cb {
2102 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2105 /* Length of frag0. */
2106 unsigned int frag0_len;
2108 /* This indicates where we are processing relative to skb->data. */
2111 /* This is non-zero if the packet cannot be merged with the new skb. */
2114 /* Save the IP ID here and check when we get to the transport layer */
2117 /* Number of segments aggregated. */
2120 /* Start offset for remote checksum offload */
2121 u16 gro_remcsum_start;
2123 /* jiffies when first packet was created/queued */
2126 /* Used in ipv6_gro_receive() and foo-over-udp */
2129 /* This is non-zero if the packet may be of the same flow. */
2132 /* Used in tunnel GRO receive */
2135 /* GRO checksum is valid */
2138 /* Number of checksums via CHECKSUM_UNNECESSARY */
2143 #define NAPI_GRO_FREE 1
2144 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2146 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2149 /* Used in GRE, set in fou/gue_gro_receive */
2152 /* Used to determine if flush_id can be ignored */
2155 /* Number of gro_receive callbacks this packet already went through */
2156 u8 recursion_counter:4;
2160 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2163 /* used in skb_gro_receive() slow path */
2164 struct sk_buff *last;
2167 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2169 #define GRO_RECURSION_LIMIT 15
2170 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2172 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2175 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2176 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2177 struct sk_buff **head,
2178 struct sk_buff *skb)
2180 if (unlikely(gro_recursion_inc_test(skb))) {
2181 NAPI_GRO_CB(skb)->flush |= 1;
2185 return cb(head, skb);
2188 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2190 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2192 struct sk_buff **head,
2193 struct sk_buff *skb)
2195 if (unlikely(gro_recursion_inc_test(skb))) {
2196 NAPI_GRO_CB(skb)->flush |= 1;
2200 return cb(sk, head, skb);
2203 struct packet_type {
2204 __be16 type; /* This is really htons(ether_type). */
2205 struct net_device *dev; /* NULL is wildcarded here */
2206 int (*func) (struct sk_buff *,
2207 struct net_device *,
2208 struct packet_type *,
2209 struct net_device *);
2210 bool (*id_match)(struct packet_type *ptype,
2212 struct net *af_packet_net;
2213 void *af_packet_priv;
2214 struct list_head list;
2217 struct offload_callbacks {
2218 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2219 netdev_features_t features);
2220 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2221 struct sk_buff *skb);
2222 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2225 struct packet_offload {
2226 __be16 type; /* This is really htons(ether_type). */
2228 struct offload_callbacks callbacks;
2229 struct list_head list;
2232 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2233 struct pcpu_sw_netstats {
2238 struct u64_stats_sync syncp;
2241 #define __netdev_alloc_pcpu_stats(type, gfp) \
2243 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2246 for_each_possible_cpu(__cpu) { \
2247 typeof(type) *stat; \
2248 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2249 u64_stats_init(&stat->syncp); \
2255 #define netdev_alloc_pcpu_stats(type) \
2256 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2258 enum netdev_lag_tx_type {
2259 NETDEV_LAG_TX_TYPE_UNKNOWN,
2260 NETDEV_LAG_TX_TYPE_RANDOM,
2261 NETDEV_LAG_TX_TYPE_BROADCAST,
2262 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2263 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2264 NETDEV_LAG_TX_TYPE_HASH,
2267 struct netdev_lag_upper_info {
2268 enum netdev_lag_tx_type tx_type;
2271 struct netdev_lag_lower_state_info {
2276 #include <linux/notifier.h>
2278 /* netdevice notifier chain. Please remember to update the rtnetlink
2279 * notification exclusion list in rtnetlink_event() when adding new
2282 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2283 #define NETDEV_DOWN 0x0002
2284 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2285 detected a hardware crash and restarted
2286 - we can use this eg to kick tcp sessions
2288 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2289 #define NETDEV_REGISTER 0x0005
2290 #define NETDEV_UNREGISTER 0x0006
2291 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2292 #define NETDEV_CHANGEADDR 0x0008
2293 #define NETDEV_GOING_DOWN 0x0009
2294 #define NETDEV_CHANGENAME 0x000A
2295 #define NETDEV_FEAT_CHANGE 0x000B
2296 #define NETDEV_BONDING_FAILOVER 0x000C
2297 #define NETDEV_PRE_UP 0x000D
2298 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2299 #define NETDEV_POST_TYPE_CHANGE 0x000F
2300 #define NETDEV_POST_INIT 0x0010
2301 #define NETDEV_UNREGISTER_FINAL 0x0011
2302 #define NETDEV_RELEASE 0x0012
2303 #define NETDEV_NOTIFY_PEERS 0x0013
2304 #define NETDEV_JOIN 0x0014
2305 #define NETDEV_CHANGEUPPER 0x0015
2306 #define NETDEV_RESEND_IGMP 0x0016
2307 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2308 #define NETDEV_CHANGEINFODATA 0x0018
2309 #define NETDEV_BONDING_INFO 0x0019
2310 #define NETDEV_PRECHANGEUPPER 0x001A
2311 #define NETDEV_CHANGELOWERSTATE 0x001B
2312 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2313 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2314 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2316 int register_netdevice_notifier(struct notifier_block *nb);
2317 int unregister_netdevice_notifier(struct notifier_block *nb);
2319 struct netdev_notifier_info {
2320 struct net_device *dev;
2323 struct netdev_notifier_info_ext {
2324 struct netdev_notifier_info info; /* must be first */
2330 struct netdev_notifier_change_info {
2331 struct netdev_notifier_info info; /* must be first */
2332 unsigned int flags_changed;
2335 struct netdev_notifier_changeupper_info {
2336 struct netdev_notifier_info info; /* must be first */
2337 struct net_device *upper_dev; /* new upper dev */
2338 bool master; /* is upper dev master */
2339 bool linking; /* is the notification for link or unlink */
2340 void *upper_info; /* upper dev info */
2343 struct netdev_notifier_changelowerstate_info {
2344 struct netdev_notifier_info info; /* must be first */
2345 void *lower_state_info; /* is lower dev state */
2348 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2349 struct net_device *dev)
2354 static inline struct net_device *
2355 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2360 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2363 extern rwlock_t dev_base_lock; /* Device list lock */
2365 #define for_each_netdev(net, d) \
2366 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2367 #define for_each_netdev_reverse(net, d) \
2368 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2369 #define for_each_netdev_rcu(net, d) \
2370 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2371 #define for_each_netdev_safe(net, d, n) \
2372 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2373 #define for_each_netdev_continue(net, d) \
2374 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2375 #define for_each_netdev_continue_rcu(net, d) \
2376 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2377 #define for_each_netdev_in_bond_rcu(bond, slave) \
2378 for_each_netdev_rcu(&init_net, slave) \
2379 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2380 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2382 static inline struct net_device *next_net_device(struct net_device *dev)
2384 struct list_head *lh;
2388 lh = dev->dev_list.next;
2389 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2392 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2394 struct list_head *lh;
2398 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2399 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2402 static inline struct net_device *first_net_device(struct net *net)
2404 return list_empty(&net->dev_base_head) ? NULL :
2405 net_device_entry(net->dev_base_head.next);
2408 static inline struct net_device *first_net_device_rcu(struct net *net)
2410 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2412 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2415 int netdev_boot_setup_check(struct net_device *dev);
2416 unsigned long netdev_boot_base(const char *prefix, int unit);
2417 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2418 const char *hwaddr);
2419 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2420 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2421 void dev_add_pack(struct packet_type *pt);
2422 void dev_remove_pack(struct packet_type *pt);
2423 void __dev_remove_pack(struct packet_type *pt);
2424 void dev_add_offload(struct packet_offload *po);
2425 void dev_remove_offload(struct packet_offload *po);
2427 int dev_get_iflink(const struct net_device *dev);
2428 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2429 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2430 unsigned short mask);
2431 struct net_device *dev_get_by_name(struct net *net, const char *name);
2432 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2433 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2434 int dev_alloc_name(struct net_device *dev, const char *name);
2435 int dev_open(struct net_device *dev);
2436 void dev_close(struct net_device *dev);
2437 void dev_close_many(struct list_head *head, bool unlink);
2438 void dev_disable_lro(struct net_device *dev);
2439 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2440 int dev_queue_xmit(struct sk_buff *skb);
2441 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2442 int register_netdevice(struct net_device *dev);
2443 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2444 void unregister_netdevice_many(struct list_head *head);
2445 static inline void unregister_netdevice(struct net_device *dev)
2447 unregister_netdevice_queue(dev, NULL);
2450 int netdev_refcnt_read(const struct net_device *dev);
2451 void free_netdev(struct net_device *dev);
2452 void netdev_freemem(struct net_device *dev);
2453 void synchronize_net(void);
2454 int init_dummy_netdev(struct net_device *dev);
2456 DECLARE_PER_CPU(int, xmit_recursion);
2457 #define XMIT_RECURSION_LIMIT 8
2459 static inline int dev_recursion_level(void)
2461 return this_cpu_read(xmit_recursion);
2464 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2465 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2466 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2467 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2468 int netdev_get_name(struct net *net, char *name, int ifindex);
2469 int dev_restart(struct net_device *dev);
2470 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2472 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2474 return NAPI_GRO_CB(skb)->data_offset;
2477 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2479 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2482 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2484 NAPI_GRO_CB(skb)->data_offset += len;
2487 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2488 unsigned int offset)
2490 return NAPI_GRO_CB(skb)->frag0 + offset;
2493 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2495 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2498 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2500 NAPI_GRO_CB(skb)->frag0 = NULL;
2501 NAPI_GRO_CB(skb)->frag0_len = 0;
2504 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2505 unsigned int offset)
2507 if (!pskb_may_pull(skb, hlen))
2510 skb_gro_frag0_invalidate(skb);
2511 return skb->data + offset;
2514 static inline void *skb_gro_network_header(struct sk_buff *skb)
2516 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2517 skb_network_offset(skb);
2520 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2521 const void *start, unsigned int len)
2523 if (NAPI_GRO_CB(skb)->csum_valid)
2524 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2525 csum_partial(start, len, 0));
2528 /* GRO checksum functions. These are logical equivalents of the normal
2529 * checksum functions (in skbuff.h) except that they operate on the GRO
2530 * offsets and fields in sk_buff.
2533 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2535 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2537 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2540 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2544 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2545 skb_checksum_start_offset(skb) <
2546 skb_gro_offset(skb)) &&
2547 !skb_at_gro_remcsum_start(skb) &&
2548 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2549 (!zero_okay || check));
2552 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2555 if (NAPI_GRO_CB(skb)->csum_valid &&
2556 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2559 NAPI_GRO_CB(skb)->csum = psum;
2561 return __skb_gro_checksum_complete(skb);
2564 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2566 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2567 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2568 NAPI_GRO_CB(skb)->csum_cnt--;
2570 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2571 * verified a new top level checksum or an encapsulated one
2572 * during GRO. This saves work if we fallback to normal path.
2574 __skb_incr_checksum_unnecessary(skb);
2578 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2581 __sum16 __ret = 0; \
2582 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2583 __ret = __skb_gro_checksum_validate_complete(skb, \
2584 compute_pseudo(skb, proto)); \
2586 skb_gro_incr_csum_unnecessary(skb); \
2590 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2591 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2593 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2595 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2597 #define skb_gro_checksum_simple_validate(skb) \
2598 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2600 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2602 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2603 !NAPI_GRO_CB(skb)->csum_valid);
2606 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2607 __sum16 check, __wsum pseudo)
2609 NAPI_GRO_CB(skb)->csum = ~pseudo;
2610 NAPI_GRO_CB(skb)->csum_valid = 1;
2613 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2615 if (__skb_gro_checksum_convert_check(skb)) \
2616 __skb_gro_checksum_convert(skb, check, \
2617 compute_pseudo(skb, proto)); \
2620 struct gro_remcsum {
2625 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2631 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2632 unsigned int off, size_t hdrlen,
2633 int start, int offset,
2634 struct gro_remcsum *grc,
2638 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2640 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2643 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2647 ptr = skb_gro_header_fast(skb, off);
2648 if (skb_gro_header_hard(skb, off + plen)) {
2649 ptr = skb_gro_header_slow(skb, off + plen, off);
2654 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2657 /* Adjust skb->csum since we changed the packet */
2658 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2660 grc->offset = off + hdrlen + offset;
2666 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2667 struct gro_remcsum *grc)
2670 size_t plen = grc->offset + sizeof(u16);
2675 ptr = skb_gro_header_fast(skb, grc->offset);
2676 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2677 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2682 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2685 #ifdef CONFIG_XFRM_OFFLOAD
2686 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2688 if (PTR_ERR(pp) != -EINPROGRESS)
2689 NAPI_GRO_CB(skb)->flush |= flush;
2691 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2692 struct sk_buff **pp,
2694 struct gro_remcsum *grc)
2696 if (PTR_ERR(pp) != -EINPROGRESS) {
2697 NAPI_GRO_CB(skb)->flush |= flush;
2698 skb_gro_remcsum_cleanup(skb, grc);
2699 skb->remcsum_offload = 0;
2703 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2705 NAPI_GRO_CB(skb)->flush |= flush;
2707 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2708 struct sk_buff **pp,
2710 struct gro_remcsum *grc)
2712 NAPI_GRO_CB(skb)->flush |= flush;
2713 skb_gro_remcsum_cleanup(skb, grc);
2714 skb->remcsum_offload = 0;
2718 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2719 unsigned short type,
2720 const void *daddr, const void *saddr,
2723 if (!dev->header_ops || !dev->header_ops->create)
2726 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2729 static inline int dev_parse_header(const struct sk_buff *skb,
2730 unsigned char *haddr)
2732 const struct net_device *dev = skb->dev;
2734 if (!dev->header_ops || !dev->header_ops->parse)
2736 return dev->header_ops->parse(skb, haddr);
2739 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2741 const struct net_device *dev = skb->dev;
2743 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2745 return dev->header_ops->parse_protocol(skb);
2748 /* ll_header must have at least hard_header_len allocated */
2749 static inline bool dev_validate_header(const struct net_device *dev,
2750 char *ll_header, int len)
2752 if (likely(len >= dev->hard_header_len))
2754 if (len < dev->min_header_len)
2757 if (capable(CAP_SYS_RAWIO)) {
2758 memset(ll_header + len, 0, dev->hard_header_len - len);
2762 if (dev->header_ops && dev->header_ops->validate)
2763 return dev->header_ops->validate(ll_header, len);
2768 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2769 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2770 static inline int unregister_gifconf(unsigned int family)
2772 return register_gifconf(family, NULL);
2775 #ifdef CONFIG_NET_FLOW_LIMIT
2776 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2777 struct sd_flow_limit {
2779 unsigned int num_buckets;
2780 unsigned int history_head;
2781 u16 history[FLOW_LIMIT_HISTORY];
2785 extern int netdev_flow_limit_table_len;
2786 #endif /* CONFIG_NET_FLOW_LIMIT */
2789 * Incoming packets are placed on per-CPU queues
2791 struct softnet_data {
2792 struct list_head poll_list;
2793 struct sk_buff_head process_queue;
2796 unsigned int processed;
2797 unsigned int time_squeeze;
2798 unsigned int received_rps;
2800 struct softnet_data *rps_ipi_list;
2802 #ifdef CONFIG_NET_FLOW_LIMIT
2803 struct sd_flow_limit __rcu *flow_limit;
2805 struct Qdisc *output_queue;
2806 struct Qdisc **output_queue_tailp;
2807 struct sk_buff *completion_queue;
2810 /* input_queue_head should be written by cpu owning this struct,
2811 * and only read by other cpus. Worth using a cache line.
2813 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2815 /* Elements below can be accessed between CPUs for RPS/RFS */
2816 call_single_data_t csd ____cacheline_aligned_in_smp;
2817 struct softnet_data *rps_ipi_next;
2819 unsigned int input_queue_tail;
2821 unsigned int dropped;
2822 struct sk_buff_head input_pkt_queue;
2823 struct napi_struct backlog;
2827 static inline void input_queue_head_incr(struct softnet_data *sd)
2830 sd->input_queue_head++;
2834 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2835 unsigned int *qtail)
2838 *qtail = ++sd->input_queue_tail;
2842 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2844 void __netif_schedule(struct Qdisc *q);
2845 void netif_schedule_queue(struct netdev_queue *txq);
2847 static inline void netif_tx_schedule_all(struct net_device *dev)
2851 for (i = 0; i < dev->num_tx_queues; i++)
2852 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2855 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2857 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2861 * netif_start_queue - allow transmit
2862 * @dev: network device
2864 * Allow upper layers to call the device hard_start_xmit routine.
2866 static inline void netif_start_queue(struct net_device *dev)
2868 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2871 static inline void netif_tx_start_all_queues(struct net_device *dev)
2875 for (i = 0; i < dev->num_tx_queues; i++) {
2876 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2877 netif_tx_start_queue(txq);
2881 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2884 * netif_wake_queue - restart transmit
2885 * @dev: network device
2887 * Allow upper layers to call the device hard_start_xmit routine.
2888 * Used for flow control when transmit resources are available.
2890 static inline void netif_wake_queue(struct net_device *dev)
2892 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2895 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2899 for (i = 0; i < dev->num_tx_queues; i++) {
2900 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2901 netif_tx_wake_queue(txq);
2905 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2907 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2911 * netif_stop_queue - stop transmitted packets
2912 * @dev: network device
2914 * Stop upper layers calling the device hard_start_xmit routine.
2915 * Used for flow control when transmit resources are unavailable.
2917 static inline void netif_stop_queue(struct net_device *dev)
2919 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2922 void netif_tx_stop_all_queues(struct net_device *dev);
2924 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2926 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2930 * netif_queue_stopped - test if transmit queue is flowblocked
2931 * @dev: network device
2933 * Test if transmit queue on device is currently unable to send.
2935 static inline bool netif_queue_stopped(const struct net_device *dev)
2937 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2940 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2942 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2946 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2948 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2952 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2954 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2958 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2959 * @dev_queue: pointer to transmit queue
2961 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2962 * to give appropriate hint to the CPU.
2964 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2967 prefetchw(&dev_queue->dql.num_queued);
2972 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2973 * @dev_queue: pointer to transmit queue
2975 * BQL enabled drivers might use this helper in their TX completion path,
2976 * to give appropriate hint to the CPU.
2978 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2981 prefetchw(&dev_queue->dql.limit);
2985 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2989 dql_queued(&dev_queue->dql, bytes);
2991 if (likely(dql_avail(&dev_queue->dql) >= 0))
2994 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2997 * The XOFF flag must be set before checking the dql_avail below,
2998 * because in netdev_tx_completed_queue we update the dql_completed
2999 * before checking the XOFF flag.
3003 /* check again in case another CPU has just made room avail */
3004 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3005 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3010 * netdev_sent_queue - report the number of bytes queued to hardware
3011 * @dev: network device
3012 * @bytes: number of bytes queued to the hardware device queue
3014 * Report the number of bytes queued for sending/completion to the network
3015 * device hardware queue. @bytes should be a good approximation and should
3016 * exactly match netdev_completed_queue() @bytes
3018 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3020 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3023 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3024 unsigned int pkts, unsigned int bytes)
3027 if (unlikely(!bytes))
3030 dql_completed(&dev_queue->dql, bytes);
3033 * Without the memory barrier there is a small possiblity that
3034 * netdev_tx_sent_queue will miss the update and cause the queue to
3035 * be stopped forever
3039 if (dql_avail(&dev_queue->dql) < 0)
3042 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3043 netif_schedule_queue(dev_queue);
3048 * netdev_completed_queue - report bytes and packets completed by device
3049 * @dev: network device
3050 * @pkts: actual number of packets sent over the medium
3051 * @bytes: actual number of bytes sent over the medium
3053 * Report the number of bytes and packets transmitted by the network device
3054 * hardware queue over the physical medium, @bytes must exactly match the
3055 * @bytes amount passed to netdev_sent_queue()
3057 static inline void netdev_completed_queue(struct net_device *dev,
3058 unsigned int pkts, unsigned int bytes)
3060 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3063 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3066 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3072 * netdev_reset_queue - reset the packets and bytes count of a network device
3073 * @dev_queue: network device
3075 * Reset the bytes and packet count of a network device and clear the
3076 * software flow control OFF bit for this network device
3078 static inline void netdev_reset_queue(struct net_device *dev_queue)
3080 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3084 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3085 * @dev: network device
3086 * @queue_index: given tx queue index
3088 * Returns 0 if given tx queue index >= number of device tx queues,
3089 * otherwise returns the originally passed tx queue index.
3091 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3093 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3094 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3095 dev->name, queue_index,
3096 dev->real_num_tx_queues);
3104 * netif_running - test if up
3105 * @dev: network device
3107 * Test if the device has been brought up.
3109 static inline bool netif_running(const struct net_device *dev)
3111 return test_bit(__LINK_STATE_START, &dev->state);
3115 * Routines to manage the subqueues on a device. We only need start,
3116 * stop, and a check if it's stopped. All other device management is
3117 * done at the overall netdevice level.
3118 * Also test the device if we're multiqueue.
3122 * netif_start_subqueue - allow sending packets on subqueue
3123 * @dev: network device
3124 * @queue_index: sub queue index
3126 * Start individual transmit queue of a device with multiple transmit queues.
3128 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3130 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3132 netif_tx_start_queue(txq);
3136 * netif_stop_subqueue - stop sending packets on subqueue
3137 * @dev: network device
3138 * @queue_index: sub queue index
3140 * Stop individual transmit queue of a device with multiple transmit queues.
3142 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3144 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3145 netif_tx_stop_queue(txq);
3149 * netif_subqueue_stopped - test status of subqueue
3150 * @dev: network device
3151 * @queue_index: sub queue index
3153 * Check individual transmit queue of a device with multiple transmit queues.
3155 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3158 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3160 return netif_tx_queue_stopped(txq);
3163 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3164 struct sk_buff *skb)
3166 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3170 * netif_wake_subqueue - allow sending packets on subqueue
3171 * @dev: network device
3172 * @queue_index: sub queue index
3174 * Resume individual transmit queue of a device with multiple transmit queues.
3176 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3178 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3180 netif_tx_wake_queue(txq);
3184 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3187 static inline int netif_set_xps_queue(struct net_device *dev,
3188 const struct cpumask *mask,
3195 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3196 unsigned int num_tx_queues);
3199 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3200 * as a distribution range limit for the returned value.
3202 static inline u16 skb_tx_hash(const struct net_device *dev,
3203 struct sk_buff *skb)
3205 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3209 * netif_is_multiqueue - test if device has multiple transmit queues
3210 * @dev: network device
3212 * Check if device has multiple transmit queues
3214 static inline bool netif_is_multiqueue(const struct net_device *dev)
3216 return dev->num_tx_queues > 1;
3219 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3222 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3224 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3232 static inline unsigned int get_netdev_rx_queue_index(
3233 struct netdev_rx_queue *queue)
3235 struct net_device *dev = queue->dev;
3236 int index = queue - dev->_rx;
3238 BUG_ON(index >= dev->num_rx_queues);
3243 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3244 int netif_get_num_default_rss_queues(void);
3246 enum skb_free_reason {
3247 SKB_REASON_CONSUMED,
3251 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3252 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3255 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3256 * interrupt context or with hardware interrupts being disabled.
3257 * (in_irq() || irqs_disabled())
3259 * We provide four helpers that can be used in following contexts :
3261 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3262 * replacing kfree_skb(skb)
3264 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3265 * Typically used in place of consume_skb(skb) in TX completion path
3267 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3268 * replacing kfree_skb(skb)
3270 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3271 * and consumed a packet. Used in place of consume_skb(skb)
3273 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3275 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3278 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3280 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3283 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3285 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3288 static inline void dev_consume_skb_any(struct sk_buff *skb)
3290 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3293 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3294 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3295 int netif_rx(struct sk_buff *skb);
3296 int netif_rx_ni(struct sk_buff *skb);
3297 int netif_receive_skb(struct sk_buff *skb);
3298 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3299 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3300 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3301 gro_result_t napi_gro_frags(struct napi_struct *napi);
3302 struct packet_offload *gro_find_receive_by_type(__be16 type);
3303 struct packet_offload *gro_find_complete_by_type(__be16 type);
3305 static inline void napi_free_frags(struct napi_struct *napi)
3307 kfree_skb(napi->skb);
3311 bool netdev_is_rx_handler_busy(struct net_device *dev);
3312 int netdev_rx_handler_register(struct net_device *dev,
3313 rx_handler_func_t *rx_handler,
3314 void *rx_handler_data);
3315 void netdev_rx_handler_unregister(struct net_device *dev);
3317 bool dev_valid_name(const char *name);
3318 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3319 int dev_ethtool(struct net *net, struct ifreq *);
3320 unsigned int dev_get_flags(const struct net_device *);
3321 int __dev_change_flags(struct net_device *, unsigned int flags);
3322 int dev_change_flags(struct net_device *, unsigned int);
3323 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3324 unsigned int gchanges);
3325 int dev_change_name(struct net_device *, const char *);
3326 int dev_set_alias(struct net_device *, const char *, size_t);
3327 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3328 int __dev_set_mtu(struct net_device *, int);
3329 int dev_set_mtu(struct net_device *, int);
3330 int dev_validate_mtu(struct net_device *dev, int mtu);
3331 void dev_set_group(struct net_device *, int);
3332 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3333 int dev_change_carrier(struct net_device *, bool new_carrier);
3334 int dev_get_phys_port_id(struct net_device *dev,
3335 struct netdev_phys_item_id *ppid);
3336 int dev_get_phys_port_name(struct net_device *dev,
3337 char *name, size_t len);
3338 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3339 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3340 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3341 struct netdev_queue *txq, int *ret);
3343 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3344 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3346 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3348 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3349 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3350 bool is_skb_forwardable(const struct net_device *dev,
3351 const struct sk_buff *skb);
3353 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3354 struct sk_buff *skb)
3356 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3357 unlikely(!is_skb_forwardable(dev, skb))) {
3358 atomic_long_inc(&dev->rx_dropped);
3363 skb_scrub_packet(skb, true);
3368 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3370 extern int netdev_budget;
3371 extern unsigned int netdev_budget_usecs;
3373 /* Called by rtnetlink.c:rtnl_unlock() */
3374 void netdev_run_todo(void);
3377 * dev_put - release reference to device
3378 * @dev: network device
3380 * Release reference to device to allow it to be freed.
3382 static inline void dev_put(struct net_device *dev)
3385 this_cpu_dec(*dev->pcpu_refcnt);
3389 * dev_hold - get reference to device
3390 * @dev: network device
3392 * Hold reference to device to keep it from being freed.
3394 static inline void dev_hold(struct net_device *dev)
3397 this_cpu_inc(*dev->pcpu_refcnt);
3400 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3401 * and _off may be called from IRQ context, but it is caller
3402 * who is responsible for serialization of these calls.
3404 * The name carrier is inappropriate, these functions should really be
3405 * called netif_lowerlayer_*() because they represent the state of any
3406 * kind of lower layer not just hardware media.
3409 void linkwatch_init_dev(struct net_device *dev);
3410 void linkwatch_fire_event(struct net_device *dev);
3411 void linkwatch_forget_dev(struct net_device *dev);
3414 * netif_carrier_ok - test if carrier present
3415 * @dev: network device
3417 * Check if carrier is present on device
3419 static inline bool netif_carrier_ok(const struct net_device *dev)
3421 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3424 unsigned long dev_trans_start(struct net_device *dev);
3426 void __netdev_watchdog_up(struct net_device *dev);
3428 void netif_carrier_on(struct net_device *dev);
3430 void netif_carrier_off(struct net_device *dev);
3433 * netif_dormant_on - mark device as dormant.
3434 * @dev: network device
3436 * Mark device as dormant (as per RFC2863).
3438 * The dormant state indicates that the relevant interface is not
3439 * actually in a condition to pass packets (i.e., it is not 'up') but is
3440 * in a "pending" state, waiting for some external event. For "on-
3441 * demand" interfaces, this new state identifies the situation where the
3442 * interface is waiting for events to place it in the up state.
3444 static inline void netif_dormant_on(struct net_device *dev)
3446 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3447 linkwatch_fire_event(dev);
3451 * netif_dormant_off - set device as not dormant.
3452 * @dev: network device
3454 * Device is not in dormant state.
3456 static inline void netif_dormant_off(struct net_device *dev)
3458 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3459 linkwatch_fire_event(dev);
3463 * netif_dormant - test if device is dormant
3464 * @dev: network device
3466 * Check if device is dormant.
3468 static inline bool netif_dormant(const struct net_device *dev)
3470 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3475 * netif_oper_up - test if device is operational
3476 * @dev: network device
3478 * Check if carrier is operational
3480 static inline bool netif_oper_up(const struct net_device *dev)
3482 return (dev->operstate == IF_OPER_UP ||
3483 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3487 * netif_device_present - is device available or removed
3488 * @dev: network device
3490 * Check if device has not been removed from system.
3492 static inline bool netif_device_present(struct net_device *dev)
3494 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3497 void netif_device_detach(struct net_device *dev);
3499 void netif_device_attach(struct net_device *dev);
3502 * Network interface message level settings
3506 NETIF_MSG_DRV = 0x0001,
3507 NETIF_MSG_PROBE = 0x0002,
3508 NETIF_MSG_LINK = 0x0004,
3509 NETIF_MSG_TIMER = 0x0008,
3510 NETIF_MSG_IFDOWN = 0x0010,
3511 NETIF_MSG_IFUP = 0x0020,
3512 NETIF_MSG_RX_ERR = 0x0040,
3513 NETIF_MSG_TX_ERR = 0x0080,
3514 NETIF_MSG_TX_QUEUED = 0x0100,
3515 NETIF_MSG_INTR = 0x0200,
3516 NETIF_MSG_TX_DONE = 0x0400,
3517 NETIF_MSG_RX_STATUS = 0x0800,
3518 NETIF_MSG_PKTDATA = 0x1000,
3519 NETIF_MSG_HW = 0x2000,
3520 NETIF_MSG_WOL = 0x4000,
3523 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3524 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3525 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3526 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3527 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3528 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3529 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3530 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3531 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3532 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3533 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3534 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3535 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3536 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3537 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3539 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3542 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3543 return default_msg_enable_bits;
3544 if (debug_value == 0) /* no output */
3546 /* set low N bits */
3547 return (1U << debug_value) - 1;
3550 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3552 spin_lock(&txq->_xmit_lock);
3553 txq->xmit_lock_owner = cpu;
3556 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3558 __acquire(&txq->_xmit_lock);
3562 static inline void __netif_tx_release(struct netdev_queue *txq)
3564 __release(&txq->_xmit_lock);
3567 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3569 spin_lock_bh(&txq->_xmit_lock);
3570 txq->xmit_lock_owner = smp_processor_id();
3573 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3575 bool ok = spin_trylock(&txq->_xmit_lock);
3577 txq->xmit_lock_owner = smp_processor_id();
3581 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3583 txq->xmit_lock_owner = -1;
3584 spin_unlock(&txq->_xmit_lock);
3587 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3589 txq->xmit_lock_owner = -1;
3590 spin_unlock_bh(&txq->_xmit_lock);
3593 static inline void txq_trans_update(struct netdev_queue *txq)
3595 if (txq->xmit_lock_owner != -1)
3596 txq->trans_start = jiffies;
3599 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3600 static inline void netif_trans_update(struct net_device *dev)
3602 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3604 if (txq->trans_start != jiffies)
3605 txq->trans_start = jiffies;
3609 * netif_tx_lock - grab network device transmit lock
3610 * @dev: network device
3612 * Get network device transmit lock
3614 static inline void netif_tx_lock(struct net_device *dev)
3619 spin_lock(&dev->tx_global_lock);
3620 cpu = smp_processor_id();
3621 for (i = 0; i < dev->num_tx_queues; i++) {
3622 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3624 /* We are the only thread of execution doing a
3625 * freeze, but we have to grab the _xmit_lock in
3626 * order to synchronize with threads which are in
3627 * the ->hard_start_xmit() handler and already
3628 * checked the frozen bit.
3630 __netif_tx_lock(txq, cpu);
3631 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3632 __netif_tx_unlock(txq);
3636 static inline void netif_tx_lock_bh(struct net_device *dev)
3642 static inline void netif_tx_unlock(struct net_device *dev)
3646 for (i = 0; i < dev->num_tx_queues; i++) {
3647 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3649 /* No need to grab the _xmit_lock here. If the
3650 * queue is not stopped for another reason, we
3653 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3654 netif_schedule_queue(txq);
3656 spin_unlock(&dev->tx_global_lock);
3659 static inline void netif_tx_unlock_bh(struct net_device *dev)
3661 netif_tx_unlock(dev);
3665 #define HARD_TX_LOCK(dev, txq, cpu) { \
3666 if ((dev->features & NETIF_F_LLTX) == 0) { \
3667 __netif_tx_lock(txq, cpu); \
3669 __netif_tx_acquire(txq); \
3673 #define HARD_TX_TRYLOCK(dev, txq) \
3674 (((dev->features & NETIF_F_LLTX) == 0) ? \
3675 __netif_tx_trylock(txq) : \
3676 __netif_tx_acquire(txq))
3678 #define HARD_TX_UNLOCK(dev, txq) { \
3679 if ((dev->features & NETIF_F_LLTX) == 0) { \
3680 __netif_tx_unlock(txq); \
3682 __netif_tx_release(txq); \
3686 static inline void netif_tx_disable(struct net_device *dev)
3692 cpu = smp_processor_id();
3693 spin_lock(&dev->tx_global_lock);
3694 for (i = 0; i < dev->num_tx_queues; i++) {
3695 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3697 __netif_tx_lock(txq, cpu);
3698 netif_tx_stop_queue(txq);
3699 __netif_tx_unlock(txq);
3701 spin_unlock(&dev->tx_global_lock);
3705 static inline void netif_addr_lock(struct net_device *dev)
3707 spin_lock(&dev->addr_list_lock);
3710 static inline void netif_addr_lock_nested(struct net_device *dev)
3712 int subclass = SINGLE_DEPTH_NESTING;
3714 if (dev->netdev_ops->ndo_get_lock_subclass)
3715 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3717 spin_lock_nested(&dev->addr_list_lock, subclass);
3720 static inline void netif_addr_lock_bh(struct net_device *dev)
3722 spin_lock_bh(&dev->addr_list_lock);
3725 static inline void netif_addr_unlock(struct net_device *dev)
3727 spin_unlock(&dev->addr_list_lock);
3730 static inline void netif_addr_unlock_bh(struct net_device *dev)
3732 spin_unlock_bh(&dev->addr_list_lock);
3736 * dev_addrs walker. Should be used only for read access. Call with
3737 * rcu_read_lock held.
3739 #define for_each_dev_addr(dev, ha) \
3740 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3742 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3744 void ether_setup(struct net_device *dev);
3746 /* Support for loadable net-drivers */
3747 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3748 unsigned char name_assign_type,
3749 void (*setup)(struct net_device *),
3750 unsigned int txqs, unsigned int rxqs);
3751 int dev_get_valid_name(struct net *net, struct net_device *dev,
3754 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3755 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3757 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3758 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3761 int register_netdev(struct net_device *dev);
3762 void unregister_netdev(struct net_device *dev);
3764 /* General hardware address lists handling functions */
3765 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3766 struct netdev_hw_addr_list *from_list, int addr_len);
3767 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3768 struct netdev_hw_addr_list *from_list, int addr_len);
3769 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3770 struct net_device *dev,
3771 int (*sync)(struct net_device *, const unsigned char *),
3772 int (*unsync)(struct net_device *,
3773 const unsigned char *));
3774 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3775 struct net_device *dev,
3776 int (*unsync)(struct net_device *,
3777 const unsigned char *));
3778 void __hw_addr_init(struct netdev_hw_addr_list *list);
3780 /* Functions used for device addresses handling */
3781 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3782 unsigned char addr_type);
3783 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3784 unsigned char addr_type);
3785 void dev_addr_flush(struct net_device *dev);
3786 int dev_addr_init(struct net_device *dev);
3788 /* Functions used for unicast addresses handling */
3789 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3790 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3791 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3792 int dev_uc_sync(struct net_device *to, struct net_device *from);
3793 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3794 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3795 void dev_uc_flush(struct net_device *dev);
3796 void dev_uc_init(struct net_device *dev);
3799 * __dev_uc_sync - Synchonize device's unicast list
3800 * @dev: device to sync
3801 * @sync: function to call if address should be added
3802 * @unsync: function to call if address should be removed
3804 * Add newly added addresses to the interface, and release
3805 * addresses that have been deleted.
3807 static inline int __dev_uc_sync(struct net_device *dev,
3808 int (*sync)(struct net_device *,
3809 const unsigned char *),
3810 int (*unsync)(struct net_device *,
3811 const unsigned char *))
3813 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3817 * __dev_uc_unsync - Remove synchronized addresses from device
3818 * @dev: device to sync
3819 * @unsync: function to call if address should be removed
3821 * Remove all addresses that were added to the device by dev_uc_sync().
3823 static inline void __dev_uc_unsync(struct net_device *dev,
3824 int (*unsync)(struct net_device *,
3825 const unsigned char *))
3827 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3830 /* Functions used for multicast addresses handling */
3831 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3832 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3833 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3834 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3835 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3836 int dev_mc_sync(struct net_device *to, struct net_device *from);
3837 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3838 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3839 void dev_mc_flush(struct net_device *dev);
3840 void dev_mc_init(struct net_device *dev);
3843 * __dev_mc_sync - Synchonize device's multicast list
3844 * @dev: device to sync
3845 * @sync: function to call if address should be added
3846 * @unsync: function to call if address should be removed
3848 * Add newly added addresses to the interface, and release
3849 * addresses that have been deleted.
3851 static inline int __dev_mc_sync(struct net_device *dev,
3852 int (*sync)(struct net_device *,
3853 const unsigned char *),
3854 int (*unsync)(struct net_device *,
3855 const unsigned char *))
3857 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3861 * __dev_mc_unsync - Remove synchronized addresses from device
3862 * @dev: device to sync
3863 * @unsync: function to call if address should be removed
3865 * Remove all addresses that were added to the device by dev_mc_sync().
3867 static inline void __dev_mc_unsync(struct net_device *dev,
3868 int (*unsync)(struct net_device *,
3869 const unsigned char *))
3871 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3874 /* Functions used for secondary unicast and multicast support */
3875 void dev_set_rx_mode(struct net_device *dev);
3876 void __dev_set_rx_mode(struct net_device *dev);
3877 int dev_set_promiscuity(struct net_device *dev, int inc);
3878 int dev_set_allmulti(struct net_device *dev, int inc);
3879 void netdev_state_change(struct net_device *dev);
3880 void netdev_notify_peers(struct net_device *dev);
3881 void netdev_features_change(struct net_device *dev);
3882 /* Load a device via the kmod */
3883 void dev_load(struct net *net, const char *name);
3884 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3885 struct rtnl_link_stats64 *storage);
3886 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3887 const struct net_device_stats *netdev_stats);
3889 extern int netdev_max_backlog;
3890 extern int netdev_tstamp_prequeue;
3891 extern int weight_p;
3892 extern int dev_weight_rx_bias;
3893 extern int dev_weight_tx_bias;
3894 extern int dev_rx_weight;
3895 extern int dev_tx_weight;
3897 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3898 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3899 struct list_head **iter);
3900 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3901 struct list_head **iter);
3903 /* iterate through upper list, must be called under RCU read lock */
3904 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3905 for (iter = &(dev)->adj_list.upper, \
3906 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3908 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3910 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3911 int (*fn)(struct net_device *upper_dev,
3915 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3916 struct net_device *upper_dev);
3918 bool netdev_has_any_upper_dev(struct net_device *dev);
3920 void *netdev_lower_get_next_private(struct net_device *dev,
3921 struct list_head **iter);
3922 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3923 struct list_head **iter);
3925 #define netdev_for_each_lower_private(dev, priv, iter) \
3926 for (iter = (dev)->adj_list.lower.next, \
3927 priv = netdev_lower_get_next_private(dev, &(iter)); \
3929 priv = netdev_lower_get_next_private(dev, &(iter)))
3931 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3932 for (iter = &(dev)->adj_list.lower, \
3933 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3935 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3937 void *netdev_lower_get_next(struct net_device *dev,
3938 struct list_head **iter);
3940 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3941 for (iter = (dev)->adj_list.lower.next, \
3942 ldev = netdev_lower_get_next(dev, &(iter)); \
3944 ldev = netdev_lower_get_next(dev, &(iter)))
3946 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3947 struct list_head **iter);
3948 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3949 struct list_head **iter);
3951 int netdev_walk_all_lower_dev(struct net_device *dev,
3952 int (*fn)(struct net_device *lower_dev,
3955 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3956 int (*fn)(struct net_device *lower_dev,
3960 void *netdev_adjacent_get_private(struct list_head *adj_list);
3961 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3962 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3963 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3964 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3965 int netdev_master_upper_dev_link(struct net_device *dev,
3966 struct net_device *upper_dev,
3967 void *upper_priv, void *upper_info);
3968 void netdev_upper_dev_unlink(struct net_device *dev,
3969 struct net_device *upper_dev);
3970 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3971 void *netdev_lower_dev_get_private(struct net_device *dev,
3972 struct net_device *lower_dev);
3973 void netdev_lower_state_changed(struct net_device *lower_dev,
3974 void *lower_state_info);
3976 /* RSS keys are 40 or 52 bytes long */
3977 #define NETDEV_RSS_KEY_LEN 52
3978 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3979 void netdev_rss_key_fill(void *buffer, size_t len);
3981 int dev_get_nest_level(struct net_device *dev);
3982 int skb_checksum_help(struct sk_buff *skb);
3983 int skb_crc32c_csum_help(struct sk_buff *skb);
3984 int skb_csum_hwoffload_help(struct sk_buff *skb,
3985 const netdev_features_t features);
3987 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3988 netdev_features_t features, bool tx_path);
3989 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3990 netdev_features_t features);
3992 struct netdev_bonding_info {
3997 struct netdev_notifier_bonding_info {
3998 struct netdev_notifier_info info; /* must be first */
3999 struct netdev_bonding_info bonding_info;
4002 void netdev_bonding_info_change(struct net_device *dev,
4003 struct netdev_bonding_info *bonding_info);
4006 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4008 return __skb_gso_segment(skb, features, true);
4010 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4012 static inline bool can_checksum_protocol(netdev_features_t features,
4015 if (protocol == htons(ETH_P_FCOE))
4016 return !!(features & NETIF_F_FCOE_CRC);
4018 /* Assume this is an IP checksum (not SCTP CRC) */
4020 if (features & NETIF_F_HW_CSUM) {
4021 /* Can checksum everything */
4026 case htons(ETH_P_IP):
4027 return !!(features & NETIF_F_IP_CSUM);
4028 case htons(ETH_P_IPV6):
4029 return !!(features & NETIF_F_IPV6_CSUM);
4036 void netdev_rx_csum_fault(struct net_device *dev);
4038 static inline void netdev_rx_csum_fault(struct net_device *dev)
4042 /* rx skb timestamps */
4043 void net_enable_timestamp(void);
4044 void net_disable_timestamp(void);
4046 #ifdef CONFIG_PROC_FS
4047 int __init dev_proc_init(void);
4049 #define dev_proc_init() 0
4052 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4053 struct sk_buff *skb, struct net_device *dev,
4056 skb->xmit_more = more ? 1 : 0;
4057 return ops->ndo_start_xmit(skb, dev);
4060 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4061 struct netdev_queue *txq, bool more)
4063 const struct net_device_ops *ops = dev->netdev_ops;
4066 rc = __netdev_start_xmit(ops, skb, dev, more);
4067 if (rc == NETDEV_TX_OK)
4068 txq_trans_update(txq);
4073 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4075 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4078 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4080 return netdev_class_create_file_ns(class_attr, NULL);
4083 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4085 netdev_class_remove_file_ns(class_attr, NULL);
4088 extern const struct kobj_ns_type_operations net_ns_type_operations;
4090 const char *netdev_drivername(const struct net_device *dev);
4092 void linkwatch_run_queue(void);
4094 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4095 netdev_features_t f2)
4097 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4098 if (f1 & NETIF_F_HW_CSUM)
4099 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4101 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4107 static inline netdev_features_t netdev_get_wanted_features(
4108 struct net_device *dev)
4110 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4112 netdev_features_t netdev_increment_features(netdev_features_t all,
4113 netdev_features_t one, netdev_features_t mask);
4115 /* Allow TSO being used on stacked device :
4116 * Performing the GSO segmentation before last device
4117 * is a performance improvement.
4119 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4120 netdev_features_t mask)
4122 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4125 int __netdev_update_features(struct net_device *dev);
4126 void netdev_update_features(struct net_device *dev);
4127 void netdev_change_features(struct net_device *dev);
4129 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4130 struct net_device *dev);
4132 netdev_features_t passthru_features_check(struct sk_buff *skb,
4133 struct net_device *dev,
4134 netdev_features_t features);
4135 netdev_features_t netif_skb_features(struct sk_buff *skb);
4137 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4139 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4141 /* check flags correspondence */
4142 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4143 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4144 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4145 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4146 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4147 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4148 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4149 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4150 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4151 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4152 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4153 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4154 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4155 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4156 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4157 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4158 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4160 return (features & feature) == feature;
4163 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4165 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4166 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4169 static inline bool netif_needs_gso(struct sk_buff *skb,
4170 netdev_features_t features)
4172 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4173 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4174 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4177 static inline void netif_set_gso_max_size(struct net_device *dev,
4180 dev->gso_max_size = size;
4183 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4184 int pulled_hlen, u16 mac_offset,
4187 skb->protocol = protocol;
4188 skb->encapsulation = 1;
4189 skb_push(skb, pulled_hlen);
4190 skb_reset_transport_header(skb);
4191 skb->mac_header = mac_offset;
4192 skb->network_header = skb->mac_header + mac_len;
4193 skb->mac_len = mac_len;
4196 static inline bool netif_is_macsec(const struct net_device *dev)
4198 return dev->priv_flags & IFF_MACSEC;
4201 static inline bool netif_is_macvlan(const struct net_device *dev)
4203 return dev->priv_flags & IFF_MACVLAN;
4206 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4208 return dev->priv_flags & IFF_MACVLAN_PORT;
4211 static inline bool netif_is_ipvlan(const struct net_device *dev)
4213 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4216 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4218 return dev->priv_flags & IFF_IPVLAN_MASTER;
4221 static inline bool netif_is_bond_master(const struct net_device *dev)
4223 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4226 static inline bool netif_is_bond_slave(const struct net_device *dev)
4228 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4231 static inline bool netif_supports_nofcs(struct net_device *dev)
4233 return dev->priv_flags & IFF_SUPP_NOFCS;
4236 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4238 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4241 static inline bool netif_is_l3_master(const struct net_device *dev)
4243 return dev->priv_flags & IFF_L3MDEV_MASTER;
4246 static inline bool netif_is_l3_slave(const struct net_device *dev)
4248 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4251 static inline bool netif_is_bridge_master(const struct net_device *dev)
4253 return dev->priv_flags & IFF_EBRIDGE;
4256 static inline bool netif_is_bridge_port(const struct net_device *dev)
4258 return dev->priv_flags & IFF_BRIDGE_PORT;
4261 static inline bool netif_is_ovs_master(const struct net_device *dev)
4263 return dev->priv_flags & IFF_OPENVSWITCH;
4266 static inline bool netif_is_ovs_port(const struct net_device *dev)
4268 return dev->priv_flags & IFF_OVS_DATAPATH;
4271 static inline bool netif_is_team_master(const struct net_device *dev)
4273 return dev->priv_flags & IFF_TEAM;
4276 static inline bool netif_is_team_port(const struct net_device *dev)
4278 return dev->priv_flags & IFF_TEAM_PORT;
4281 static inline bool netif_is_lag_master(const struct net_device *dev)
4283 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4286 static inline bool netif_is_lag_port(const struct net_device *dev)
4288 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4291 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4293 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4296 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4297 static inline void netif_keep_dst(struct net_device *dev)
4299 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4302 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4303 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4305 /* TODO: reserve and use an additional IFF bit, if we get more users */
4306 return dev->priv_flags & IFF_MACSEC;
4309 extern struct pernet_operations __net_initdata loopback_net_ops;
4311 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4313 /* netdev_printk helpers, similar to dev_printk */
4315 static inline const char *netdev_name(const struct net_device *dev)
4317 if (!dev->name[0] || strchr(dev->name, '%'))
4318 return "(unnamed net_device)";
4322 static inline bool netdev_unregistering(const struct net_device *dev)
4324 return dev->reg_state == NETREG_UNREGISTERING;
4327 static inline const char *netdev_reg_state(const struct net_device *dev)
4329 switch (dev->reg_state) {
4330 case NETREG_UNINITIALIZED: return " (uninitialized)";
4331 case NETREG_REGISTERED: return "";
4332 case NETREG_UNREGISTERING: return " (unregistering)";
4333 case NETREG_UNREGISTERED: return " (unregistered)";
4334 case NETREG_RELEASED: return " (released)";
4335 case NETREG_DUMMY: return " (dummy)";
4338 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4339 return " (unknown)";
4343 void netdev_printk(const char *level, const struct net_device *dev,
4344 const char *format, ...);
4346 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4348 void netdev_alert(const struct net_device *dev, const char *format, ...);
4350 void netdev_crit(const struct net_device *dev, const char *format, ...);
4352 void netdev_err(const struct net_device *dev, const char *format, ...);
4354 void netdev_warn(const struct net_device *dev, const char *format, ...);
4356 void netdev_notice(const struct net_device *dev, const char *format, ...);
4358 void netdev_info(const struct net_device *dev, const char *format, ...);
4360 #define MODULE_ALIAS_NETDEV(device) \
4361 MODULE_ALIAS("netdev-" device)
4363 #if defined(CONFIG_DYNAMIC_DEBUG)
4364 #define netdev_dbg(__dev, format, args...) \
4366 dynamic_netdev_dbg(__dev, format, ##args); \
4368 #elif defined(DEBUG)
4369 #define netdev_dbg(__dev, format, args...) \
4370 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4372 #define netdev_dbg(__dev, format, args...) \
4375 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4379 #if defined(VERBOSE_DEBUG)
4380 #define netdev_vdbg netdev_dbg
4383 #define netdev_vdbg(dev, format, args...) \
4386 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4392 * netdev_WARN() acts like dev_printk(), but with the key difference
4393 * of using a WARN/WARN_ON to get the message out, including the
4394 * file/line information and a backtrace.
4396 #define netdev_WARN(dev, format, args...) \
4397 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4398 netdev_reg_state(dev), ##args)
4400 /* netif printk helpers, similar to netdev_printk */
4402 #define netif_printk(priv, type, level, dev, fmt, args...) \
4404 if (netif_msg_##type(priv)) \
4405 netdev_printk(level, (dev), fmt, ##args); \
4408 #define netif_level(level, priv, type, dev, fmt, args...) \
4410 if (netif_msg_##type(priv)) \
4411 netdev_##level(dev, fmt, ##args); \
4414 #define netif_emerg(priv, type, dev, fmt, args...) \
4415 netif_level(emerg, priv, type, dev, fmt, ##args)
4416 #define netif_alert(priv, type, dev, fmt, args...) \
4417 netif_level(alert, priv, type, dev, fmt, ##args)
4418 #define netif_crit(priv, type, dev, fmt, args...) \
4419 netif_level(crit, priv, type, dev, fmt, ##args)
4420 #define netif_err(priv, type, dev, fmt, args...) \
4421 netif_level(err, priv, type, dev, fmt, ##args)
4422 #define netif_warn(priv, type, dev, fmt, args...) \
4423 netif_level(warn, priv, type, dev, fmt, ##args)
4424 #define netif_notice(priv, type, dev, fmt, args...) \
4425 netif_level(notice, priv, type, dev, fmt, ##args)
4426 #define netif_info(priv, type, dev, fmt, args...) \
4427 netif_level(info, priv, type, dev, fmt, ##args)
4429 #if defined(CONFIG_DYNAMIC_DEBUG)
4430 #define netif_dbg(priv, type, netdev, format, args...) \
4432 if (netif_msg_##type(priv)) \
4433 dynamic_netdev_dbg(netdev, format, ##args); \
4435 #elif defined(DEBUG)
4436 #define netif_dbg(priv, type, dev, format, args...) \
4437 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4439 #define netif_dbg(priv, type, dev, format, args...) \
4442 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4447 /* if @cond then downgrade to debug, else print at @level */
4448 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4451 netif_dbg(priv, type, netdev, fmt, ##args); \
4453 netif_ ## level(priv, type, netdev, fmt, ##args); \
4456 #if defined(VERBOSE_DEBUG)
4457 #define netif_vdbg netif_dbg
4459 #define netif_vdbg(priv, type, dev, format, args...) \
4462 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4468 * The list of packet types we will receive (as opposed to discard)
4469 * and the routines to invoke.
4471 * Why 16. Because with 16 the only overlap we get on a hash of the
4472 * low nibble of the protocol value is RARP/SNAP/X.25.
4474 * NOTE: That is no longer true with the addition of VLAN tags. Not
4475 * sure which should go first, but I bet it won't make much
4476 * difference if we are running VLANs. The good news is that
4477 * this protocol won't be in the list unless compiled in, so
4478 * the average user (w/out VLANs) will not be adversely affected.
4494 #define PTYPE_HASH_SIZE (16)
4495 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4497 #endif /* _LINUX_NETDEVICE_H */