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+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
274 bool (*validate)(const char *ll_header, unsigned int len);
275 __be16 (*parse_protocol)(const struct sk_buff *skb);
278 /* These flag bits are private to the generic network queueing
279 * layer; they may not be explicitly referenced by any other
283 enum netdev_state_t {
285 __LINK_STATE_PRESENT,
286 __LINK_STATE_NOCARRIER,
287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
293 * This structure holds boot-time configured netdevice settings. They
294 * are then used in the device probing.
296 struct netdev_boot_setup {
300 #define NETDEV_BOOT_SETUP_MAX 8
302 int __init netdev_boot_setup(char *str);
305 * Structure for NAPI scheduling similar to tasklet but with weighting
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
311 * to the per-CPU poll_list, and whoever clears that bit
312 * can remove from the list right before clearing the bit.
314 struct list_head poll_list;
318 unsigned int gro_count;
319 int (*poll)(struct napi_struct *, int);
320 #ifdef CONFIG_NETPOLL
323 struct net_device *dev;
324 struct sk_buff *gro_list;
326 struct hrtimer timer;
327 struct list_head dev_list;
328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
333 NAPI_STATE_SCHED, /* Poll is scheduled */
334 NAPI_STATE_MISSED, /* reschedule a napi */
335 NAPI_STATE_DISABLE, /* Disable pending */
336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
360 typedef enum gro_result gro_result_t;
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
384 * If the rx_handler consumed the skb in some way, it should return
385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
386 * the skb to be delivered in some other way.
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
392 * If the rx_handler decides the skb should be ignored, it should return
393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
394 * are registered on exact device (ptype->dev == skb->dev).
396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
397 * delivered, it should return RX_HANDLER_PASS.
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
403 enum rx_handler_result {
409 typedef enum rx_handler_result rx_handler_result_t;
410 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
412 void __napi_schedule(struct napi_struct *n);
413 void __napi_schedule_irqoff(struct napi_struct *n);
415 static inline bool napi_disable_pending(struct napi_struct *n)
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
420 bool napi_schedule_prep(struct napi_struct *n);
423 * napi_schedule - schedule NAPI poll
426 * Schedule NAPI poll routine to be called if it is not already
429 static inline void napi_schedule(struct napi_struct *n)
431 if (napi_schedule_prep(n))
436 * napi_schedule_irqoff - schedule NAPI poll
439 * Variant of napi_schedule(), assuming hard irqs are masked.
441 static inline void napi_schedule_irqoff(struct napi_struct *n)
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
447 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
448 static inline bool napi_reschedule(struct napi_struct *napi)
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
457 bool napi_complete_done(struct napi_struct *n, int work_done);
459 * napi_complete - NAPI processing complete
462 * Mark NAPI processing as complete.
463 * Consider using napi_complete_done() instead.
464 * Return false if device should avoid rearming interrupts.
466 static inline bool napi_complete(struct napi_struct *n)
468 return napi_complete_done(n, 0);
472 * napi_hash_del - remove a NAPI from global table
473 * @napi: NAPI context
475 * Warning: caller must observe RCU grace period
476 * before freeing memory containing @napi, if
477 * this function returns true.
478 * Note: core networking stack automatically calls it
479 * from netif_napi_del().
480 * Drivers might want to call this helper to combine all
481 * the needed RCU grace periods into a single one.
483 bool napi_hash_del(struct napi_struct *napi);
486 * napi_disable - prevent NAPI from scheduling
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
492 void napi_disable(struct napi_struct *n);
495 * napi_enable - enable NAPI scheduling
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
501 static inline void napi_enable(struct napi_struct *n)
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
504 smp_mb__before_atomic();
505 clear_bit(NAPI_STATE_SCHED, &n->state);
506 clear_bit(NAPI_STATE_NPSVC, &n->state);
510 * napi_synchronize - wait until NAPI is not running
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
517 static inline void napi_synchronize(const struct napi_struct *n)
519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
526 enum netdev_queue_state_t {
527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
529 __QUEUE_STATE_FROZEN,
532 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
536 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
539 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
552 struct netdev_queue {
556 struct net_device *dev;
557 struct Qdisc __rcu *qdisc;
558 struct Qdisc *qdisc_sleeping;
562 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
565 unsigned long tx_maxrate;
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
570 unsigned long trans_timeout;
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
577 * Time (in jiffies) of last Tx
579 unsigned long trans_start;
586 } ____cacheline_aligned_in_smp;
588 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
590 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 * This structure holds an RPS map which can be of variable length. The
607 * map is an array of CPUs.
614 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618 * tail pointer for that CPU's input queue at the time of last enqueue, and
619 * a hardware filter index.
621 struct rps_dev_flow {
624 unsigned int last_qtail;
626 #define RPS_NO_FILTER 0xffff
629 * The rps_dev_flow_table structure contains a table of flow mappings.
631 struct rps_dev_flow_table {
634 struct rps_dev_flow flows[0];
636 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
637 ((_num) * sizeof(struct rps_dev_flow)))
640 * The rps_sock_flow_table contains mappings of flows to the last CPU
641 * on which they were processed by the application (set in recvmsg).
642 * Each entry is a 32bit value. Upper part is the high-order bits
643 * of flow hash, lower part is CPU number.
644 * rps_cpu_mask is used to partition the space, depending on number of
645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
647 * meaning we use 32-6=26 bits for the hash.
649 struct rps_sock_flow_table {
652 u32 ents[0] ____cacheline_aligned_in_smp;
654 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
656 #define RPS_NO_CPU 0xffff
658 extern u32 rps_cpu_mask;
659 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
661 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
665 unsigned int index = hash & table->mask;
666 u32 val = hash & ~rps_cpu_mask;
668 /* We only give a hint, preemption can change CPU under us */
669 val |= raw_smp_processor_id();
671 if (table->ents[index] != val)
672 table->ents[index] = val;
676 #ifdef CONFIG_RFS_ACCEL
677 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
680 #endif /* CONFIG_RPS */
682 /* This structure contains an instance of an RX queue. */
683 struct netdev_rx_queue {
685 struct rps_map __rcu *rps_map;
686 struct rps_dev_flow_table __rcu *rps_flow_table;
689 struct net_device *dev;
690 } ____cacheline_aligned_in_smp;
693 * RX queue sysfs structures and functions.
695 struct rx_queue_attribute {
696 struct attribute attr;
697 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
698 ssize_t (*store)(struct netdev_rx_queue *queue,
699 const char *buf, size_t len);
704 * This structure holds an XPS map which can be of variable length. The
705 * map is an array of queues.
709 unsigned int alloc_len;
713 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
714 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715 - sizeof(struct xps_map)) / sizeof(u16))
718 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 struct xps_dev_maps {
722 struct xps_map __rcu *cpu_map[0];
724 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
726 #endif /* CONFIG_XPS */
728 #define TC_MAX_QUEUE 16
729 #define TC_BITMASK 15
730 /* HW offloaded queuing disciplines txq count and offset maps */
731 struct netdev_tc_txq {
736 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738 * This structure is to hold information about the device
739 * configured to run FCoE protocol stack.
741 struct netdev_fcoe_hbainfo {
742 char manufacturer[64];
743 char serial_number[64];
744 char hardware_version[64];
745 char driver_version[64];
746 char optionrom_version[64];
747 char firmware_version[64];
749 char model_description[256];
753 #define MAX_PHYS_ITEM_ID_LEN 32
755 /* This structure holds a unique identifier to identify some
756 * physical item (port for example) used by a netdevice.
758 struct netdev_phys_item_id {
759 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
760 unsigned char id_len;
763 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
764 struct netdev_phys_item_id *b)
766 return a->id_len == b->id_len &&
767 memcmp(a->id, b->id, a->id_len) == 0;
770 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
777 TC_SETUP_CLSMATCHALL,
781 /* These structures hold the attributes of xdp state that are being passed
782 * to the netdevice through the xdp op.
784 enum xdp_netdev_command {
785 /* Set or clear a bpf program used in the earliest stages of packet
786 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
787 * is responsible for calling bpf_prog_put on any old progs that are
788 * stored. In case of error, the callee need not release the new prog
789 * reference, but on success it takes ownership and must bpf_prog_put
790 * when it is no longer used.
794 /* Check if a bpf program is set on the device. The callee should
795 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
796 * is equivalent to XDP_ATTACHED_DRV.
801 struct netlink_ext_ack;
804 enum xdp_netdev_command command;
809 struct bpf_prog *prog;
810 struct netlink_ext_ack *extack;
820 #ifdef CONFIG_XFRM_OFFLOAD
822 int (*xdo_dev_state_add) (struct xfrm_state *x);
823 void (*xdo_dev_state_delete) (struct xfrm_state *x);
824 void (*xdo_dev_state_free) (struct xfrm_state *x);
825 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
826 struct xfrm_state *x);
831 * This structure defines the management hooks for network devices.
832 * The following hooks can be defined; unless noted otherwise, they are
833 * optional and can be filled with a null pointer.
835 * int (*ndo_init)(struct net_device *dev);
836 * This function is called once when a network device is registered.
837 * The network device can use this for any late stage initialization
838 * or semantic validation. It can fail with an error code which will
839 * be propagated back to register_netdev.
841 * void (*ndo_uninit)(struct net_device *dev);
842 * This function is called when device is unregistered or when registration
843 * fails. It is not called if init fails.
845 * int (*ndo_open)(struct net_device *dev);
846 * This function is called when a network device transitions to the up
849 * int (*ndo_stop)(struct net_device *dev);
850 * This function is called when a network device transitions to the down
853 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
854 * struct net_device *dev);
855 * Called when a packet needs to be transmitted.
856 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
857 * the queue before that can happen; it's for obsolete devices and weird
858 * corner cases, but the stack really does a non-trivial amount
859 * of useless work if you return NETDEV_TX_BUSY.
860 * Required; cannot be NULL.
862 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
863 * struct net_device *dev
864 * netdev_features_t features);
865 * Called by core transmit path to determine if device is capable of
866 * performing offload operations on a given packet. This is to give
867 * the device an opportunity to implement any restrictions that cannot
868 * be otherwise expressed by feature flags. The check is called with
869 * the set of features that the stack has calculated and it returns
870 * those the driver believes to be appropriate.
872 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
873 * void *accel_priv, select_queue_fallback_t fallback);
874 * Called to decide which queue to use when device supports multiple
877 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
878 * This function is called to allow device receiver to make
879 * changes to configuration when multicast or promiscuous is enabled.
881 * void (*ndo_set_rx_mode)(struct net_device *dev);
882 * This function is called device changes address list filtering.
883 * If driver handles unicast address filtering, it should set
884 * IFF_UNICAST_FLT in its priv_flags.
886 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
887 * This function is called when the Media Access Control address
888 * needs to be changed. If this interface is not defined, the
889 * MAC address can not be changed.
891 * int (*ndo_validate_addr)(struct net_device *dev);
892 * Test if Media Access Control address is valid for the device.
894 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
895 * Called when a user requests an ioctl which can't be handled by
896 * the generic interface code. If not defined ioctls return
897 * not supported error code.
899 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
900 * Used to set network devices bus interface parameters. This interface
901 * is retained for legacy reasons; new devices should use the bus
902 * interface (PCI) for low level management.
904 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
905 * Called when a user wants to change the Maximum Transfer Unit
908 * void (*ndo_tx_timeout)(struct net_device *dev);
909 * Callback used when the transmitter has not made any progress
910 * for dev->watchdog ticks.
912 * void (*ndo_get_stats64)(struct net_device *dev,
913 * struct rtnl_link_stats64 *storage);
914 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
915 * Called when a user wants to get the network device usage
916 * statistics. Drivers must do one of the following:
917 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
918 * rtnl_link_stats64 structure passed by the caller.
919 * 2. Define @ndo_get_stats to update a net_device_stats structure
920 * (which should normally be dev->stats) and return a pointer to
921 * it. The structure may be changed asynchronously only if each
922 * field is written atomically.
923 * 3. Update dev->stats asynchronously and atomically, and define
926 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
927 * Return true if this device supports offload stats of this attr_id.
929 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
931 * Get statistics for offload operations by attr_id. Write it into the
934 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
935 * If device supports VLAN filtering this function is called when a
936 * VLAN id is registered.
938 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
939 * If device supports VLAN filtering this function is called when a
940 * VLAN id is unregistered.
942 * void (*ndo_poll_controller)(struct net_device *dev);
944 * SR-IOV management functions.
945 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
946 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
947 * u8 qos, __be16 proto);
948 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
950 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
951 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
952 * int (*ndo_get_vf_config)(struct net_device *dev,
953 * int vf, struct ifla_vf_info *ivf);
954 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
955 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
956 * struct nlattr *port[]);
958 * Enable or disable the VF ability to query its RSS Redirection Table and
959 * Hash Key. This is needed since on some devices VF share this information
960 * with PF and querying it may introduce a theoretical security risk.
961 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
962 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
963 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
965 * Called to setup any 'tc' scheduler, classifier or action on @dev.
966 * This is always called from the stack with the rtnl lock held and netif
967 * tx queues stopped. This allows the netdevice to perform queue
970 * Fiber Channel over Ethernet (FCoE) offload functions.
971 * int (*ndo_fcoe_enable)(struct net_device *dev);
972 * Called when the FCoE protocol stack wants to start using LLD for FCoE
973 * so the underlying device can perform whatever needed configuration or
974 * initialization to support acceleration of FCoE traffic.
976 * int (*ndo_fcoe_disable)(struct net_device *dev);
977 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
978 * so the underlying device can perform whatever needed clean-ups to
979 * stop supporting acceleration of FCoE traffic.
981 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
982 * struct scatterlist *sgl, unsigned int sgc);
983 * Called when the FCoE Initiator wants to initialize an I/O that
984 * is a possible candidate for Direct Data Placement (DDP). The LLD can
985 * perform necessary setup and returns 1 to indicate the device is set up
986 * successfully to perform DDP on this I/O, otherwise this returns 0.
988 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
989 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
990 * indicated by the FC exchange id 'xid', so the underlying device can
991 * clean up and reuse resources for later DDP requests.
993 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
994 * struct scatterlist *sgl, unsigned int sgc);
995 * Called when the FCoE Target wants to initialize an I/O that
996 * is a possible candidate for Direct Data Placement (DDP). The LLD can
997 * perform necessary setup and returns 1 to indicate the device is set up
998 * successfully to perform DDP on this I/O, otherwise this returns 0.
1000 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1001 * struct netdev_fcoe_hbainfo *hbainfo);
1002 * Called when the FCoE Protocol stack wants information on the underlying
1003 * device. This information is utilized by the FCoE protocol stack to
1004 * register attributes with Fiber Channel management service as per the
1005 * FC-GS Fabric Device Management Information(FDMI) specification.
1007 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1008 * Called when the underlying device wants to override default World Wide
1009 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1010 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1011 * protocol stack to use.
1014 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1015 * u16 rxq_index, u32 flow_id);
1016 * Set hardware filter for RFS. rxq_index is the target queue index;
1017 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1018 * Return the filter ID on success, or a negative error code.
1020 * Slave management functions (for bridge, bonding, etc).
1021 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1022 * Called to make another netdev an underling.
1024 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1025 * Called to release previously enslaved netdev.
1027 * Feature/offload setting functions.
1028 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1029 * netdev_features_t features);
1030 * Adjusts the requested feature flags according to device-specific
1031 * constraints, and returns the resulting flags. Must not modify
1034 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1035 * Called to update device configuration to new features. Passed
1036 * feature set might be less than what was returned by ndo_fix_features()).
1037 * Must return >0 or -errno if it changed dev->features itself.
1039 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1040 * struct net_device *dev,
1041 * const unsigned char *addr, u16 vid, u16 flags)
1042 * Adds an FDB entry to dev for addr.
1043 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1044 * struct net_device *dev,
1045 * const unsigned char *addr, u16 vid)
1046 * Deletes the FDB entry from dev coresponding to addr.
1047 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1048 * struct net_device *dev, struct net_device *filter_dev,
1050 * Used to add FDB entries to dump requests. Implementers should add
1051 * entries to skb and update idx with the number of entries.
1053 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1055 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1056 * struct net_device *dev, u32 filter_mask,
1058 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1061 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1062 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1063 * which do not represent real hardware may define this to allow their
1064 * userspace components to manage their virtual carrier state. Devices
1065 * that determine carrier state from physical hardware properties (eg
1066 * network cables) or protocol-dependent mechanisms (eg
1067 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1069 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1070 * struct netdev_phys_item_id *ppid);
1071 * Called to get ID of physical port of this device. If driver does
1072 * not implement this, it is assumed that the hw is not able to have
1073 * multiple net devices on single physical port.
1075 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1076 * struct udp_tunnel_info *ti);
1077 * Called by UDP tunnel to notify a driver about the UDP port and socket
1078 * address family that a UDP tunnel is listnening to. It is called only
1079 * when a new port starts listening. The operation is protected by the
1082 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1083 * struct udp_tunnel_info *ti);
1084 * Called by UDP tunnel to notify the driver about a UDP port and socket
1085 * address family that the UDP tunnel is not listening to anymore. The
1086 * operation is protected by the RTNL.
1088 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1089 * struct net_device *dev)
1090 * Called by upper layer devices to accelerate switching or other
1091 * station functionality into hardware. 'pdev is the lowerdev
1092 * to use for the offload and 'dev' is the net device that will
1093 * back the offload. Returns a pointer to the private structure
1094 * the upper layer will maintain.
1095 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1096 * Called by upper layer device to delete the station created
1097 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1098 * the station and priv is the structure returned by the add
1100 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1101 * int queue_index, u32 maxrate);
1102 * Called when a user wants to set a max-rate limitation of specific
1104 * int (*ndo_get_iflink)(const struct net_device *dev);
1105 * Called to get the iflink value of this device.
1106 * void (*ndo_change_proto_down)(struct net_device *dev,
1108 * This function is used to pass protocol port error state information
1109 * to the switch driver. The switch driver can react to the proto_down
1110 * by doing a phys down on the associated switch port.
1111 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1112 * This function is used to get egress tunnel information for given skb.
1113 * This is useful for retrieving outer tunnel header parameters while
1115 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1116 * This function is used to specify the headroom that the skb must
1117 * consider when allocation skb during packet reception. Setting
1118 * appropriate rx headroom value allows avoiding skb head copy on
1119 * forward. Setting a negative value resets the rx headroom to the
1121 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1122 * This function is used to set or query state related to XDP on the
1123 * netdevice. See definition of enum xdp_netdev_command for details.
1124 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1125 * This function is used to submit a XDP packet for transmit on a
1127 * void (*ndo_xdp_flush)(struct net_device *dev);
1128 * This function is used to inform the driver to flush a particular
1129 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1131 struct net_device_ops {
1132 int (*ndo_init)(struct net_device *dev);
1133 void (*ndo_uninit)(struct net_device *dev);
1134 int (*ndo_open)(struct net_device *dev);
1135 int (*ndo_stop)(struct net_device *dev);
1136 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1137 struct net_device *dev);
1138 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1139 struct net_device *dev,
1140 netdev_features_t features);
1141 u16 (*ndo_select_queue)(struct net_device *dev,
1142 struct sk_buff *skb,
1144 select_queue_fallback_t fallback);
1145 void (*ndo_change_rx_flags)(struct net_device *dev,
1147 void (*ndo_set_rx_mode)(struct net_device *dev);
1148 int (*ndo_set_mac_address)(struct net_device *dev,
1150 int (*ndo_validate_addr)(struct net_device *dev);
1151 int (*ndo_do_ioctl)(struct net_device *dev,
1152 struct ifreq *ifr, int cmd);
1153 int (*ndo_set_config)(struct net_device *dev,
1155 int (*ndo_change_mtu)(struct net_device *dev,
1157 int (*ndo_neigh_setup)(struct net_device *dev,
1158 struct neigh_parms *);
1159 void (*ndo_tx_timeout) (struct net_device *dev);
1161 void (*ndo_get_stats64)(struct net_device *dev,
1162 struct rtnl_link_stats64 *storage);
1163 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1164 int (*ndo_get_offload_stats)(int attr_id,
1165 const struct net_device *dev,
1167 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1169 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1170 __be16 proto, u16 vid);
1171 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1172 __be16 proto, u16 vid);
1173 #ifdef CONFIG_NET_POLL_CONTROLLER
1174 void (*ndo_poll_controller)(struct net_device *dev);
1175 int (*ndo_netpoll_setup)(struct net_device *dev,
1176 struct netpoll_info *info);
1177 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1179 int (*ndo_set_vf_mac)(struct net_device *dev,
1180 int queue, u8 *mac);
1181 int (*ndo_set_vf_vlan)(struct net_device *dev,
1182 int queue, u16 vlan,
1183 u8 qos, __be16 proto);
1184 int (*ndo_set_vf_rate)(struct net_device *dev,
1185 int vf, int min_tx_rate,
1187 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1188 int vf, bool setting);
1189 int (*ndo_set_vf_trust)(struct net_device *dev,
1190 int vf, bool setting);
1191 int (*ndo_get_vf_config)(struct net_device *dev,
1193 struct ifla_vf_info *ivf);
1194 int (*ndo_set_vf_link_state)(struct net_device *dev,
1195 int vf, int link_state);
1196 int (*ndo_get_vf_stats)(struct net_device *dev,
1198 struct ifla_vf_stats
1200 int (*ndo_set_vf_port)(struct net_device *dev,
1202 struct nlattr *port[]);
1203 int (*ndo_get_vf_port)(struct net_device *dev,
1204 int vf, struct sk_buff *skb);
1205 int (*ndo_set_vf_guid)(struct net_device *dev,
1208 int (*ndo_set_vf_rss_query_en)(
1209 struct net_device *dev,
1210 int vf, bool setting);
1211 int (*ndo_setup_tc)(struct net_device *dev,
1212 enum tc_setup_type type,
1214 #if IS_ENABLED(CONFIG_FCOE)
1215 int (*ndo_fcoe_enable)(struct net_device *dev);
1216 int (*ndo_fcoe_disable)(struct net_device *dev);
1217 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1219 struct scatterlist *sgl,
1221 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1223 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1225 struct scatterlist *sgl,
1227 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1228 struct netdev_fcoe_hbainfo *hbainfo);
1231 #if IS_ENABLED(CONFIG_LIBFCOE)
1232 #define NETDEV_FCOE_WWNN 0
1233 #define NETDEV_FCOE_WWPN 1
1234 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1235 u64 *wwn, int type);
1238 #ifdef CONFIG_RFS_ACCEL
1239 int (*ndo_rx_flow_steer)(struct net_device *dev,
1240 const struct sk_buff *skb,
1244 int (*ndo_add_slave)(struct net_device *dev,
1245 struct net_device *slave_dev);
1246 int (*ndo_del_slave)(struct net_device *dev,
1247 struct net_device *slave_dev);
1248 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1249 netdev_features_t features);
1250 int (*ndo_set_features)(struct net_device *dev,
1251 netdev_features_t features);
1252 int (*ndo_neigh_construct)(struct net_device *dev,
1253 struct neighbour *n);
1254 void (*ndo_neigh_destroy)(struct net_device *dev,
1255 struct neighbour *n);
1257 int (*ndo_fdb_add)(struct ndmsg *ndm,
1258 struct nlattr *tb[],
1259 struct net_device *dev,
1260 const unsigned char *addr,
1263 int (*ndo_fdb_del)(struct ndmsg *ndm,
1264 struct nlattr *tb[],
1265 struct net_device *dev,
1266 const unsigned char *addr,
1268 int (*ndo_fdb_dump)(struct sk_buff *skb,
1269 struct netlink_callback *cb,
1270 struct net_device *dev,
1271 struct net_device *filter_dev,
1274 int (*ndo_bridge_setlink)(struct net_device *dev,
1275 struct nlmsghdr *nlh,
1277 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1279 struct net_device *dev,
1282 int (*ndo_bridge_dellink)(struct net_device *dev,
1283 struct nlmsghdr *nlh,
1285 int (*ndo_change_carrier)(struct net_device *dev,
1287 int (*ndo_get_phys_port_id)(struct net_device *dev,
1288 struct netdev_phys_item_id *ppid);
1289 int (*ndo_get_phys_port_name)(struct net_device *dev,
1290 char *name, size_t len);
1291 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1292 struct udp_tunnel_info *ti);
1293 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1294 struct udp_tunnel_info *ti);
1295 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1296 struct net_device *dev);
1297 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1300 int (*ndo_get_lock_subclass)(struct net_device *dev);
1301 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1304 int (*ndo_get_iflink)(const struct net_device *dev);
1305 int (*ndo_change_proto_down)(struct net_device *dev,
1307 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1308 struct sk_buff *skb);
1309 void (*ndo_set_rx_headroom)(struct net_device *dev,
1310 int needed_headroom);
1311 int (*ndo_xdp)(struct net_device *dev,
1312 struct netdev_xdp *xdp);
1313 int (*ndo_xdp_xmit)(struct net_device *dev,
1314 struct xdp_buff *xdp);
1315 void (*ndo_xdp_flush)(struct net_device *dev);
1319 * enum net_device_priv_flags - &struct net_device priv_flags
1321 * These are the &struct net_device, they are only set internally
1322 * by drivers and used in the kernel. These flags are invisible to
1323 * userspace; this means that the order of these flags can change
1324 * during any kernel release.
1326 * You should have a pretty good reason to be extending these flags.
1328 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1329 * @IFF_EBRIDGE: Ethernet bridging device
1330 * @IFF_BONDING: bonding master or slave
1331 * @IFF_ISATAP: ISATAP interface (RFC4214)
1332 * @IFF_WAN_HDLC: WAN HDLC device
1333 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1335 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1336 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1337 * @IFF_MACVLAN_PORT: device used as macvlan port
1338 * @IFF_BRIDGE_PORT: device used as bridge port
1339 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1340 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1341 * @IFF_UNICAST_FLT: Supports unicast filtering
1342 * @IFF_TEAM_PORT: device used as team port
1343 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1344 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1345 * change when it's running
1346 * @IFF_MACVLAN: Macvlan device
1347 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1348 * underlying stacked devices
1349 * @IFF_IPVLAN_MASTER: IPvlan master device
1350 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1351 * @IFF_L3MDEV_MASTER: device is an L3 master device
1352 * @IFF_NO_QUEUE: device can run without qdisc attached
1353 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1354 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1355 * @IFF_TEAM: device is a team device
1356 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1357 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1358 * entity (i.e. the master device for bridged veth)
1359 * @IFF_MACSEC: device is a MACsec device
1360 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1362 enum netdev_priv_flags {
1363 IFF_802_1Q_VLAN = 1<<0,
1367 IFF_WAN_HDLC = 1<<4,
1368 IFF_XMIT_DST_RELEASE = 1<<5,
1369 IFF_DONT_BRIDGE = 1<<6,
1370 IFF_DISABLE_NETPOLL = 1<<7,
1371 IFF_MACVLAN_PORT = 1<<8,
1372 IFF_BRIDGE_PORT = 1<<9,
1373 IFF_OVS_DATAPATH = 1<<10,
1374 IFF_TX_SKB_SHARING = 1<<11,
1375 IFF_UNICAST_FLT = 1<<12,
1376 IFF_TEAM_PORT = 1<<13,
1377 IFF_SUPP_NOFCS = 1<<14,
1378 IFF_LIVE_ADDR_CHANGE = 1<<15,
1379 IFF_MACVLAN = 1<<16,
1380 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1381 IFF_IPVLAN_MASTER = 1<<18,
1382 IFF_IPVLAN_SLAVE = 1<<19,
1383 IFF_L3MDEV_MASTER = 1<<20,
1384 IFF_NO_QUEUE = 1<<21,
1385 IFF_OPENVSWITCH = 1<<22,
1386 IFF_L3MDEV_SLAVE = 1<<23,
1388 IFF_RXFH_CONFIGURED = 1<<25,
1389 IFF_PHONY_HEADROOM = 1<<26,
1391 IFF_L3MDEV_RX_HANDLER = 1<<28,
1394 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1395 #define IFF_EBRIDGE IFF_EBRIDGE
1396 #define IFF_BONDING IFF_BONDING
1397 #define IFF_ISATAP IFF_ISATAP
1398 #define IFF_WAN_HDLC IFF_WAN_HDLC
1399 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1400 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1401 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1402 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1403 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1404 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1405 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1406 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1407 #define IFF_TEAM_PORT IFF_TEAM_PORT
1408 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1409 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1410 #define IFF_MACVLAN IFF_MACVLAN
1411 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1412 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1413 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1414 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1415 #define IFF_NO_QUEUE IFF_NO_QUEUE
1416 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1417 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1418 #define IFF_TEAM IFF_TEAM
1419 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1420 #define IFF_MACSEC IFF_MACSEC
1421 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1424 * struct net_device - The DEVICE structure.
1426 * Actually, this whole structure is a big mistake. It mixes I/O
1427 * data with strictly "high-level" data, and it has to know about
1428 * almost every data structure used in the INET module.
1430 * @name: This is the first field of the "visible" part of this structure
1431 * (i.e. as seen by users in the "Space.c" file). It is the name
1434 * @name_hlist: Device name hash chain, please keep it close to name[]
1435 * @ifalias: SNMP alias
1436 * @mem_end: Shared memory end
1437 * @mem_start: Shared memory start
1438 * @base_addr: Device I/O address
1439 * @irq: Device IRQ number
1441 * @carrier_changes: Stats to monitor carrier on<->off transitions
1443 * @state: Generic network queuing layer state, see netdev_state_t
1444 * @dev_list: The global list of network devices
1445 * @napi_list: List entry used for polling NAPI devices
1446 * @unreg_list: List entry when we are unregistering the
1447 * device; see the function unregister_netdev
1448 * @close_list: List entry used when we are closing the device
1449 * @ptype_all: Device-specific packet handlers for all protocols
1450 * @ptype_specific: Device-specific, protocol-specific packet handlers
1452 * @adj_list: Directly linked devices, like slaves for bonding
1453 * @features: Currently active device features
1454 * @hw_features: User-changeable features
1456 * @wanted_features: User-requested features
1457 * @vlan_features: Mask of features inheritable by VLAN devices
1459 * @hw_enc_features: Mask of features inherited by encapsulating devices
1460 * This field indicates what encapsulation
1461 * offloads the hardware is capable of doing,
1462 * and drivers will need to set them appropriately.
1464 * @mpls_features: Mask of features inheritable by MPLS
1466 * @ifindex: interface index
1467 * @group: The group the device belongs to
1469 * @stats: Statistics struct, which was left as a legacy, use
1470 * rtnl_link_stats64 instead
1472 * @rx_dropped: Dropped packets by core network,
1473 * do not use this in drivers
1474 * @tx_dropped: Dropped packets by core network,
1475 * do not use this in drivers
1476 * @rx_nohandler: nohandler dropped packets by core network on
1477 * inactive devices, do not use this in drivers
1479 * @wireless_handlers: List of functions to handle Wireless Extensions,
1481 * see <net/iw_handler.h> for details.
1482 * @wireless_data: Instance data managed by the core of wireless extensions
1484 * @netdev_ops: Includes several pointers to callbacks,
1485 * if one wants to override the ndo_*() functions
1486 * @ethtool_ops: Management operations
1487 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1488 * discovery handling. Necessary for e.g. 6LoWPAN.
1489 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1490 * of Layer 2 headers.
1492 * @flags: Interface flags (a la BSD)
1493 * @priv_flags: Like 'flags' but invisible to userspace,
1494 * see if.h for the definitions
1495 * @gflags: Global flags ( kept as legacy )
1496 * @padded: How much padding added by alloc_netdev()
1497 * @operstate: RFC2863 operstate
1498 * @link_mode: Mapping policy to operstate
1499 * @if_port: Selectable AUI, TP, ...
1501 * @mtu: Interface MTU value
1502 * @min_mtu: Interface Minimum MTU value
1503 * @max_mtu: Interface Maximum MTU value
1504 * @type: Interface hardware type
1505 * @hard_header_len: Maximum hardware header length.
1506 * @min_header_len: Minimum hardware header length
1508 * @needed_headroom: Extra headroom the hardware may need, but not in all
1509 * cases can this be guaranteed
1510 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1511 * cases can this be guaranteed. Some cases also use
1512 * LL_MAX_HEADER instead to allocate the skb
1514 * interface address info:
1516 * @perm_addr: Permanent hw address
1517 * @addr_assign_type: Hw address assignment type
1518 * @addr_len: Hardware address length
1519 * @neigh_priv_len: Used in neigh_alloc()
1520 * @dev_id: Used to differentiate devices that share
1521 * the same link layer address
1522 * @dev_port: Used to differentiate devices that share
1524 * @addr_list_lock: XXX: need comments on this one
1525 * @uc_promisc: Counter that indicates promiscuous mode
1526 * has been enabled due to the need to listen to
1527 * additional unicast addresses in a device that
1528 * does not implement ndo_set_rx_mode()
1529 * @uc: unicast mac addresses
1530 * @mc: multicast mac addresses
1531 * @dev_addrs: list of device hw addresses
1532 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1533 * @promiscuity: Number of times the NIC is told to work in
1534 * promiscuous mode; if it becomes 0 the NIC will
1535 * exit promiscuous mode
1536 * @allmulti: Counter, enables or disables allmulticast mode
1538 * @vlan_info: VLAN info
1539 * @dsa_ptr: dsa specific data
1540 * @tipc_ptr: TIPC specific data
1541 * @atalk_ptr: AppleTalk link
1542 * @ip_ptr: IPv4 specific data
1543 * @dn_ptr: DECnet specific data
1544 * @ip6_ptr: IPv6 specific data
1545 * @ax25_ptr: AX.25 specific data
1546 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1548 * @dev_addr: Hw address (before bcast,
1549 * because most packets are unicast)
1551 * @_rx: Array of RX queues
1552 * @num_rx_queues: Number of RX queues
1553 * allocated at register_netdev() time
1554 * @real_num_rx_queues: Number of RX queues currently active in device
1556 * @rx_handler: handler for received packets
1557 * @rx_handler_data: XXX: need comments on this one
1558 * @ingress_queue: XXX: need comments on this one
1559 * @broadcast: hw bcast address
1561 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1562 * indexed by RX queue number. Assigned by driver.
1563 * This must only be set if the ndo_rx_flow_steer
1564 * operation is defined
1565 * @index_hlist: Device index hash chain
1567 * @_tx: Array of TX queues
1568 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1569 * @real_num_tx_queues: Number of TX queues currently active in device
1570 * @qdisc: Root qdisc from userspace point of view
1571 * @tx_queue_len: Max frames per queue allowed
1572 * @tx_global_lock: XXX: need comments on this one
1574 * @xps_maps: XXX: need comments on this one
1576 * @watchdog_timeo: Represents the timeout that is used by
1577 * the watchdog (see dev_watchdog())
1578 * @watchdog_timer: List of timers
1580 * @pcpu_refcnt: Number of references to this device
1581 * @todo_list: Delayed register/unregister
1582 * @link_watch_list: XXX: need comments on this one
1584 * @reg_state: Register/unregister state machine
1585 * @dismantle: Device is going to be freed
1586 * @rtnl_link_state: This enum represents the phases of creating
1589 * @needs_free_netdev: Should unregister perform free_netdev?
1590 * @priv_destructor: Called from unregister
1591 * @npinfo: XXX: need comments on this one
1592 * @nd_net: Network namespace this network device is inside
1594 * @ml_priv: Mid-layer private
1595 * @lstats: Loopback statistics
1596 * @tstats: Tunnel statistics
1597 * @dstats: Dummy statistics
1598 * @vstats: Virtual ethernet statistics
1603 * @dev: Class/net/name entry
1604 * @sysfs_groups: Space for optional device, statistics and wireless
1607 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1608 * @rtnl_link_ops: Rtnl_link_ops
1610 * @gso_max_size: Maximum size of generic segmentation offload
1611 * @gso_max_segs: Maximum number of segments that can be passed to the
1614 * @dcbnl_ops: Data Center Bridging netlink ops
1615 * @num_tc: Number of traffic classes in the net device
1616 * @tc_to_txq: XXX: need comments on this one
1617 * @prio_tc_map: XXX: need comments on this one
1619 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1621 * @priomap: XXX: need comments on this one
1622 * @phydev: Physical device may attach itself
1623 * for hardware timestamping
1625 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1626 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1628 * @proto_down: protocol port state information can be sent to the
1629 * switch driver and used to set the phys state of the
1632 * FIXME: cleanup struct net_device such that network protocol info
1637 char name[IFNAMSIZ];
1638 struct hlist_node name_hlist;
1641 * I/O specific fields
1642 * FIXME: Merge these and struct ifmap into one
1644 unsigned long mem_end;
1645 unsigned long mem_start;
1646 unsigned long base_addr;
1649 atomic_t carrier_changes;
1652 * Some hardware also needs these fields (state,dev_list,
1653 * napi_list,unreg_list,close_list) but they are not
1654 * part of the usual set specified in Space.c.
1657 unsigned long state;
1659 struct list_head dev_list;
1660 struct list_head napi_list;
1661 struct list_head unreg_list;
1662 struct list_head close_list;
1663 struct list_head ptype_all;
1664 struct list_head ptype_specific;
1667 struct list_head upper;
1668 struct list_head lower;
1671 netdev_features_t features;
1672 netdev_features_t hw_features;
1673 netdev_features_t wanted_features;
1674 netdev_features_t vlan_features;
1675 netdev_features_t hw_enc_features;
1676 netdev_features_t mpls_features;
1677 netdev_features_t gso_partial_features;
1682 struct net_device_stats stats;
1684 atomic_long_t rx_dropped;
1685 atomic_long_t tx_dropped;
1686 atomic_long_t rx_nohandler;
1688 #ifdef CONFIG_WIRELESS_EXT
1689 const struct iw_handler_def *wireless_handlers;
1690 struct iw_public_data *wireless_data;
1692 const struct net_device_ops *netdev_ops;
1693 const struct ethtool_ops *ethtool_ops;
1694 #ifdef CONFIG_NET_SWITCHDEV
1695 const struct switchdev_ops *switchdev_ops;
1697 #ifdef CONFIG_NET_L3_MASTER_DEV
1698 const struct l3mdev_ops *l3mdev_ops;
1700 #if IS_ENABLED(CONFIG_IPV6)
1701 const struct ndisc_ops *ndisc_ops;
1705 const struct xfrmdev_ops *xfrmdev_ops;
1708 const struct header_ops *header_ops;
1711 unsigned int priv_flags;
1713 unsigned short gflags;
1714 unsigned short padded;
1716 unsigned char operstate;
1717 unsigned char link_mode;
1719 unsigned char if_port;
1722 /* Note : dev->mtu is often read without holding a lock.
1723 * Writers usually hold RTNL.
1724 * It is recommended to use READ_ONCE() to annotate the reads,
1725 * and to use WRITE_ONCE() to annotate the writes.
1728 unsigned int min_mtu;
1729 unsigned int max_mtu;
1730 unsigned short type;
1731 unsigned short hard_header_len;
1732 unsigned char min_header_len;
1734 unsigned short needed_headroom;
1735 unsigned short needed_tailroom;
1737 /* Interface address info. */
1738 unsigned char perm_addr[MAX_ADDR_LEN];
1739 unsigned char addr_assign_type;
1740 unsigned char addr_len;
1741 unsigned short neigh_priv_len;
1742 unsigned short dev_id;
1743 unsigned short dev_port;
1744 spinlock_t addr_list_lock;
1745 unsigned char name_assign_type;
1747 struct netdev_hw_addr_list uc;
1748 struct netdev_hw_addr_list mc;
1749 struct netdev_hw_addr_list dev_addrs;
1752 struct kset *queues_kset;
1754 unsigned int promiscuity;
1755 unsigned int allmulti;
1758 /* Protocol-specific pointers */
1760 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1761 struct vlan_info __rcu *vlan_info;
1763 #if IS_ENABLED(CONFIG_NET_DSA)
1764 struct dsa_switch_tree *dsa_ptr;
1766 #if IS_ENABLED(CONFIG_TIPC)
1767 struct tipc_bearer __rcu *tipc_ptr;
1770 struct in_device __rcu *ip_ptr;
1771 struct dn_dev __rcu *dn_ptr;
1772 struct inet6_dev __rcu *ip6_ptr;
1774 struct wireless_dev *ieee80211_ptr;
1775 struct wpan_dev *ieee802154_ptr;
1776 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1777 struct mpls_dev __rcu *mpls_ptr;
1781 * Cache lines mostly used on receive path (including eth_type_trans())
1783 /* Interface address info used in eth_type_trans() */
1784 unsigned char *dev_addr;
1787 struct netdev_rx_queue *_rx;
1789 unsigned int num_rx_queues;
1790 unsigned int real_num_rx_queues;
1793 struct bpf_prog __rcu *xdp_prog;
1794 unsigned long gro_flush_timeout;
1795 rx_handler_func_t __rcu *rx_handler;
1796 void __rcu *rx_handler_data;
1798 #ifdef CONFIG_NET_CLS_ACT
1799 struct tcf_proto __rcu *ingress_cl_list;
1801 struct netdev_queue __rcu *ingress_queue;
1802 #ifdef CONFIG_NETFILTER_INGRESS
1803 struct nf_hook_entries __rcu *nf_hooks_ingress;
1806 unsigned char broadcast[MAX_ADDR_LEN];
1807 #ifdef CONFIG_RFS_ACCEL
1808 struct cpu_rmap *rx_cpu_rmap;
1810 struct hlist_node index_hlist;
1813 * Cache lines mostly used on transmit path
1815 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1816 unsigned int num_tx_queues;
1817 unsigned int real_num_tx_queues;
1818 struct Qdisc *qdisc;
1819 #ifdef CONFIG_NET_SCHED
1820 DECLARE_HASHTABLE (qdisc_hash, 4);
1822 unsigned int tx_queue_len;
1823 spinlock_t tx_global_lock;
1827 struct xps_dev_maps __rcu *xps_maps;
1829 #ifdef CONFIG_NET_CLS_ACT
1830 struct tcf_proto __rcu *egress_cl_list;
1833 /* These may be needed for future network-power-down code. */
1834 struct timer_list watchdog_timer;
1836 int __percpu *pcpu_refcnt;
1837 struct list_head todo_list;
1839 struct list_head link_watch_list;
1841 enum { NETREG_UNINITIALIZED=0,
1842 NETREG_REGISTERED, /* completed register_netdevice */
1843 NETREG_UNREGISTERING, /* called unregister_netdevice */
1844 NETREG_UNREGISTERED, /* completed unregister todo */
1845 NETREG_RELEASED, /* called free_netdev */
1846 NETREG_DUMMY, /* dummy device for NAPI poll */
1852 RTNL_LINK_INITIALIZED,
1853 RTNL_LINK_INITIALIZING,
1854 } rtnl_link_state:16;
1856 bool needs_free_netdev;
1857 void (*priv_destructor)(struct net_device *dev);
1859 #ifdef CONFIG_NETPOLL
1860 struct netpoll_info __rcu *npinfo;
1863 possible_net_t nd_net;
1865 /* mid-layer private */
1868 struct pcpu_lstats __percpu *lstats;
1869 struct pcpu_sw_netstats __percpu *tstats;
1870 struct pcpu_dstats __percpu *dstats;
1871 struct pcpu_vstats __percpu *vstats;
1874 #if IS_ENABLED(CONFIG_GARP)
1875 struct garp_port __rcu *garp_port;
1877 #if IS_ENABLED(CONFIG_MRP)
1878 struct mrp_port __rcu *mrp_port;
1882 const struct attribute_group *sysfs_groups[4];
1883 const struct attribute_group *sysfs_rx_queue_group;
1885 const struct rtnl_link_ops *rtnl_link_ops;
1887 /* for setting kernel sock attribute on TCP connection setup */
1888 #define GSO_MAX_SIZE 65536
1889 unsigned int gso_max_size;
1890 #define GSO_MAX_SEGS 65535
1894 const struct dcbnl_rtnl_ops *dcbnl_ops;
1897 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1898 u8 prio_tc_map[TC_BITMASK + 1];
1900 #if IS_ENABLED(CONFIG_FCOE)
1901 unsigned int fcoe_ddp_xid;
1903 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1904 struct netprio_map __rcu *priomap;
1906 struct phy_device *phydev;
1907 struct lock_class_key *qdisc_tx_busylock;
1908 struct lock_class_key *qdisc_running_key;
1911 #define to_net_dev(d) container_of(d, struct net_device, dev)
1913 static inline bool netif_elide_gro(const struct net_device *dev)
1915 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1920 #define NETDEV_ALIGN 32
1923 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1925 return dev->prio_tc_map[prio & TC_BITMASK];
1929 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1931 if (tc >= dev->num_tc)
1934 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1938 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1939 void netdev_reset_tc(struct net_device *dev);
1940 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1941 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1944 int netdev_get_num_tc(struct net_device *dev)
1950 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1953 return &dev->_tx[index];
1956 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1957 const struct sk_buff *skb)
1959 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1962 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1963 void (*f)(struct net_device *,
1964 struct netdev_queue *,
1970 for (i = 0; i < dev->num_tx_queues; i++)
1971 f(dev, &dev->_tx[i], arg);
1974 #define netdev_lockdep_set_classes(dev) \
1976 static struct lock_class_key qdisc_tx_busylock_key; \
1977 static struct lock_class_key qdisc_running_key; \
1978 static struct lock_class_key qdisc_xmit_lock_key; \
1979 static struct lock_class_key dev_addr_list_lock_key; \
1982 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1983 (dev)->qdisc_running_key = &qdisc_running_key; \
1984 lockdep_set_class(&(dev)->addr_list_lock, \
1985 &dev_addr_list_lock_key); \
1986 for (i = 0; i < (dev)->num_tx_queues; i++) \
1987 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1988 &qdisc_xmit_lock_key); \
1991 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1992 struct sk_buff *skb,
1995 /* returns the headroom that the master device needs to take in account
1996 * when forwarding to this dev
1998 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2000 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2003 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2005 if (dev->netdev_ops->ndo_set_rx_headroom)
2006 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2009 /* set the device rx headroom to the dev's default */
2010 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2012 netdev_set_rx_headroom(dev, -1);
2016 * Net namespace inlines
2019 struct net *dev_net(const struct net_device *dev)
2021 return read_pnet(&dev->nd_net);
2025 void dev_net_set(struct net_device *dev, struct net *net)
2027 write_pnet(&dev->nd_net, net);
2031 * netdev_priv - access network device private data
2032 * @dev: network device
2034 * Get network device private data
2036 static inline void *netdev_priv(const struct net_device *dev)
2038 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2041 /* Set the sysfs physical device reference for the network logical device
2042 * if set prior to registration will cause a symlink during initialization.
2044 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2046 /* Set the sysfs device type for the network logical device to allow
2047 * fine-grained identification of different network device types. For
2048 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2050 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2052 /* Default NAPI poll() weight
2053 * Device drivers are strongly advised to not use bigger value
2055 #define NAPI_POLL_WEIGHT 64
2058 * netif_napi_add - initialize a NAPI context
2059 * @dev: network device
2060 * @napi: NAPI context
2061 * @poll: polling function
2062 * @weight: default weight
2064 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2065 * *any* of the other NAPI-related functions.
2067 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2068 int (*poll)(struct napi_struct *, int), int weight);
2071 * netif_tx_napi_add - initialize a NAPI context
2072 * @dev: network device
2073 * @napi: NAPI context
2074 * @poll: polling function
2075 * @weight: default weight
2077 * This variant of netif_napi_add() should be used from drivers using NAPI
2078 * to exclusively poll a TX queue.
2079 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2081 static inline void netif_tx_napi_add(struct net_device *dev,
2082 struct napi_struct *napi,
2083 int (*poll)(struct napi_struct *, int),
2086 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2087 netif_napi_add(dev, napi, poll, weight);
2091 * netif_napi_del - remove a NAPI context
2092 * @napi: NAPI context
2094 * netif_napi_del() removes a NAPI context from the network device NAPI list
2096 void netif_napi_del(struct napi_struct *napi);
2098 struct napi_gro_cb {
2099 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2102 /* Length of frag0. */
2103 unsigned int frag0_len;
2105 /* This indicates where we are processing relative to skb->data. */
2108 /* This is non-zero if the packet cannot be merged with the new skb. */
2111 /* Save the IP ID here and check when we get to the transport layer */
2114 /* Number of segments aggregated. */
2117 /* Start offset for remote checksum offload */
2118 u16 gro_remcsum_start;
2120 /* jiffies when first packet was created/queued */
2123 /* Used in ipv6_gro_receive() and foo-over-udp */
2126 /* This is non-zero if the packet may be of the same flow. */
2129 /* Used in tunnel GRO receive */
2132 /* GRO checksum is valid */
2135 /* Number of checksums via CHECKSUM_UNNECESSARY */
2140 #define NAPI_GRO_FREE 1
2141 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2143 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2146 /* Used in GRE, set in fou/gue_gro_receive */
2149 /* Used to determine if flush_id can be ignored */
2152 /* Number of gro_receive callbacks this packet already went through */
2153 u8 recursion_counter:4;
2157 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2160 /* used in skb_gro_receive() slow path */
2161 struct sk_buff *last;
2164 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2166 #define GRO_RECURSION_LIMIT 15
2167 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2169 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2172 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2173 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2174 struct sk_buff **head,
2175 struct sk_buff *skb)
2177 if (unlikely(gro_recursion_inc_test(skb))) {
2178 NAPI_GRO_CB(skb)->flush |= 1;
2182 return cb(head, skb);
2185 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2187 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2189 struct sk_buff **head,
2190 struct sk_buff *skb)
2192 if (unlikely(gro_recursion_inc_test(skb))) {
2193 NAPI_GRO_CB(skb)->flush |= 1;
2197 return cb(sk, head, skb);
2200 struct packet_type {
2201 __be16 type; /* This is really htons(ether_type). */
2202 struct net_device *dev; /* NULL is wildcarded here */
2203 int (*func) (struct sk_buff *,
2204 struct net_device *,
2205 struct packet_type *,
2206 struct net_device *);
2207 bool (*id_match)(struct packet_type *ptype,
2209 void *af_packet_priv;
2210 struct list_head list;
2213 struct offload_callbacks {
2214 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2215 netdev_features_t features);
2216 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2217 struct sk_buff *skb);
2218 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2221 struct packet_offload {
2222 __be16 type; /* This is really htons(ether_type). */
2224 struct offload_callbacks callbacks;
2225 struct list_head list;
2228 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2229 struct pcpu_sw_netstats {
2234 struct u64_stats_sync syncp;
2237 #define __netdev_alloc_pcpu_stats(type, gfp) \
2239 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2242 for_each_possible_cpu(__cpu) { \
2243 typeof(type) *stat; \
2244 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2245 u64_stats_init(&stat->syncp); \
2251 #define netdev_alloc_pcpu_stats(type) \
2252 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2254 enum netdev_lag_tx_type {
2255 NETDEV_LAG_TX_TYPE_UNKNOWN,
2256 NETDEV_LAG_TX_TYPE_RANDOM,
2257 NETDEV_LAG_TX_TYPE_BROADCAST,
2258 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2259 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2260 NETDEV_LAG_TX_TYPE_HASH,
2263 struct netdev_lag_upper_info {
2264 enum netdev_lag_tx_type tx_type;
2267 struct netdev_lag_lower_state_info {
2272 #include <linux/notifier.h>
2274 /* netdevice notifier chain. Please remember to update the rtnetlink
2275 * notification exclusion list in rtnetlink_event() when adding new
2278 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2279 #define NETDEV_DOWN 0x0002
2280 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2281 detected a hardware crash and restarted
2282 - we can use this eg to kick tcp sessions
2284 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2285 #define NETDEV_REGISTER 0x0005
2286 #define NETDEV_UNREGISTER 0x0006
2287 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2288 #define NETDEV_CHANGEADDR 0x0008
2289 #define NETDEV_GOING_DOWN 0x0009
2290 #define NETDEV_CHANGENAME 0x000A
2291 #define NETDEV_FEAT_CHANGE 0x000B
2292 #define NETDEV_BONDING_FAILOVER 0x000C
2293 #define NETDEV_PRE_UP 0x000D
2294 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2295 #define NETDEV_POST_TYPE_CHANGE 0x000F
2296 #define NETDEV_POST_INIT 0x0010
2297 #define NETDEV_UNREGISTER_FINAL 0x0011
2298 #define NETDEV_RELEASE 0x0012
2299 #define NETDEV_NOTIFY_PEERS 0x0013
2300 #define NETDEV_JOIN 0x0014
2301 #define NETDEV_CHANGEUPPER 0x0015
2302 #define NETDEV_RESEND_IGMP 0x0016
2303 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2304 #define NETDEV_CHANGEINFODATA 0x0018
2305 #define NETDEV_BONDING_INFO 0x0019
2306 #define NETDEV_PRECHANGEUPPER 0x001A
2307 #define NETDEV_CHANGELOWERSTATE 0x001B
2308 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2309 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2310 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2312 int register_netdevice_notifier(struct notifier_block *nb);
2313 int unregister_netdevice_notifier(struct notifier_block *nb);
2315 struct netdev_notifier_info {
2316 struct net_device *dev;
2319 struct netdev_notifier_info_ext {
2320 struct netdev_notifier_info info; /* must be first */
2326 struct netdev_notifier_change_info {
2327 struct netdev_notifier_info info; /* must be first */
2328 unsigned int flags_changed;
2331 struct netdev_notifier_changeupper_info {
2332 struct netdev_notifier_info info; /* must be first */
2333 struct net_device *upper_dev; /* new upper dev */
2334 bool master; /* is upper dev master */
2335 bool linking; /* is the notification for link or unlink */
2336 void *upper_info; /* upper dev info */
2339 struct netdev_notifier_changelowerstate_info {
2340 struct netdev_notifier_info info; /* must be first */
2341 void *lower_state_info; /* is lower dev state */
2344 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2345 struct net_device *dev)
2350 static inline struct net_device *
2351 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2356 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2359 extern rwlock_t dev_base_lock; /* Device list lock */
2361 #define for_each_netdev(net, d) \
2362 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2363 #define for_each_netdev_reverse(net, d) \
2364 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2365 #define for_each_netdev_rcu(net, d) \
2366 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2367 #define for_each_netdev_safe(net, d, n) \
2368 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2369 #define for_each_netdev_continue(net, d) \
2370 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2371 #define for_each_netdev_continue_rcu(net, d) \
2372 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2373 #define for_each_netdev_in_bond_rcu(bond, slave) \
2374 for_each_netdev_rcu(&init_net, slave) \
2375 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2376 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2378 static inline struct net_device *next_net_device(struct net_device *dev)
2380 struct list_head *lh;
2384 lh = dev->dev_list.next;
2385 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2388 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2390 struct list_head *lh;
2394 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2395 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2398 static inline struct net_device *first_net_device(struct net *net)
2400 return list_empty(&net->dev_base_head) ? NULL :
2401 net_device_entry(net->dev_base_head.next);
2404 static inline struct net_device *first_net_device_rcu(struct net *net)
2406 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2408 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2411 int netdev_boot_setup_check(struct net_device *dev);
2412 unsigned long netdev_boot_base(const char *prefix, int unit);
2413 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2414 const char *hwaddr);
2415 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2416 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2417 void dev_add_pack(struct packet_type *pt);
2418 void dev_remove_pack(struct packet_type *pt);
2419 void __dev_remove_pack(struct packet_type *pt);
2420 void dev_add_offload(struct packet_offload *po);
2421 void dev_remove_offload(struct packet_offload *po);
2423 int dev_get_iflink(const struct net_device *dev);
2424 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2425 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2426 unsigned short mask);
2427 struct net_device *dev_get_by_name(struct net *net, const char *name);
2428 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2429 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2430 int dev_alloc_name(struct net_device *dev, const char *name);
2431 int dev_open(struct net_device *dev);
2432 void dev_close(struct net_device *dev);
2433 void dev_close_many(struct list_head *head, bool unlink);
2434 void dev_disable_lro(struct net_device *dev);
2435 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2436 int dev_queue_xmit(struct sk_buff *skb);
2437 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2438 int register_netdevice(struct net_device *dev);
2439 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2440 void unregister_netdevice_many(struct list_head *head);
2441 static inline void unregister_netdevice(struct net_device *dev)
2443 unregister_netdevice_queue(dev, NULL);
2446 int netdev_refcnt_read(const struct net_device *dev);
2447 void free_netdev(struct net_device *dev);
2448 void netdev_freemem(struct net_device *dev);
2449 void synchronize_net(void);
2450 int init_dummy_netdev(struct net_device *dev);
2452 DECLARE_PER_CPU(int, xmit_recursion);
2453 #define XMIT_RECURSION_LIMIT 8
2455 static inline int dev_recursion_level(void)
2457 return this_cpu_read(xmit_recursion);
2460 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2461 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2462 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2463 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2464 int netdev_get_name(struct net *net, char *name, int ifindex);
2465 int dev_restart(struct net_device *dev);
2466 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2468 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2470 return NAPI_GRO_CB(skb)->data_offset;
2473 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2475 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2478 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2480 NAPI_GRO_CB(skb)->data_offset += len;
2483 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2484 unsigned int offset)
2486 return NAPI_GRO_CB(skb)->frag0 + offset;
2489 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2491 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2494 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2496 NAPI_GRO_CB(skb)->frag0 = NULL;
2497 NAPI_GRO_CB(skb)->frag0_len = 0;
2500 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2501 unsigned int offset)
2503 if (!pskb_may_pull(skb, hlen))
2506 skb_gro_frag0_invalidate(skb);
2507 return skb->data + offset;
2510 static inline void *skb_gro_network_header(struct sk_buff *skb)
2512 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2513 skb_network_offset(skb);
2516 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2517 const void *start, unsigned int len)
2519 if (NAPI_GRO_CB(skb)->csum_valid)
2520 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2521 csum_partial(start, len, 0));
2524 /* GRO checksum functions. These are logical equivalents of the normal
2525 * checksum functions (in skbuff.h) except that they operate on the GRO
2526 * offsets and fields in sk_buff.
2529 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2531 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2533 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2536 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2540 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2541 skb_checksum_start_offset(skb) <
2542 skb_gro_offset(skb)) &&
2543 !skb_at_gro_remcsum_start(skb) &&
2544 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2545 (!zero_okay || check));
2548 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2551 if (NAPI_GRO_CB(skb)->csum_valid &&
2552 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2555 NAPI_GRO_CB(skb)->csum = psum;
2557 return __skb_gro_checksum_complete(skb);
2560 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2562 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2563 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2564 NAPI_GRO_CB(skb)->csum_cnt--;
2566 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2567 * verified a new top level checksum or an encapsulated one
2568 * during GRO. This saves work if we fallback to normal path.
2570 __skb_incr_checksum_unnecessary(skb);
2574 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2577 __sum16 __ret = 0; \
2578 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2579 __ret = __skb_gro_checksum_validate_complete(skb, \
2580 compute_pseudo(skb, proto)); \
2582 skb_gro_incr_csum_unnecessary(skb); \
2586 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2587 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2589 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2591 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2593 #define skb_gro_checksum_simple_validate(skb) \
2594 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2596 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2598 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2599 !NAPI_GRO_CB(skb)->csum_valid);
2602 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2603 __sum16 check, __wsum pseudo)
2605 NAPI_GRO_CB(skb)->csum = ~pseudo;
2606 NAPI_GRO_CB(skb)->csum_valid = 1;
2609 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2611 if (__skb_gro_checksum_convert_check(skb)) \
2612 __skb_gro_checksum_convert(skb, check, \
2613 compute_pseudo(skb, proto)); \
2616 struct gro_remcsum {
2621 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2627 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2628 unsigned int off, size_t hdrlen,
2629 int start, int offset,
2630 struct gro_remcsum *grc,
2634 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2636 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2639 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2643 ptr = skb_gro_header_fast(skb, off);
2644 if (skb_gro_header_hard(skb, off + plen)) {
2645 ptr = skb_gro_header_slow(skb, off + plen, off);
2650 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2653 /* Adjust skb->csum since we changed the packet */
2654 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2656 grc->offset = off + hdrlen + offset;
2662 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2663 struct gro_remcsum *grc)
2666 size_t plen = grc->offset + sizeof(u16);
2671 ptr = skb_gro_header_fast(skb, grc->offset);
2672 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2673 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2678 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2681 #ifdef CONFIG_XFRM_OFFLOAD
2682 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2684 if (PTR_ERR(pp) != -EINPROGRESS)
2685 NAPI_GRO_CB(skb)->flush |= flush;
2687 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2688 struct sk_buff **pp,
2690 struct gro_remcsum *grc)
2692 if (PTR_ERR(pp) != -EINPROGRESS) {
2693 NAPI_GRO_CB(skb)->flush |= flush;
2694 skb_gro_remcsum_cleanup(skb, grc);
2695 skb->remcsum_offload = 0;
2699 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2701 NAPI_GRO_CB(skb)->flush |= flush;
2703 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2704 struct sk_buff **pp,
2706 struct gro_remcsum *grc)
2708 NAPI_GRO_CB(skb)->flush |= flush;
2709 skb_gro_remcsum_cleanup(skb, grc);
2710 skb->remcsum_offload = 0;
2714 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2715 unsigned short type,
2716 const void *daddr, const void *saddr,
2719 if (!dev->header_ops || !dev->header_ops->create)
2722 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2725 static inline int dev_parse_header(const struct sk_buff *skb,
2726 unsigned char *haddr)
2728 const struct net_device *dev = skb->dev;
2730 if (!dev->header_ops || !dev->header_ops->parse)
2732 return dev->header_ops->parse(skb, haddr);
2735 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2737 const struct net_device *dev = skb->dev;
2739 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2741 return dev->header_ops->parse_protocol(skb);
2744 /* ll_header must have at least hard_header_len allocated */
2745 static inline bool dev_validate_header(const struct net_device *dev,
2746 char *ll_header, int len)
2748 if (likely(len >= dev->hard_header_len))
2750 if (len < dev->min_header_len)
2753 if (capable(CAP_SYS_RAWIO)) {
2754 memset(ll_header + len, 0, dev->hard_header_len - len);
2758 if (dev->header_ops && dev->header_ops->validate)
2759 return dev->header_ops->validate(ll_header, len);
2764 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2765 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2766 static inline int unregister_gifconf(unsigned int family)
2768 return register_gifconf(family, NULL);
2771 #ifdef CONFIG_NET_FLOW_LIMIT
2772 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2773 struct sd_flow_limit {
2775 unsigned int num_buckets;
2776 unsigned int history_head;
2777 u16 history[FLOW_LIMIT_HISTORY];
2781 extern int netdev_flow_limit_table_len;
2782 #endif /* CONFIG_NET_FLOW_LIMIT */
2785 * Incoming packets are placed on per-CPU queues
2787 struct softnet_data {
2788 struct list_head poll_list;
2789 struct sk_buff_head process_queue;
2792 unsigned int processed;
2793 unsigned int time_squeeze;
2794 unsigned int received_rps;
2796 struct softnet_data *rps_ipi_list;
2798 #ifdef CONFIG_NET_FLOW_LIMIT
2799 struct sd_flow_limit __rcu *flow_limit;
2801 struct Qdisc *output_queue;
2802 struct Qdisc **output_queue_tailp;
2803 struct sk_buff *completion_queue;
2806 /* input_queue_head should be written by cpu owning this struct,
2807 * and only read by other cpus. Worth using a cache line.
2809 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2811 /* Elements below can be accessed between CPUs for RPS/RFS */
2812 call_single_data_t csd ____cacheline_aligned_in_smp;
2813 struct softnet_data *rps_ipi_next;
2815 unsigned int input_queue_tail;
2817 unsigned int dropped;
2818 struct sk_buff_head input_pkt_queue;
2819 struct napi_struct backlog;
2823 static inline void input_queue_head_incr(struct softnet_data *sd)
2826 sd->input_queue_head++;
2830 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2831 unsigned int *qtail)
2834 *qtail = ++sd->input_queue_tail;
2838 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2840 void __netif_schedule(struct Qdisc *q);
2841 void netif_schedule_queue(struct netdev_queue *txq);
2843 static inline void netif_tx_schedule_all(struct net_device *dev)
2847 for (i = 0; i < dev->num_tx_queues; i++)
2848 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2851 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2853 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2857 * netif_start_queue - allow transmit
2858 * @dev: network device
2860 * Allow upper layers to call the device hard_start_xmit routine.
2862 static inline void netif_start_queue(struct net_device *dev)
2864 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2867 static inline void netif_tx_start_all_queues(struct net_device *dev)
2871 for (i = 0; i < dev->num_tx_queues; i++) {
2872 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2873 netif_tx_start_queue(txq);
2877 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2880 * netif_wake_queue - restart transmit
2881 * @dev: network device
2883 * Allow upper layers to call the device hard_start_xmit routine.
2884 * Used for flow control when transmit resources are available.
2886 static inline void netif_wake_queue(struct net_device *dev)
2888 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2891 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2895 for (i = 0; i < dev->num_tx_queues; i++) {
2896 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2897 netif_tx_wake_queue(txq);
2901 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2903 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2907 * netif_stop_queue - stop transmitted packets
2908 * @dev: network device
2910 * Stop upper layers calling the device hard_start_xmit routine.
2911 * Used for flow control when transmit resources are unavailable.
2913 static inline void netif_stop_queue(struct net_device *dev)
2915 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2918 void netif_tx_stop_all_queues(struct net_device *dev);
2920 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2922 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2926 * netif_queue_stopped - test if transmit queue is flowblocked
2927 * @dev: network device
2929 * Test if transmit queue on device is currently unable to send.
2931 static inline bool netif_queue_stopped(const struct net_device *dev)
2933 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2936 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2938 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2942 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2944 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2948 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2950 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2954 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2955 * @dev_queue: pointer to transmit queue
2957 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2958 * to give appropriate hint to the CPU.
2960 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2963 prefetchw(&dev_queue->dql.num_queued);
2968 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2969 * @dev_queue: pointer to transmit queue
2971 * BQL enabled drivers might use this helper in their TX completion path,
2972 * to give appropriate hint to the CPU.
2974 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2977 prefetchw(&dev_queue->dql.limit);
2981 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2985 dql_queued(&dev_queue->dql, bytes);
2987 if (likely(dql_avail(&dev_queue->dql) >= 0))
2990 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2993 * The XOFF flag must be set before checking the dql_avail below,
2994 * because in netdev_tx_completed_queue we update the dql_completed
2995 * before checking the XOFF flag.
2999 /* check again in case another CPU has just made room avail */
3000 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3001 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3006 * netdev_sent_queue - report the number of bytes queued to hardware
3007 * @dev: network device
3008 * @bytes: number of bytes queued to the hardware device queue
3010 * Report the number of bytes queued for sending/completion to the network
3011 * device hardware queue. @bytes should be a good approximation and should
3012 * exactly match netdev_completed_queue() @bytes
3014 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3016 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3019 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3020 unsigned int pkts, unsigned int bytes)
3023 if (unlikely(!bytes))
3026 dql_completed(&dev_queue->dql, bytes);
3029 * Without the memory barrier there is a small possiblity that
3030 * netdev_tx_sent_queue will miss the update and cause the queue to
3031 * be stopped forever
3035 if (dql_avail(&dev_queue->dql) < 0)
3038 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3039 netif_schedule_queue(dev_queue);
3044 * netdev_completed_queue - report bytes and packets completed by device
3045 * @dev: network device
3046 * @pkts: actual number of packets sent over the medium
3047 * @bytes: actual number of bytes sent over the medium
3049 * Report the number of bytes and packets transmitted by the network device
3050 * hardware queue over the physical medium, @bytes must exactly match the
3051 * @bytes amount passed to netdev_sent_queue()
3053 static inline void netdev_completed_queue(struct net_device *dev,
3054 unsigned int pkts, unsigned int bytes)
3056 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3059 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3062 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3068 * netdev_reset_queue - reset the packets and bytes count of a network device
3069 * @dev_queue: network device
3071 * Reset the bytes and packet count of a network device and clear the
3072 * software flow control OFF bit for this network device
3074 static inline void netdev_reset_queue(struct net_device *dev_queue)
3076 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3080 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3081 * @dev: network device
3082 * @queue_index: given tx queue index
3084 * Returns 0 if given tx queue index >= number of device tx queues,
3085 * otherwise returns the originally passed tx queue index.
3087 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3089 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3090 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3091 dev->name, queue_index,
3092 dev->real_num_tx_queues);
3100 * netif_running - test if up
3101 * @dev: network device
3103 * Test if the device has been brought up.
3105 static inline bool netif_running(const struct net_device *dev)
3107 return test_bit(__LINK_STATE_START, &dev->state);
3111 * Routines to manage the subqueues on a device. We only need start,
3112 * stop, and a check if it's stopped. All other device management is
3113 * done at the overall netdevice level.
3114 * Also test the device if we're multiqueue.
3118 * netif_start_subqueue - allow sending packets on subqueue
3119 * @dev: network device
3120 * @queue_index: sub queue index
3122 * Start individual transmit queue of a device with multiple transmit queues.
3124 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3126 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3128 netif_tx_start_queue(txq);
3132 * netif_stop_subqueue - stop sending packets on subqueue
3133 * @dev: network device
3134 * @queue_index: sub queue index
3136 * Stop individual transmit queue of a device with multiple transmit queues.
3138 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3140 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3141 netif_tx_stop_queue(txq);
3145 * netif_subqueue_stopped - test status of subqueue
3146 * @dev: network device
3147 * @queue_index: sub queue index
3149 * Check individual transmit queue of a device with multiple transmit queues.
3151 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3154 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3156 return netif_tx_queue_stopped(txq);
3159 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3160 struct sk_buff *skb)
3162 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3166 * netif_wake_subqueue - allow sending packets on subqueue
3167 * @dev: network device
3168 * @queue_index: sub queue index
3170 * Resume individual transmit queue of a device with multiple transmit queues.
3172 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3174 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3176 netif_tx_wake_queue(txq);
3180 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3183 static inline int netif_set_xps_queue(struct net_device *dev,
3184 const struct cpumask *mask,
3191 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3192 unsigned int num_tx_queues);
3195 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3196 * as a distribution range limit for the returned value.
3198 static inline u16 skb_tx_hash(const struct net_device *dev,
3199 struct sk_buff *skb)
3201 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3205 * netif_is_multiqueue - test if device has multiple transmit queues
3206 * @dev: network device
3208 * Check if device has multiple transmit queues
3210 static inline bool netif_is_multiqueue(const struct net_device *dev)
3212 return dev->num_tx_queues > 1;
3215 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3218 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3220 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3228 static inline unsigned int get_netdev_rx_queue_index(
3229 struct netdev_rx_queue *queue)
3231 struct net_device *dev = queue->dev;
3232 int index = queue - dev->_rx;
3234 BUG_ON(index >= dev->num_rx_queues);
3239 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3240 int netif_get_num_default_rss_queues(void);
3242 enum skb_free_reason {
3243 SKB_REASON_CONSUMED,
3247 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3248 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3251 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3252 * interrupt context or with hardware interrupts being disabled.
3253 * (in_irq() || irqs_disabled())
3255 * We provide four helpers that can be used in following contexts :
3257 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3258 * replacing kfree_skb(skb)
3260 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3261 * Typically used in place of consume_skb(skb) in TX completion path
3263 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3264 * replacing kfree_skb(skb)
3266 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3267 * and consumed a packet. Used in place of consume_skb(skb)
3269 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3271 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3274 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3276 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3279 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3281 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3284 static inline void dev_consume_skb_any(struct sk_buff *skb)
3286 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3289 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3290 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3291 int netif_rx(struct sk_buff *skb);
3292 int netif_rx_ni(struct sk_buff *skb);
3293 int netif_receive_skb(struct sk_buff *skb);
3294 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3295 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3296 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3297 gro_result_t napi_gro_frags(struct napi_struct *napi);
3298 struct packet_offload *gro_find_receive_by_type(__be16 type);
3299 struct packet_offload *gro_find_complete_by_type(__be16 type);
3301 static inline void napi_free_frags(struct napi_struct *napi)
3303 kfree_skb(napi->skb);
3307 bool netdev_is_rx_handler_busy(struct net_device *dev);
3308 int netdev_rx_handler_register(struct net_device *dev,
3309 rx_handler_func_t *rx_handler,
3310 void *rx_handler_data);
3311 void netdev_rx_handler_unregister(struct net_device *dev);
3313 bool dev_valid_name(const char *name);
3314 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3315 int dev_ethtool(struct net *net, struct ifreq *);
3316 unsigned int dev_get_flags(const struct net_device *);
3317 int __dev_change_flags(struct net_device *, unsigned int flags);
3318 int dev_change_flags(struct net_device *, unsigned int);
3319 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3320 unsigned int gchanges);
3321 int dev_change_name(struct net_device *, const char *);
3322 int dev_set_alias(struct net_device *, const char *, size_t);
3323 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3324 int __dev_set_mtu(struct net_device *, int);
3325 int dev_set_mtu(struct net_device *, int);
3326 int dev_validate_mtu(struct net_device *dev, int mtu);
3327 void dev_set_group(struct net_device *, int);
3328 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3329 int dev_change_carrier(struct net_device *, bool new_carrier);
3330 int dev_get_phys_port_id(struct net_device *dev,
3331 struct netdev_phys_item_id *ppid);
3332 int dev_get_phys_port_name(struct net_device *dev,
3333 char *name, size_t len);
3334 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3335 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3336 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3337 struct netdev_queue *txq, int *ret);
3339 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3340 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3342 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3344 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3345 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3346 bool is_skb_forwardable(const struct net_device *dev,
3347 const struct sk_buff *skb);
3349 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3350 struct sk_buff *skb)
3352 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3353 unlikely(!is_skb_forwardable(dev, skb))) {
3354 atomic_long_inc(&dev->rx_dropped);
3359 skb_scrub_packet(skb, true);
3364 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3366 extern int netdev_budget;
3367 extern unsigned int netdev_budget_usecs;
3369 /* Called by rtnetlink.c:rtnl_unlock() */
3370 void netdev_run_todo(void);
3373 * dev_put - release reference to device
3374 * @dev: network device
3376 * Release reference to device to allow it to be freed.
3378 static inline void dev_put(struct net_device *dev)
3380 this_cpu_dec(*dev->pcpu_refcnt);
3384 * dev_hold - get reference to device
3385 * @dev: network device
3387 * Hold reference to device to keep it from being freed.
3389 static inline void dev_hold(struct net_device *dev)
3391 this_cpu_inc(*dev->pcpu_refcnt);
3394 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3395 * and _off may be called from IRQ context, but it is caller
3396 * who is responsible for serialization of these calls.
3398 * The name carrier is inappropriate, these functions should really be
3399 * called netif_lowerlayer_*() because they represent the state of any
3400 * kind of lower layer not just hardware media.
3403 void linkwatch_init_dev(struct net_device *dev);
3404 void linkwatch_fire_event(struct net_device *dev);
3405 void linkwatch_forget_dev(struct net_device *dev);
3408 * netif_carrier_ok - test if carrier present
3409 * @dev: network device
3411 * Check if carrier is present on device
3413 static inline bool netif_carrier_ok(const struct net_device *dev)
3415 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3418 unsigned long dev_trans_start(struct net_device *dev);
3420 void __netdev_watchdog_up(struct net_device *dev);
3422 void netif_carrier_on(struct net_device *dev);
3424 void netif_carrier_off(struct net_device *dev);
3427 * netif_dormant_on - mark device as dormant.
3428 * @dev: network device
3430 * Mark device as dormant (as per RFC2863).
3432 * The dormant state indicates that the relevant interface is not
3433 * actually in a condition to pass packets (i.e., it is not 'up') but is
3434 * in a "pending" state, waiting for some external event. For "on-
3435 * demand" interfaces, this new state identifies the situation where the
3436 * interface is waiting for events to place it in the up state.
3438 static inline void netif_dormant_on(struct net_device *dev)
3440 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3441 linkwatch_fire_event(dev);
3445 * netif_dormant_off - set device as not dormant.
3446 * @dev: network device
3448 * Device is not in dormant state.
3450 static inline void netif_dormant_off(struct net_device *dev)
3452 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3453 linkwatch_fire_event(dev);
3457 * netif_dormant - test if device is dormant
3458 * @dev: network device
3460 * Check if device is dormant.
3462 static inline bool netif_dormant(const struct net_device *dev)
3464 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3469 * netif_oper_up - test if device is operational
3470 * @dev: network device
3472 * Check if carrier is operational
3474 static inline bool netif_oper_up(const struct net_device *dev)
3476 return (dev->operstate == IF_OPER_UP ||
3477 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3481 * netif_device_present - is device available or removed
3482 * @dev: network device
3484 * Check if device has not been removed from system.
3486 static inline bool netif_device_present(struct net_device *dev)
3488 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3491 void netif_device_detach(struct net_device *dev);
3493 void netif_device_attach(struct net_device *dev);
3496 * Network interface message level settings
3500 NETIF_MSG_DRV = 0x0001,
3501 NETIF_MSG_PROBE = 0x0002,
3502 NETIF_MSG_LINK = 0x0004,
3503 NETIF_MSG_TIMER = 0x0008,
3504 NETIF_MSG_IFDOWN = 0x0010,
3505 NETIF_MSG_IFUP = 0x0020,
3506 NETIF_MSG_RX_ERR = 0x0040,
3507 NETIF_MSG_TX_ERR = 0x0080,
3508 NETIF_MSG_TX_QUEUED = 0x0100,
3509 NETIF_MSG_INTR = 0x0200,
3510 NETIF_MSG_TX_DONE = 0x0400,
3511 NETIF_MSG_RX_STATUS = 0x0800,
3512 NETIF_MSG_PKTDATA = 0x1000,
3513 NETIF_MSG_HW = 0x2000,
3514 NETIF_MSG_WOL = 0x4000,
3517 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3518 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3519 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3520 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3521 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3522 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3523 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3524 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3525 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3526 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3527 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3528 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3529 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3530 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3531 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3533 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3536 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3537 return default_msg_enable_bits;
3538 if (debug_value == 0) /* no output */
3540 /* set low N bits */
3541 return (1U << debug_value) - 1;
3544 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3546 spin_lock(&txq->_xmit_lock);
3547 txq->xmit_lock_owner = cpu;
3550 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3552 __acquire(&txq->_xmit_lock);
3556 static inline void __netif_tx_release(struct netdev_queue *txq)
3558 __release(&txq->_xmit_lock);
3561 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3563 spin_lock_bh(&txq->_xmit_lock);
3564 txq->xmit_lock_owner = smp_processor_id();
3567 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3569 bool ok = spin_trylock(&txq->_xmit_lock);
3571 txq->xmit_lock_owner = smp_processor_id();
3575 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3577 txq->xmit_lock_owner = -1;
3578 spin_unlock(&txq->_xmit_lock);
3581 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3583 txq->xmit_lock_owner = -1;
3584 spin_unlock_bh(&txq->_xmit_lock);
3587 static inline void txq_trans_update(struct netdev_queue *txq)
3589 if (txq->xmit_lock_owner != -1)
3590 txq->trans_start = jiffies;
3593 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3594 static inline void netif_trans_update(struct net_device *dev)
3596 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3598 if (txq->trans_start != jiffies)
3599 txq->trans_start = jiffies;
3603 * netif_tx_lock - grab network device transmit lock
3604 * @dev: network device
3606 * Get network device transmit lock
3608 static inline void netif_tx_lock(struct net_device *dev)
3613 spin_lock(&dev->tx_global_lock);
3614 cpu = smp_processor_id();
3615 for (i = 0; i < dev->num_tx_queues; i++) {
3616 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3618 /* We are the only thread of execution doing a
3619 * freeze, but we have to grab the _xmit_lock in
3620 * order to synchronize with threads which are in
3621 * the ->hard_start_xmit() handler and already
3622 * checked the frozen bit.
3624 __netif_tx_lock(txq, cpu);
3625 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3626 __netif_tx_unlock(txq);
3630 static inline void netif_tx_lock_bh(struct net_device *dev)
3636 static inline void netif_tx_unlock(struct net_device *dev)
3640 for (i = 0; i < dev->num_tx_queues; i++) {
3641 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3643 /* No need to grab the _xmit_lock here. If the
3644 * queue is not stopped for another reason, we
3647 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3648 netif_schedule_queue(txq);
3650 spin_unlock(&dev->tx_global_lock);
3653 static inline void netif_tx_unlock_bh(struct net_device *dev)
3655 netif_tx_unlock(dev);
3659 #define HARD_TX_LOCK(dev, txq, cpu) { \
3660 if ((dev->features & NETIF_F_LLTX) == 0) { \
3661 __netif_tx_lock(txq, cpu); \
3663 __netif_tx_acquire(txq); \
3667 #define HARD_TX_TRYLOCK(dev, txq) \
3668 (((dev->features & NETIF_F_LLTX) == 0) ? \
3669 __netif_tx_trylock(txq) : \
3670 __netif_tx_acquire(txq))
3672 #define HARD_TX_UNLOCK(dev, txq) { \
3673 if ((dev->features & NETIF_F_LLTX) == 0) { \
3674 __netif_tx_unlock(txq); \
3676 __netif_tx_release(txq); \
3680 static inline void netif_tx_disable(struct net_device *dev)
3686 cpu = smp_processor_id();
3687 spin_lock(&dev->tx_global_lock);
3688 for (i = 0; i < dev->num_tx_queues; i++) {
3689 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3691 __netif_tx_lock(txq, cpu);
3692 netif_tx_stop_queue(txq);
3693 __netif_tx_unlock(txq);
3695 spin_unlock(&dev->tx_global_lock);
3699 static inline void netif_addr_lock(struct net_device *dev)
3701 spin_lock(&dev->addr_list_lock);
3704 static inline void netif_addr_lock_nested(struct net_device *dev)
3706 int subclass = SINGLE_DEPTH_NESTING;
3708 if (dev->netdev_ops->ndo_get_lock_subclass)
3709 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3711 spin_lock_nested(&dev->addr_list_lock, subclass);
3714 static inline void netif_addr_lock_bh(struct net_device *dev)
3716 spin_lock_bh(&dev->addr_list_lock);
3719 static inline void netif_addr_unlock(struct net_device *dev)
3721 spin_unlock(&dev->addr_list_lock);
3724 static inline void netif_addr_unlock_bh(struct net_device *dev)
3726 spin_unlock_bh(&dev->addr_list_lock);
3730 * dev_addrs walker. Should be used only for read access. Call with
3731 * rcu_read_lock held.
3733 #define for_each_dev_addr(dev, ha) \
3734 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3736 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3738 void ether_setup(struct net_device *dev);
3740 /* Support for loadable net-drivers */
3741 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3742 unsigned char name_assign_type,
3743 void (*setup)(struct net_device *),
3744 unsigned int txqs, unsigned int rxqs);
3745 int dev_get_valid_name(struct net *net, struct net_device *dev,
3748 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3749 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3751 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3752 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3755 int register_netdev(struct net_device *dev);
3756 void unregister_netdev(struct net_device *dev);
3758 /* General hardware address lists handling functions */
3759 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3760 struct netdev_hw_addr_list *from_list, int addr_len);
3761 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3762 struct netdev_hw_addr_list *from_list, int addr_len);
3763 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3764 struct net_device *dev,
3765 int (*sync)(struct net_device *, const unsigned char *),
3766 int (*unsync)(struct net_device *,
3767 const unsigned char *));
3768 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3769 struct net_device *dev,
3770 int (*unsync)(struct net_device *,
3771 const unsigned char *));
3772 void __hw_addr_init(struct netdev_hw_addr_list *list);
3774 /* Functions used for device addresses handling */
3775 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3776 unsigned char addr_type);
3777 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3778 unsigned char addr_type);
3779 void dev_addr_flush(struct net_device *dev);
3780 int dev_addr_init(struct net_device *dev);
3782 /* Functions used for unicast addresses handling */
3783 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3784 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3785 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3786 int dev_uc_sync(struct net_device *to, struct net_device *from);
3787 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3788 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3789 void dev_uc_flush(struct net_device *dev);
3790 void dev_uc_init(struct net_device *dev);
3793 * __dev_uc_sync - Synchonize device's unicast list
3794 * @dev: device to sync
3795 * @sync: function to call if address should be added
3796 * @unsync: function to call if address should be removed
3798 * Add newly added addresses to the interface, and release
3799 * addresses that have been deleted.
3801 static inline int __dev_uc_sync(struct net_device *dev,
3802 int (*sync)(struct net_device *,
3803 const unsigned char *),
3804 int (*unsync)(struct net_device *,
3805 const unsigned char *))
3807 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3811 * __dev_uc_unsync - Remove synchronized addresses from device
3812 * @dev: device to sync
3813 * @unsync: function to call if address should be removed
3815 * Remove all addresses that were added to the device by dev_uc_sync().
3817 static inline void __dev_uc_unsync(struct net_device *dev,
3818 int (*unsync)(struct net_device *,
3819 const unsigned char *))
3821 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3824 /* Functions used for multicast addresses handling */
3825 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3826 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3827 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3828 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3829 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3830 int dev_mc_sync(struct net_device *to, struct net_device *from);
3831 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3832 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3833 void dev_mc_flush(struct net_device *dev);
3834 void dev_mc_init(struct net_device *dev);
3837 * __dev_mc_sync - Synchonize device's multicast list
3838 * @dev: device to sync
3839 * @sync: function to call if address should be added
3840 * @unsync: function to call if address should be removed
3842 * Add newly added addresses to the interface, and release
3843 * addresses that have been deleted.
3845 static inline int __dev_mc_sync(struct net_device *dev,
3846 int (*sync)(struct net_device *,
3847 const unsigned char *),
3848 int (*unsync)(struct net_device *,
3849 const unsigned char *))
3851 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3855 * __dev_mc_unsync - Remove synchronized addresses from device
3856 * @dev: device to sync
3857 * @unsync: function to call if address should be removed
3859 * Remove all addresses that were added to the device by dev_mc_sync().
3861 static inline void __dev_mc_unsync(struct net_device *dev,
3862 int (*unsync)(struct net_device *,
3863 const unsigned char *))
3865 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3868 /* Functions used for secondary unicast and multicast support */
3869 void dev_set_rx_mode(struct net_device *dev);
3870 void __dev_set_rx_mode(struct net_device *dev);
3871 int dev_set_promiscuity(struct net_device *dev, int inc);
3872 int dev_set_allmulti(struct net_device *dev, int inc);
3873 void netdev_state_change(struct net_device *dev);
3874 void netdev_notify_peers(struct net_device *dev);
3875 void netdev_features_change(struct net_device *dev);
3876 /* Load a device via the kmod */
3877 void dev_load(struct net *net, const char *name);
3878 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3879 struct rtnl_link_stats64 *storage);
3880 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3881 const struct net_device_stats *netdev_stats);
3883 extern int netdev_max_backlog;
3884 extern int netdev_tstamp_prequeue;
3885 extern int weight_p;
3886 extern int dev_weight_rx_bias;
3887 extern int dev_weight_tx_bias;
3888 extern int dev_rx_weight;
3889 extern int dev_tx_weight;
3891 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3892 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3893 struct list_head **iter);
3894 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3895 struct list_head **iter);
3897 /* iterate through upper list, must be called under RCU read lock */
3898 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3899 for (iter = &(dev)->adj_list.upper, \
3900 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3902 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3904 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3905 int (*fn)(struct net_device *upper_dev,
3909 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3910 struct net_device *upper_dev);
3912 bool netdev_has_any_upper_dev(struct net_device *dev);
3914 void *netdev_lower_get_next_private(struct net_device *dev,
3915 struct list_head **iter);
3916 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3917 struct list_head **iter);
3919 #define netdev_for_each_lower_private(dev, priv, iter) \
3920 for (iter = (dev)->adj_list.lower.next, \
3921 priv = netdev_lower_get_next_private(dev, &(iter)); \
3923 priv = netdev_lower_get_next_private(dev, &(iter)))
3925 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3926 for (iter = &(dev)->adj_list.lower, \
3927 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3929 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3931 void *netdev_lower_get_next(struct net_device *dev,
3932 struct list_head **iter);
3934 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3935 for (iter = (dev)->adj_list.lower.next, \
3936 ldev = netdev_lower_get_next(dev, &(iter)); \
3938 ldev = netdev_lower_get_next(dev, &(iter)))
3940 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3941 struct list_head **iter);
3942 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3943 struct list_head **iter);
3945 int netdev_walk_all_lower_dev(struct net_device *dev,
3946 int (*fn)(struct net_device *lower_dev,
3949 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3950 int (*fn)(struct net_device *lower_dev,
3954 void *netdev_adjacent_get_private(struct list_head *adj_list);
3955 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3956 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3957 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3958 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3959 int netdev_master_upper_dev_link(struct net_device *dev,
3960 struct net_device *upper_dev,
3961 void *upper_priv, void *upper_info);
3962 void netdev_upper_dev_unlink(struct net_device *dev,
3963 struct net_device *upper_dev);
3964 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3965 void *netdev_lower_dev_get_private(struct net_device *dev,
3966 struct net_device *lower_dev);
3967 void netdev_lower_state_changed(struct net_device *lower_dev,
3968 void *lower_state_info);
3970 /* RSS keys are 40 or 52 bytes long */
3971 #define NETDEV_RSS_KEY_LEN 52
3972 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3973 void netdev_rss_key_fill(void *buffer, size_t len);
3975 int dev_get_nest_level(struct net_device *dev);
3976 int skb_checksum_help(struct sk_buff *skb);
3977 int skb_crc32c_csum_help(struct sk_buff *skb);
3978 int skb_csum_hwoffload_help(struct sk_buff *skb,
3979 const netdev_features_t features);
3981 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3982 netdev_features_t features, bool tx_path);
3983 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3984 netdev_features_t features);
3986 struct netdev_bonding_info {
3991 struct netdev_notifier_bonding_info {
3992 struct netdev_notifier_info info; /* must be first */
3993 struct netdev_bonding_info bonding_info;
3996 void netdev_bonding_info_change(struct net_device *dev,
3997 struct netdev_bonding_info *bonding_info);
4000 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4002 return __skb_gso_segment(skb, features, true);
4004 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4006 static inline bool can_checksum_protocol(netdev_features_t features,
4009 if (protocol == htons(ETH_P_FCOE))
4010 return !!(features & NETIF_F_FCOE_CRC);
4012 /* Assume this is an IP checksum (not SCTP CRC) */
4014 if (features & NETIF_F_HW_CSUM) {
4015 /* Can checksum everything */
4020 case htons(ETH_P_IP):
4021 return !!(features & NETIF_F_IP_CSUM);
4022 case htons(ETH_P_IPV6):
4023 return !!(features & NETIF_F_IPV6_CSUM);
4030 void netdev_rx_csum_fault(struct net_device *dev);
4032 static inline void netdev_rx_csum_fault(struct net_device *dev)
4036 /* rx skb timestamps */
4037 void net_enable_timestamp(void);
4038 void net_disable_timestamp(void);
4040 #ifdef CONFIG_PROC_FS
4041 int __init dev_proc_init(void);
4043 #define dev_proc_init() 0
4046 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4047 struct sk_buff *skb, struct net_device *dev,
4050 skb->xmit_more = more ? 1 : 0;
4051 return ops->ndo_start_xmit(skb, dev);
4054 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4055 struct netdev_queue *txq, bool more)
4057 const struct net_device_ops *ops = dev->netdev_ops;
4060 rc = __netdev_start_xmit(ops, skb, dev, more);
4061 if (rc == NETDEV_TX_OK)
4062 txq_trans_update(txq);
4067 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4069 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4072 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4074 return netdev_class_create_file_ns(class_attr, NULL);
4077 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4079 netdev_class_remove_file_ns(class_attr, NULL);
4082 extern const struct kobj_ns_type_operations net_ns_type_operations;
4084 const char *netdev_drivername(const struct net_device *dev);
4086 void linkwatch_run_queue(void);
4088 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4089 netdev_features_t f2)
4091 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4092 if (f1 & NETIF_F_HW_CSUM)
4093 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4095 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4101 static inline netdev_features_t netdev_get_wanted_features(
4102 struct net_device *dev)
4104 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4106 netdev_features_t netdev_increment_features(netdev_features_t all,
4107 netdev_features_t one, netdev_features_t mask);
4109 /* Allow TSO being used on stacked device :
4110 * Performing the GSO segmentation before last device
4111 * is a performance improvement.
4113 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4114 netdev_features_t mask)
4116 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4119 int __netdev_update_features(struct net_device *dev);
4120 void netdev_update_features(struct net_device *dev);
4121 void netdev_change_features(struct net_device *dev);
4123 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4124 struct net_device *dev);
4126 netdev_features_t passthru_features_check(struct sk_buff *skb,
4127 struct net_device *dev,
4128 netdev_features_t features);
4129 netdev_features_t netif_skb_features(struct sk_buff *skb);
4131 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4133 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4135 /* check flags correspondence */
4136 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4137 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4138 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4139 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4140 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4141 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4142 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4143 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4144 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4145 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4146 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4147 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4148 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4149 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4150 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4151 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4152 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4154 return (features & feature) == feature;
4157 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4159 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4160 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4163 static inline bool netif_needs_gso(struct sk_buff *skb,
4164 netdev_features_t features)
4166 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4167 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4168 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4171 static inline void netif_set_gso_max_size(struct net_device *dev,
4174 dev->gso_max_size = size;
4177 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4178 int pulled_hlen, u16 mac_offset,
4181 skb->protocol = protocol;
4182 skb->encapsulation = 1;
4183 skb_push(skb, pulled_hlen);
4184 skb_reset_transport_header(skb);
4185 skb->mac_header = mac_offset;
4186 skb->network_header = skb->mac_header + mac_len;
4187 skb->mac_len = mac_len;
4190 static inline bool netif_is_macsec(const struct net_device *dev)
4192 return dev->priv_flags & IFF_MACSEC;
4195 static inline bool netif_is_macvlan(const struct net_device *dev)
4197 return dev->priv_flags & IFF_MACVLAN;
4200 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4202 return dev->priv_flags & IFF_MACVLAN_PORT;
4205 static inline bool netif_is_ipvlan(const struct net_device *dev)
4207 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4210 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4212 return dev->priv_flags & IFF_IPVLAN_MASTER;
4215 static inline bool netif_is_bond_master(const struct net_device *dev)
4217 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4220 static inline bool netif_is_bond_slave(const struct net_device *dev)
4222 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4225 static inline bool netif_supports_nofcs(struct net_device *dev)
4227 return dev->priv_flags & IFF_SUPP_NOFCS;
4230 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4232 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4235 static inline bool netif_is_l3_master(const struct net_device *dev)
4237 return dev->priv_flags & IFF_L3MDEV_MASTER;
4240 static inline bool netif_is_l3_slave(const struct net_device *dev)
4242 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4245 static inline bool netif_is_bridge_master(const struct net_device *dev)
4247 return dev->priv_flags & IFF_EBRIDGE;
4250 static inline bool netif_is_bridge_port(const struct net_device *dev)
4252 return dev->priv_flags & IFF_BRIDGE_PORT;
4255 static inline bool netif_is_ovs_master(const struct net_device *dev)
4257 return dev->priv_flags & IFF_OPENVSWITCH;
4260 static inline bool netif_is_ovs_port(const struct net_device *dev)
4262 return dev->priv_flags & IFF_OVS_DATAPATH;
4265 static inline bool netif_is_team_master(const struct net_device *dev)
4267 return dev->priv_flags & IFF_TEAM;
4270 static inline bool netif_is_team_port(const struct net_device *dev)
4272 return dev->priv_flags & IFF_TEAM_PORT;
4275 static inline bool netif_is_lag_master(const struct net_device *dev)
4277 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4280 static inline bool netif_is_lag_port(const struct net_device *dev)
4282 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4285 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4287 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4290 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4291 static inline void netif_keep_dst(struct net_device *dev)
4293 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4296 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4297 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4299 /* TODO: reserve and use an additional IFF bit, if we get more users */
4300 return dev->priv_flags & IFF_MACSEC;
4303 extern struct pernet_operations __net_initdata loopback_net_ops;
4305 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4307 /* netdev_printk helpers, similar to dev_printk */
4309 static inline const char *netdev_name(const struct net_device *dev)
4311 if (!dev->name[0] || strchr(dev->name, '%'))
4312 return "(unnamed net_device)";
4316 static inline bool netdev_unregistering(const struct net_device *dev)
4318 return dev->reg_state == NETREG_UNREGISTERING;
4321 static inline const char *netdev_reg_state(const struct net_device *dev)
4323 switch (dev->reg_state) {
4324 case NETREG_UNINITIALIZED: return " (uninitialized)";
4325 case NETREG_REGISTERED: return "";
4326 case NETREG_UNREGISTERING: return " (unregistering)";
4327 case NETREG_UNREGISTERED: return " (unregistered)";
4328 case NETREG_RELEASED: return " (released)";
4329 case NETREG_DUMMY: return " (dummy)";
4332 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4333 return " (unknown)";
4337 void netdev_printk(const char *level, const struct net_device *dev,
4338 const char *format, ...);
4340 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4342 void netdev_alert(const struct net_device *dev, const char *format, ...);
4344 void netdev_crit(const struct net_device *dev, const char *format, ...);
4346 void netdev_err(const struct net_device *dev, const char *format, ...);
4348 void netdev_warn(const struct net_device *dev, const char *format, ...);
4350 void netdev_notice(const struct net_device *dev, const char *format, ...);
4352 void netdev_info(const struct net_device *dev, const char *format, ...);
4354 #define MODULE_ALIAS_NETDEV(device) \
4355 MODULE_ALIAS("netdev-" device)
4357 #if defined(CONFIG_DYNAMIC_DEBUG)
4358 #define netdev_dbg(__dev, format, args...) \
4360 dynamic_netdev_dbg(__dev, format, ##args); \
4362 #elif defined(DEBUG)
4363 #define netdev_dbg(__dev, format, args...) \
4364 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4366 #define netdev_dbg(__dev, format, args...) \
4369 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4373 #if defined(VERBOSE_DEBUG)
4374 #define netdev_vdbg netdev_dbg
4377 #define netdev_vdbg(dev, format, args...) \
4380 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4386 * netdev_WARN() acts like dev_printk(), but with the key difference
4387 * of using a WARN/WARN_ON to get the message out, including the
4388 * file/line information and a backtrace.
4390 #define netdev_WARN(dev, format, args...) \
4391 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4392 netdev_reg_state(dev), ##args)
4394 /* netif printk helpers, similar to netdev_printk */
4396 #define netif_printk(priv, type, level, dev, fmt, args...) \
4398 if (netif_msg_##type(priv)) \
4399 netdev_printk(level, (dev), fmt, ##args); \
4402 #define netif_level(level, priv, type, dev, fmt, args...) \
4404 if (netif_msg_##type(priv)) \
4405 netdev_##level(dev, fmt, ##args); \
4408 #define netif_emerg(priv, type, dev, fmt, args...) \
4409 netif_level(emerg, priv, type, dev, fmt, ##args)
4410 #define netif_alert(priv, type, dev, fmt, args...) \
4411 netif_level(alert, priv, type, dev, fmt, ##args)
4412 #define netif_crit(priv, type, dev, fmt, args...) \
4413 netif_level(crit, priv, type, dev, fmt, ##args)
4414 #define netif_err(priv, type, dev, fmt, args...) \
4415 netif_level(err, priv, type, dev, fmt, ##args)
4416 #define netif_warn(priv, type, dev, fmt, args...) \
4417 netif_level(warn, priv, type, dev, fmt, ##args)
4418 #define netif_notice(priv, type, dev, fmt, args...) \
4419 netif_level(notice, priv, type, dev, fmt, ##args)
4420 #define netif_info(priv, type, dev, fmt, args...) \
4421 netif_level(info, priv, type, dev, fmt, ##args)
4423 #if defined(CONFIG_DYNAMIC_DEBUG)
4424 #define netif_dbg(priv, type, netdev, format, args...) \
4426 if (netif_msg_##type(priv)) \
4427 dynamic_netdev_dbg(netdev, format, ##args); \
4429 #elif defined(DEBUG)
4430 #define netif_dbg(priv, type, dev, format, args...) \
4431 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4433 #define netif_dbg(priv, type, dev, format, args...) \
4436 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4441 /* if @cond then downgrade to debug, else print at @level */
4442 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4445 netif_dbg(priv, type, netdev, fmt, ##args); \
4447 netif_ ## level(priv, type, netdev, fmt, ##args); \
4450 #if defined(VERBOSE_DEBUG)
4451 #define netif_vdbg netif_dbg
4453 #define netif_vdbg(priv, type, dev, format, args...) \
4456 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4462 * The list of packet types we will receive (as opposed to discard)
4463 * and the routines to invoke.
4465 * Why 16. Because with 16 the only overlap we get on a hash of the
4466 * low nibble of the protocol value is RARP/SNAP/X.25.
4468 * NOTE: That is no longer true with the addition of VLAN tags. Not
4469 * sure which should go first, but I bet it won't make much
4470 * difference if we are running VLANs. The good news is that
4471 * this protocol won't be in the list unless compiled in, so
4472 * the average user (w/out VLANs) will not be adversely affected.
4488 #define PTYPE_HASH_SIZE (16)
4489 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4491 #endif /* _LINUX_NETDEVICE_H */