2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
44 #include <net/dcbnl.h>
46 #include <net/netprio_cgroup.h>
48 #include <linux/netdev_features.h>
49 #include <linux/neighbour.h>
50 #include <uapi/linux/netdevice.h>
51 #include <uapi/linux/if_bonding.h>
52 #include <uapi/linux/pkt_cls.h>
53 #include <linux/hashtable.h>
58 struct dsa_switch_tree;
62 /* 802.15.4 specific */
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
70 void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
78 * Transmit return codes: transmit return codes originate from three different
81 * - qdisc return codes
82 * - driver transmit return codes
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91 * others are propagated to higher layers.
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK 0xf0
110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 typedef enum netdev_tx netdev_tx_t;
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 static inline bool dev_xmit_complete(int rc)
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 if (likely(rc < NET_XMIT_MASK))
135 * Compute the worst-case header length according to the protocols
139 #if defined(CONFIG_HYPERV_NET)
140 # define LL_MAX_HEADER 128
141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
145 # define LL_MAX_HEADER 96
148 # define LL_MAX_HEADER 32
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 extern struct static_key rfs_needed;
203 struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
207 #define NETDEV_HW_ADDR_T_LAN 1
208 #define NETDEV_HW_ADDR_T_SAN 2
209 #define NETDEV_HW_ADDR_T_SLAVE 3
210 #define NETDEV_HW_ADDR_T_UNICAST 4
211 #define NETDEV_HW_ADDR_T_MULTICAST 5
216 struct rcu_head rcu_head;
219 struct netdev_hw_addr_list {
220 struct list_head list;
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
262 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
264 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
265 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
270 const void *saddr, unsigned int len);
271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
276 bool (*validate)(const char *ll_header, unsigned int len);
277 __be16 (*parse_protocol)(const struct sk_buff *skb);
280 /* These flag bits are private to the generic network queueing
281 * layer; they may not be explicitly referenced by any other
285 enum netdev_state_t {
287 __LINK_STATE_PRESENT,
288 __LINK_STATE_NOCARRIER,
289 __LINK_STATE_LINKWATCH_PENDING,
290 __LINK_STATE_DORMANT,
295 * This structure holds boot-time configured netdevice settings. They
296 * are then used in the device probing.
298 struct netdev_boot_setup {
302 #define NETDEV_BOOT_SETUP_MAX 8
304 int __init netdev_boot_setup(char *str);
307 * Structure for NAPI scheduling similar to tasklet but with weighting
310 /* The poll_list must only be managed by the entity which
311 * changes the state of the NAPI_STATE_SCHED bit. This means
312 * whoever atomically sets that bit can add this napi_struct
313 * to the per-CPU poll_list, and whoever clears that bit
314 * can remove from the list right before clearing the bit.
316 struct list_head poll_list;
320 unsigned int gro_count;
321 int (*poll)(struct napi_struct *, int);
322 #ifdef CONFIG_NETPOLL
325 struct net_device *dev;
326 struct sk_buff *gro_list;
328 struct hrtimer timer;
329 struct list_head dev_list;
330 struct hlist_node napi_hash_node;
331 unsigned int napi_id;
335 NAPI_STATE_SCHED, /* Poll is scheduled */
336 NAPI_STATE_MISSED, /* reschedule a napi */
337 NAPI_STATE_DISABLE, /* Disable pending */
338 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
339 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
340 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
341 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
345 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
346 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
347 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
348 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
349 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
350 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
351 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
362 typedef enum gro_result gro_result_t;
365 * enum rx_handler_result - Possible return values for rx_handlers.
366 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
368 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
369 * case skb->dev was changed by rx_handler.
370 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
371 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
373 * rx_handlers are functions called from inside __netif_receive_skb(), to do
374 * special processing of the skb, prior to delivery to protocol handlers.
376 * Currently, a net_device can only have a single rx_handler registered. Trying
377 * to register a second rx_handler will return -EBUSY.
379 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
380 * To unregister a rx_handler on a net_device, use
381 * netdev_rx_handler_unregister().
383 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
386 * If the rx_handler consumed the skb in some way, it should return
387 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
388 * the skb to be delivered in some other way.
390 * If the rx_handler changed skb->dev, to divert the skb to another
391 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
392 * new device will be called if it exists.
394 * If the rx_handler decides the skb should be ignored, it should return
395 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
396 * are registered on exact device (ptype->dev == skb->dev).
398 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
399 * delivered, it should return RX_HANDLER_PASS.
401 * A device without a registered rx_handler will behave as if rx_handler
402 * returned RX_HANDLER_PASS.
405 enum rx_handler_result {
411 typedef enum rx_handler_result rx_handler_result_t;
412 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
414 void __napi_schedule(struct napi_struct *n);
415 void __napi_schedule_irqoff(struct napi_struct *n);
417 static inline bool napi_disable_pending(struct napi_struct *n)
419 return test_bit(NAPI_STATE_DISABLE, &n->state);
422 bool napi_schedule_prep(struct napi_struct *n);
425 * napi_schedule - schedule NAPI poll
428 * Schedule NAPI poll routine to be called if it is not already
431 static inline void napi_schedule(struct napi_struct *n)
433 if (napi_schedule_prep(n))
438 * napi_schedule_irqoff - schedule NAPI poll
441 * Variant of napi_schedule(), assuming hard irqs are masked.
443 static inline void napi_schedule_irqoff(struct napi_struct *n)
445 if (napi_schedule_prep(n))
446 __napi_schedule_irqoff(n);
449 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
450 static inline bool napi_reschedule(struct napi_struct *napi)
452 if (napi_schedule_prep(napi)) {
453 __napi_schedule(napi);
459 bool napi_complete_done(struct napi_struct *n, int work_done);
461 * napi_complete - NAPI processing complete
464 * Mark NAPI processing as complete.
465 * Consider using napi_complete_done() instead.
466 * Return false if device should avoid rearming interrupts.
468 static inline bool napi_complete(struct napi_struct *n)
470 return napi_complete_done(n, 0);
474 * napi_hash_del - remove a NAPI from global table
475 * @napi: NAPI context
477 * Warning: caller must observe RCU grace period
478 * before freeing memory containing @napi, if
479 * this function returns true.
480 * Note: core networking stack automatically calls it
481 * from netif_napi_del().
482 * Drivers might want to call this helper to combine all
483 * the needed RCU grace periods into a single one.
485 bool napi_hash_del(struct napi_struct *napi);
488 * napi_disable - prevent NAPI from scheduling
491 * Stop NAPI from being scheduled on this context.
492 * Waits till any outstanding processing completes.
494 void napi_disable(struct napi_struct *n);
497 * napi_enable - enable NAPI scheduling
500 * Resume NAPI from being scheduled on this context.
501 * Must be paired with napi_disable.
503 static inline void napi_enable(struct napi_struct *n)
505 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
506 smp_mb__before_atomic();
507 clear_bit(NAPI_STATE_SCHED, &n->state);
508 clear_bit(NAPI_STATE_NPSVC, &n->state);
512 * napi_synchronize - wait until NAPI is not running
515 * Wait until NAPI is done being scheduled on this context.
516 * Waits till any outstanding processing completes but
517 * does not disable future activations.
519 static inline void napi_synchronize(const struct napi_struct *n)
521 if (IS_ENABLED(CONFIG_SMP))
522 while (test_bit(NAPI_STATE_SCHED, &n->state))
528 enum netdev_queue_state_t {
529 __QUEUE_STATE_DRV_XOFF,
530 __QUEUE_STATE_STACK_XOFF,
531 __QUEUE_STATE_FROZEN,
534 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
535 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
538 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
539 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
541 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
545 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
546 * netif_tx_* functions below are used to manipulate this flag. The
547 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
548 * queue independently. The netif_xmit_*stopped functions below are called
549 * to check if the queue has been stopped by the driver or stack (either
550 * of the XOFF bits are set in the state). Drivers should not need to call
551 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 struct netdev_queue {
558 struct net_device *dev;
559 struct Qdisc __rcu *qdisc;
560 struct Qdisc *qdisc_sleeping;
564 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 unsigned long tx_maxrate;
569 * Number of TX timeouts for this queue
570 * (/sys/class/net/DEV/Q/trans_timeout)
572 unsigned long trans_timeout;
576 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
579 * Time (in jiffies) of last Tx
581 unsigned long trans_start;
588 } ____cacheline_aligned_in_smp;
590 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
592 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 * This structure holds an RPS map which can be of variable length. The
609 * map is an array of CPUs.
616 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
619 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
620 * tail pointer for that CPU's input queue at the time of last enqueue, and
621 * a hardware filter index.
623 struct rps_dev_flow {
626 unsigned int last_qtail;
628 #define RPS_NO_FILTER 0xffff
631 * The rps_dev_flow_table structure contains a table of flow mappings.
633 struct rps_dev_flow_table {
636 struct rps_dev_flow flows[0];
638 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
639 ((_num) * sizeof(struct rps_dev_flow)))
642 * The rps_sock_flow_table contains mappings of flows to the last CPU
643 * on which they were processed by the application (set in recvmsg).
644 * Each entry is a 32bit value. Upper part is the high-order bits
645 * of flow hash, lower part is CPU number.
646 * rps_cpu_mask is used to partition the space, depending on number of
647 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
648 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
649 * meaning we use 32-6=26 bits for the hash.
651 struct rps_sock_flow_table {
654 u32 ents[0] ____cacheline_aligned_in_smp;
656 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
658 #define RPS_NO_CPU 0xffff
660 extern u32 rps_cpu_mask;
661 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
663 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 unsigned int index = hash & table->mask;
668 u32 val = hash & ~rps_cpu_mask;
670 /* We only give a hint, preemption can change CPU under us */
671 val |= raw_smp_processor_id();
673 if (table->ents[index] != val)
674 table->ents[index] = val;
678 #ifdef CONFIG_RFS_ACCEL
679 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
682 #endif /* CONFIG_RPS */
684 /* This structure contains an instance of an RX queue. */
685 struct netdev_rx_queue {
687 struct rps_map __rcu *rps_map;
688 struct rps_dev_flow_table __rcu *rps_flow_table;
691 struct net_device *dev;
692 } ____cacheline_aligned_in_smp;
695 * RX queue sysfs structures and functions.
697 struct rx_queue_attribute {
698 struct attribute attr;
699 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
700 ssize_t (*store)(struct netdev_rx_queue *queue,
701 const char *buf, size_t len);
706 * This structure holds an XPS map which can be of variable length. The
707 * map is an array of queues.
711 unsigned int alloc_len;
715 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
716 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
717 - sizeof(struct xps_map)) / sizeof(u16))
720 * This structure holds all XPS maps for device. Maps are indexed by CPU.
722 struct xps_dev_maps {
724 struct xps_map __rcu *cpu_map[0];
726 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
727 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
728 #endif /* CONFIG_XPS */
730 #define TC_MAX_QUEUE 16
731 #define TC_BITMASK 15
732 /* HW offloaded queuing disciplines txq count and offset maps */
733 struct netdev_tc_txq {
738 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
740 * This structure is to hold information about the device
741 * configured to run FCoE protocol stack.
743 struct netdev_fcoe_hbainfo {
744 char manufacturer[64];
745 char serial_number[64];
746 char hardware_version[64];
747 char driver_version[64];
748 char optionrom_version[64];
749 char firmware_version[64];
751 char model_description[256];
755 #define MAX_PHYS_ITEM_ID_LEN 32
757 /* This structure holds a unique identifier to identify some
758 * physical item (port for example) used by a netdevice.
760 struct netdev_phys_item_id {
761 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
762 unsigned char id_len;
765 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
766 struct netdev_phys_item_id *b)
768 return a->id_len == b->id_len &&
769 memcmp(a->id, b->id, a->id_len) == 0;
772 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
773 struct sk_buff *skb);
779 TC_SETUP_CLSMATCHALL,
783 /* These structures hold the attributes of xdp state that are being passed
784 * to the netdevice through the xdp op.
786 enum xdp_netdev_command {
787 /* Set or clear a bpf program used in the earliest stages of packet
788 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
789 * is responsible for calling bpf_prog_put on any old progs that are
790 * stored. In case of error, the callee need not release the new prog
791 * reference, but on success it takes ownership and must bpf_prog_put
792 * when it is no longer used.
796 /* Check if a bpf program is set on the device. The callee should
797 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
798 * is equivalent to XDP_ATTACHED_DRV.
803 struct netlink_ext_ack;
806 enum xdp_netdev_command command;
811 struct bpf_prog *prog;
812 struct netlink_ext_ack *extack;
822 #ifdef CONFIG_XFRM_OFFLOAD
824 int (*xdo_dev_state_add) (struct xfrm_state *x);
825 void (*xdo_dev_state_delete) (struct xfrm_state *x);
826 void (*xdo_dev_state_free) (struct xfrm_state *x);
827 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
828 struct xfrm_state *x);
833 * This structure defines the management hooks for network devices.
834 * The following hooks can be defined; unless noted otherwise, they are
835 * optional and can be filled with a null pointer.
837 * int (*ndo_init)(struct net_device *dev);
838 * This function is called once when a network device is registered.
839 * The network device can use this for any late stage initialization
840 * or semantic validation. It can fail with an error code which will
841 * be propagated back to register_netdev.
843 * void (*ndo_uninit)(struct net_device *dev);
844 * This function is called when device is unregistered or when registration
845 * fails. It is not called if init fails.
847 * int (*ndo_open)(struct net_device *dev);
848 * This function is called when a network device transitions to the up
851 * int (*ndo_stop)(struct net_device *dev);
852 * This function is called when a network device transitions to the down
855 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
856 * struct net_device *dev);
857 * Called when a packet needs to be transmitted.
858 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
859 * the queue before that can happen; it's for obsolete devices and weird
860 * corner cases, but the stack really does a non-trivial amount
861 * of useless work if you return NETDEV_TX_BUSY.
862 * Required; cannot be NULL.
864 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
865 * struct net_device *dev
866 * netdev_features_t features);
867 * Called by core transmit path to determine if device is capable of
868 * performing offload operations on a given packet. This is to give
869 * the device an opportunity to implement any restrictions that cannot
870 * be otherwise expressed by feature flags. The check is called with
871 * the set of features that the stack has calculated and it returns
872 * those the driver believes to be appropriate.
874 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
875 * void *accel_priv, select_queue_fallback_t fallback);
876 * Called to decide which queue to use when device supports multiple
879 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
880 * This function is called to allow device receiver to make
881 * changes to configuration when multicast or promiscuous is enabled.
883 * void (*ndo_set_rx_mode)(struct net_device *dev);
884 * This function is called device changes address list filtering.
885 * If driver handles unicast address filtering, it should set
886 * IFF_UNICAST_FLT in its priv_flags.
888 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
889 * This function is called when the Media Access Control address
890 * needs to be changed. If this interface is not defined, the
891 * MAC address can not be changed.
893 * int (*ndo_validate_addr)(struct net_device *dev);
894 * Test if Media Access Control address is valid for the device.
896 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
897 * Called when a user requests an ioctl which can't be handled by
898 * the generic interface code. If not defined ioctls return
899 * not supported error code.
901 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
902 * Used to set network devices bus interface parameters. This interface
903 * is retained for legacy reasons; new devices should use the bus
904 * interface (PCI) for low level management.
906 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
907 * Called when a user wants to change the Maximum Transfer Unit
910 * void (*ndo_tx_timeout)(struct net_device *dev);
911 * Callback used when the transmitter has not made any progress
912 * for dev->watchdog ticks.
914 * void (*ndo_get_stats64)(struct net_device *dev,
915 * struct rtnl_link_stats64 *storage);
916 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
917 * Called when a user wants to get the network device usage
918 * statistics. Drivers must do one of the following:
919 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
920 * rtnl_link_stats64 structure passed by the caller.
921 * 2. Define @ndo_get_stats to update a net_device_stats structure
922 * (which should normally be dev->stats) and return a pointer to
923 * it. The structure may be changed asynchronously only if each
924 * field is written atomically.
925 * 3. Update dev->stats asynchronously and atomically, and define
928 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
929 * Return true if this device supports offload stats of this attr_id.
931 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
933 * Get statistics for offload operations by attr_id. Write it into the
936 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
937 * If device supports VLAN filtering this function is called when a
938 * VLAN id is registered.
940 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
941 * If device supports VLAN filtering this function is called when a
942 * VLAN id is unregistered.
944 * void (*ndo_poll_controller)(struct net_device *dev);
946 * SR-IOV management functions.
947 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
948 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
949 * u8 qos, __be16 proto);
950 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
952 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
953 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
954 * int (*ndo_get_vf_config)(struct net_device *dev,
955 * int vf, struct ifla_vf_info *ivf);
956 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
957 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
958 * struct nlattr *port[]);
960 * Enable or disable the VF ability to query its RSS Redirection Table and
961 * Hash Key. This is needed since on some devices VF share this information
962 * with PF and querying it may introduce a theoretical security risk.
963 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
964 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
965 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
967 * Called to setup any 'tc' scheduler, classifier or action on @dev.
968 * This is always called from the stack with the rtnl lock held and netif
969 * tx queues stopped. This allows the netdevice to perform queue
972 * Fiber Channel over Ethernet (FCoE) offload functions.
973 * int (*ndo_fcoe_enable)(struct net_device *dev);
974 * Called when the FCoE protocol stack wants to start using LLD for FCoE
975 * so the underlying device can perform whatever needed configuration or
976 * initialization to support acceleration of FCoE traffic.
978 * int (*ndo_fcoe_disable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
980 * so the underlying device can perform whatever needed clean-ups to
981 * stop supporting acceleration of FCoE traffic.
983 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
984 * struct scatterlist *sgl, unsigned int sgc);
985 * Called when the FCoE Initiator wants to initialize an I/O that
986 * is a possible candidate for Direct Data Placement (DDP). The LLD can
987 * perform necessary setup and returns 1 to indicate the device is set up
988 * successfully to perform DDP on this I/O, otherwise this returns 0.
990 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
991 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
992 * indicated by the FC exchange id 'xid', so the underlying device can
993 * clean up and reuse resources for later DDP requests.
995 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
996 * struct scatterlist *sgl, unsigned int sgc);
997 * Called when the FCoE Target wants to initialize an I/O that
998 * is a possible candidate for Direct Data Placement (DDP). The LLD can
999 * perform necessary setup and returns 1 to indicate the device is set up
1000 * successfully to perform DDP on this I/O, otherwise this returns 0.
1002 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1003 * struct netdev_fcoe_hbainfo *hbainfo);
1004 * Called when the FCoE Protocol stack wants information on the underlying
1005 * device. This information is utilized by the FCoE protocol stack to
1006 * register attributes with Fiber Channel management service as per the
1007 * FC-GS Fabric Device Management Information(FDMI) specification.
1009 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1010 * Called when the underlying device wants to override default World Wide
1011 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1012 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1013 * protocol stack to use.
1016 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1017 * u16 rxq_index, u32 flow_id);
1018 * Set hardware filter for RFS. rxq_index is the target queue index;
1019 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1020 * Return the filter ID on success, or a negative error code.
1022 * Slave management functions (for bridge, bonding, etc).
1023 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1024 * Called to make another netdev an underling.
1026 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1027 * Called to release previously enslaved netdev.
1029 * Feature/offload setting functions.
1030 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1031 * netdev_features_t features);
1032 * Adjusts the requested feature flags according to device-specific
1033 * constraints, and returns the resulting flags. Must not modify
1036 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1037 * Called to update device configuration to new features. Passed
1038 * feature set might be less than what was returned by ndo_fix_features()).
1039 * Must return >0 or -errno if it changed dev->features itself.
1041 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1042 * struct net_device *dev,
1043 * const unsigned char *addr, u16 vid, u16 flags)
1044 * Adds an FDB entry to dev for addr.
1045 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1046 * struct net_device *dev,
1047 * const unsigned char *addr, u16 vid)
1048 * Deletes the FDB entry from dev coresponding to addr.
1049 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1050 * struct net_device *dev, struct net_device *filter_dev,
1052 * Used to add FDB entries to dump requests. Implementers should add
1053 * entries to skb and update idx with the number of entries.
1055 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1057 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1058 * struct net_device *dev, u32 filter_mask,
1060 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1063 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1064 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1065 * which do not represent real hardware may define this to allow their
1066 * userspace components to manage their virtual carrier state. Devices
1067 * that determine carrier state from physical hardware properties (eg
1068 * network cables) or protocol-dependent mechanisms (eg
1069 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1071 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1072 * struct netdev_phys_item_id *ppid);
1073 * Called to get ID of physical port of this device. If driver does
1074 * not implement this, it is assumed that the hw is not able to have
1075 * multiple net devices on single physical port.
1077 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1078 * struct udp_tunnel_info *ti);
1079 * Called by UDP tunnel to notify a driver about the UDP port and socket
1080 * address family that a UDP tunnel is listnening to. It is called only
1081 * when a new port starts listening. The operation is protected by the
1084 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1085 * struct udp_tunnel_info *ti);
1086 * Called by UDP tunnel to notify the driver about a UDP port and socket
1087 * address family that the UDP tunnel is not listening to anymore. The
1088 * operation is protected by the RTNL.
1090 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1091 * struct net_device *dev)
1092 * Called by upper layer devices to accelerate switching or other
1093 * station functionality into hardware. 'pdev is the lowerdev
1094 * to use for the offload and 'dev' is the net device that will
1095 * back the offload. Returns a pointer to the private structure
1096 * the upper layer will maintain.
1097 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1098 * Called by upper layer device to delete the station created
1099 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1100 * the station and priv is the structure returned by the add
1102 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1103 * int queue_index, u32 maxrate);
1104 * Called when a user wants to set a max-rate limitation of specific
1106 * int (*ndo_get_iflink)(const struct net_device *dev);
1107 * Called to get the iflink value of this device.
1108 * void (*ndo_change_proto_down)(struct net_device *dev,
1110 * This function is used to pass protocol port error state information
1111 * to the switch driver. The switch driver can react to the proto_down
1112 * by doing a phys down on the associated switch port.
1113 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1114 * This function is used to get egress tunnel information for given skb.
1115 * This is useful for retrieving outer tunnel header parameters while
1117 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1118 * This function is used to specify the headroom that the skb must
1119 * consider when allocation skb during packet reception. Setting
1120 * appropriate rx headroom value allows avoiding skb head copy on
1121 * forward. Setting a negative value resets the rx headroom to the
1123 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1124 * This function is used to set or query state related to XDP on the
1125 * netdevice. See definition of enum xdp_netdev_command for details.
1126 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1127 * This function is used to submit a XDP packet for transmit on a
1129 * void (*ndo_xdp_flush)(struct net_device *dev);
1130 * This function is used to inform the driver to flush a particular
1131 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1133 struct net_device_ops {
1134 int (*ndo_init)(struct net_device *dev);
1135 void (*ndo_uninit)(struct net_device *dev);
1136 int (*ndo_open)(struct net_device *dev);
1137 int (*ndo_stop)(struct net_device *dev);
1138 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1139 struct net_device *dev);
1140 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1141 struct net_device *dev,
1142 netdev_features_t features);
1143 u16 (*ndo_select_queue)(struct net_device *dev,
1144 struct sk_buff *skb,
1146 select_queue_fallback_t fallback);
1147 void (*ndo_change_rx_flags)(struct net_device *dev,
1149 void (*ndo_set_rx_mode)(struct net_device *dev);
1150 int (*ndo_set_mac_address)(struct net_device *dev,
1152 int (*ndo_validate_addr)(struct net_device *dev);
1153 int (*ndo_do_ioctl)(struct net_device *dev,
1154 struct ifreq *ifr, int cmd);
1155 int (*ndo_set_config)(struct net_device *dev,
1157 int (*ndo_change_mtu)(struct net_device *dev,
1159 int (*ndo_neigh_setup)(struct net_device *dev,
1160 struct neigh_parms *);
1161 void (*ndo_tx_timeout) (struct net_device *dev);
1163 void (*ndo_get_stats64)(struct net_device *dev,
1164 struct rtnl_link_stats64 *storage);
1165 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1166 int (*ndo_get_offload_stats)(int attr_id,
1167 const struct net_device *dev,
1169 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1171 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1172 __be16 proto, u16 vid);
1173 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1174 __be16 proto, u16 vid);
1175 #ifdef CONFIG_NET_POLL_CONTROLLER
1176 void (*ndo_poll_controller)(struct net_device *dev);
1177 int (*ndo_netpoll_setup)(struct net_device *dev,
1178 struct netpoll_info *info);
1179 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1181 int (*ndo_set_vf_mac)(struct net_device *dev,
1182 int queue, u8 *mac);
1183 int (*ndo_set_vf_vlan)(struct net_device *dev,
1184 int queue, u16 vlan,
1185 u8 qos, __be16 proto);
1186 int (*ndo_set_vf_rate)(struct net_device *dev,
1187 int vf, int min_tx_rate,
1189 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1190 int vf, bool setting);
1191 int (*ndo_set_vf_trust)(struct net_device *dev,
1192 int vf, bool setting);
1193 int (*ndo_get_vf_config)(struct net_device *dev,
1195 struct ifla_vf_info *ivf);
1196 int (*ndo_set_vf_link_state)(struct net_device *dev,
1197 int vf, int link_state);
1198 int (*ndo_get_vf_stats)(struct net_device *dev,
1200 struct ifla_vf_stats
1202 int (*ndo_set_vf_port)(struct net_device *dev,
1204 struct nlattr *port[]);
1205 int (*ndo_get_vf_port)(struct net_device *dev,
1206 int vf, struct sk_buff *skb);
1207 int (*ndo_set_vf_guid)(struct net_device *dev,
1210 int (*ndo_set_vf_rss_query_en)(
1211 struct net_device *dev,
1212 int vf, bool setting);
1213 int (*ndo_setup_tc)(struct net_device *dev,
1214 enum tc_setup_type type,
1216 #if IS_ENABLED(CONFIG_FCOE)
1217 int (*ndo_fcoe_enable)(struct net_device *dev);
1218 int (*ndo_fcoe_disable)(struct net_device *dev);
1219 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1221 struct scatterlist *sgl,
1223 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1225 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1227 struct scatterlist *sgl,
1229 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1230 struct netdev_fcoe_hbainfo *hbainfo);
1233 #if IS_ENABLED(CONFIG_LIBFCOE)
1234 #define NETDEV_FCOE_WWNN 0
1235 #define NETDEV_FCOE_WWPN 1
1236 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1237 u64 *wwn, int type);
1240 #ifdef CONFIG_RFS_ACCEL
1241 int (*ndo_rx_flow_steer)(struct net_device *dev,
1242 const struct sk_buff *skb,
1246 int (*ndo_add_slave)(struct net_device *dev,
1247 struct net_device *slave_dev);
1248 int (*ndo_del_slave)(struct net_device *dev,
1249 struct net_device *slave_dev);
1250 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1251 netdev_features_t features);
1252 int (*ndo_set_features)(struct net_device *dev,
1253 netdev_features_t features);
1254 int (*ndo_neigh_construct)(struct net_device *dev,
1255 struct neighbour *n);
1256 void (*ndo_neigh_destroy)(struct net_device *dev,
1257 struct neighbour *n);
1259 int (*ndo_fdb_add)(struct ndmsg *ndm,
1260 struct nlattr *tb[],
1261 struct net_device *dev,
1262 const unsigned char *addr,
1265 int (*ndo_fdb_del)(struct ndmsg *ndm,
1266 struct nlattr *tb[],
1267 struct net_device *dev,
1268 const unsigned char *addr,
1270 int (*ndo_fdb_dump)(struct sk_buff *skb,
1271 struct netlink_callback *cb,
1272 struct net_device *dev,
1273 struct net_device *filter_dev,
1276 int (*ndo_bridge_setlink)(struct net_device *dev,
1277 struct nlmsghdr *nlh,
1279 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1281 struct net_device *dev,
1284 int (*ndo_bridge_dellink)(struct net_device *dev,
1285 struct nlmsghdr *nlh,
1287 int (*ndo_change_carrier)(struct net_device *dev,
1289 int (*ndo_get_phys_port_id)(struct net_device *dev,
1290 struct netdev_phys_item_id *ppid);
1291 int (*ndo_get_phys_port_name)(struct net_device *dev,
1292 char *name, size_t len);
1293 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1294 struct udp_tunnel_info *ti);
1295 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1296 struct udp_tunnel_info *ti);
1297 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1298 struct net_device *dev);
1299 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1302 int (*ndo_get_lock_subclass)(struct net_device *dev);
1303 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1306 int (*ndo_get_iflink)(const struct net_device *dev);
1307 int (*ndo_change_proto_down)(struct net_device *dev,
1309 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1310 struct sk_buff *skb);
1311 void (*ndo_set_rx_headroom)(struct net_device *dev,
1312 int needed_headroom);
1313 int (*ndo_xdp)(struct net_device *dev,
1314 struct netdev_xdp *xdp);
1315 int (*ndo_xdp_xmit)(struct net_device *dev,
1316 struct xdp_buff *xdp);
1317 void (*ndo_xdp_flush)(struct net_device *dev);
1321 * enum net_device_priv_flags - &struct net_device priv_flags
1323 * These are the &struct net_device, they are only set internally
1324 * by drivers and used in the kernel. These flags are invisible to
1325 * userspace; this means that the order of these flags can change
1326 * during any kernel release.
1328 * You should have a pretty good reason to be extending these flags.
1330 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1331 * @IFF_EBRIDGE: Ethernet bridging device
1332 * @IFF_BONDING: bonding master or slave
1333 * @IFF_ISATAP: ISATAP interface (RFC4214)
1334 * @IFF_WAN_HDLC: WAN HDLC device
1335 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1337 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1338 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1339 * @IFF_MACVLAN_PORT: device used as macvlan port
1340 * @IFF_BRIDGE_PORT: device used as bridge port
1341 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1342 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1343 * @IFF_UNICAST_FLT: Supports unicast filtering
1344 * @IFF_TEAM_PORT: device used as team port
1345 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1346 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1347 * change when it's running
1348 * @IFF_MACVLAN: Macvlan device
1349 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1350 * underlying stacked devices
1351 * @IFF_IPVLAN_MASTER: IPvlan master device
1352 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1353 * @IFF_L3MDEV_MASTER: device is an L3 master device
1354 * @IFF_NO_QUEUE: device can run without qdisc attached
1355 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1356 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1357 * @IFF_TEAM: device is a team device
1358 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1359 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1360 * entity (i.e. the master device for bridged veth)
1361 * @IFF_MACSEC: device is a MACsec device
1362 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1364 enum netdev_priv_flags {
1365 IFF_802_1Q_VLAN = 1<<0,
1369 IFF_WAN_HDLC = 1<<4,
1370 IFF_XMIT_DST_RELEASE = 1<<5,
1371 IFF_DONT_BRIDGE = 1<<6,
1372 IFF_DISABLE_NETPOLL = 1<<7,
1373 IFF_MACVLAN_PORT = 1<<8,
1374 IFF_BRIDGE_PORT = 1<<9,
1375 IFF_OVS_DATAPATH = 1<<10,
1376 IFF_TX_SKB_SHARING = 1<<11,
1377 IFF_UNICAST_FLT = 1<<12,
1378 IFF_TEAM_PORT = 1<<13,
1379 IFF_SUPP_NOFCS = 1<<14,
1380 IFF_LIVE_ADDR_CHANGE = 1<<15,
1381 IFF_MACVLAN = 1<<16,
1382 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1383 IFF_IPVLAN_MASTER = 1<<18,
1384 IFF_IPVLAN_SLAVE = 1<<19,
1385 IFF_L3MDEV_MASTER = 1<<20,
1386 IFF_NO_QUEUE = 1<<21,
1387 IFF_OPENVSWITCH = 1<<22,
1388 IFF_L3MDEV_SLAVE = 1<<23,
1390 IFF_RXFH_CONFIGURED = 1<<25,
1391 IFF_PHONY_HEADROOM = 1<<26,
1393 IFF_L3MDEV_RX_HANDLER = 1<<28,
1396 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1397 #define IFF_EBRIDGE IFF_EBRIDGE
1398 #define IFF_BONDING IFF_BONDING
1399 #define IFF_ISATAP IFF_ISATAP
1400 #define IFF_WAN_HDLC IFF_WAN_HDLC
1401 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1402 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1403 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1404 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1405 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1406 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1407 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1408 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1409 #define IFF_TEAM_PORT IFF_TEAM_PORT
1410 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1411 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1412 #define IFF_MACVLAN IFF_MACVLAN
1413 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1414 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1415 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1416 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1417 #define IFF_NO_QUEUE IFF_NO_QUEUE
1418 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1419 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1420 #define IFF_TEAM IFF_TEAM
1421 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1422 #define IFF_MACSEC IFF_MACSEC
1423 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1426 * struct net_device - The DEVICE structure.
1428 * Actually, this whole structure is a big mistake. It mixes I/O
1429 * data with strictly "high-level" data, and it has to know about
1430 * almost every data structure used in the INET module.
1432 * @name: This is the first field of the "visible" part of this structure
1433 * (i.e. as seen by users in the "Space.c" file). It is the name
1436 * @name_hlist: Device name hash chain, please keep it close to name[]
1437 * @ifalias: SNMP alias
1438 * @mem_end: Shared memory end
1439 * @mem_start: Shared memory start
1440 * @base_addr: Device I/O address
1441 * @irq: Device IRQ number
1443 * @carrier_changes: Stats to monitor carrier on<->off transitions
1445 * @state: Generic network queuing layer state, see netdev_state_t
1446 * @dev_list: The global list of network devices
1447 * @napi_list: List entry used for polling NAPI devices
1448 * @unreg_list: List entry when we are unregistering the
1449 * device; see the function unregister_netdev
1450 * @close_list: List entry used when we are closing the device
1451 * @ptype_all: Device-specific packet handlers for all protocols
1452 * @ptype_specific: Device-specific, protocol-specific packet handlers
1454 * @adj_list: Directly linked devices, like slaves for bonding
1455 * @features: Currently active device features
1456 * @hw_features: User-changeable features
1458 * @wanted_features: User-requested features
1459 * @vlan_features: Mask of features inheritable by VLAN devices
1461 * @hw_enc_features: Mask of features inherited by encapsulating devices
1462 * This field indicates what encapsulation
1463 * offloads the hardware is capable of doing,
1464 * and drivers will need to set them appropriately.
1466 * @mpls_features: Mask of features inheritable by MPLS
1468 * @ifindex: interface index
1469 * @group: The group the device belongs to
1471 * @stats: Statistics struct, which was left as a legacy, use
1472 * rtnl_link_stats64 instead
1474 * @rx_dropped: Dropped packets by core network,
1475 * do not use this in drivers
1476 * @tx_dropped: Dropped packets by core network,
1477 * do not use this in drivers
1478 * @rx_nohandler: nohandler dropped packets by core network on
1479 * inactive devices, do not use this in drivers
1481 * @wireless_handlers: List of functions to handle Wireless Extensions,
1483 * see <net/iw_handler.h> for details.
1484 * @wireless_data: Instance data managed by the core of wireless extensions
1486 * @netdev_ops: Includes several pointers to callbacks,
1487 * if one wants to override the ndo_*() functions
1488 * @ethtool_ops: Management operations
1489 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1490 * discovery handling. Necessary for e.g. 6LoWPAN.
1491 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1492 * of Layer 2 headers.
1494 * @flags: Interface flags (a la BSD)
1495 * @priv_flags: Like 'flags' but invisible to userspace,
1496 * see if.h for the definitions
1497 * @gflags: Global flags ( kept as legacy )
1498 * @padded: How much padding added by alloc_netdev()
1499 * @operstate: RFC2863 operstate
1500 * @link_mode: Mapping policy to operstate
1501 * @if_port: Selectable AUI, TP, ...
1503 * @mtu: Interface MTU value
1504 * @min_mtu: Interface Minimum MTU value
1505 * @max_mtu: Interface Maximum MTU value
1506 * @type: Interface hardware type
1507 * @hard_header_len: Maximum hardware header length.
1508 * @min_header_len: Minimum hardware header length
1510 * @needed_headroom: Extra headroom the hardware may need, but not in all
1511 * cases can this be guaranteed
1512 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1513 * cases can this be guaranteed. Some cases also use
1514 * LL_MAX_HEADER instead to allocate the skb
1516 * interface address info:
1518 * @perm_addr: Permanent hw address
1519 * @addr_assign_type: Hw address assignment type
1520 * @addr_len: Hardware address length
1521 * @neigh_priv_len: Used in neigh_alloc()
1522 * @dev_id: Used to differentiate devices that share
1523 * the same link layer address
1524 * @dev_port: Used to differentiate devices that share
1526 * @addr_list_lock: XXX: need comments on this one
1527 * @uc_promisc: Counter that indicates promiscuous mode
1528 * has been enabled due to the need to listen to
1529 * additional unicast addresses in a device that
1530 * does not implement ndo_set_rx_mode()
1531 * @uc: unicast mac addresses
1532 * @mc: multicast mac addresses
1533 * @dev_addrs: list of device hw addresses
1534 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1535 * @promiscuity: Number of times the NIC is told to work in
1536 * promiscuous mode; if it becomes 0 the NIC will
1537 * exit promiscuous mode
1538 * @allmulti: Counter, enables or disables allmulticast mode
1540 * @vlan_info: VLAN info
1541 * @dsa_ptr: dsa specific data
1542 * @tipc_ptr: TIPC specific data
1543 * @atalk_ptr: AppleTalk link
1544 * @ip_ptr: IPv4 specific data
1545 * @dn_ptr: DECnet specific data
1546 * @ip6_ptr: IPv6 specific data
1547 * @ax25_ptr: AX.25 specific data
1548 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1550 * @dev_addr: Hw address (before bcast,
1551 * because most packets are unicast)
1553 * @_rx: Array of RX queues
1554 * @num_rx_queues: Number of RX queues
1555 * allocated at register_netdev() time
1556 * @real_num_rx_queues: Number of RX queues currently active in device
1558 * @rx_handler: handler for received packets
1559 * @rx_handler_data: XXX: need comments on this one
1560 * @ingress_queue: XXX: need comments on this one
1561 * @broadcast: hw bcast address
1563 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1564 * indexed by RX queue number. Assigned by driver.
1565 * This must only be set if the ndo_rx_flow_steer
1566 * operation is defined
1567 * @index_hlist: Device index hash chain
1569 * @_tx: Array of TX queues
1570 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1571 * @real_num_tx_queues: Number of TX queues currently active in device
1572 * @qdisc: Root qdisc from userspace point of view
1573 * @tx_queue_len: Max frames per queue allowed
1574 * @tx_global_lock: XXX: need comments on this one
1576 * @xps_maps: XXX: need comments on this one
1578 * @watchdog_timeo: Represents the timeout that is used by
1579 * the watchdog (see dev_watchdog())
1580 * @watchdog_timer: List of timers
1582 * @pcpu_refcnt: Number of references to this device
1583 * @todo_list: Delayed register/unregister
1584 * @link_watch_list: XXX: need comments on this one
1586 * @reg_state: Register/unregister state machine
1587 * @dismantle: Device is going to be freed
1588 * @rtnl_link_state: This enum represents the phases of creating
1591 * @needs_free_netdev: Should unregister perform free_netdev?
1592 * @priv_destructor: Called from unregister
1593 * @npinfo: XXX: need comments on this one
1594 * @nd_net: Network namespace this network device is inside
1596 * @ml_priv: Mid-layer private
1597 * @lstats: Loopback statistics
1598 * @tstats: Tunnel statistics
1599 * @dstats: Dummy statistics
1600 * @vstats: Virtual ethernet statistics
1605 * @dev: Class/net/name entry
1606 * @sysfs_groups: Space for optional device, statistics and wireless
1609 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1610 * @rtnl_link_ops: Rtnl_link_ops
1612 * @gso_max_size: Maximum size of generic segmentation offload
1613 * @gso_max_segs: Maximum number of segments that can be passed to the
1616 * @dcbnl_ops: Data Center Bridging netlink ops
1617 * @num_tc: Number of traffic classes in the net device
1618 * @tc_to_txq: XXX: need comments on this one
1619 * @prio_tc_map: XXX: need comments on this one
1621 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1623 * @priomap: XXX: need comments on this one
1624 * @phydev: Physical device may attach itself
1625 * for hardware timestamping
1627 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1628 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1630 * @proto_down: protocol port state information can be sent to the
1631 * switch driver and used to set the phys state of the
1634 * FIXME: cleanup struct net_device such that network protocol info
1639 char name[IFNAMSIZ];
1640 struct hlist_node name_hlist;
1643 * I/O specific fields
1644 * FIXME: Merge these and struct ifmap into one
1646 unsigned long mem_end;
1647 unsigned long mem_start;
1648 unsigned long base_addr;
1651 atomic_t carrier_changes;
1654 * Some hardware also needs these fields (state,dev_list,
1655 * napi_list,unreg_list,close_list) but they are not
1656 * part of the usual set specified in Space.c.
1659 unsigned long state;
1661 struct list_head dev_list;
1662 struct list_head napi_list;
1663 struct list_head unreg_list;
1664 struct list_head close_list;
1665 struct list_head ptype_all;
1666 struct list_head ptype_specific;
1669 struct list_head upper;
1670 struct list_head lower;
1673 netdev_features_t features;
1674 netdev_features_t hw_features;
1675 netdev_features_t wanted_features;
1676 netdev_features_t vlan_features;
1677 netdev_features_t hw_enc_features;
1678 netdev_features_t mpls_features;
1679 netdev_features_t gso_partial_features;
1684 struct net_device_stats stats;
1686 atomic_long_t rx_dropped;
1687 atomic_long_t tx_dropped;
1688 atomic_long_t rx_nohandler;
1690 #ifdef CONFIG_WIRELESS_EXT
1691 const struct iw_handler_def *wireless_handlers;
1692 struct iw_public_data *wireless_data;
1694 const struct net_device_ops *netdev_ops;
1695 const struct ethtool_ops *ethtool_ops;
1696 #ifdef CONFIG_NET_SWITCHDEV
1697 const struct switchdev_ops *switchdev_ops;
1699 #ifdef CONFIG_NET_L3_MASTER_DEV
1700 const struct l3mdev_ops *l3mdev_ops;
1702 #if IS_ENABLED(CONFIG_IPV6)
1703 const struct ndisc_ops *ndisc_ops;
1707 const struct xfrmdev_ops *xfrmdev_ops;
1710 const struct header_ops *header_ops;
1713 unsigned int priv_flags;
1715 unsigned short gflags;
1716 unsigned short padded;
1718 unsigned char operstate;
1719 unsigned char link_mode;
1721 unsigned char if_port;
1724 /* Note : dev->mtu is often read without holding a lock.
1725 * Writers usually hold RTNL.
1726 * It is recommended to use READ_ONCE() to annotate the reads,
1727 * and to use WRITE_ONCE() to annotate the writes.
1730 unsigned int min_mtu;
1731 unsigned int max_mtu;
1732 unsigned short type;
1733 unsigned short hard_header_len;
1734 unsigned char min_header_len;
1736 unsigned short needed_headroom;
1737 unsigned short needed_tailroom;
1739 /* Interface address info. */
1740 unsigned char perm_addr[MAX_ADDR_LEN];
1741 unsigned char addr_assign_type;
1742 unsigned char addr_len;
1743 unsigned short neigh_priv_len;
1744 unsigned short dev_id;
1745 unsigned short dev_port;
1746 spinlock_t addr_list_lock;
1747 unsigned char name_assign_type;
1749 struct netdev_hw_addr_list uc;
1750 struct netdev_hw_addr_list mc;
1751 struct netdev_hw_addr_list dev_addrs;
1754 struct kset *queues_kset;
1756 unsigned int promiscuity;
1757 unsigned int allmulti;
1760 /* Protocol-specific pointers */
1762 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1763 struct vlan_info __rcu *vlan_info;
1765 #if IS_ENABLED(CONFIG_NET_DSA)
1766 struct dsa_switch_tree *dsa_ptr;
1768 #if IS_ENABLED(CONFIG_TIPC)
1769 struct tipc_bearer __rcu *tipc_ptr;
1772 struct in_device __rcu *ip_ptr;
1773 struct dn_dev __rcu *dn_ptr;
1774 struct inet6_dev __rcu *ip6_ptr;
1776 struct wireless_dev *ieee80211_ptr;
1777 struct wpan_dev *ieee802154_ptr;
1778 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1779 struct mpls_dev __rcu *mpls_ptr;
1783 * Cache lines mostly used on receive path (including eth_type_trans())
1785 /* Interface address info used in eth_type_trans() */
1786 unsigned char *dev_addr;
1789 struct netdev_rx_queue *_rx;
1791 unsigned int num_rx_queues;
1792 unsigned int real_num_rx_queues;
1795 struct bpf_prog __rcu *xdp_prog;
1796 unsigned long gro_flush_timeout;
1797 rx_handler_func_t __rcu *rx_handler;
1798 void __rcu *rx_handler_data;
1800 #ifdef CONFIG_NET_CLS_ACT
1801 struct tcf_proto __rcu *ingress_cl_list;
1803 struct netdev_queue __rcu *ingress_queue;
1804 #ifdef CONFIG_NETFILTER_INGRESS
1805 struct nf_hook_entries __rcu *nf_hooks_ingress;
1808 unsigned char broadcast[MAX_ADDR_LEN];
1809 #ifdef CONFIG_RFS_ACCEL
1810 struct cpu_rmap *rx_cpu_rmap;
1812 struct hlist_node index_hlist;
1815 * Cache lines mostly used on transmit path
1817 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1818 unsigned int num_tx_queues;
1819 unsigned int real_num_tx_queues;
1820 struct Qdisc *qdisc;
1821 #ifdef CONFIG_NET_SCHED
1822 DECLARE_HASHTABLE (qdisc_hash, 4);
1824 unsigned int tx_queue_len;
1825 spinlock_t tx_global_lock;
1829 struct xps_dev_maps __rcu *xps_maps;
1831 #ifdef CONFIG_NET_CLS_ACT
1832 struct tcf_proto __rcu *egress_cl_list;
1835 /* These may be needed for future network-power-down code. */
1836 struct timer_list watchdog_timer;
1838 int __percpu *pcpu_refcnt;
1839 struct list_head todo_list;
1841 struct list_head link_watch_list;
1843 enum { NETREG_UNINITIALIZED=0,
1844 NETREG_REGISTERED, /* completed register_netdevice */
1845 NETREG_UNREGISTERING, /* called unregister_netdevice */
1846 NETREG_UNREGISTERED, /* completed unregister todo */
1847 NETREG_RELEASED, /* called free_netdev */
1848 NETREG_DUMMY, /* dummy device for NAPI poll */
1854 RTNL_LINK_INITIALIZED,
1855 RTNL_LINK_INITIALIZING,
1856 } rtnl_link_state:16;
1858 bool needs_free_netdev;
1859 void (*priv_destructor)(struct net_device *dev);
1861 #ifdef CONFIG_NETPOLL
1862 struct netpoll_info __rcu *npinfo;
1865 possible_net_t nd_net;
1867 /* mid-layer private */
1870 struct pcpu_lstats __percpu *lstats;
1871 struct pcpu_sw_netstats __percpu *tstats;
1872 struct pcpu_dstats __percpu *dstats;
1873 struct pcpu_vstats __percpu *vstats;
1876 #if IS_ENABLED(CONFIG_GARP)
1877 struct garp_port __rcu *garp_port;
1879 #if IS_ENABLED(CONFIG_MRP)
1880 struct mrp_port __rcu *mrp_port;
1884 const struct attribute_group *sysfs_groups[4];
1885 const struct attribute_group *sysfs_rx_queue_group;
1887 const struct rtnl_link_ops *rtnl_link_ops;
1889 /* for setting kernel sock attribute on TCP connection setup */
1890 #define GSO_MAX_SIZE 65536
1891 unsigned int gso_max_size;
1892 #define GSO_MAX_SEGS 65535
1896 const struct dcbnl_rtnl_ops *dcbnl_ops;
1899 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1900 u8 prio_tc_map[TC_BITMASK + 1];
1902 #if IS_ENABLED(CONFIG_FCOE)
1903 unsigned int fcoe_ddp_xid;
1905 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1906 struct netprio_map __rcu *priomap;
1908 struct phy_device *phydev;
1909 struct lock_class_key *qdisc_tx_busylock;
1910 struct lock_class_key *qdisc_running_key;
1913 #define to_net_dev(d) container_of(d, struct net_device, dev)
1915 static inline bool netif_elide_gro(const struct net_device *dev)
1917 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1922 #define NETDEV_ALIGN 32
1925 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1927 return dev->prio_tc_map[prio & TC_BITMASK];
1931 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1933 if (tc >= dev->num_tc)
1936 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1940 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1941 void netdev_reset_tc(struct net_device *dev);
1942 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1943 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1946 int netdev_get_num_tc(struct net_device *dev)
1952 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1955 return &dev->_tx[index];
1958 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1959 const struct sk_buff *skb)
1961 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1964 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1965 void (*f)(struct net_device *,
1966 struct netdev_queue *,
1972 for (i = 0; i < dev->num_tx_queues; i++)
1973 f(dev, &dev->_tx[i], arg);
1976 #define netdev_lockdep_set_classes(dev) \
1978 static struct lock_class_key qdisc_tx_busylock_key; \
1979 static struct lock_class_key qdisc_running_key; \
1980 static struct lock_class_key qdisc_xmit_lock_key; \
1981 static struct lock_class_key dev_addr_list_lock_key; \
1984 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1985 (dev)->qdisc_running_key = &qdisc_running_key; \
1986 lockdep_set_class(&(dev)->addr_list_lock, \
1987 &dev_addr_list_lock_key); \
1988 for (i = 0; i < (dev)->num_tx_queues; i++) \
1989 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1990 &qdisc_xmit_lock_key); \
1993 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1994 struct sk_buff *skb,
1997 /* returns the headroom that the master device needs to take in account
1998 * when forwarding to this dev
2000 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2002 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2005 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2007 if (dev->netdev_ops->ndo_set_rx_headroom)
2008 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2011 /* set the device rx headroom to the dev's default */
2012 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2014 netdev_set_rx_headroom(dev, -1);
2018 * Net namespace inlines
2021 struct net *dev_net(const struct net_device *dev)
2023 return read_pnet(&dev->nd_net);
2027 void dev_net_set(struct net_device *dev, struct net *net)
2029 write_pnet(&dev->nd_net, net);
2033 * netdev_priv - access network device private data
2034 * @dev: network device
2036 * Get network device private data
2038 static inline void *netdev_priv(const struct net_device *dev)
2040 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2043 /* Set the sysfs physical device reference for the network logical device
2044 * if set prior to registration will cause a symlink during initialization.
2046 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2048 /* Set the sysfs device type for the network logical device to allow
2049 * fine-grained identification of different network device types. For
2050 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2052 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2054 /* Default NAPI poll() weight
2055 * Device drivers are strongly advised to not use bigger value
2057 #define NAPI_POLL_WEIGHT 64
2060 * netif_napi_add - initialize a NAPI context
2061 * @dev: network device
2062 * @napi: NAPI context
2063 * @poll: polling function
2064 * @weight: default weight
2066 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2067 * *any* of the other NAPI-related functions.
2069 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2070 int (*poll)(struct napi_struct *, int), int weight);
2073 * netif_tx_napi_add - initialize a NAPI context
2074 * @dev: network device
2075 * @napi: NAPI context
2076 * @poll: polling function
2077 * @weight: default weight
2079 * This variant of netif_napi_add() should be used from drivers using NAPI
2080 * to exclusively poll a TX queue.
2081 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2083 static inline void netif_tx_napi_add(struct net_device *dev,
2084 struct napi_struct *napi,
2085 int (*poll)(struct napi_struct *, int),
2088 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2089 netif_napi_add(dev, napi, poll, weight);
2093 * netif_napi_del - remove a NAPI context
2094 * @napi: NAPI context
2096 * netif_napi_del() removes a NAPI context from the network device NAPI list
2098 void netif_napi_del(struct napi_struct *napi);
2100 struct napi_gro_cb {
2101 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2104 /* Length of frag0. */
2105 unsigned int frag0_len;
2107 /* This indicates where we are processing relative to skb->data. */
2110 /* This is non-zero if the packet cannot be merged with the new skb. */
2113 /* Save the IP ID here and check when we get to the transport layer */
2116 /* Number of segments aggregated. */
2119 /* Start offset for remote checksum offload */
2120 u16 gro_remcsum_start;
2122 /* jiffies when first packet was created/queued */
2125 /* Used in ipv6_gro_receive() and foo-over-udp */
2128 /* This is non-zero if the packet may be of the same flow. */
2131 /* Used in tunnel GRO receive */
2134 /* GRO checksum is valid */
2137 /* Number of checksums via CHECKSUM_UNNECESSARY */
2142 #define NAPI_GRO_FREE 1
2143 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2145 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2148 /* Used in GRE, set in fou/gue_gro_receive */
2151 /* Used to determine if flush_id can be ignored */
2154 /* Number of gro_receive callbacks this packet already went through */
2155 u8 recursion_counter:4;
2159 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2162 /* used in skb_gro_receive() slow path */
2163 struct sk_buff *last;
2166 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2168 #define GRO_RECURSION_LIMIT 15
2169 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2171 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2174 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2175 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2176 struct sk_buff **head,
2177 struct sk_buff *skb)
2179 if (unlikely(gro_recursion_inc_test(skb))) {
2180 NAPI_GRO_CB(skb)->flush |= 1;
2184 return cb(head, skb);
2187 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2189 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2191 struct sk_buff **head,
2192 struct sk_buff *skb)
2194 if (unlikely(gro_recursion_inc_test(skb))) {
2195 NAPI_GRO_CB(skb)->flush |= 1;
2199 return cb(sk, head, skb);
2202 struct packet_type {
2203 __be16 type; /* This is really htons(ether_type). */
2204 struct net_device *dev; /* NULL is wildcarded here */
2205 int (*func) (struct sk_buff *,
2206 struct net_device *,
2207 struct packet_type *,
2208 struct net_device *);
2209 bool (*id_match)(struct packet_type *ptype,
2211 struct net *af_packet_net;
2212 void *af_packet_priv;
2213 struct list_head list;
2216 struct offload_callbacks {
2217 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2218 netdev_features_t features);
2219 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2220 struct sk_buff *skb);
2221 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2224 struct packet_offload {
2225 __be16 type; /* This is really htons(ether_type). */
2227 struct offload_callbacks callbacks;
2228 struct list_head list;
2231 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2232 struct pcpu_sw_netstats {
2237 struct u64_stats_sync syncp;
2240 #define __netdev_alloc_pcpu_stats(type, gfp) \
2242 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2245 for_each_possible_cpu(__cpu) { \
2246 typeof(type) *stat; \
2247 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2248 u64_stats_init(&stat->syncp); \
2254 #define netdev_alloc_pcpu_stats(type) \
2255 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2257 enum netdev_lag_tx_type {
2258 NETDEV_LAG_TX_TYPE_UNKNOWN,
2259 NETDEV_LAG_TX_TYPE_RANDOM,
2260 NETDEV_LAG_TX_TYPE_BROADCAST,
2261 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2262 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2263 NETDEV_LAG_TX_TYPE_HASH,
2266 struct netdev_lag_upper_info {
2267 enum netdev_lag_tx_type tx_type;
2270 struct netdev_lag_lower_state_info {
2275 #include <linux/notifier.h>
2277 /* netdevice notifier chain. Please remember to update the rtnetlink
2278 * notification exclusion list in rtnetlink_event() when adding new
2281 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2282 #define NETDEV_DOWN 0x0002
2283 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2284 detected a hardware crash and restarted
2285 - we can use this eg to kick tcp sessions
2287 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2288 #define NETDEV_REGISTER 0x0005
2289 #define NETDEV_UNREGISTER 0x0006
2290 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2291 #define NETDEV_CHANGEADDR 0x0008
2292 #define NETDEV_GOING_DOWN 0x0009
2293 #define NETDEV_CHANGENAME 0x000A
2294 #define NETDEV_FEAT_CHANGE 0x000B
2295 #define NETDEV_BONDING_FAILOVER 0x000C
2296 #define NETDEV_PRE_UP 0x000D
2297 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2298 #define NETDEV_POST_TYPE_CHANGE 0x000F
2299 #define NETDEV_POST_INIT 0x0010
2300 #define NETDEV_UNREGISTER_FINAL 0x0011
2301 #define NETDEV_RELEASE 0x0012
2302 #define NETDEV_NOTIFY_PEERS 0x0013
2303 #define NETDEV_JOIN 0x0014
2304 #define NETDEV_CHANGEUPPER 0x0015
2305 #define NETDEV_RESEND_IGMP 0x0016
2306 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2307 #define NETDEV_CHANGEINFODATA 0x0018
2308 #define NETDEV_BONDING_INFO 0x0019
2309 #define NETDEV_PRECHANGEUPPER 0x001A
2310 #define NETDEV_CHANGELOWERSTATE 0x001B
2311 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2312 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2313 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2315 int register_netdevice_notifier(struct notifier_block *nb);
2316 int unregister_netdevice_notifier(struct notifier_block *nb);
2318 struct netdev_notifier_info {
2319 struct net_device *dev;
2322 struct netdev_notifier_info_ext {
2323 struct netdev_notifier_info info; /* must be first */
2329 struct netdev_notifier_change_info {
2330 struct netdev_notifier_info info; /* must be first */
2331 unsigned int flags_changed;
2334 struct netdev_notifier_changeupper_info {
2335 struct netdev_notifier_info info; /* must be first */
2336 struct net_device *upper_dev; /* new upper dev */
2337 bool master; /* is upper dev master */
2338 bool linking; /* is the notification for link or unlink */
2339 void *upper_info; /* upper dev info */
2342 struct netdev_notifier_changelowerstate_info {
2343 struct netdev_notifier_info info; /* must be first */
2344 void *lower_state_info; /* is lower dev state */
2347 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2348 struct net_device *dev)
2353 static inline struct net_device *
2354 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2359 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2362 extern rwlock_t dev_base_lock; /* Device list lock */
2364 #define for_each_netdev(net, d) \
2365 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2366 #define for_each_netdev_reverse(net, d) \
2367 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2368 #define for_each_netdev_rcu(net, d) \
2369 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2370 #define for_each_netdev_safe(net, d, n) \
2371 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2372 #define for_each_netdev_continue(net, d) \
2373 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2374 #define for_each_netdev_continue_rcu(net, d) \
2375 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2376 #define for_each_netdev_in_bond_rcu(bond, slave) \
2377 for_each_netdev_rcu(&init_net, slave) \
2378 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2379 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2381 static inline struct net_device *next_net_device(struct net_device *dev)
2383 struct list_head *lh;
2387 lh = dev->dev_list.next;
2388 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2391 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2393 struct list_head *lh;
2397 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2398 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2401 static inline struct net_device *first_net_device(struct net *net)
2403 return list_empty(&net->dev_base_head) ? NULL :
2404 net_device_entry(net->dev_base_head.next);
2407 static inline struct net_device *first_net_device_rcu(struct net *net)
2409 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2411 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2414 int netdev_boot_setup_check(struct net_device *dev);
2415 unsigned long netdev_boot_base(const char *prefix, int unit);
2416 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2417 const char *hwaddr);
2418 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2419 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2420 void dev_add_pack(struct packet_type *pt);
2421 void dev_remove_pack(struct packet_type *pt);
2422 void __dev_remove_pack(struct packet_type *pt);
2423 void dev_add_offload(struct packet_offload *po);
2424 void dev_remove_offload(struct packet_offload *po);
2426 int dev_get_iflink(const struct net_device *dev);
2427 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2428 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2429 unsigned short mask);
2430 struct net_device *dev_get_by_name(struct net *net, const char *name);
2431 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2432 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2433 int dev_alloc_name(struct net_device *dev, const char *name);
2434 int dev_open(struct net_device *dev);
2435 void dev_close(struct net_device *dev);
2436 void dev_close_many(struct list_head *head, bool unlink);
2437 void dev_disable_lro(struct net_device *dev);
2438 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2439 int dev_queue_xmit(struct sk_buff *skb);
2440 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2441 int register_netdevice(struct net_device *dev);
2442 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2443 void unregister_netdevice_many(struct list_head *head);
2444 static inline void unregister_netdevice(struct net_device *dev)
2446 unregister_netdevice_queue(dev, NULL);
2449 int netdev_refcnt_read(const struct net_device *dev);
2450 void free_netdev(struct net_device *dev);
2451 void netdev_freemem(struct net_device *dev);
2452 void synchronize_net(void);
2453 int init_dummy_netdev(struct net_device *dev);
2455 DECLARE_PER_CPU(int, xmit_recursion);
2456 #define XMIT_RECURSION_LIMIT 8
2458 static inline int dev_recursion_level(void)
2460 return this_cpu_read(xmit_recursion);
2463 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2464 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2465 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2466 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2467 int netdev_get_name(struct net *net, char *name, int ifindex);
2468 int dev_restart(struct net_device *dev);
2469 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2471 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2473 return NAPI_GRO_CB(skb)->data_offset;
2476 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2478 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2481 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2483 NAPI_GRO_CB(skb)->data_offset += len;
2486 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2487 unsigned int offset)
2489 return NAPI_GRO_CB(skb)->frag0 + offset;
2492 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2494 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2497 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2499 NAPI_GRO_CB(skb)->frag0 = NULL;
2500 NAPI_GRO_CB(skb)->frag0_len = 0;
2503 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2504 unsigned int offset)
2506 if (!pskb_may_pull(skb, hlen))
2509 skb_gro_frag0_invalidate(skb);
2510 return skb->data + offset;
2513 static inline void *skb_gro_network_header(struct sk_buff *skb)
2515 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2516 skb_network_offset(skb);
2519 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2520 const void *start, unsigned int len)
2522 if (NAPI_GRO_CB(skb)->csum_valid)
2523 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2524 csum_partial(start, len, 0));
2527 /* GRO checksum functions. These are logical equivalents of the normal
2528 * checksum functions (in skbuff.h) except that they operate on the GRO
2529 * offsets and fields in sk_buff.
2532 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2534 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2536 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2539 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2543 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2544 skb_checksum_start_offset(skb) <
2545 skb_gro_offset(skb)) &&
2546 !skb_at_gro_remcsum_start(skb) &&
2547 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2548 (!zero_okay || check));
2551 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2554 if (NAPI_GRO_CB(skb)->csum_valid &&
2555 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2558 NAPI_GRO_CB(skb)->csum = psum;
2560 return __skb_gro_checksum_complete(skb);
2563 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2565 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2566 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2567 NAPI_GRO_CB(skb)->csum_cnt--;
2569 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2570 * verified a new top level checksum or an encapsulated one
2571 * during GRO. This saves work if we fallback to normal path.
2573 __skb_incr_checksum_unnecessary(skb);
2577 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2580 __sum16 __ret = 0; \
2581 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2582 __ret = __skb_gro_checksum_validate_complete(skb, \
2583 compute_pseudo(skb, proto)); \
2585 skb_gro_incr_csum_unnecessary(skb); \
2589 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2590 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2592 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2594 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2596 #define skb_gro_checksum_simple_validate(skb) \
2597 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2599 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2601 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2602 !NAPI_GRO_CB(skb)->csum_valid);
2605 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2606 __sum16 check, __wsum pseudo)
2608 NAPI_GRO_CB(skb)->csum = ~pseudo;
2609 NAPI_GRO_CB(skb)->csum_valid = 1;
2612 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2614 if (__skb_gro_checksum_convert_check(skb)) \
2615 __skb_gro_checksum_convert(skb, check, \
2616 compute_pseudo(skb, proto)); \
2619 struct gro_remcsum {
2624 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2630 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2631 unsigned int off, size_t hdrlen,
2632 int start, int offset,
2633 struct gro_remcsum *grc,
2637 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2639 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2642 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2646 ptr = skb_gro_header_fast(skb, off);
2647 if (skb_gro_header_hard(skb, off + plen)) {
2648 ptr = skb_gro_header_slow(skb, off + plen, off);
2653 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2656 /* Adjust skb->csum since we changed the packet */
2657 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2659 grc->offset = off + hdrlen + offset;
2665 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2666 struct gro_remcsum *grc)
2669 size_t plen = grc->offset + sizeof(u16);
2674 ptr = skb_gro_header_fast(skb, grc->offset);
2675 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2676 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2681 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2684 #ifdef CONFIG_XFRM_OFFLOAD
2685 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2687 if (PTR_ERR(pp) != -EINPROGRESS)
2688 NAPI_GRO_CB(skb)->flush |= flush;
2690 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2691 struct sk_buff **pp,
2693 struct gro_remcsum *grc)
2695 if (PTR_ERR(pp) != -EINPROGRESS) {
2696 NAPI_GRO_CB(skb)->flush |= flush;
2697 skb_gro_remcsum_cleanup(skb, grc);
2698 skb->remcsum_offload = 0;
2702 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2704 NAPI_GRO_CB(skb)->flush |= flush;
2706 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2707 struct sk_buff **pp,
2709 struct gro_remcsum *grc)
2711 NAPI_GRO_CB(skb)->flush |= flush;
2712 skb_gro_remcsum_cleanup(skb, grc);
2713 skb->remcsum_offload = 0;
2717 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2718 unsigned short type,
2719 const void *daddr, const void *saddr,
2722 if (!dev->header_ops || !dev->header_ops->create)
2725 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2728 static inline int dev_parse_header(const struct sk_buff *skb,
2729 unsigned char *haddr)
2731 const struct net_device *dev = skb->dev;
2733 if (!dev->header_ops || !dev->header_ops->parse)
2735 return dev->header_ops->parse(skb, haddr);
2738 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2740 const struct net_device *dev = skb->dev;
2742 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2744 return dev->header_ops->parse_protocol(skb);
2747 /* ll_header must have at least hard_header_len allocated */
2748 static inline bool dev_validate_header(const struct net_device *dev,
2749 char *ll_header, int len)
2751 if (likely(len >= dev->hard_header_len))
2753 if (len < dev->min_header_len)
2756 if (capable(CAP_SYS_RAWIO)) {
2757 memset(ll_header + len, 0, dev->hard_header_len - len);
2761 if (dev->header_ops && dev->header_ops->validate)
2762 return dev->header_ops->validate(ll_header, len);
2767 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2768 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2769 static inline int unregister_gifconf(unsigned int family)
2771 return register_gifconf(family, NULL);
2774 #ifdef CONFIG_NET_FLOW_LIMIT
2775 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2776 struct sd_flow_limit {
2778 unsigned int num_buckets;
2779 unsigned int history_head;
2780 u16 history[FLOW_LIMIT_HISTORY];
2784 extern int netdev_flow_limit_table_len;
2785 #endif /* CONFIG_NET_FLOW_LIMIT */
2788 * Incoming packets are placed on per-CPU queues
2790 struct softnet_data {
2791 struct list_head poll_list;
2792 struct sk_buff_head process_queue;
2795 unsigned int processed;
2796 unsigned int time_squeeze;
2797 unsigned int received_rps;
2799 struct softnet_data *rps_ipi_list;
2801 #ifdef CONFIG_NET_FLOW_LIMIT
2802 struct sd_flow_limit __rcu *flow_limit;
2804 struct Qdisc *output_queue;
2805 struct Qdisc **output_queue_tailp;
2806 struct sk_buff *completion_queue;
2809 /* input_queue_head should be written by cpu owning this struct,
2810 * and only read by other cpus. Worth using a cache line.
2812 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2814 /* Elements below can be accessed between CPUs for RPS/RFS */
2815 call_single_data_t csd ____cacheline_aligned_in_smp;
2816 struct softnet_data *rps_ipi_next;
2818 unsigned int input_queue_tail;
2820 unsigned int dropped;
2821 struct sk_buff_head input_pkt_queue;
2822 struct napi_struct backlog;
2826 static inline void input_queue_head_incr(struct softnet_data *sd)
2829 sd->input_queue_head++;
2833 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2834 unsigned int *qtail)
2837 *qtail = ++sd->input_queue_tail;
2841 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2843 void __netif_schedule(struct Qdisc *q);
2844 void netif_schedule_queue(struct netdev_queue *txq);
2846 static inline void netif_tx_schedule_all(struct net_device *dev)
2850 for (i = 0; i < dev->num_tx_queues; i++)
2851 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2854 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2856 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2860 * netif_start_queue - allow transmit
2861 * @dev: network device
2863 * Allow upper layers to call the device hard_start_xmit routine.
2865 static inline void netif_start_queue(struct net_device *dev)
2867 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2870 static inline void netif_tx_start_all_queues(struct net_device *dev)
2874 for (i = 0; i < dev->num_tx_queues; i++) {
2875 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2876 netif_tx_start_queue(txq);
2880 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2883 * netif_wake_queue - restart transmit
2884 * @dev: network device
2886 * Allow upper layers to call the device hard_start_xmit routine.
2887 * Used for flow control when transmit resources are available.
2889 static inline void netif_wake_queue(struct net_device *dev)
2891 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2894 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2898 for (i = 0; i < dev->num_tx_queues; i++) {
2899 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2900 netif_tx_wake_queue(txq);
2904 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2906 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2910 * netif_stop_queue - stop transmitted packets
2911 * @dev: network device
2913 * Stop upper layers calling the device hard_start_xmit routine.
2914 * Used for flow control when transmit resources are unavailable.
2916 static inline void netif_stop_queue(struct net_device *dev)
2918 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2921 void netif_tx_stop_all_queues(struct net_device *dev);
2923 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2925 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2929 * netif_queue_stopped - test if transmit queue is flowblocked
2930 * @dev: network device
2932 * Test if transmit queue on device is currently unable to send.
2934 static inline bool netif_queue_stopped(const struct net_device *dev)
2936 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2939 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2941 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2945 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2947 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2951 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2953 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2957 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2958 * @dev_queue: pointer to transmit queue
2960 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2961 * to give appropriate hint to the CPU.
2963 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2966 prefetchw(&dev_queue->dql.num_queued);
2971 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2972 * @dev_queue: pointer to transmit queue
2974 * BQL enabled drivers might use this helper in their TX completion path,
2975 * to give appropriate hint to the CPU.
2977 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2980 prefetchw(&dev_queue->dql.limit);
2984 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2988 dql_queued(&dev_queue->dql, bytes);
2990 if (likely(dql_avail(&dev_queue->dql) >= 0))
2993 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2996 * The XOFF flag must be set before checking the dql_avail below,
2997 * because in netdev_tx_completed_queue we update the dql_completed
2998 * before checking the XOFF flag.
3002 /* check again in case another CPU has just made room avail */
3003 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3004 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3009 * netdev_sent_queue - report the number of bytes queued to hardware
3010 * @dev: network device
3011 * @bytes: number of bytes queued to the hardware device queue
3013 * Report the number of bytes queued for sending/completion to the network
3014 * device hardware queue. @bytes should be a good approximation and should
3015 * exactly match netdev_completed_queue() @bytes
3017 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3019 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3022 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3023 unsigned int pkts, unsigned int bytes)
3026 if (unlikely(!bytes))
3029 dql_completed(&dev_queue->dql, bytes);
3032 * Without the memory barrier there is a small possiblity that
3033 * netdev_tx_sent_queue will miss the update and cause the queue to
3034 * be stopped forever
3038 if (dql_avail(&dev_queue->dql) < 0)
3041 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3042 netif_schedule_queue(dev_queue);
3047 * netdev_completed_queue - report bytes and packets completed by device
3048 * @dev: network device
3049 * @pkts: actual number of packets sent over the medium
3050 * @bytes: actual number of bytes sent over the medium
3052 * Report the number of bytes and packets transmitted by the network device
3053 * hardware queue over the physical medium, @bytes must exactly match the
3054 * @bytes amount passed to netdev_sent_queue()
3056 static inline void netdev_completed_queue(struct net_device *dev,
3057 unsigned int pkts, unsigned int bytes)
3059 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3062 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3065 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3071 * netdev_reset_queue - reset the packets and bytes count of a network device
3072 * @dev_queue: network device
3074 * Reset the bytes and packet count of a network device and clear the
3075 * software flow control OFF bit for this network device
3077 static inline void netdev_reset_queue(struct net_device *dev_queue)
3079 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3083 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3084 * @dev: network device
3085 * @queue_index: given tx queue index
3087 * Returns 0 if given tx queue index >= number of device tx queues,
3088 * otherwise returns the originally passed tx queue index.
3090 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3092 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3093 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3094 dev->name, queue_index,
3095 dev->real_num_tx_queues);
3103 * netif_running - test if up
3104 * @dev: network device
3106 * Test if the device has been brought up.
3108 static inline bool netif_running(const struct net_device *dev)
3110 return test_bit(__LINK_STATE_START, &dev->state);
3114 * Routines to manage the subqueues on a device. We only need start,
3115 * stop, and a check if it's stopped. All other device management is
3116 * done at the overall netdevice level.
3117 * Also test the device if we're multiqueue.
3121 * netif_start_subqueue - allow sending packets on subqueue
3122 * @dev: network device
3123 * @queue_index: sub queue index
3125 * Start individual transmit queue of a device with multiple transmit queues.
3127 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3129 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3131 netif_tx_start_queue(txq);
3135 * netif_stop_subqueue - stop sending packets on subqueue
3136 * @dev: network device
3137 * @queue_index: sub queue index
3139 * Stop individual transmit queue of a device with multiple transmit queues.
3141 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3143 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3144 netif_tx_stop_queue(txq);
3148 * netif_subqueue_stopped - test status of subqueue
3149 * @dev: network device
3150 * @queue_index: sub queue index
3152 * Check individual transmit queue of a device with multiple transmit queues.
3154 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3157 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3159 return netif_tx_queue_stopped(txq);
3162 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3163 struct sk_buff *skb)
3165 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3169 * netif_wake_subqueue - allow sending packets on subqueue
3170 * @dev: network device
3171 * @queue_index: sub queue index
3173 * Resume individual transmit queue of a device with multiple transmit queues.
3175 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3177 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3179 netif_tx_wake_queue(txq);
3183 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3186 static inline int netif_set_xps_queue(struct net_device *dev,
3187 const struct cpumask *mask,
3194 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3195 unsigned int num_tx_queues);
3198 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3199 * as a distribution range limit for the returned value.
3201 static inline u16 skb_tx_hash(const struct net_device *dev,
3202 struct sk_buff *skb)
3204 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3208 * netif_is_multiqueue - test if device has multiple transmit queues
3209 * @dev: network device
3211 * Check if device has multiple transmit queues
3213 static inline bool netif_is_multiqueue(const struct net_device *dev)
3215 return dev->num_tx_queues > 1;
3218 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3221 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3223 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3231 static inline unsigned int get_netdev_rx_queue_index(
3232 struct netdev_rx_queue *queue)
3234 struct net_device *dev = queue->dev;
3235 int index = queue - dev->_rx;
3237 BUG_ON(index >= dev->num_rx_queues);
3242 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3243 int netif_get_num_default_rss_queues(void);
3245 enum skb_free_reason {
3246 SKB_REASON_CONSUMED,
3250 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3251 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3254 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3255 * interrupt context or with hardware interrupts being disabled.
3256 * (in_irq() || irqs_disabled())
3258 * We provide four helpers that can be used in following contexts :
3260 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3261 * replacing kfree_skb(skb)
3263 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3264 * Typically used in place of consume_skb(skb) in TX completion path
3266 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3267 * replacing kfree_skb(skb)
3269 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3270 * and consumed a packet. Used in place of consume_skb(skb)
3272 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3274 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3277 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3279 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3282 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3284 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3287 static inline void dev_consume_skb_any(struct sk_buff *skb)
3289 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3292 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3293 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3294 int netif_rx(struct sk_buff *skb);
3295 int netif_rx_ni(struct sk_buff *skb);
3296 int netif_receive_skb(struct sk_buff *skb);
3297 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3298 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3299 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3300 gro_result_t napi_gro_frags(struct napi_struct *napi);
3301 struct packet_offload *gro_find_receive_by_type(__be16 type);
3302 struct packet_offload *gro_find_complete_by_type(__be16 type);
3304 static inline void napi_free_frags(struct napi_struct *napi)
3306 kfree_skb(napi->skb);
3310 bool netdev_is_rx_handler_busy(struct net_device *dev);
3311 int netdev_rx_handler_register(struct net_device *dev,
3312 rx_handler_func_t *rx_handler,
3313 void *rx_handler_data);
3314 void netdev_rx_handler_unregister(struct net_device *dev);
3316 bool dev_valid_name(const char *name);
3317 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3318 int dev_ethtool(struct net *net, struct ifreq *);
3319 unsigned int dev_get_flags(const struct net_device *);
3320 int __dev_change_flags(struct net_device *, unsigned int flags);
3321 int dev_change_flags(struct net_device *, unsigned int);
3322 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3323 unsigned int gchanges);
3324 int dev_change_name(struct net_device *, const char *);
3325 int dev_set_alias(struct net_device *, const char *, size_t);
3326 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3327 int __dev_set_mtu(struct net_device *, int);
3328 int dev_set_mtu(struct net_device *, int);
3329 int dev_validate_mtu(struct net_device *dev, int mtu);
3330 void dev_set_group(struct net_device *, int);
3331 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3332 int dev_change_carrier(struct net_device *, bool new_carrier);
3333 int dev_get_phys_port_id(struct net_device *dev,
3334 struct netdev_phys_item_id *ppid);
3335 int dev_get_phys_port_name(struct net_device *dev,
3336 char *name, size_t len);
3337 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3338 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3339 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3340 struct netdev_queue *txq, int *ret);
3342 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3343 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3345 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3347 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3348 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3349 bool is_skb_forwardable(const struct net_device *dev,
3350 const struct sk_buff *skb);
3352 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3353 struct sk_buff *skb)
3355 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3356 unlikely(!is_skb_forwardable(dev, skb))) {
3357 atomic_long_inc(&dev->rx_dropped);
3362 skb_scrub_packet(skb, true);
3367 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3369 extern int netdev_budget;
3370 extern unsigned int netdev_budget_usecs;
3372 /* Called by rtnetlink.c:rtnl_unlock() */
3373 void netdev_run_todo(void);
3376 * dev_put - release reference to device
3377 * @dev: network device
3379 * Release reference to device to allow it to be freed.
3381 static inline void dev_put(struct net_device *dev)
3384 this_cpu_dec(*dev->pcpu_refcnt);
3388 * dev_hold - get reference to device
3389 * @dev: network device
3391 * Hold reference to device to keep it from being freed.
3393 static inline void dev_hold(struct net_device *dev)
3396 this_cpu_inc(*dev->pcpu_refcnt);
3399 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3400 * and _off may be called from IRQ context, but it is caller
3401 * who is responsible for serialization of these calls.
3403 * The name carrier is inappropriate, these functions should really be
3404 * called netif_lowerlayer_*() because they represent the state of any
3405 * kind of lower layer not just hardware media.
3408 void linkwatch_init_dev(struct net_device *dev);
3409 void linkwatch_fire_event(struct net_device *dev);
3410 void linkwatch_forget_dev(struct net_device *dev);
3413 * netif_carrier_ok - test if carrier present
3414 * @dev: network device
3416 * Check if carrier is present on device
3418 static inline bool netif_carrier_ok(const struct net_device *dev)
3420 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3423 unsigned long dev_trans_start(struct net_device *dev);
3425 void __netdev_watchdog_up(struct net_device *dev);
3427 void netif_carrier_on(struct net_device *dev);
3429 void netif_carrier_off(struct net_device *dev);
3432 * netif_dormant_on - mark device as dormant.
3433 * @dev: network device
3435 * Mark device as dormant (as per RFC2863).
3437 * The dormant state indicates that the relevant interface is not
3438 * actually in a condition to pass packets (i.e., it is not 'up') but is
3439 * in a "pending" state, waiting for some external event. For "on-
3440 * demand" interfaces, this new state identifies the situation where the
3441 * interface is waiting for events to place it in the up state.
3443 static inline void netif_dormant_on(struct net_device *dev)
3445 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3446 linkwatch_fire_event(dev);
3450 * netif_dormant_off - set device as not dormant.
3451 * @dev: network device
3453 * Device is not in dormant state.
3455 static inline void netif_dormant_off(struct net_device *dev)
3457 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3458 linkwatch_fire_event(dev);
3462 * netif_dormant - test if device is dormant
3463 * @dev: network device
3465 * Check if device is dormant.
3467 static inline bool netif_dormant(const struct net_device *dev)
3469 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3474 * netif_oper_up - test if device is operational
3475 * @dev: network device
3477 * Check if carrier is operational
3479 static inline bool netif_oper_up(const struct net_device *dev)
3481 return (dev->operstate == IF_OPER_UP ||
3482 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3486 * netif_device_present - is device available or removed
3487 * @dev: network device
3489 * Check if device has not been removed from system.
3491 static inline bool netif_device_present(struct net_device *dev)
3493 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3496 void netif_device_detach(struct net_device *dev);
3498 void netif_device_attach(struct net_device *dev);
3501 * Network interface message level settings
3505 NETIF_MSG_DRV = 0x0001,
3506 NETIF_MSG_PROBE = 0x0002,
3507 NETIF_MSG_LINK = 0x0004,
3508 NETIF_MSG_TIMER = 0x0008,
3509 NETIF_MSG_IFDOWN = 0x0010,
3510 NETIF_MSG_IFUP = 0x0020,
3511 NETIF_MSG_RX_ERR = 0x0040,
3512 NETIF_MSG_TX_ERR = 0x0080,
3513 NETIF_MSG_TX_QUEUED = 0x0100,
3514 NETIF_MSG_INTR = 0x0200,
3515 NETIF_MSG_TX_DONE = 0x0400,
3516 NETIF_MSG_RX_STATUS = 0x0800,
3517 NETIF_MSG_PKTDATA = 0x1000,
3518 NETIF_MSG_HW = 0x2000,
3519 NETIF_MSG_WOL = 0x4000,
3522 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3523 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3524 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3525 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3526 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3527 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3528 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3529 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3530 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3531 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3532 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3533 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3534 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3535 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3536 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3538 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3541 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3542 return default_msg_enable_bits;
3543 if (debug_value == 0) /* no output */
3545 /* set low N bits */
3546 return (1U << debug_value) - 1;
3549 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3551 spin_lock(&txq->_xmit_lock);
3552 txq->xmit_lock_owner = cpu;
3555 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3557 __acquire(&txq->_xmit_lock);
3561 static inline void __netif_tx_release(struct netdev_queue *txq)
3563 __release(&txq->_xmit_lock);
3566 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3568 spin_lock_bh(&txq->_xmit_lock);
3569 txq->xmit_lock_owner = smp_processor_id();
3572 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3574 bool ok = spin_trylock(&txq->_xmit_lock);
3576 txq->xmit_lock_owner = smp_processor_id();
3580 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3582 txq->xmit_lock_owner = -1;
3583 spin_unlock(&txq->_xmit_lock);
3586 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3588 txq->xmit_lock_owner = -1;
3589 spin_unlock_bh(&txq->_xmit_lock);
3592 static inline void txq_trans_update(struct netdev_queue *txq)
3594 if (txq->xmit_lock_owner != -1)
3595 txq->trans_start = jiffies;
3598 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3599 static inline void netif_trans_update(struct net_device *dev)
3601 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3603 if (txq->trans_start != jiffies)
3604 txq->trans_start = jiffies;
3608 * netif_tx_lock - grab network device transmit lock
3609 * @dev: network device
3611 * Get network device transmit lock
3613 static inline void netif_tx_lock(struct net_device *dev)
3618 spin_lock(&dev->tx_global_lock);
3619 cpu = smp_processor_id();
3620 for (i = 0; i < dev->num_tx_queues; i++) {
3621 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3623 /* We are the only thread of execution doing a
3624 * freeze, but we have to grab the _xmit_lock in
3625 * order to synchronize with threads which are in
3626 * the ->hard_start_xmit() handler and already
3627 * checked the frozen bit.
3629 __netif_tx_lock(txq, cpu);
3630 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3631 __netif_tx_unlock(txq);
3635 static inline void netif_tx_lock_bh(struct net_device *dev)
3641 static inline void netif_tx_unlock(struct net_device *dev)
3645 for (i = 0; i < dev->num_tx_queues; i++) {
3646 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3648 /* No need to grab the _xmit_lock here. If the
3649 * queue is not stopped for another reason, we
3652 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3653 netif_schedule_queue(txq);
3655 spin_unlock(&dev->tx_global_lock);
3658 static inline void netif_tx_unlock_bh(struct net_device *dev)
3660 netif_tx_unlock(dev);
3664 #define HARD_TX_LOCK(dev, txq, cpu) { \
3665 if ((dev->features & NETIF_F_LLTX) == 0) { \
3666 __netif_tx_lock(txq, cpu); \
3668 __netif_tx_acquire(txq); \
3672 #define HARD_TX_TRYLOCK(dev, txq) \
3673 (((dev->features & NETIF_F_LLTX) == 0) ? \
3674 __netif_tx_trylock(txq) : \
3675 __netif_tx_acquire(txq))
3677 #define HARD_TX_UNLOCK(dev, txq) { \
3678 if ((dev->features & NETIF_F_LLTX) == 0) { \
3679 __netif_tx_unlock(txq); \
3681 __netif_tx_release(txq); \
3685 static inline void netif_tx_disable(struct net_device *dev)
3691 cpu = smp_processor_id();
3692 spin_lock(&dev->tx_global_lock);
3693 for (i = 0; i < dev->num_tx_queues; i++) {
3694 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3696 __netif_tx_lock(txq, cpu);
3697 netif_tx_stop_queue(txq);
3698 __netif_tx_unlock(txq);
3700 spin_unlock(&dev->tx_global_lock);
3704 static inline void netif_addr_lock(struct net_device *dev)
3706 spin_lock(&dev->addr_list_lock);
3709 static inline void netif_addr_lock_nested(struct net_device *dev)
3711 int subclass = SINGLE_DEPTH_NESTING;
3713 if (dev->netdev_ops->ndo_get_lock_subclass)
3714 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3716 spin_lock_nested(&dev->addr_list_lock, subclass);
3719 static inline void netif_addr_lock_bh(struct net_device *dev)
3721 spin_lock_bh(&dev->addr_list_lock);
3724 static inline void netif_addr_unlock(struct net_device *dev)
3726 spin_unlock(&dev->addr_list_lock);
3729 static inline void netif_addr_unlock_bh(struct net_device *dev)
3731 spin_unlock_bh(&dev->addr_list_lock);
3735 * dev_addrs walker. Should be used only for read access. Call with
3736 * rcu_read_lock held.
3738 #define for_each_dev_addr(dev, ha) \
3739 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3741 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3743 void ether_setup(struct net_device *dev);
3745 /* Support for loadable net-drivers */
3746 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3747 unsigned char name_assign_type,
3748 void (*setup)(struct net_device *),
3749 unsigned int txqs, unsigned int rxqs);
3750 int dev_get_valid_name(struct net *net, struct net_device *dev,
3753 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3754 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3756 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3757 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3760 int register_netdev(struct net_device *dev);
3761 void unregister_netdev(struct net_device *dev);
3763 /* General hardware address lists handling functions */
3764 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3765 struct netdev_hw_addr_list *from_list, int addr_len);
3766 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3767 struct netdev_hw_addr_list *from_list, int addr_len);
3768 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3769 struct net_device *dev,
3770 int (*sync)(struct net_device *, const unsigned char *),
3771 int (*unsync)(struct net_device *,
3772 const unsigned char *));
3773 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3774 struct net_device *dev,
3775 int (*unsync)(struct net_device *,
3776 const unsigned char *));
3777 void __hw_addr_init(struct netdev_hw_addr_list *list);
3779 /* Functions used for device addresses handling */
3780 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3781 unsigned char addr_type);
3782 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3783 unsigned char addr_type);
3784 void dev_addr_flush(struct net_device *dev);
3785 int dev_addr_init(struct net_device *dev);
3787 /* Functions used for unicast addresses handling */
3788 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3789 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3790 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3791 int dev_uc_sync(struct net_device *to, struct net_device *from);
3792 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3793 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3794 void dev_uc_flush(struct net_device *dev);
3795 void dev_uc_init(struct net_device *dev);
3798 * __dev_uc_sync - Synchonize device's unicast list
3799 * @dev: device to sync
3800 * @sync: function to call if address should be added
3801 * @unsync: function to call if address should be removed
3803 * Add newly added addresses to the interface, and release
3804 * addresses that have been deleted.
3806 static inline int __dev_uc_sync(struct net_device *dev,
3807 int (*sync)(struct net_device *,
3808 const unsigned char *),
3809 int (*unsync)(struct net_device *,
3810 const unsigned char *))
3812 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3816 * __dev_uc_unsync - Remove synchronized addresses from device
3817 * @dev: device to sync
3818 * @unsync: function to call if address should be removed
3820 * Remove all addresses that were added to the device by dev_uc_sync().
3822 static inline void __dev_uc_unsync(struct net_device *dev,
3823 int (*unsync)(struct net_device *,
3824 const unsigned char *))
3826 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3829 /* Functions used for multicast addresses handling */
3830 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3831 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3832 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3833 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3834 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3835 int dev_mc_sync(struct net_device *to, struct net_device *from);
3836 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3837 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3838 void dev_mc_flush(struct net_device *dev);
3839 void dev_mc_init(struct net_device *dev);
3842 * __dev_mc_sync - Synchonize device's multicast list
3843 * @dev: device to sync
3844 * @sync: function to call if address should be added
3845 * @unsync: function to call if address should be removed
3847 * Add newly added addresses to the interface, and release
3848 * addresses that have been deleted.
3850 static inline int __dev_mc_sync(struct net_device *dev,
3851 int (*sync)(struct net_device *,
3852 const unsigned char *),
3853 int (*unsync)(struct net_device *,
3854 const unsigned char *))
3856 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3860 * __dev_mc_unsync - Remove synchronized addresses from device
3861 * @dev: device to sync
3862 * @unsync: function to call if address should be removed
3864 * Remove all addresses that were added to the device by dev_mc_sync().
3866 static inline void __dev_mc_unsync(struct net_device *dev,
3867 int (*unsync)(struct net_device *,
3868 const unsigned char *))
3870 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3873 /* Functions used for secondary unicast and multicast support */
3874 void dev_set_rx_mode(struct net_device *dev);
3875 void __dev_set_rx_mode(struct net_device *dev);
3876 int dev_set_promiscuity(struct net_device *dev, int inc);
3877 int dev_set_allmulti(struct net_device *dev, int inc);
3878 void netdev_state_change(struct net_device *dev);
3879 void netdev_notify_peers(struct net_device *dev);
3880 void netdev_features_change(struct net_device *dev);
3881 /* Load a device via the kmod */
3882 void dev_load(struct net *net, const char *name);
3883 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3884 struct rtnl_link_stats64 *storage);
3885 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3886 const struct net_device_stats *netdev_stats);
3888 extern int netdev_max_backlog;
3889 extern int netdev_tstamp_prequeue;
3890 extern int weight_p;
3891 extern int dev_weight_rx_bias;
3892 extern int dev_weight_tx_bias;
3893 extern int dev_rx_weight;
3894 extern int dev_tx_weight;
3896 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3897 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3898 struct list_head **iter);
3899 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3900 struct list_head **iter);
3902 /* iterate through upper list, must be called under RCU read lock */
3903 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3904 for (iter = &(dev)->adj_list.upper, \
3905 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3907 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3909 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3910 int (*fn)(struct net_device *upper_dev,
3914 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3915 struct net_device *upper_dev);
3917 bool netdev_has_any_upper_dev(struct net_device *dev);
3919 void *netdev_lower_get_next_private(struct net_device *dev,
3920 struct list_head **iter);
3921 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3922 struct list_head **iter);
3924 #define netdev_for_each_lower_private(dev, priv, iter) \
3925 for (iter = (dev)->adj_list.lower.next, \
3926 priv = netdev_lower_get_next_private(dev, &(iter)); \
3928 priv = netdev_lower_get_next_private(dev, &(iter)))
3930 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3931 for (iter = &(dev)->adj_list.lower, \
3932 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3934 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3936 void *netdev_lower_get_next(struct net_device *dev,
3937 struct list_head **iter);
3939 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3940 for (iter = (dev)->adj_list.lower.next, \
3941 ldev = netdev_lower_get_next(dev, &(iter)); \
3943 ldev = netdev_lower_get_next(dev, &(iter)))
3945 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3946 struct list_head **iter);
3947 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3948 struct list_head **iter);
3950 int netdev_walk_all_lower_dev(struct net_device *dev,
3951 int (*fn)(struct net_device *lower_dev,
3954 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3955 int (*fn)(struct net_device *lower_dev,
3959 void *netdev_adjacent_get_private(struct list_head *adj_list);
3960 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3961 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3962 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3963 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3964 int netdev_master_upper_dev_link(struct net_device *dev,
3965 struct net_device *upper_dev,
3966 void *upper_priv, void *upper_info);
3967 void netdev_upper_dev_unlink(struct net_device *dev,
3968 struct net_device *upper_dev);
3969 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3970 void *netdev_lower_dev_get_private(struct net_device *dev,
3971 struct net_device *lower_dev);
3972 void netdev_lower_state_changed(struct net_device *lower_dev,
3973 void *lower_state_info);
3975 /* RSS keys are 40 or 52 bytes long */
3976 #define NETDEV_RSS_KEY_LEN 52
3977 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3978 void netdev_rss_key_fill(void *buffer, size_t len);
3980 int dev_get_nest_level(struct net_device *dev);
3981 int skb_checksum_help(struct sk_buff *skb);
3982 int skb_crc32c_csum_help(struct sk_buff *skb);
3983 int skb_csum_hwoffload_help(struct sk_buff *skb,
3984 const netdev_features_t features);
3986 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3987 netdev_features_t features, bool tx_path);
3988 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3989 netdev_features_t features);
3991 struct netdev_bonding_info {
3996 struct netdev_notifier_bonding_info {
3997 struct netdev_notifier_info info; /* must be first */
3998 struct netdev_bonding_info bonding_info;
4001 void netdev_bonding_info_change(struct net_device *dev,
4002 struct netdev_bonding_info *bonding_info);
4005 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4007 return __skb_gso_segment(skb, features, true);
4009 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4011 static inline bool can_checksum_protocol(netdev_features_t features,
4014 if (protocol == htons(ETH_P_FCOE))
4015 return !!(features & NETIF_F_FCOE_CRC);
4017 /* Assume this is an IP checksum (not SCTP CRC) */
4019 if (features & NETIF_F_HW_CSUM) {
4020 /* Can checksum everything */
4025 case htons(ETH_P_IP):
4026 return !!(features & NETIF_F_IP_CSUM);
4027 case htons(ETH_P_IPV6):
4028 return !!(features & NETIF_F_IPV6_CSUM);
4035 void netdev_rx_csum_fault(struct net_device *dev);
4037 static inline void netdev_rx_csum_fault(struct net_device *dev)
4041 /* rx skb timestamps */
4042 void net_enable_timestamp(void);
4043 void net_disable_timestamp(void);
4045 #ifdef CONFIG_PROC_FS
4046 int __init dev_proc_init(void);
4048 #define dev_proc_init() 0
4051 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4052 struct sk_buff *skb, struct net_device *dev,
4055 skb->xmit_more = more ? 1 : 0;
4056 return ops->ndo_start_xmit(skb, dev);
4059 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4060 struct netdev_queue *txq, bool more)
4062 const struct net_device_ops *ops = dev->netdev_ops;
4065 rc = __netdev_start_xmit(ops, skb, dev, more);
4066 if (rc == NETDEV_TX_OK)
4067 txq_trans_update(txq);
4072 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4074 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4077 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4079 return netdev_class_create_file_ns(class_attr, NULL);
4082 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4084 netdev_class_remove_file_ns(class_attr, NULL);
4087 extern const struct kobj_ns_type_operations net_ns_type_operations;
4089 const char *netdev_drivername(const struct net_device *dev);
4091 void linkwatch_run_queue(void);
4093 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4094 netdev_features_t f2)
4096 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4097 if (f1 & NETIF_F_HW_CSUM)
4098 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4100 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4106 static inline netdev_features_t netdev_get_wanted_features(
4107 struct net_device *dev)
4109 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4111 netdev_features_t netdev_increment_features(netdev_features_t all,
4112 netdev_features_t one, netdev_features_t mask);
4114 /* Allow TSO being used on stacked device :
4115 * Performing the GSO segmentation before last device
4116 * is a performance improvement.
4118 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4119 netdev_features_t mask)
4121 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4124 int __netdev_update_features(struct net_device *dev);
4125 void netdev_update_features(struct net_device *dev);
4126 void netdev_change_features(struct net_device *dev);
4128 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4129 struct net_device *dev);
4131 netdev_features_t passthru_features_check(struct sk_buff *skb,
4132 struct net_device *dev,
4133 netdev_features_t features);
4134 netdev_features_t netif_skb_features(struct sk_buff *skb);
4136 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4138 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4140 /* check flags correspondence */
4141 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4142 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4143 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4144 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4145 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4146 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4147 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4148 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4149 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4150 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4151 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4152 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4153 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4154 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4155 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4156 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4157 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4159 return (features & feature) == feature;
4162 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4164 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4165 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4168 static inline bool netif_needs_gso(struct sk_buff *skb,
4169 netdev_features_t features)
4171 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4172 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4173 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4176 static inline void netif_set_gso_max_size(struct net_device *dev,
4179 dev->gso_max_size = size;
4182 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4183 int pulled_hlen, u16 mac_offset,
4186 skb->protocol = protocol;
4187 skb->encapsulation = 1;
4188 skb_push(skb, pulled_hlen);
4189 skb_reset_transport_header(skb);
4190 skb->mac_header = mac_offset;
4191 skb->network_header = skb->mac_header + mac_len;
4192 skb->mac_len = mac_len;
4195 static inline bool netif_is_macsec(const struct net_device *dev)
4197 return dev->priv_flags & IFF_MACSEC;
4200 static inline bool netif_is_macvlan(const struct net_device *dev)
4202 return dev->priv_flags & IFF_MACVLAN;
4205 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4207 return dev->priv_flags & IFF_MACVLAN_PORT;
4210 static inline bool netif_is_ipvlan(const struct net_device *dev)
4212 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4215 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4217 return dev->priv_flags & IFF_IPVLAN_MASTER;
4220 static inline bool netif_is_bond_master(const struct net_device *dev)
4222 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4225 static inline bool netif_is_bond_slave(const struct net_device *dev)
4227 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4230 static inline bool netif_supports_nofcs(struct net_device *dev)
4232 return dev->priv_flags & IFF_SUPP_NOFCS;
4235 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4237 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4240 static inline bool netif_is_l3_master(const struct net_device *dev)
4242 return dev->priv_flags & IFF_L3MDEV_MASTER;
4245 static inline bool netif_is_l3_slave(const struct net_device *dev)
4247 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4250 static inline bool netif_is_bridge_master(const struct net_device *dev)
4252 return dev->priv_flags & IFF_EBRIDGE;
4255 static inline bool netif_is_bridge_port(const struct net_device *dev)
4257 return dev->priv_flags & IFF_BRIDGE_PORT;
4260 static inline bool netif_is_ovs_master(const struct net_device *dev)
4262 return dev->priv_flags & IFF_OPENVSWITCH;
4265 static inline bool netif_is_ovs_port(const struct net_device *dev)
4267 return dev->priv_flags & IFF_OVS_DATAPATH;
4270 static inline bool netif_is_team_master(const struct net_device *dev)
4272 return dev->priv_flags & IFF_TEAM;
4275 static inline bool netif_is_team_port(const struct net_device *dev)
4277 return dev->priv_flags & IFF_TEAM_PORT;
4280 static inline bool netif_is_lag_master(const struct net_device *dev)
4282 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4285 static inline bool netif_is_lag_port(const struct net_device *dev)
4287 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4290 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4292 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4295 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4296 static inline void netif_keep_dst(struct net_device *dev)
4298 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4301 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4302 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4304 /* TODO: reserve and use an additional IFF bit, if we get more users */
4305 return dev->priv_flags & IFF_MACSEC;
4308 extern struct pernet_operations __net_initdata loopback_net_ops;
4310 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4312 /* netdev_printk helpers, similar to dev_printk */
4314 static inline const char *netdev_name(const struct net_device *dev)
4316 if (!dev->name[0] || strchr(dev->name, '%'))
4317 return "(unnamed net_device)";
4321 static inline bool netdev_unregistering(const struct net_device *dev)
4323 return dev->reg_state == NETREG_UNREGISTERING;
4326 static inline const char *netdev_reg_state(const struct net_device *dev)
4328 switch (dev->reg_state) {
4329 case NETREG_UNINITIALIZED: return " (uninitialized)";
4330 case NETREG_REGISTERED: return "";
4331 case NETREG_UNREGISTERING: return " (unregistering)";
4332 case NETREG_UNREGISTERED: return " (unregistered)";
4333 case NETREG_RELEASED: return " (released)";
4334 case NETREG_DUMMY: return " (dummy)";
4337 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4338 return " (unknown)";
4342 void netdev_printk(const char *level, const struct net_device *dev,
4343 const char *format, ...);
4345 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4347 void netdev_alert(const struct net_device *dev, const char *format, ...);
4349 void netdev_crit(const struct net_device *dev, const char *format, ...);
4351 void netdev_err(const struct net_device *dev, const char *format, ...);
4353 void netdev_warn(const struct net_device *dev, const char *format, ...);
4355 void netdev_notice(const struct net_device *dev, const char *format, ...);
4357 void netdev_info(const struct net_device *dev, const char *format, ...);
4359 #define MODULE_ALIAS_NETDEV(device) \
4360 MODULE_ALIAS("netdev-" device)
4362 #if defined(CONFIG_DYNAMIC_DEBUG)
4363 #define netdev_dbg(__dev, format, args...) \
4365 dynamic_netdev_dbg(__dev, format, ##args); \
4367 #elif defined(DEBUG)
4368 #define netdev_dbg(__dev, format, args...) \
4369 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4371 #define netdev_dbg(__dev, format, args...) \
4374 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4378 #if defined(VERBOSE_DEBUG)
4379 #define netdev_vdbg netdev_dbg
4382 #define netdev_vdbg(dev, format, args...) \
4385 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4391 * netdev_WARN() acts like dev_printk(), but with the key difference
4392 * of using a WARN/WARN_ON to get the message out, including the
4393 * file/line information and a backtrace.
4395 #define netdev_WARN(dev, format, args...) \
4396 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4397 netdev_reg_state(dev), ##args)
4399 /* netif printk helpers, similar to netdev_printk */
4401 #define netif_printk(priv, type, level, dev, fmt, args...) \
4403 if (netif_msg_##type(priv)) \
4404 netdev_printk(level, (dev), fmt, ##args); \
4407 #define netif_level(level, priv, type, dev, fmt, args...) \
4409 if (netif_msg_##type(priv)) \
4410 netdev_##level(dev, fmt, ##args); \
4413 #define netif_emerg(priv, type, dev, fmt, args...) \
4414 netif_level(emerg, priv, type, dev, fmt, ##args)
4415 #define netif_alert(priv, type, dev, fmt, args...) \
4416 netif_level(alert, priv, type, dev, fmt, ##args)
4417 #define netif_crit(priv, type, dev, fmt, args...) \
4418 netif_level(crit, priv, type, dev, fmt, ##args)
4419 #define netif_err(priv, type, dev, fmt, args...) \
4420 netif_level(err, priv, type, dev, fmt, ##args)
4421 #define netif_warn(priv, type, dev, fmt, args...) \
4422 netif_level(warn, priv, type, dev, fmt, ##args)
4423 #define netif_notice(priv, type, dev, fmt, args...) \
4424 netif_level(notice, priv, type, dev, fmt, ##args)
4425 #define netif_info(priv, type, dev, fmt, args...) \
4426 netif_level(info, priv, type, dev, fmt, ##args)
4428 #if defined(CONFIG_DYNAMIC_DEBUG)
4429 #define netif_dbg(priv, type, netdev, format, args...) \
4431 if (netif_msg_##type(priv)) \
4432 dynamic_netdev_dbg(netdev, format, ##args); \
4434 #elif defined(DEBUG)
4435 #define netif_dbg(priv, type, dev, format, args...) \
4436 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4438 #define netif_dbg(priv, type, dev, format, args...) \
4441 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4446 /* if @cond then downgrade to debug, else print at @level */
4447 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4450 netif_dbg(priv, type, netdev, fmt, ##args); \
4452 netif_ ## level(priv, type, netdev, fmt, ##args); \
4455 #if defined(VERBOSE_DEBUG)
4456 #define netif_vdbg netif_dbg
4458 #define netif_vdbg(priv, type, dev, format, args...) \
4461 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4467 * The list of packet types we will receive (as opposed to discard)
4468 * and the routines to invoke.
4470 * Why 16. Because with 16 the only overlap we get on a hash of the
4471 * low nibble of the protocol value is RARP/SNAP/X.25.
4473 * NOTE: That is no longer true with the addition of VLAN tags. Not
4474 * sure which should go first, but I bet it won't make much
4475 * difference if we are running VLANs. The good news is that
4476 * this protocol won't be in the list unless compiled in, so
4477 * the average user (w/out VLANs) will not be adversely affected.
4493 #define PTYPE_HASH_SIZE (16)
4494 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4496 #endif /* _LINUX_NETDEVICE_H */