2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
45 #include <net/dcbnl.h>
47 #include <net/netprio_cgroup.h>
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
59 struct dsa_switch_tree;
63 /* 802.15.4 specific */
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
70 void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
78 * Transmit return codes: transmit return codes originate from three different
81 * - qdisc return codes
82 * - driver transmit return codes
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91 * others are propagated to higher layers.
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK 0xf0
110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 typedef enum netdev_tx netdev_tx_t;
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 static inline bool dev_xmit_complete(int rc)
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 if (likely(rc < NET_XMIT_MASK))
135 * Compute the worst-case header length according to the protocols
139 #if defined(CONFIG_HYPERV_NET)
140 # define LL_MAX_HEADER 128
141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
145 # define LL_MAX_HEADER 96
148 # define LL_MAX_HEADER 32
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 extern struct static_key rfs_needed;
203 struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
207 #define NETDEV_HW_ADDR_T_LAN 1
208 #define NETDEV_HW_ADDR_T_SAN 2
209 #define NETDEV_HW_ADDR_T_SLAVE 3
210 #define NETDEV_HW_ADDR_T_UNICAST 4
211 #define NETDEV_HW_ADDR_T_MULTICAST 5
216 struct rcu_head rcu_head;
219 struct netdev_hw_addr_list {
220 struct list_head list;
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
274 bool (*validate)(const char *ll_header, unsigned int len);
277 /* These flag bits are private to the generic network queueing
278 * layer; they may not be explicitly referenced by any other
282 enum netdev_state_t {
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
292 * This structure holds boot-time configured netdevice settings. They
293 * are then used in the device probing.
295 struct netdev_boot_setup {
299 #define NETDEV_BOOT_SETUP_MAX 8
301 int __init netdev_boot_setup(char *str);
304 * Structure for NAPI scheduling similar to tasklet but with weighting
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-CPU poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
313 struct list_head poll_list;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
322 struct net_device *dev;
323 struct sk_buff *gro_list;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_MISSED, /* reschedule a napi */
334 NAPI_STATE_DISABLE, /* Disable pending */
335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
359 typedef enum gro_result gro_result_t;
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
383 * If the rx_handler consumed the skb in some way, it should return
384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
385 * the skb to be delivered in some other way.
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
391 * If the rx_handler decides the skb should be ignored, it should return
392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
393 * are registered on exact device (ptype->dev == skb->dev).
395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
396 * delivered, it should return RX_HANDLER_PASS.
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
402 enum rx_handler_result {
408 typedef enum rx_handler_result rx_handler_result_t;
409 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
411 void __napi_schedule(struct napi_struct *n);
412 void __napi_schedule_irqoff(struct napi_struct *n);
414 static inline bool napi_disable_pending(struct napi_struct *n)
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
419 bool napi_schedule_prep(struct napi_struct *n);
422 * napi_schedule - schedule NAPI poll
425 * Schedule NAPI poll routine to be called if it is not already
428 static inline void napi_schedule(struct napi_struct *n)
430 if (napi_schedule_prep(n))
435 * napi_schedule_irqoff - schedule NAPI poll
438 * Variant of napi_schedule(), assuming hard irqs are masked.
440 static inline void napi_schedule_irqoff(struct napi_struct *n)
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
447 static inline bool napi_reschedule(struct napi_struct *napi)
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
456 bool napi_complete_done(struct napi_struct *n, int work_done);
458 * napi_complete - NAPI processing complete
461 * Mark NAPI processing as complete.
462 * Consider using napi_complete_done() instead.
463 * Return false if device should avoid rearming interrupts.
465 static inline bool napi_complete(struct napi_struct *n)
467 return napi_complete_done(n, 0);
471 * napi_hash_del - remove a NAPI from global table
472 * @napi: NAPI context
474 * Warning: caller must observe RCU grace period
475 * before freeing memory containing @napi, if
476 * this function returns true.
477 * Note: core networking stack automatically calls it
478 * from netif_napi_del().
479 * Drivers might want to call this helper to combine all
480 * the needed RCU grace periods into a single one.
482 bool napi_hash_del(struct napi_struct *napi);
485 * napi_disable - prevent NAPI from scheduling
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
491 void napi_disable(struct napi_struct *n);
494 * napi_enable - enable NAPI scheduling
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
500 static inline void napi_enable(struct napi_struct *n)
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
503 smp_mb__before_atomic();
504 clear_bit(NAPI_STATE_SCHED, &n->state);
505 clear_bit(NAPI_STATE_NPSVC, &n->state);
509 * napi_synchronize - wait until NAPI is not running
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct *n)
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
525 enum netdev_queue_state_t {
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
528 __QUEUE_STATE_FROZEN,
531 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
551 struct netdev_queue {
555 struct net_device *dev;
556 struct Qdisc __rcu *qdisc;
557 struct Qdisc *qdisc_sleeping;
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
564 unsigned long tx_maxrate;
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
569 unsigned long trans_timeout;
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
576 * Time (in jiffies) of last Tx
578 unsigned long trans_start;
585 } ____cacheline_aligned_in_smp;
587 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
613 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
620 struct rps_dev_flow {
623 unsigned int last_qtail;
625 #define RPS_NO_FILTER 0xffff
628 * The rps_dev_flow_table structure contains a table of flow mappings.
630 struct rps_dev_flow_table {
633 struct rps_dev_flow flows[0];
635 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
636 ((_num) * sizeof(struct rps_dev_flow)))
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
643 * rps_cpu_mask is used to partition the space, depending on number of
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
646 * meaning we use 32-6=26 bits for the hash.
648 struct rps_sock_flow_table {
651 u32 ents[0] ____cacheline_aligned_in_smp;
653 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
655 #define RPS_NO_CPU 0xffff
657 extern u32 rps_cpu_mask;
658 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
667 /* We only give a hint, preemption can change CPU under us */
668 val |= raw_smp_processor_id();
670 if (table->ents[index] != val)
671 table->ents[index] = val;
675 #ifdef CONFIG_RFS_ACCEL
676 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
679 #endif /* CONFIG_RPS */
681 /* This structure contains an instance of an RX queue. */
682 struct netdev_rx_queue {
684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
688 struct net_device *dev;
689 } ____cacheline_aligned_in_smp;
692 * RX queue sysfs structures and functions.
694 struct rx_queue_attribute {
695 struct attribute attr;
696 ssize_t (*show)(struct netdev_rx_queue *queue,
697 struct rx_queue_attribute *attr, char *buf);
698 ssize_t (*store)(struct netdev_rx_queue *queue,
699 struct rx_queue_attribute *attr, const char *buf, size_t len);
704 * This structure holds an XPS map which can be of variable length. The
705 * map is an array of queues.
709 unsigned int alloc_len;
713 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
714 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715 - sizeof(struct xps_map)) / sizeof(u16))
718 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 struct xps_dev_maps {
722 struct xps_map __rcu *cpu_map[0];
724 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
726 #endif /* CONFIG_XPS */
728 #define TC_MAX_QUEUE 16
729 #define TC_BITMASK 15
730 /* HW offloaded queuing disciplines txq count and offset maps */
731 struct netdev_tc_txq {
736 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738 * This structure is to hold information about the device
739 * configured to run FCoE protocol stack.
741 struct netdev_fcoe_hbainfo {
742 char manufacturer[64];
743 char serial_number[64];
744 char hardware_version[64];
745 char driver_version[64];
746 char optionrom_version[64];
747 char firmware_version[64];
749 char model_description[256];
753 #define MAX_PHYS_ITEM_ID_LEN 32
755 /* This structure holds a unique identifier to identify some
756 * physical item (port for example) used by a netdevice.
758 struct netdev_phys_item_id {
759 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
760 unsigned char id_len;
763 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
764 struct netdev_phys_item_id *b)
766 return a->id_len == b->id_len &&
767 memcmp(a->id, b->id, a->id_len) == 0;
770 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
773 /* These structures hold the attributes of qdisc and classifiers
774 * that are being passed to the netdevice through the setup_tc op.
784 struct tc_cls_u32_offload;
786 struct tc_to_netdev {
789 struct tc_cls_u32_offload *cls_u32;
790 struct tc_cls_flower_offload *cls_flower;
791 struct tc_cls_matchall_offload *cls_mall;
792 struct tc_cls_bpf_offload *cls_bpf;
793 struct tc_mqprio_qopt *mqprio;
798 /* These structures hold the attributes of xdp state that are being passed
799 * to the netdevice through the xdp op.
801 enum xdp_netdev_command {
802 /* Set or clear a bpf program used in the earliest stages of packet
803 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
804 * is responsible for calling bpf_prog_put on any old progs that are
805 * stored. In case of error, the callee need not release the new prog
806 * reference, but on success it takes ownership and must bpf_prog_put
807 * when it is no longer used.
811 /* Check if a bpf program is set on the device. The callee should
812 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
813 * is equivalent to XDP_ATTACHED_DRV.
818 struct netlink_ext_ack;
821 enum xdp_netdev_command command;
826 struct bpf_prog *prog;
827 struct netlink_ext_ack *extack;
837 #ifdef CONFIG_XFRM_OFFLOAD
839 int (*xdo_dev_state_add) (struct xfrm_state *x);
840 void (*xdo_dev_state_delete) (struct xfrm_state *x);
841 void (*xdo_dev_state_free) (struct xfrm_state *x);
842 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
843 struct xfrm_state *x);
848 * This structure defines the management hooks for network devices.
849 * The following hooks can be defined; unless noted otherwise, they are
850 * optional and can be filled with a null pointer.
852 * int (*ndo_init)(struct net_device *dev);
853 * This function is called once when a network device is registered.
854 * The network device can use this for any late stage initialization
855 * or semantic validation. It can fail with an error code which will
856 * be propagated back to register_netdev.
858 * void (*ndo_uninit)(struct net_device *dev);
859 * This function is called when device is unregistered or when registration
860 * fails. It is not called if init fails.
862 * int (*ndo_open)(struct net_device *dev);
863 * This function is called when a network device transitions to the up
866 * int (*ndo_stop)(struct net_device *dev);
867 * This function is called when a network device transitions to the down
870 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
871 * struct net_device *dev);
872 * Called when a packet needs to be transmitted.
873 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
874 * the queue before that can happen; it's for obsolete devices and weird
875 * corner cases, but the stack really does a non-trivial amount
876 * of useless work if you return NETDEV_TX_BUSY.
877 * Required; cannot be NULL.
879 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
880 * struct net_device *dev
881 * netdev_features_t features);
882 * Called by core transmit path to determine if device is capable of
883 * performing offload operations on a given packet. This is to give
884 * the device an opportunity to implement any restrictions that cannot
885 * be otherwise expressed by feature flags. The check is called with
886 * the set of features that the stack has calculated and it returns
887 * those the driver believes to be appropriate.
889 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
890 * void *accel_priv, select_queue_fallback_t fallback);
891 * Called to decide which queue to use when device supports multiple
894 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
895 * This function is called to allow device receiver to make
896 * changes to configuration when multicast or promiscuous is enabled.
898 * void (*ndo_set_rx_mode)(struct net_device *dev);
899 * This function is called device changes address list filtering.
900 * If driver handles unicast address filtering, it should set
901 * IFF_UNICAST_FLT in its priv_flags.
903 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
904 * This function is called when the Media Access Control address
905 * needs to be changed. If this interface is not defined, the
906 * MAC address can not be changed.
908 * int (*ndo_validate_addr)(struct net_device *dev);
909 * Test if Media Access Control address is valid for the device.
911 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
912 * Called when a user requests an ioctl which can't be handled by
913 * the generic interface code. If not defined ioctls return
914 * not supported error code.
916 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
917 * Used to set network devices bus interface parameters. This interface
918 * is retained for legacy reasons; new devices should use the bus
919 * interface (PCI) for low level management.
921 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
922 * Called when a user wants to change the Maximum Transfer Unit
925 * void (*ndo_tx_timeout)(struct net_device *dev);
926 * Callback used when the transmitter has not made any progress
927 * for dev->watchdog ticks.
929 * void (*ndo_get_stats64)(struct net_device *dev,
930 * struct rtnl_link_stats64 *storage);
931 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
932 * Called when a user wants to get the network device usage
933 * statistics. Drivers must do one of the following:
934 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
935 * rtnl_link_stats64 structure passed by the caller.
936 * 2. Define @ndo_get_stats to update a net_device_stats structure
937 * (which should normally be dev->stats) and return a pointer to
938 * it. The structure may be changed asynchronously only if each
939 * field is written atomically.
940 * 3. Update dev->stats asynchronously and atomically, and define
943 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
944 * Return true if this device supports offload stats of this attr_id.
946 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
948 * Get statistics for offload operations by attr_id. Write it into the
951 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
952 * If device supports VLAN filtering this function is called when a
953 * VLAN id is registered.
955 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
956 * If device supports VLAN filtering this function is called when a
957 * VLAN id is unregistered.
959 * void (*ndo_poll_controller)(struct net_device *dev);
961 * SR-IOV management functions.
962 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
963 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
964 * u8 qos, __be16 proto);
965 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
967 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
968 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
969 * int (*ndo_get_vf_config)(struct net_device *dev,
970 * int vf, struct ifla_vf_info *ivf);
971 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
972 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
973 * struct nlattr *port[]);
975 * Enable or disable the VF ability to query its RSS Redirection Table and
976 * Hash Key. This is needed since on some devices VF share this information
977 * with PF and querying it may introduce a theoretical security risk.
978 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
979 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
980 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, u32 chain_index,
981 * __be16 protocol, struct tc_to_netdev *tc);
982 * Called to setup any 'tc' scheduler, classifier or action on @dev.
983 * This is always called from the stack with the rtnl lock held and netif
984 * tx queues stopped. This allows the netdevice to perform queue
987 * Fiber Channel over Ethernet (FCoE) offload functions.
988 * int (*ndo_fcoe_enable)(struct net_device *dev);
989 * Called when the FCoE protocol stack wants to start using LLD for FCoE
990 * so the underlying device can perform whatever needed configuration or
991 * initialization to support acceleration of FCoE traffic.
993 * int (*ndo_fcoe_disable)(struct net_device *dev);
994 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
995 * so the underlying device can perform whatever needed clean-ups to
996 * stop supporting acceleration of FCoE traffic.
998 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
999 * struct scatterlist *sgl, unsigned int sgc);
1000 * Called when the FCoE Initiator wants to initialize an I/O that
1001 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1002 * perform necessary setup and returns 1 to indicate the device is set up
1003 * successfully to perform DDP on this I/O, otherwise this returns 0.
1005 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1006 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1007 * indicated by the FC exchange id 'xid', so the underlying device can
1008 * clean up and reuse resources for later DDP requests.
1010 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1011 * struct scatterlist *sgl, unsigned int sgc);
1012 * Called when the FCoE Target wants to initialize an I/O that
1013 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1014 * perform necessary setup and returns 1 to indicate the device is set up
1015 * successfully to perform DDP on this I/O, otherwise this returns 0.
1017 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1018 * struct netdev_fcoe_hbainfo *hbainfo);
1019 * Called when the FCoE Protocol stack wants information on the underlying
1020 * device. This information is utilized by the FCoE protocol stack to
1021 * register attributes with Fiber Channel management service as per the
1022 * FC-GS Fabric Device Management Information(FDMI) specification.
1024 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1025 * Called when the underlying device wants to override default World Wide
1026 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1027 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1028 * protocol stack to use.
1031 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1032 * u16 rxq_index, u32 flow_id);
1033 * Set hardware filter for RFS. rxq_index is the target queue index;
1034 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1035 * Return the filter ID on success, or a negative error code.
1037 * Slave management functions (for bridge, bonding, etc).
1038 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1039 * Called to make another netdev an underling.
1041 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1042 * Called to release previously enslaved netdev.
1044 * Feature/offload setting functions.
1045 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1046 * netdev_features_t features);
1047 * Adjusts the requested feature flags according to device-specific
1048 * constraints, and returns the resulting flags. Must not modify
1051 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1052 * Called to update device configuration to new features. Passed
1053 * feature set might be less than what was returned by ndo_fix_features()).
1054 * Must return >0 or -errno if it changed dev->features itself.
1056 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1057 * struct net_device *dev,
1058 * const unsigned char *addr, u16 vid, u16 flags)
1059 * Adds an FDB entry to dev for addr.
1060 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1061 * struct net_device *dev,
1062 * const unsigned char *addr, u16 vid)
1063 * Deletes the FDB entry from dev coresponding to addr.
1064 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1065 * struct net_device *dev, struct net_device *filter_dev,
1067 * Used to add FDB entries to dump requests. Implementers should add
1068 * entries to skb and update idx with the number of entries.
1070 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1072 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1073 * struct net_device *dev, u32 filter_mask,
1075 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1078 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1079 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1080 * which do not represent real hardware may define this to allow their
1081 * userspace components to manage their virtual carrier state. Devices
1082 * that determine carrier state from physical hardware properties (eg
1083 * network cables) or protocol-dependent mechanisms (eg
1084 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1086 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1087 * struct netdev_phys_item_id *ppid);
1088 * Called to get ID of physical port of this device. If driver does
1089 * not implement this, it is assumed that the hw is not able to have
1090 * multiple net devices on single physical port.
1092 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1093 * struct udp_tunnel_info *ti);
1094 * Called by UDP tunnel to notify a driver about the UDP port and socket
1095 * address family that a UDP tunnel is listnening to. It is called only
1096 * when a new port starts listening. The operation is protected by the
1099 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1100 * struct udp_tunnel_info *ti);
1101 * Called by UDP tunnel to notify the driver about a UDP port and socket
1102 * address family that the UDP tunnel is not listening to anymore. The
1103 * operation is protected by the RTNL.
1105 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1106 * struct net_device *dev)
1107 * Called by upper layer devices to accelerate switching or other
1108 * station functionality into hardware. 'pdev is the lowerdev
1109 * to use for the offload and 'dev' is the net device that will
1110 * back the offload. Returns a pointer to the private structure
1111 * the upper layer will maintain.
1112 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1113 * Called by upper layer device to delete the station created
1114 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1115 * the station and priv is the structure returned by the add
1117 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1118 * struct net_device *dev,
1120 * Callback to use for xmit over the accelerated station. This
1121 * is used in place of ndo_start_xmit on accelerated net
1123 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1124 * int queue_index, u32 maxrate);
1125 * Called when a user wants to set a max-rate limitation of specific
1127 * int (*ndo_get_iflink)(const struct net_device *dev);
1128 * Called to get the iflink value of this device.
1129 * void (*ndo_change_proto_down)(struct net_device *dev,
1131 * This function is used to pass protocol port error state information
1132 * to the switch driver. The switch driver can react to the proto_down
1133 * by doing a phys down on the associated switch port.
1134 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1135 * This function is used to get egress tunnel information for given skb.
1136 * This is useful for retrieving outer tunnel header parameters while
1138 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1139 * This function is used to specify the headroom that the skb must
1140 * consider when allocation skb during packet reception. Setting
1141 * appropriate rx headroom value allows avoiding skb head copy on
1142 * forward. Setting a negative value resets the rx headroom to the
1144 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1145 * This function is used to set or query state related to XDP on the
1146 * netdevice. See definition of enum xdp_netdev_command for details.
1149 struct net_device_ops {
1150 int (*ndo_init)(struct net_device *dev);
1151 void (*ndo_uninit)(struct net_device *dev);
1152 int (*ndo_open)(struct net_device *dev);
1153 int (*ndo_stop)(struct net_device *dev);
1154 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1155 struct net_device *dev);
1156 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1157 struct net_device *dev,
1158 netdev_features_t features);
1159 u16 (*ndo_select_queue)(struct net_device *dev,
1160 struct sk_buff *skb,
1162 select_queue_fallback_t fallback);
1163 void (*ndo_change_rx_flags)(struct net_device *dev,
1165 void (*ndo_set_rx_mode)(struct net_device *dev);
1166 int (*ndo_set_mac_address)(struct net_device *dev,
1168 int (*ndo_validate_addr)(struct net_device *dev);
1169 int (*ndo_do_ioctl)(struct net_device *dev,
1170 struct ifreq *ifr, int cmd);
1171 int (*ndo_set_config)(struct net_device *dev,
1173 int (*ndo_change_mtu)(struct net_device *dev,
1175 int (*ndo_neigh_setup)(struct net_device *dev,
1176 struct neigh_parms *);
1177 void (*ndo_tx_timeout) (struct net_device *dev);
1179 void (*ndo_get_stats64)(struct net_device *dev,
1180 struct rtnl_link_stats64 *storage);
1181 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1182 int (*ndo_get_offload_stats)(int attr_id,
1183 const struct net_device *dev,
1185 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1187 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1188 __be16 proto, u16 vid);
1189 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1190 __be16 proto, u16 vid);
1191 #ifdef CONFIG_NET_POLL_CONTROLLER
1192 void (*ndo_poll_controller)(struct net_device *dev);
1193 int (*ndo_netpoll_setup)(struct net_device *dev,
1194 struct netpoll_info *info);
1195 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1197 int (*ndo_set_vf_mac)(struct net_device *dev,
1198 int queue, u8 *mac);
1199 int (*ndo_set_vf_vlan)(struct net_device *dev,
1200 int queue, u16 vlan,
1201 u8 qos, __be16 proto);
1202 int (*ndo_set_vf_rate)(struct net_device *dev,
1203 int vf, int min_tx_rate,
1205 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1206 int vf, bool setting);
1207 int (*ndo_set_vf_trust)(struct net_device *dev,
1208 int vf, bool setting);
1209 int (*ndo_get_vf_config)(struct net_device *dev,
1211 struct ifla_vf_info *ivf);
1212 int (*ndo_set_vf_link_state)(struct net_device *dev,
1213 int vf, int link_state);
1214 int (*ndo_get_vf_stats)(struct net_device *dev,
1216 struct ifla_vf_stats
1218 int (*ndo_set_vf_port)(struct net_device *dev,
1220 struct nlattr *port[]);
1221 int (*ndo_get_vf_port)(struct net_device *dev,
1222 int vf, struct sk_buff *skb);
1223 int (*ndo_set_vf_guid)(struct net_device *dev,
1226 int (*ndo_set_vf_rss_query_en)(
1227 struct net_device *dev,
1228 int vf, bool setting);
1229 int (*ndo_setup_tc)(struct net_device *dev,
1230 u32 handle, u32 chain_index,
1232 struct tc_to_netdev *tc);
1233 #if IS_ENABLED(CONFIG_FCOE)
1234 int (*ndo_fcoe_enable)(struct net_device *dev);
1235 int (*ndo_fcoe_disable)(struct net_device *dev);
1236 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1238 struct scatterlist *sgl,
1240 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1242 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1244 struct scatterlist *sgl,
1246 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1247 struct netdev_fcoe_hbainfo *hbainfo);
1250 #if IS_ENABLED(CONFIG_LIBFCOE)
1251 #define NETDEV_FCOE_WWNN 0
1252 #define NETDEV_FCOE_WWPN 1
1253 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1254 u64 *wwn, int type);
1257 #ifdef CONFIG_RFS_ACCEL
1258 int (*ndo_rx_flow_steer)(struct net_device *dev,
1259 const struct sk_buff *skb,
1263 int (*ndo_add_slave)(struct net_device *dev,
1264 struct net_device *slave_dev);
1265 int (*ndo_del_slave)(struct net_device *dev,
1266 struct net_device *slave_dev);
1267 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1268 netdev_features_t features);
1269 int (*ndo_set_features)(struct net_device *dev,
1270 netdev_features_t features);
1271 int (*ndo_neigh_construct)(struct net_device *dev,
1272 struct neighbour *n);
1273 void (*ndo_neigh_destroy)(struct net_device *dev,
1274 struct neighbour *n);
1276 int (*ndo_fdb_add)(struct ndmsg *ndm,
1277 struct nlattr *tb[],
1278 struct net_device *dev,
1279 const unsigned char *addr,
1282 int (*ndo_fdb_del)(struct ndmsg *ndm,
1283 struct nlattr *tb[],
1284 struct net_device *dev,
1285 const unsigned char *addr,
1287 int (*ndo_fdb_dump)(struct sk_buff *skb,
1288 struct netlink_callback *cb,
1289 struct net_device *dev,
1290 struct net_device *filter_dev,
1293 int (*ndo_bridge_setlink)(struct net_device *dev,
1294 struct nlmsghdr *nlh,
1296 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1298 struct net_device *dev,
1301 int (*ndo_bridge_dellink)(struct net_device *dev,
1302 struct nlmsghdr *nlh,
1304 int (*ndo_change_carrier)(struct net_device *dev,
1306 int (*ndo_get_phys_port_id)(struct net_device *dev,
1307 struct netdev_phys_item_id *ppid);
1308 int (*ndo_get_phys_port_name)(struct net_device *dev,
1309 char *name, size_t len);
1310 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1311 struct udp_tunnel_info *ti);
1312 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1313 struct udp_tunnel_info *ti);
1314 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1315 struct net_device *dev);
1316 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1319 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1320 struct net_device *dev,
1322 int (*ndo_get_lock_subclass)(struct net_device *dev);
1323 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1326 int (*ndo_get_iflink)(const struct net_device *dev);
1327 int (*ndo_change_proto_down)(struct net_device *dev,
1329 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1330 struct sk_buff *skb);
1331 void (*ndo_set_rx_headroom)(struct net_device *dev,
1332 int needed_headroom);
1333 int (*ndo_xdp)(struct net_device *dev,
1334 struct netdev_xdp *xdp);
1338 * enum net_device_priv_flags - &struct net_device priv_flags
1340 * These are the &struct net_device, they are only set internally
1341 * by drivers and used in the kernel. These flags are invisible to
1342 * userspace; this means that the order of these flags can change
1343 * during any kernel release.
1345 * You should have a pretty good reason to be extending these flags.
1347 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1348 * @IFF_EBRIDGE: Ethernet bridging device
1349 * @IFF_BONDING: bonding master or slave
1350 * @IFF_ISATAP: ISATAP interface (RFC4214)
1351 * @IFF_WAN_HDLC: WAN HDLC device
1352 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1354 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1355 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1356 * @IFF_MACVLAN_PORT: device used as macvlan port
1357 * @IFF_BRIDGE_PORT: device used as bridge port
1358 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1359 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1360 * @IFF_UNICAST_FLT: Supports unicast filtering
1361 * @IFF_TEAM_PORT: device used as team port
1362 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1363 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1364 * change when it's running
1365 * @IFF_MACVLAN: Macvlan device
1366 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1367 * underlying stacked devices
1368 * @IFF_IPVLAN_MASTER: IPvlan master device
1369 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1370 * @IFF_L3MDEV_MASTER: device is an L3 master device
1371 * @IFF_NO_QUEUE: device can run without qdisc attached
1372 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1373 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1374 * @IFF_TEAM: device is a team device
1375 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1376 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1377 * entity (i.e. the master device for bridged veth)
1378 * @IFF_MACSEC: device is a MACsec device
1380 enum netdev_priv_flags {
1381 IFF_802_1Q_VLAN = 1<<0,
1385 IFF_WAN_HDLC = 1<<4,
1386 IFF_XMIT_DST_RELEASE = 1<<5,
1387 IFF_DONT_BRIDGE = 1<<6,
1388 IFF_DISABLE_NETPOLL = 1<<7,
1389 IFF_MACVLAN_PORT = 1<<8,
1390 IFF_BRIDGE_PORT = 1<<9,
1391 IFF_OVS_DATAPATH = 1<<10,
1392 IFF_TX_SKB_SHARING = 1<<11,
1393 IFF_UNICAST_FLT = 1<<12,
1394 IFF_TEAM_PORT = 1<<13,
1395 IFF_SUPP_NOFCS = 1<<14,
1396 IFF_LIVE_ADDR_CHANGE = 1<<15,
1397 IFF_MACVLAN = 1<<16,
1398 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1399 IFF_IPVLAN_MASTER = 1<<18,
1400 IFF_IPVLAN_SLAVE = 1<<19,
1401 IFF_L3MDEV_MASTER = 1<<20,
1402 IFF_NO_QUEUE = 1<<21,
1403 IFF_OPENVSWITCH = 1<<22,
1404 IFF_L3MDEV_SLAVE = 1<<23,
1406 IFF_RXFH_CONFIGURED = 1<<25,
1407 IFF_PHONY_HEADROOM = 1<<26,
1411 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1412 #define IFF_EBRIDGE IFF_EBRIDGE
1413 #define IFF_BONDING IFF_BONDING
1414 #define IFF_ISATAP IFF_ISATAP
1415 #define IFF_WAN_HDLC IFF_WAN_HDLC
1416 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1417 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1418 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1419 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1420 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1421 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1422 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1423 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1424 #define IFF_TEAM_PORT IFF_TEAM_PORT
1425 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1426 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1427 #define IFF_MACVLAN IFF_MACVLAN
1428 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1429 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1430 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1431 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1432 #define IFF_NO_QUEUE IFF_NO_QUEUE
1433 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1434 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1435 #define IFF_TEAM IFF_TEAM
1436 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1437 #define IFF_MACSEC IFF_MACSEC
1440 * struct net_device - The DEVICE structure.
1441 * Actually, this whole structure is a big mistake. It mixes I/O
1442 * data with strictly "high-level" data, and it has to know about
1443 * almost every data structure used in the INET module.
1445 * @name: This is the first field of the "visible" part of this structure
1446 * (i.e. as seen by users in the "Space.c" file). It is the name
1449 * @name_hlist: Device name hash chain, please keep it close to name[]
1450 * @ifalias: SNMP alias
1451 * @mem_end: Shared memory end
1452 * @mem_start: Shared memory start
1453 * @base_addr: Device I/O address
1454 * @irq: Device IRQ number
1456 * @carrier_changes: Stats to monitor carrier on<->off transitions
1458 * @state: Generic network queuing layer state, see netdev_state_t
1459 * @dev_list: The global list of network devices
1460 * @napi_list: List entry used for polling NAPI devices
1461 * @unreg_list: List entry when we are unregistering the
1462 * device; see the function unregister_netdev
1463 * @close_list: List entry used when we are closing the device
1464 * @ptype_all: Device-specific packet handlers for all protocols
1465 * @ptype_specific: Device-specific, protocol-specific packet handlers
1467 * @adj_list: Directly linked devices, like slaves for bonding
1468 * @features: Currently active device features
1469 * @hw_features: User-changeable features
1471 * @wanted_features: User-requested features
1472 * @vlan_features: Mask of features inheritable by VLAN devices
1474 * @hw_enc_features: Mask of features inherited by encapsulating devices
1475 * This field indicates what encapsulation
1476 * offloads the hardware is capable of doing,
1477 * and drivers will need to set them appropriately.
1479 * @mpls_features: Mask of features inheritable by MPLS
1481 * @ifindex: interface index
1482 * @group: The group the device belongs to
1484 * @stats: Statistics struct, which was left as a legacy, use
1485 * rtnl_link_stats64 instead
1487 * @rx_dropped: Dropped packets by core network,
1488 * do not use this in drivers
1489 * @tx_dropped: Dropped packets by core network,
1490 * do not use this in drivers
1491 * @rx_nohandler: nohandler dropped packets by core network on
1492 * inactive devices, do not use this in drivers
1494 * @wireless_handlers: List of functions to handle Wireless Extensions,
1496 * see <net/iw_handler.h> for details.
1497 * @wireless_data: Instance data managed by the core of wireless extensions
1499 * @netdev_ops: Includes several pointers to callbacks,
1500 * if one wants to override the ndo_*() functions
1501 * @ethtool_ops: Management operations
1502 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1503 * discovery handling. Necessary for e.g. 6LoWPAN.
1504 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1505 * of Layer 2 headers.
1507 * @flags: Interface flags (a la BSD)
1508 * @priv_flags: Like 'flags' but invisible to userspace,
1509 * see if.h for the definitions
1510 * @gflags: Global flags ( kept as legacy )
1511 * @padded: How much padding added by alloc_netdev()
1512 * @operstate: RFC2863 operstate
1513 * @link_mode: Mapping policy to operstate
1514 * @if_port: Selectable AUI, TP, ...
1516 * @mtu: Interface MTU value
1517 * @min_mtu: Interface Minimum MTU value
1518 * @max_mtu: Interface Maximum MTU value
1519 * @type: Interface hardware type
1520 * @hard_header_len: Maximum hardware header length.
1521 * @min_header_len: Minimum hardware header length
1523 * @needed_headroom: Extra headroom the hardware may need, but not in all
1524 * cases can this be guaranteed
1525 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1526 * cases can this be guaranteed. Some cases also use
1527 * LL_MAX_HEADER instead to allocate the skb
1529 * interface address info:
1531 * @perm_addr: Permanent hw address
1532 * @addr_assign_type: Hw address assignment type
1533 * @addr_len: Hardware address length
1534 * @neigh_priv_len: Used in neigh_alloc()
1535 * @dev_id: Used to differentiate devices that share
1536 * the same link layer address
1537 * @dev_port: Used to differentiate devices that share
1539 * @addr_list_lock: XXX: need comments on this one
1540 * @uc_promisc: Counter that indicates promiscuous mode
1541 * has been enabled due to the need to listen to
1542 * additional unicast addresses in a device that
1543 * does not implement ndo_set_rx_mode()
1544 * @uc: unicast mac addresses
1545 * @mc: multicast mac addresses
1546 * @dev_addrs: list of device hw addresses
1547 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1548 * @promiscuity: Number of times the NIC is told to work in
1549 * promiscuous mode; if it becomes 0 the NIC will
1550 * exit promiscuous mode
1551 * @allmulti: Counter, enables or disables allmulticast mode
1553 * @vlan_info: VLAN info
1554 * @dsa_ptr: dsa specific data
1555 * @tipc_ptr: TIPC specific data
1556 * @atalk_ptr: AppleTalk link
1557 * @ip_ptr: IPv4 specific data
1558 * @dn_ptr: DECnet specific data
1559 * @ip6_ptr: IPv6 specific data
1560 * @ax25_ptr: AX.25 specific data
1561 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1563 * @dev_addr: Hw address (before bcast,
1564 * because most packets are unicast)
1566 * @_rx: Array of RX queues
1567 * @num_rx_queues: Number of RX queues
1568 * allocated at register_netdev() time
1569 * @real_num_rx_queues: Number of RX queues currently active in device
1571 * @rx_handler: handler for received packets
1572 * @rx_handler_data: XXX: need comments on this one
1573 * @ingress_queue: XXX: need comments on this one
1574 * @broadcast: hw bcast address
1576 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1577 * indexed by RX queue number. Assigned by driver.
1578 * This must only be set if the ndo_rx_flow_steer
1579 * operation is defined
1580 * @index_hlist: Device index hash chain
1582 * @_tx: Array of TX queues
1583 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1584 * @real_num_tx_queues: Number of TX queues currently active in device
1585 * @qdisc: Root qdisc from userspace point of view
1586 * @tx_queue_len: Max frames per queue allowed
1587 * @tx_global_lock: XXX: need comments on this one
1589 * @xps_maps: XXX: need comments on this one
1591 * @watchdog_timeo: Represents the timeout that is used by
1592 * the watchdog (see dev_watchdog())
1593 * @watchdog_timer: List of timers
1595 * @pcpu_refcnt: Number of references to this device
1596 * @todo_list: Delayed register/unregister
1597 * @link_watch_list: XXX: need comments on this one
1599 * @reg_state: Register/unregister state machine
1600 * @dismantle: Device is going to be freed
1601 * @rtnl_link_state: This enum represents the phases of creating
1604 * @needs_free_netdev: Should unregister perform free_netdev?
1605 * @priv_destructor: Called from unregister
1606 * @npinfo: XXX: need comments on this one
1607 * @nd_net: Network namespace this network device is inside
1609 * @ml_priv: Mid-layer private
1610 * @lstats: Loopback statistics
1611 * @tstats: Tunnel statistics
1612 * @dstats: Dummy statistics
1613 * @vstats: Virtual ethernet statistics
1618 * @dev: Class/net/name entry
1619 * @sysfs_groups: Space for optional device, statistics and wireless
1622 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1623 * @rtnl_link_ops: Rtnl_link_ops
1625 * @gso_max_size: Maximum size of generic segmentation offload
1626 * @gso_max_segs: Maximum number of segments that can be passed to the
1629 * @dcbnl_ops: Data Center Bridging netlink ops
1630 * @num_tc: Number of traffic classes in the net device
1631 * @tc_to_txq: XXX: need comments on this one
1632 * @prio_tc_map: XXX: need comments on this one
1634 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1636 * @priomap: XXX: need comments on this one
1637 * @phydev: Physical device may attach itself
1638 * for hardware timestamping
1640 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1641 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1643 * @proto_down: protocol port state information can be sent to the
1644 * switch driver and used to set the phys state of the
1647 * FIXME: cleanup struct net_device such that network protocol info
1652 char name[IFNAMSIZ];
1653 struct hlist_node name_hlist;
1656 * I/O specific fields
1657 * FIXME: Merge these and struct ifmap into one
1659 unsigned long mem_end;
1660 unsigned long mem_start;
1661 unsigned long base_addr;
1664 atomic_t carrier_changes;
1667 * Some hardware also needs these fields (state,dev_list,
1668 * napi_list,unreg_list,close_list) but they are not
1669 * part of the usual set specified in Space.c.
1672 unsigned long state;
1674 struct list_head dev_list;
1675 struct list_head napi_list;
1676 struct list_head unreg_list;
1677 struct list_head close_list;
1678 struct list_head ptype_all;
1679 struct list_head ptype_specific;
1682 struct list_head upper;
1683 struct list_head lower;
1686 netdev_features_t features;
1687 netdev_features_t hw_features;
1688 netdev_features_t wanted_features;
1689 netdev_features_t vlan_features;
1690 netdev_features_t hw_enc_features;
1691 netdev_features_t mpls_features;
1692 netdev_features_t gso_partial_features;
1697 struct net_device_stats stats;
1699 atomic_long_t rx_dropped;
1700 atomic_long_t tx_dropped;
1701 atomic_long_t rx_nohandler;
1703 #ifdef CONFIG_WIRELESS_EXT
1704 const struct iw_handler_def *wireless_handlers;
1705 struct iw_public_data *wireless_data;
1707 const struct net_device_ops *netdev_ops;
1708 const struct ethtool_ops *ethtool_ops;
1709 #ifdef CONFIG_NET_SWITCHDEV
1710 const struct switchdev_ops *switchdev_ops;
1712 #ifdef CONFIG_NET_L3_MASTER_DEV
1713 const struct l3mdev_ops *l3mdev_ops;
1715 #if IS_ENABLED(CONFIG_IPV6)
1716 const struct ndisc_ops *ndisc_ops;
1720 const struct xfrmdev_ops *xfrmdev_ops;
1723 const struct header_ops *header_ops;
1726 unsigned int priv_flags;
1728 unsigned short gflags;
1729 unsigned short padded;
1731 unsigned char operstate;
1732 unsigned char link_mode;
1734 unsigned char if_port;
1738 unsigned int min_mtu;
1739 unsigned int max_mtu;
1740 unsigned short type;
1741 unsigned short hard_header_len;
1742 unsigned char min_header_len;
1744 unsigned short needed_headroom;
1745 unsigned short needed_tailroom;
1747 /* Interface address info. */
1748 unsigned char perm_addr[MAX_ADDR_LEN];
1749 unsigned char addr_assign_type;
1750 unsigned char addr_len;
1751 unsigned short neigh_priv_len;
1752 unsigned short dev_id;
1753 unsigned short dev_port;
1754 spinlock_t addr_list_lock;
1755 unsigned char name_assign_type;
1757 struct netdev_hw_addr_list uc;
1758 struct netdev_hw_addr_list mc;
1759 struct netdev_hw_addr_list dev_addrs;
1762 struct kset *queues_kset;
1764 unsigned int promiscuity;
1765 unsigned int allmulti;
1768 /* Protocol-specific pointers */
1770 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1771 struct vlan_info __rcu *vlan_info;
1773 #if IS_ENABLED(CONFIG_NET_DSA)
1774 struct dsa_switch_tree *dsa_ptr;
1776 #if IS_ENABLED(CONFIG_TIPC)
1777 struct tipc_bearer __rcu *tipc_ptr;
1780 struct in_device __rcu *ip_ptr;
1781 struct dn_dev __rcu *dn_ptr;
1782 struct inet6_dev __rcu *ip6_ptr;
1784 struct wireless_dev *ieee80211_ptr;
1785 struct wpan_dev *ieee802154_ptr;
1786 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1787 struct mpls_dev __rcu *mpls_ptr;
1791 * Cache lines mostly used on receive path (including eth_type_trans())
1793 /* Interface address info used in eth_type_trans() */
1794 unsigned char *dev_addr;
1797 struct netdev_rx_queue *_rx;
1799 unsigned int num_rx_queues;
1800 unsigned int real_num_rx_queues;
1803 struct bpf_prog __rcu *xdp_prog;
1804 unsigned long gro_flush_timeout;
1805 rx_handler_func_t __rcu *rx_handler;
1806 void __rcu *rx_handler_data;
1808 #ifdef CONFIG_NET_CLS_ACT
1809 struct tcf_proto __rcu *ingress_cl_list;
1811 struct netdev_queue __rcu *ingress_queue;
1812 #ifdef CONFIG_NETFILTER_INGRESS
1813 struct nf_hook_entry __rcu *nf_hooks_ingress;
1816 unsigned char broadcast[MAX_ADDR_LEN];
1817 #ifdef CONFIG_RFS_ACCEL
1818 struct cpu_rmap *rx_cpu_rmap;
1820 struct hlist_node index_hlist;
1823 * Cache lines mostly used on transmit path
1825 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1826 unsigned int num_tx_queues;
1827 unsigned int real_num_tx_queues;
1828 struct Qdisc *qdisc;
1829 #ifdef CONFIG_NET_SCHED
1830 DECLARE_HASHTABLE (qdisc_hash, 4);
1832 unsigned int tx_queue_len;
1833 spinlock_t tx_global_lock;
1837 struct xps_dev_maps __rcu *xps_maps;
1839 #ifdef CONFIG_NET_CLS_ACT
1840 struct tcf_proto __rcu *egress_cl_list;
1843 /* These may be needed for future network-power-down code. */
1844 struct timer_list watchdog_timer;
1846 int __percpu *pcpu_refcnt;
1847 struct list_head todo_list;
1849 struct list_head link_watch_list;
1851 enum { NETREG_UNINITIALIZED=0,
1852 NETREG_REGISTERED, /* completed register_netdevice */
1853 NETREG_UNREGISTERING, /* called unregister_netdevice */
1854 NETREG_UNREGISTERED, /* completed unregister todo */
1855 NETREG_RELEASED, /* called free_netdev */
1856 NETREG_DUMMY, /* dummy device for NAPI poll */
1862 RTNL_LINK_INITIALIZED,
1863 RTNL_LINK_INITIALIZING,
1864 } rtnl_link_state:16;
1866 bool needs_free_netdev;
1867 void (*priv_destructor)(struct net_device *dev);
1869 #ifdef CONFIG_NETPOLL
1870 struct netpoll_info __rcu *npinfo;
1873 possible_net_t nd_net;
1875 /* mid-layer private */
1878 struct pcpu_lstats __percpu *lstats;
1879 struct pcpu_sw_netstats __percpu *tstats;
1880 struct pcpu_dstats __percpu *dstats;
1881 struct pcpu_vstats __percpu *vstats;
1884 #if IS_ENABLED(CONFIG_GARP)
1885 struct garp_port __rcu *garp_port;
1887 #if IS_ENABLED(CONFIG_MRP)
1888 struct mrp_port __rcu *mrp_port;
1892 const struct attribute_group *sysfs_groups[4];
1893 const struct attribute_group *sysfs_rx_queue_group;
1895 const struct rtnl_link_ops *rtnl_link_ops;
1897 /* for setting kernel sock attribute on TCP connection setup */
1898 #define GSO_MAX_SIZE 65536
1899 unsigned int gso_max_size;
1900 #define GSO_MAX_SEGS 65535
1904 const struct dcbnl_rtnl_ops *dcbnl_ops;
1907 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1908 u8 prio_tc_map[TC_BITMASK + 1];
1910 #if IS_ENABLED(CONFIG_FCOE)
1911 unsigned int fcoe_ddp_xid;
1913 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1914 struct netprio_map __rcu *priomap;
1916 struct phy_device *phydev;
1917 struct lock_class_key *qdisc_tx_busylock;
1918 struct lock_class_key *qdisc_running_key;
1921 #define to_net_dev(d) container_of(d, struct net_device, dev)
1923 static inline bool netif_elide_gro(const struct net_device *dev)
1925 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1930 #define NETDEV_ALIGN 32
1933 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1935 return dev->prio_tc_map[prio & TC_BITMASK];
1939 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1941 if (tc >= dev->num_tc)
1944 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1948 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1949 void netdev_reset_tc(struct net_device *dev);
1950 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1951 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1954 int netdev_get_num_tc(struct net_device *dev)
1960 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1963 return &dev->_tx[index];
1966 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1967 const struct sk_buff *skb)
1969 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1972 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1973 void (*f)(struct net_device *,
1974 struct netdev_queue *,
1980 for (i = 0; i < dev->num_tx_queues; i++)
1981 f(dev, &dev->_tx[i], arg);
1984 #define netdev_lockdep_set_classes(dev) \
1986 static struct lock_class_key qdisc_tx_busylock_key; \
1987 static struct lock_class_key qdisc_running_key; \
1988 static struct lock_class_key qdisc_xmit_lock_key; \
1989 static struct lock_class_key dev_addr_list_lock_key; \
1992 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1993 (dev)->qdisc_running_key = &qdisc_running_key; \
1994 lockdep_set_class(&(dev)->addr_list_lock, \
1995 &dev_addr_list_lock_key); \
1996 for (i = 0; i < (dev)->num_tx_queues; i++) \
1997 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1998 &qdisc_xmit_lock_key); \
2001 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2002 struct sk_buff *skb,
2005 /* returns the headroom that the master device needs to take in account
2006 * when forwarding to this dev
2008 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2010 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2013 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2015 if (dev->netdev_ops->ndo_set_rx_headroom)
2016 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2019 /* set the device rx headroom to the dev's default */
2020 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2022 netdev_set_rx_headroom(dev, -1);
2026 * Net namespace inlines
2029 struct net *dev_net(const struct net_device *dev)
2031 return read_pnet(&dev->nd_net);
2035 void dev_net_set(struct net_device *dev, struct net *net)
2037 write_pnet(&dev->nd_net, net);
2041 * netdev_priv - access network device private data
2042 * @dev: network device
2044 * Get network device private data
2046 static inline void *netdev_priv(const struct net_device *dev)
2048 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2051 /* Set the sysfs physical device reference for the network logical device
2052 * if set prior to registration will cause a symlink during initialization.
2054 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2056 /* Set the sysfs device type for the network logical device to allow
2057 * fine-grained identification of different network device types. For
2058 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2060 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2062 /* Default NAPI poll() weight
2063 * Device drivers are strongly advised to not use bigger value
2065 #define NAPI_POLL_WEIGHT 64
2068 * netif_napi_add - initialize a NAPI context
2069 * @dev: network device
2070 * @napi: NAPI context
2071 * @poll: polling function
2072 * @weight: default weight
2074 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2075 * *any* of the other NAPI-related functions.
2077 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2078 int (*poll)(struct napi_struct *, int), int weight);
2081 * netif_tx_napi_add - initialize a NAPI context
2082 * @dev: network device
2083 * @napi: NAPI context
2084 * @poll: polling function
2085 * @weight: default weight
2087 * This variant of netif_napi_add() should be used from drivers using NAPI
2088 * to exclusively poll a TX queue.
2089 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2091 static inline void netif_tx_napi_add(struct net_device *dev,
2092 struct napi_struct *napi,
2093 int (*poll)(struct napi_struct *, int),
2096 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2097 netif_napi_add(dev, napi, poll, weight);
2101 * netif_napi_del - remove a NAPI context
2102 * @napi: NAPI context
2104 * netif_napi_del() removes a NAPI context from the network device NAPI list
2106 void netif_napi_del(struct napi_struct *napi);
2108 struct napi_gro_cb {
2109 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2112 /* Length of frag0. */
2113 unsigned int frag0_len;
2115 /* This indicates where we are processing relative to skb->data. */
2118 /* This is non-zero if the packet cannot be merged with the new skb. */
2121 /* Save the IP ID here and check when we get to the transport layer */
2124 /* Number of segments aggregated. */
2127 /* Start offset for remote checksum offload */
2128 u16 gro_remcsum_start;
2130 /* jiffies when first packet was created/queued */
2133 /* Used in ipv6_gro_receive() and foo-over-udp */
2136 /* This is non-zero if the packet may be of the same flow. */
2139 /* Used in tunnel GRO receive */
2142 /* GRO checksum is valid */
2145 /* Number of checksums via CHECKSUM_UNNECESSARY */
2150 #define NAPI_GRO_FREE 1
2151 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2153 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2156 /* Used in GRE, set in fou/gue_gro_receive */
2159 /* Used to determine if flush_id can be ignored */
2162 /* Number of gro_receive callbacks this packet already went through */
2163 u8 recursion_counter:4;
2167 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2170 /* used in skb_gro_receive() slow path */
2171 struct sk_buff *last;
2174 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2176 #define GRO_RECURSION_LIMIT 15
2177 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2179 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2182 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2183 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2184 struct sk_buff **head,
2185 struct sk_buff *skb)
2187 if (unlikely(gro_recursion_inc_test(skb))) {
2188 NAPI_GRO_CB(skb)->flush |= 1;
2192 return cb(head, skb);
2195 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2197 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2199 struct sk_buff **head,
2200 struct sk_buff *skb)
2202 if (unlikely(gro_recursion_inc_test(skb))) {
2203 NAPI_GRO_CB(skb)->flush |= 1;
2207 return cb(sk, head, skb);
2210 struct packet_type {
2211 __be16 type; /* This is really htons(ether_type). */
2212 struct net_device *dev; /* NULL is wildcarded here */
2213 int (*func) (struct sk_buff *,
2214 struct net_device *,
2215 struct packet_type *,
2216 struct net_device *);
2217 bool (*id_match)(struct packet_type *ptype,
2219 void *af_packet_priv;
2220 struct list_head list;
2223 struct offload_callbacks {
2224 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2225 netdev_features_t features);
2226 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2227 struct sk_buff *skb);
2228 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2231 struct packet_offload {
2232 __be16 type; /* This is really htons(ether_type). */
2234 struct offload_callbacks callbacks;
2235 struct list_head list;
2238 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2239 struct pcpu_sw_netstats {
2244 struct u64_stats_sync syncp;
2247 #define __netdev_alloc_pcpu_stats(type, gfp) \
2249 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2252 for_each_possible_cpu(__cpu) { \
2253 typeof(type) *stat; \
2254 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2255 u64_stats_init(&stat->syncp); \
2261 #define netdev_alloc_pcpu_stats(type) \
2262 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2264 enum netdev_lag_tx_type {
2265 NETDEV_LAG_TX_TYPE_UNKNOWN,
2266 NETDEV_LAG_TX_TYPE_RANDOM,
2267 NETDEV_LAG_TX_TYPE_BROADCAST,
2268 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2269 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2270 NETDEV_LAG_TX_TYPE_HASH,
2273 struct netdev_lag_upper_info {
2274 enum netdev_lag_tx_type tx_type;
2277 struct netdev_lag_lower_state_info {
2282 #include <linux/notifier.h>
2284 /* netdevice notifier chain. Please remember to update the rtnetlink
2285 * notification exclusion list in rtnetlink_event() when adding new
2288 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2289 #define NETDEV_DOWN 0x0002
2290 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2291 detected a hardware crash and restarted
2292 - we can use this eg to kick tcp sessions
2294 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2295 #define NETDEV_REGISTER 0x0005
2296 #define NETDEV_UNREGISTER 0x0006
2297 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2298 #define NETDEV_CHANGEADDR 0x0008
2299 #define NETDEV_GOING_DOWN 0x0009
2300 #define NETDEV_CHANGENAME 0x000A
2301 #define NETDEV_FEAT_CHANGE 0x000B
2302 #define NETDEV_BONDING_FAILOVER 0x000C
2303 #define NETDEV_PRE_UP 0x000D
2304 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2305 #define NETDEV_POST_TYPE_CHANGE 0x000F
2306 #define NETDEV_POST_INIT 0x0010
2307 #define NETDEV_UNREGISTER_FINAL 0x0011
2308 #define NETDEV_RELEASE 0x0012
2309 #define NETDEV_NOTIFY_PEERS 0x0013
2310 #define NETDEV_JOIN 0x0014
2311 #define NETDEV_CHANGEUPPER 0x0015
2312 #define NETDEV_RESEND_IGMP 0x0016
2313 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2314 #define NETDEV_CHANGEINFODATA 0x0018
2315 #define NETDEV_BONDING_INFO 0x0019
2316 #define NETDEV_PRECHANGEUPPER 0x001A
2317 #define NETDEV_CHANGELOWERSTATE 0x001B
2318 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2319 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2321 int register_netdevice_notifier(struct notifier_block *nb);
2322 int unregister_netdevice_notifier(struct notifier_block *nb);
2324 struct netdev_notifier_info {
2325 struct net_device *dev;
2328 struct netdev_notifier_change_info {
2329 struct netdev_notifier_info info; /* must be first */
2330 unsigned int flags_changed;
2333 struct netdev_notifier_changeupper_info {
2334 struct netdev_notifier_info info; /* must be first */
2335 struct net_device *upper_dev; /* new upper dev */
2336 bool master; /* is upper dev master */
2337 bool linking; /* is the notification for link or unlink */
2338 void *upper_info; /* upper dev info */
2341 struct netdev_notifier_changelowerstate_info {
2342 struct netdev_notifier_info info; /* must be first */
2343 void *lower_state_info; /* is lower dev state */
2346 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2347 struct net_device *dev)
2352 static inline struct net_device *
2353 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2358 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2361 extern rwlock_t dev_base_lock; /* Device list lock */
2363 #define for_each_netdev(net, d) \
2364 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2365 #define for_each_netdev_reverse(net, d) \
2366 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2367 #define for_each_netdev_rcu(net, d) \
2368 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2369 #define for_each_netdev_safe(net, d, n) \
2370 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2371 #define for_each_netdev_continue(net, d) \
2372 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2373 #define for_each_netdev_continue_rcu(net, d) \
2374 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2375 #define for_each_netdev_in_bond_rcu(bond, slave) \
2376 for_each_netdev_rcu(&init_net, slave) \
2377 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2378 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2380 static inline struct net_device *next_net_device(struct net_device *dev)
2382 struct list_head *lh;
2386 lh = dev->dev_list.next;
2387 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2390 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2392 struct list_head *lh;
2396 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2397 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2400 static inline struct net_device *first_net_device(struct net *net)
2402 return list_empty(&net->dev_base_head) ? NULL :
2403 net_device_entry(net->dev_base_head.next);
2406 static inline struct net_device *first_net_device_rcu(struct net *net)
2408 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2410 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2413 int netdev_boot_setup_check(struct net_device *dev);
2414 unsigned long netdev_boot_base(const char *prefix, int unit);
2415 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2416 const char *hwaddr);
2417 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2418 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2419 void dev_add_pack(struct packet_type *pt);
2420 void dev_remove_pack(struct packet_type *pt);
2421 void __dev_remove_pack(struct packet_type *pt);
2422 void dev_add_offload(struct packet_offload *po);
2423 void dev_remove_offload(struct packet_offload *po);
2425 int dev_get_iflink(const struct net_device *dev);
2426 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2427 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2428 unsigned short mask);
2429 struct net_device *dev_get_by_name(struct net *net, const char *name);
2430 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2431 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2432 int dev_alloc_name(struct net_device *dev, const char *name);
2433 int dev_open(struct net_device *dev);
2434 int dev_close(struct net_device *dev);
2435 int dev_close_many(struct list_head *head, bool unlink);
2436 void dev_disable_lro(struct net_device *dev);
2437 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2438 int dev_queue_xmit(struct sk_buff *skb);
2439 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2440 int register_netdevice(struct net_device *dev);
2441 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2442 void unregister_netdevice_many(struct list_head *head);
2443 static inline void unregister_netdevice(struct net_device *dev)
2445 unregister_netdevice_queue(dev, NULL);
2448 int netdev_refcnt_read(const struct net_device *dev);
2449 void free_netdev(struct net_device *dev);
2450 void netdev_freemem(struct net_device *dev);
2451 void synchronize_net(void);
2452 int init_dummy_netdev(struct net_device *dev);
2454 DECLARE_PER_CPU(int, xmit_recursion);
2455 #define XMIT_RECURSION_LIMIT 10
2457 static inline int dev_recursion_level(void)
2459 return this_cpu_read(xmit_recursion);
2462 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2463 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2464 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2465 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2466 int netdev_get_name(struct net *net, char *name, int ifindex);
2467 int dev_restart(struct net_device *dev);
2468 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2470 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2472 return NAPI_GRO_CB(skb)->data_offset;
2475 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2477 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2480 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2482 NAPI_GRO_CB(skb)->data_offset += len;
2485 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2486 unsigned int offset)
2488 return NAPI_GRO_CB(skb)->frag0 + offset;
2491 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2493 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2496 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2498 NAPI_GRO_CB(skb)->frag0 = NULL;
2499 NAPI_GRO_CB(skb)->frag0_len = 0;
2502 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2503 unsigned int offset)
2505 if (!pskb_may_pull(skb, hlen))
2508 skb_gro_frag0_invalidate(skb);
2509 return skb->data + offset;
2512 static inline void *skb_gro_network_header(struct sk_buff *skb)
2514 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2515 skb_network_offset(skb);
2518 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2519 const void *start, unsigned int len)
2521 if (NAPI_GRO_CB(skb)->csum_valid)
2522 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2523 csum_partial(start, len, 0));
2526 /* GRO checksum functions. These are logical equivalents of the normal
2527 * checksum functions (in skbuff.h) except that they operate on the GRO
2528 * offsets and fields in sk_buff.
2531 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2533 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2535 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2538 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2542 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2543 skb_checksum_start_offset(skb) <
2544 skb_gro_offset(skb)) &&
2545 !skb_at_gro_remcsum_start(skb) &&
2546 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2547 (!zero_okay || check));
2550 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2553 if (NAPI_GRO_CB(skb)->csum_valid &&
2554 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2557 NAPI_GRO_CB(skb)->csum = psum;
2559 return __skb_gro_checksum_complete(skb);
2562 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2564 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2565 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2566 NAPI_GRO_CB(skb)->csum_cnt--;
2568 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2569 * verified a new top level checksum or an encapsulated one
2570 * during GRO. This saves work if we fallback to normal path.
2572 __skb_incr_checksum_unnecessary(skb);
2576 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2579 __sum16 __ret = 0; \
2580 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2581 __ret = __skb_gro_checksum_validate_complete(skb, \
2582 compute_pseudo(skb, proto)); \
2584 skb_gro_incr_csum_unnecessary(skb); \
2588 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2589 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2591 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2593 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2595 #define skb_gro_checksum_simple_validate(skb) \
2596 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2598 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2600 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2601 !NAPI_GRO_CB(skb)->csum_valid);
2604 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2605 __sum16 check, __wsum pseudo)
2607 NAPI_GRO_CB(skb)->csum = ~pseudo;
2608 NAPI_GRO_CB(skb)->csum_valid = 1;
2611 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2613 if (__skb_gro_checksum_convert_check(skb)) \
2614 __skb_gro_checksum_convert(skb, check, \
2615 compute_pseudo(skb, proto)); \
2618 struct gro_remcsum {
2623 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2629 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2630 unsigned int off, size_t hdrlen,
2631 int start, int offset,
2632 struct gro_remcsum *grc,
2636 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2638 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2641 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2645 ptr = skb_gro_header_fast(skb, off);
2646 if (skb_gro_header_hard(skb, off + plen)) {
2647 ptr = skb_gro_header_slow(skb, off + plen, off);
2652 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2655 /* Adjust skb->csum since we changed the packet */
2656 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2658 grc->offset = off + hdrlen + offset;
2664 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2665 struct gro_remcsum *grc)
2668 size_t plen = grc->offset + sizeof(u16);
2673 ptr = skb_gro_header_fast(skb, grc->offset);
2674 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2675 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2680 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2683 #ifdef CONFIG_XFRM_OFFLOAD
2684 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2686 if (PTR_ERR(pp) != -EINPROGRESS)
2687 NAPI_GRO_CB(skb)->flush |= flush;
2690 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2692 NAPI_GRO_CB(skb)->flush |= flush;
2696 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2697 unsigned short type,
2698 const void *daddr, const void *saddr,
2701 if (!dev->header_ops || !dev->header_ops->create)
2704 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2707 static inline int dev_parse_header(const struct sk_buff *skb,
2708 unsigned char *haddr)
2710 const struct net_device *dev = skb->dev;
2712 if (!dev->header_ops || !dev->header_ops->parse)
2714 return dev->header_ops->parse(skb, haddr);
2717 /* ll_header must have at least hard_header_len allocated */
2718 static inline bool dev_validate_header(const struct net_device *dev,
2719 char *ll_header, int len)
2721 if (likely(len >= dev->hard_header_len))
2723 if (len < dev->min_header_len)
2726 if (capable(CAP_SYS_RAWIO)) {
2727 memset(ll_header + len, 0, dev->hard_header_len - len);
2731 if (dev->header_ops && dev->header_ops->validate)
2732 return dev->header_ops->validate(ll_header, len);
2737 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2738 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2739 static inline int unregister_gifconf(unsigned int family)
2741 return register_gifconf(family, NULL);
2744 #ifdef CONFIG_NET_FLOW_LIMIT
2745 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2746 struct sd_flow_limit {
2748 unsigned int num_buckets;
2749 unsigned int history_head;
2750 u16 history[FLOW_LIMIT_HISTORY];
2754 extern int netdev_flow_limit_table_len;
2755 #endif /* CONFIG_NET_FLOW_LIMIT */
2758 * Incoming packets are placed on per-CPU queues
2760 struct softnet_data {
2761 struct list_head poll_list;
2762 struct sk_buff_head process_queue;
2765 unsigned int processed;
2766 unsigned int time_squeeze;
2767 unsigned int received_rps;
2769 struct softnet_data *rps_ipi_list;
2771 #ifdef CONFIG_NET_FLOW_LIMIT
2772 struct sd_flow_limit __rcu *flow_limit;
2774 struct Qdisc *output_queue;
2775 struct Qdisc **output_queue_tailp;
2776 struct sk_buff *completion_queue;
2779 /* input_queue_head should be written by cpu owning this struct,
2780 * and only read by other cpus. Worth using a cache line.
2782 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2784 /* Elements below can be accessed between CPUs for RPS/RFS */
2785 struct call_single_data csd ____cacheline_aligned_in_smp;
2786 struct softnet_data *rps_ipi_next;
2788 unsigned int input_queue_tail;
2790 unsigned int dropped;
2791 struct sk_buff_head input_pkt_queue;
2792 struct napi_struct backlog;
2796 static inline void input_queue_head_incr(struct softnet_data *sd)
2799 sd->input_queue_head++;
2803 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2804 unsigned int *qtail)
2807 *qtail = ++sd->input_queue_tail;
2811 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2813 void __netif_schedule(struct Qdisc *q);
2814 void netif_schedule_queue(struct netdev_queue *txq);
2816 static inline void netif_tx_schedule_all(struct net_device *dev)
2820 for (i = 0; i < dev->num_tx_queues; i++)
2821 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2824 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2826 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2830 * netif_start_queue - allow transmit
2831 * @dev: network device
2833 * Allow upper layers to call the device hard_start_xmit routine.
2835 static inline void netif_start_queue(struct net_device *dev)
2837 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2840 static inline void netif_tx_start_all_queues(struct net_device *dev)
2844 for (i = 0; i < dev->num_tx_queues; i++) {
2845 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2846 netif_tx_start_queue(txq);
2850 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2853 * netif_wake_queue - restart transmit
2854 * @dev: network device
2856 * Allow upper layers to call the device hard_start_xmit routine.
2857 * Used for flow control when transmit resources are available.
2859 static inline void netif_wake_queue(struct net_device *dev)
2861 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2864 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2868 for (i = 0; i < dev->num_tx_queues; i++) {
2869 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2870 netif_tx_wake_queue(txq);
2874 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2876 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2880 * netif_stop_queue - stop transmitted packets
2881 * @dev: network device
2883 * Stop upper layers calling the device hard_start_xmit routine.
2884 * Used for flow control when transmit resources are unavailable.
2886 static inline void netif_stop_queue(struct net_device *dev)
2888 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2891 void netif_tx_stop_all_queues(struct net_device *dev);
2893 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2895 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2899 * netif_queue_stopped - test if transmit queue is flowblocked
2900 * @dev: network device
2902 * Test if transmit queue on device is currently unable to send.
2904 static inline bool netif_queue_stopped(const struct net_device *dev)
2906 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2909 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2911 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2915 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2917 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2921 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2923 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2927 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2928 * @dev_queue: pointer to transmit queue
2930 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2931 * to give appropriate hint to the CPU.
2933 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2936 prefetchw(&dev_queue->dql.num_queued);
2941 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2942 * @dev_queue: pointer to transmit queue
2944 * BQL enabled drivers might use this helper in their TX completion path,
2945 * to give appropriate hint to the CPU.
2947 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2950 prefetchw(&dev_queue->dql.limit);
2954 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2958 dql_queued(&dev_queue->dql, bytes);
2960 if (likely(dql_avail(&dev_queue->dql) >= 0))
2963 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2966 * The XOFF flag must be set before checking the dql_avail below,
2967 * because in netdev_tx_completed_queue we update the dql_completed
2968 * before checking the XOFF flag.
2972 /* check again in case another CPU has just made room avail */
2973 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2974 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2979 * netdev_sent_queue - report the number of bytes queued to hardware
2980 * @dev: network device
2981 * @bytes: number of bytes queued to the hardware device queue
2983 * Report the number of bytes queued for sending/completion to the network
2984 * device hardware queue. @bytes should be a good approximation and should
2985 * exactly match netdev_completed_queue() @bytes
2987 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2989 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2992 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2993 unsigned int pkts, unsigned int bytes)
2996 if (unlikely(!bytes))
2999 dql_completed(&dev_queue->dql, bytes);
3002 * Without the memory barrier there is a small possiblity that
3003 * netdev_tx_sent_queue will miss the update and cause the queue to
3004 * be stopped forever
3008 if (dql_avail(&dev_queue->dql) < 0)
3011 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3012 netif_schedule_queue(dev_queue);
3017 * netdev_completed_queue - report bytes and packets completed by device
3018 * @dev: network device
3019 * @pkts: actual number of packets sent over the medium
3020 * @bytes: actual number of bytes sent over the medium
3022 * Report the number of bytes and packets transmitted by the network device
3023 * hardware queue over the physical medium, @bytes must exactly match the
3024 * @bytes amount passed to netdev_sent_queue()
3026 static inline void netdev_completed_queue(struct net_device *dev,
3027 unsigned int pkts, unsigned int bytes)
3029 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3032 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3035 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3041 * netdev_reset_queue - reset the packets and bytes count of a network device
3042 * @dev_queue: network device
3044 * Reset the bytes and packet count of a network device and clear the
3045 * software flow control OFF bit for this network device
3047 static inline void netdev_reset_queue(struct net_device *dev_queue)
3049 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3053 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3054 * @dev: network device
3055 * @queue_index: given tx queue index
3057 * Returns 0 if given tx queue index >= number of device tx queues,
3058 * otherwise returns the originally passed tx queue index.
3060 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3062 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3063 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3064 dev->name, queue_index,
3065 dev->real_num_tx_queues);
3073 * netif_running - test if up
3074 * @dev: network device
3076 * Test if the device has been brought up.
3078 static inline bool netif_running(const struct net_device *dev)
3080 return test_bit(__LINK_STATE_START, &dev->state);
3084 * Routines to manage the subqueues on a device. We only need start,
3085 * stop, and a check if it's stopped. All other device management is
3086 * done at the overall netdevice level.
3087 * Also test the device if we're multiqueue.
3091 * netif_start_subqueue - allow sending packets on subqueue
3092 * @dev: network device
3093 * @queue_index: sub queue index
3095 * Start individual transmit queue of a device with multiple transmit queues.
3097 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3099 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3101 netif_tx_start_queue(txq);
3105 * netif_stop_subqueue - stop sending packets on subqueue
3106 * @dev: network device
3107 * @queue_index: sub queue index
3109 * Stop individual transmit queue of a device with multiple transmit queues.
3111 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3113 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3114 netif_tx_stop_queue(txq);
3118 * netif_subqueue_stopped - test status of subqueue
3119 * @dev: network device
3120 * @queue_index: sub queue index
3122 * Check individual transmit queue of a device with multiple transmit queues.
3124 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3127 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3129 return netif_tx_queue_stopped(txq);
3132 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3133 struct sk_buff *skb)
3135 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3139 * netif_wake_subqueue - allow sending packets on subqueue
3140 * @dev: network device
3141 * @queue_index: sub queue index
3143 * Resume individual transmit queue of a device with multiple transmit queues.
3145 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3147 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3149 netif_tx_wake_queue(txq);
3153 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3156 static inline int netif_set_xps_queue(struct net_device *dev,
3157 const struct cpumask *mask,
3164 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3165 unsigned int num_tx_queues);
3168 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3169 * as a distribution range limit for the returned value.
3171 static inline u16 skb_tx_hash(const struct net_device *dev,
3172 struct sk_buff *skb)
3174 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3178 * netif_is_multiqueue - test if device has multiple transmit queues
3179 * @dev: network device
3181 * Check if device has multiple transmit queues
3183 static inline bool netif_is_multiqueue(const struct net_device *dev)
3185 return dev->num_tx_queues > 1;
3188 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3191 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3193 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3201 static inline unsigned int get_netdev_rx_queue_index(
3202 struct netdev_rx_queue *queue)
3204 struct net_device *dev = queue->dev;
3205 int index = queue - dev->_rx;
3207 BUG_ON(index >= dev->num_rx_queues);
3212 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3213 int netif_get_num_default_rss_queues(void);
3215 enum skb_free_reason {
3216 SKB_REASON_CONSUMED,
3220 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3221 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3224 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3225 * interrupt context or with hardware interrupts being disabled.
3226 * (in_irq() || irqs_disabled())
3228 * We provide four helpers that can be used in following contexts :
3230 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3231 * replacing kfree_skb(skb)
3233 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3234 * Typically used in place of consume_skb(skb) in TX completion path
3236 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3237 * replacing kfree_skb(skb)
3239 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3240 * and consumed a packet. Used in place of consume_skb(skb)
3242 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3244 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3247 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3249 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3252 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3254 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3257 static inline void dev_consume_skb_any(struct sk_buff *skb)
3259 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3262 int netif_rx(struct sk_buff *skb);
3263 int netif_rx_ni(struct sk_buff *skb);
3264 int netif_receive_skb(struct sk_buff *skb);
3265 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3266 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3267 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3268 gro_result_t napi_gro_frags(struct napi_struct *napi);
3269 struct packet_offload *gro_find_receive_by_type(__be16 type);
3270 struct packet_offload *gro_find_complete_by_type(__be16 type);
3272 static inline void napi_free_frags(struct napi_struct *napi)
3274 kfree_skb(napi->skb);
3278 bool netdev_is_rx_handler_busy(struct net_device *dev);
3279 int netdev_rx_handler_register(struct net_device *dev,
3280 rx_handler_func_t *rx_handler,
3281 void *rx_handler_data);
3282 void netdev_rx_handler_unregister(struct net_device *dev);
3284 bool dev_valid_name(const char *name);
3285 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3286 int dev_ethtool(struct net *net, struct ifreq *);
3287 unsigned int dev_get_flags(const struct net_device *);
3288 int __dev_change_flags(struct net_device *, unsigned int flags);
3289 int dev_change_flags(struct net_device *, unsigned int);
3290 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3291 unsigned int gchanges);
3292 int dev_change_name(struct net_device *, const char *);
3293 int dev_set_alias(struct net_device *, const char *, size_t);
3294 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3295 int dev_set_mtu(struct net_device *, int);
3296 void dev_set_group(struct net_device *, int);
3297 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3298 int dev_change_carrier(struct net_device *, bool new_carrier);
3299 int dev_get_phys_port_id(struct net_device *dev,
3300 struct netdev_phys_item_id *ppid);
3301 int dev_get_phys_port_name(struct net_device *dev,
3302 char *name, size_t len);
3303 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3304 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3305 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3306 struct netdev_queue *txq, int *ret);
3308 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3309 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3311 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3313 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3314 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3315 bool is_skb_forwardable(const struct net_device *dev,
3316 const struct sk_buff *skb);
3318 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3319 struct sk_buff *skb)
3321 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3322 unlikely(!is_skb_forwardable(dev, skb))) {
3323 atomic_long_inc(&dev->rx_dropped);
3328 skb_scrub_packet(skb, true);
3333 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3335 extern int netdev_budget;
3336 extern unsigned int netdev_budget_usecs;
3338 /* Called by rtnetlink.c:rtnl_unlock() */
3339 void netdev_run_todo(void);
3342 * dev_put - release reference to device
3343 * @dev: network device
3345 * Release reference to device to allow it to be freed.
3347 static inline void dev_put(struct net_device *dev)
3349 this_cpu_dec(*dev->pcpu_refcnt);
3353 * dev_hold - get reference to device
3354 * @dev: network device
3356 * Hold reference to device to keep it from being freed.
3358 static inline void dev_hold(struct net_device *dev)
3360 this_cpu_inc(*dev->pcpu_refcnt);
3363 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3364 * and _off may be called from IRQ context, but it is caller
3365 * who is responsible for serialization of these calls.
3367 * The name carrier is inappropriate, these functions should really be
3368 * called netif_lowerlayer_*() because they represent the state of any
3369 * kind of lower layer not just hardware media.
3372 void linkwatch_init_dev(struct net_device *dev);
3373 void linkwatch_fire_event(struct net_device *dev);
3374 void linkwatch_forget_dev(struct net_device *dev);
3377 * netif_carrier_ok - test if carrier present
3378 * @dev: network device
3380 * Check if carrier is present on device
3382 static inline bool netif_carrier_ok(const struct net_device *dev)
3384 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3387 unsigned long dev_trans_start(struct net_device *dev);
3389 void __netdev_watchdog_up(struct net_device *dev);
3391 void netif_carrier_on(struct net_device *dev);
3393 void netif_carrier_off(struct net_device *dev);
3396 * netif_dormant_on - mark device as dormant.
3397 * @dev: network device
3399 * Mark device as dormant (as per RFC2863).
3401 * The dormant state indicates that the relevant interface is not
3402 * actually in a condition to pass packets (i.e., it is not 'up') but is
3403 * in a "pending" state, waiting for some external event. For "on-
3404 * demand" interfaces, this new state identifies the situation where the
3405 * interface is waiting for events to place it in the up state.
3407 static inline void netif_dormant_on(struct net_device *dev)
3409 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3410 linkwatch_fire_event(dev);
3414 * netif_dormant_off - set device as not dormant.
3415 * @dev: network device
3417 * Device is not in dormant state.
3419 static inline void netif_dormant_off(struct net_device *dev)
3421 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3422 linkwatch_fire_event(dev);
3426 * netif_dormant - test if device is dormant
3427 * @dev: network device
3429 * Check if device is dormant.
3431 static inline bool netif_dormant(const struct net_device *dev)
3433 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3438 * netif_oper_up - test if device is operational
3439 * @dev: network device
3441 * Check if carrier is operational
3443 static inline bool netif_oper_up(const struct net_device *dev)
3445 return (dev->operstate == IF_OPER_UP ||
3446 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3450 * netif_device_present - is device available or removed
3451 * @dev: network device
3453 * Check if device has not been removed from system.
3455 static inline bool netif_device_present(struct net_device *dev)
3457 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3460 void netif_device_detach(struct net_device *dev);
3462 void netif_device_attach(struct net_device *dev);
3465 * Network interface message level settings
3469 NETIF_MSG_DRV = 0x0001,
3470 NETIF_MSG_PROBE = 0x0002,
3471 NETIF_MSG_LINK = 0x0004,
3472 NETIF_MSG_TIMER = 0x0008,
3473 NETIF_MSG_IFDOWN = 0x0010,
3474 NETIF_MSG_IFUP = 0x0020,
3475 NETIF_MSG_RX_ERR = 0x0040,
3476 NETIF_MSG_TX_ERR = 0x0080,
3477 NETIF_MSG_TX_QUEUED = 0x0100,
3478 NETIF_MSG_INTR = 0x0200,
3479 NETIF_MSG_TX_DONE = 0x0400,
3480 NETIF_MSG_RX_STATUS = 0x0800,
3481 NETIF_MSG_PKTDATA = 0x1000,
3482 NETIF_MSG_HW = 0x2000,
3483 NETIF_MSG_WOL = 0x4000,
3486 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3487 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3488 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3489 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3490 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3491 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3492 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3493 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3494 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3495 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3496 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3497 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3498 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3499 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3500 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3502 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3505 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3506 return default_msg_enable_bits;
3507 if (debug_value == 0) /* no output */
3509 /* set low N bits */
3510 return (1 << debug_value) - 1;
3513 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3515 spin_lock(&txq->_xmit_lock);
3516 txq->xmit_lock_owner = cpu;
3519 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3521 __acquire(&txq->_xmit_lock);
3525 static inline void __netif_tx_release(struct netdev_queue *txq)
3527 __release(&txq->_xmit_lock);
3530 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3532 spin_lock_bh(&txq->_xmit_lock);
3533 txq->xmit_lock_owner = smp_processor_id();
3536 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3538 bool ok = spin_trylock(&txq->_xmit_lock);
3540 txq->xmit_lock_owner = smp_processor_id();
3544 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3546 txq->xmit_lock_owner = -1;
3547 spin_unlock(&txq->_xmit_lock);
3550 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3552 txq->xmit_lock_owner = -1;
3553 spin_unlock_bh(&txq->_xmit_lock);
3556 static inline void txq_trans_update(struct netdev_queue *txq)
3558 if (txq->xmit_lock_owner != -1)
3559 txq->trans_start = jiffies;
3562 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3563 static inline void netif_trans_update(struct net_device *dev)
3565 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3567 if (txq->trans_start != jiffies)
3568 txq->trans_start = jiffies;
3572 * netif_tx_lock - grab network device transmit lock
3573 * @dev: network device
3575 * Get network device transmit lock
3577 static inline void netif_tx_lock(struct net_device *dev)
3582 spin_lock(&dev->tx_global_lock);
3583 cpu = smp_processor_id();
3584 for (i = 0; i < dev->num_tx_queues; i++) {
3585 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3587 /* We are the only thread of execution doing a
3588 * freeze, but we have to grab the _xmit_lock in
3589 * order to synchronize with threads which are in
3590 * the ->hard_start_xmit() handler and already
3591 * checked the frozen bit.
3593 __netif_tx_lock(txq, cpu);
3594 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3595 __netif_tx_unlock(txq);
3599 static inline void netif_tx_lock_bh(struct net_device *dev)
3605 static inline void netif_tx_unlock(struct net_device *dev)
3609 for (i = 0; i < dev->num_tx_queues; i++) {
3610 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3612 /* No need to grab the _xmit_lock here. If the
3613 * queue is not stopped for another reason, we
3616 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3617 netif_schedule_queue(txq);
3619 spin_unlock(&dev->tx_global_lock);
3622 static inline void netif_tx_unlock_bh(struct net_device *dev)
3624 netif_tx_unlock(dev);
3628 #define HARD_TX_LOCK(dev, txq, cpu) { \
3629 if ((dev->features & NETIF_F_LLTX) == 0) { \
3630 __netif_tx_lock(txq, cpu); \
3632 __netif_tx_acquire(txq); \
3636 #define HARD_TX_TRYLOCK(dev, txq) \
3637 (((dev->features & NETIF_F_LLTX) == 0) ? \
3638 __netif_tx_trylock(txq) : \
3639 __netif_tx_acquire(txq))
3641 #define HARD_TX_UNLOCK(dev, txq) { \
3642 if ((dev->features & NETIF_F_LLTX) == 0) { \
3643 __netif_tx_unlock(txq); \
3645 __netif_tx_release(txq); \
3649 static inline void netif_tx_disable(struct net_device *dev)
3655 cpu = smp_processor_id();
3656 for (i = 0; i < dev->num_tx_queues; i++) {
3657 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3659 __netif_tx_lock(txq, cpu);
3660 netif_tx_stop_queue(txq);
3661 __netif_tx_unlock(txq);
3666 static inline void netif_addr_lock(struct net_device *dev)
3668 spin_lock(&dev->addr_list_lock);
3671 static inline void netif_addr_lock_nested(struct net_device *dev)
3673 int subclass = SINGLE_DEPTH_NESTING;
3675 if (dev->netdev_ops->ndo_get_lock_subclass)
3676 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3678 spin_lock_nested(&dev->addr_list_lock, subclass);
3681 static inline void netif_addr_lock_bh(struct net_device *dev)
3683 spin_lock_bh(&dev->addr_list_lock);
3686 static inline void netif_addr_unlock(struct net_device *dev)
3688 spin_unlock(&dev->addr_list_lock);
3691 static inline void netif_addr_unlock_bh(struct net_device *dev)
3693 spin_unlock_bh(&dev->addr_list_lock);
3697 * dev_addrs walker. Should be used only for read access. Call with
3698 * rcu_read_lock held.
3700 #define for_each_dev_addr(dev, ha) \
3701 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3703 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3705 void ether_setup(struct net_device *dev);
3707 /* Support for loadable net-drivers */
3708 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3709 unsigned char name_assign_type,
3710 void (*setup)(struct net_device *),
3711 unsigned int txqs, unsigned int rxqs);
3712 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3713 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3715 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3716 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3719 int register_netdev(struct net_device *dev);
3720 void unregister_netdev(struct net_device *dev);
3722 /* General hardware address lists handling functions */
3723 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3724 struct netdev_hw_addr_list *from_list, int addr_len);
3725 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3726 struct netdev_hw_addr_list *from_list, int addr_len);
3727 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3728 struct net_device *dev,
3729 int (*sync)(struct net_device *, const unsigned char *),
3730 int (*unsync)(struct net_device *,
3731 const unsigned char *));
3732 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3733 struct net_device *dev,
3734 int (*unsync)(struct net_device *,
3735 const unsigned char *));
3736 void __hw_addr_init(struct netdev_hw_addr_list *list);
3738 /* Functions used for device addresses handling */
3739 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3740 unsigned char addr_type);
3741 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3742 unsigned char addr_type);
3743 void dev_addr_flush(struct net_device *dev);
3744 int dev_addr_init(struct net_device *dev);
3746 /* Functions used for unicast addresses handling */
3747 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3748 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3749 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3750 int dev_uc_sync(struct net_device *to, struct net_device *from);
3751 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3752 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3753 void dev_uc_flush(struct net_device *dev);
3754 void dev_uc_init(struct net_device *dev);
3757 * __dev_uc_sync - Synchonize device's unicast list
3758 * @dev: device to sync
3759 * @sync: function to call if address should be added
3760 * @unsync: function to call if address should be removed
3762 * Add newly added addresses to the interface, and release
3763 * addresses that have been deleted.
3765 static inline int __dev_uc_sync(struct net_device *dev,
3766 int (*sync)(struct net_device *,
3767 const unsigned char *),
3768 int (*unsync)(struct net_device *,
3769 const unsigned char *))
3771 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3775 * __dev_uc_unsync - Remove synchronized addresses from device
3776 * @dev: device to sync
3777 * @unsync: function to call if address should be removed
3779 * Remove all addresses that were added to the device by dev_uc_sync().
3781 static inline void __dev_uc_unsync(struct net_device *dev,
3782 int (*unsync)(struct net_device *,
3783 const unsigned char *))
3785 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3788 /* Functions used for multicast addresses handling */
3789 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3790 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3791 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3792 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3793 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3794 int dev_mc_sync(struct net_device *to, struct net_device *from);
3795 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3796 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3797 void dev_mc_flush(struct net_device *dev);
3798 void dev_mc_init(struct net_device *dev);
3801 * __dev_mc_sync - Synchonize device's multicast list
3802 * @dev: device to sync
3803 * @sync: function to call if address should be added
3804 * @unsync: function to call if address should be removed
3806 * Add newly added addresses to the interface, and release
3807 * addresses that have been deleted.
3809 static inline int __dev_mc_sync(struct net_device *dev,
3810 int (*sync)(struct net_device *,
3811 const unsigned char *),
3812 int (*unsync)(struct net_device *,
3813 const unsigned char *))
3815 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3819 * __dev_mc_unsync - Remove synchronized addresses from device
3820 * @dev: device to sync
3821 * @unsync: function to call if address should be removed
3823 * Remove all addresses that were added to the device by dev_mc_sync().
3825 static inline void __dev_mc_unsync(struct net_device *dev,
3826 int (*unsync)(struct net_device *,
3827 const unsigned char *))
3829 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3832 /* Functions used for secondary unicast and multicast support */
3833 void dev_set_rx_mode(struct net_device *dev);
3834 void __dev_set_rx_mode(struct net_device *dev);
3835 int dev_set_promiscuity(struct net_device *dev, int inc);
3836 int dev_set_allmulti(struct net_device *dev, int inc);
3837 void netdev_state_change(struct net_device *dev);
3838 void netdev_notify_peers(struct net_device *dev);
3839 void netdev_features_change(struct net_device *dev);
3840 /* Load a device via the kmod */
3841 void dev_load(struct net *net, const char *name);
3842 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3843 struct rtnl_link_stats64 *storage);
3844 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3845 const struct net_device_stats *netdev_stats);
3847 extern int netdev_max_backlog;
3848 extern int netdev_tstamp_prequeue;
3849 extern int weight_p;
3850 extern int dev_weight_rx_bias;
3851 extern int dev_weight_tx_bias;
3852 extern int dev_rx_weight;
3853 extern int dev_tx_weight;
3855 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3856 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3857 struct list_head **iter);
3858 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3859 struct list_head **iter);
3861 /* iterate through upper list, must be called under RCU read lock */
3862 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3863 for (iter = &(dev)->adj_list.upper, \
3864 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3866 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3868 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3869 int (*fn)(struct net_device *upper_dev,
3873 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3874 struct net_device *upper_dev);
3876 void *netdev_lower_get_next_private(struct net_device *dev,
3877 struct list_head **iter);
3878 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3879 struct list_head **iter);
3881 #define netdev_for_each_lower_private(dev, priv, iter) \
3882 for (iter = (dev)->adj_list.lower.next, \
3883 priv = netdev_lower_get_next_private(dev, &(iter)); \
3885 priv = netdev_lower_get_next_private(dev, &(iter)))
3887 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3888 for (iter = &(dev)->adj_list.lower, \
3889 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3891 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3893 void *netdev_lower_get_next(struct net_device *dev,
3894 struct list_head **iter);
3896 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3897 for (iter = (dev)->adj_list.lower.next, \
3898 ldev = netdev_lower_get_next(dev, &(iter)); \
3900 ldev = netdev_lower_get_next(dev, &(iter)))
3902 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3903 struct list_head **iter);
3904 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3905 struct list_head **iter);
3907 int netdev_walk_all_lower_dev(struct net_device *dev,
3908 int (*fn)(struct net_device *lower_dev,
3911 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3912 int (*fn)(struct net_device *lower_dev,
3916 void *netdev_adjacent_get_private(struct list_head *adj_list);
3917 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3918 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3919 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3920 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3921 int netdev_master_upper_dev_link(struct net_device *dev,
3922 struct net_device *upper_dev,
3923 void *upper_priv, void *upper_info);
3924 void netdev_upper_dev_unlink(struct net_device *dev,
3925 struct net_device *upper_dev);
3926 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3927 void *netdev_lower_dev_get_private(struct net_device *dev,
3928 struct net_device *lower_dev);
3929 void netdev_lower_state_changed(struct net_device *lower_dev,
3930 void *lower_state_info);
3932 /* RSS keys are 40 or 52 bytes long */
3933 #define NETDEV_RSS_KEY_LEN 52
3934 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3935 void netdev_rss_key_fill(void *buffer, size_t len);
3937 int dev_get_nest_level(struct net_device *dev);
3938 int skb_checksum_help(struct sk_buff *skb);
3939 int skb_crc32c_csum_help(struct sk_buff *skb);
3940 int skb_csum_hwoffload_help(struct sk_buff *skb,
3941 const netdev_features_t features);
3943 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3944 netdev_features_t features, bool tx_path);
3945 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3946 netdev_features_t features);
3948 struct netdev_bonding_info {
3953 struct netdev_notifier_bonding_info {
3954 struct netdev_notifier_info info; /* must be first */
3955 struct netdev_bonding_info bonding_info;
3958 void netdev_bonding_info_change(struct net_device *dev,
3959 struct netdev_bonding_info *bonding_info);
3962 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3964 return __skb_gso_segment(skb, features, true);
3966 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3968 static inline bool can_checksum_protocol(netdev_features_t features,
3971 if (protocol == htons(ETH_P_FCOE))
3972 return !!(features & NETIF_F_FCOE_CRC);
3974 /* Assume this is an IP checksum (not SCTP CRC) */
3976 if (features & NETIF_F_HW_CSUM) {
3977 /* Can checksum everything */
3982 case htons(ETH_P_IP):
3983 return !!(features & NETIF_F_IP_CSUM);
3984 case htons(ETH_P_IPV6):
3985 return !!(features & NETIF_F_IPV6_CSUM);
3992 void netdev_rx_csum_fault(struct net_device *dev);
3994 static inline void netdev_rx_csum_fault(struct net_device *dev)
3998 /* rx skb timestamps */
3999 void net_enable_timestamp(void);
4000 void net_disable_timestamp(void);
4002 #ifdef CONFIG_PROC_FS
4003 int __init dev_proc_init(void);
4005 #define dev_proc_init() 0
4008 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4009 struct sk_buff *skb, struct net_device *dev,
4012 skb->xmit_more = more ? 1 : 0;
4013 return ops->ndo_start_xmit(skb, dev);
4016 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4017 struct netdev_queue *txq, bool more)
4019 const struct net_device_ops *ops = dev->netdev_ops;
4022 rc = __netdev_start_xmit(ops, skb, dev, more);
4023 if (rc == NETDEV_TX_OK)
4024 txq_trans_update(txq);
4029 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4031 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4034 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4036 return netdev_class_create_file_ns(class_attr, NULL);
4039 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4041 netdev_class_remove_file_ns(class_attr, NULL);
4044 extern struct kobj_ns_type_operations net_ns_type_operations;
4046 const char *netdev_drivername(const struct net_device *dev);
4048 void linkwatch_run_queue(void);
4050 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4051 netdev_features_t f2)
4053 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4054 if (f1 & NETIF_F_HW_CSUM)
4055 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4057 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4063 static inline netdev_features_t netdev_get_wanted_features(
4064 struct net_device *dev)
4066 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4068 netdev_features_t netdev_increment_features(netdev_features_t all,
4069 netdev_features_t one, netdev_features_t mask);
4071 /* Allow TSO being used on stacked device :
4072 * Performing the GSO segmentation before last device
4073 * is a performance improvement.
4075 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4076 netdev_features_t mask)
4078 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4081 int __netdev_update_features(struct net_device *dev);
4082 void netdev_update_features(struct net_device *dev);
4083 void netdev_change_features(struct net_device *dev);
4085 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4086 struct net_device *dev);
4088 netdev_features_t passthru_features_check(struct sk_buff *skb,
4089 struct net_device *dev,
4090 netdev_features_t features);
4091 netdev_features_t netif_skb_features(struct sk_buff *skb);
4093 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4095 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4097 /* check flags correspondence */
4098 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4099 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4100 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4101 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4102 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4103 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4104 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4105 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4106 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4107 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4108 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4109 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4110 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4111 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4112 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4113 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4114 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4116 return (features & feature) == feature;
4119 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4121 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4122 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4125 static inline bool netif_needs_gso(struct sk_buff *skb,
4126 netdev_features_t features)
4128 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4129 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4130 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4133 static inline void netif_set_gso_max_size(struct net_device *dev,
4136 dev->gso_max_size = size;
4139 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4140 int pulled_hlen, u16 mac_offset,
4143 skb->protocol = protocol;
4144 skb->encapsulation = 1;
4145 skb_push(skb, pulled_hlen);
4146 skb_reset_transport_header(skb);
4147 skb->mac_header = mac_offset;
4148 skb->network_header = skb->mac_header + mac_len;
4149 skb->mac_len = mac_len;
4152 static inline bool netif_is_macsec(const struct net_device *dev)
4154 return dev->priv_flags & IFF_MACSEC;
4157 static inline bool netif_is_macvlan(const struct net_device *dev)
4159 return dev->priv_flags & IFF_MACVLAN;
4162 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4164 return dev->priv_flags & IFF_MACVLAN_PORT;
4167 static inline bool netif_is_ipvlan(const struct net_device *dev)
4169 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4172 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4174 return dev->priv_flags & IFF_IPVLAN_MASTER;
4177 static inline bool netif_is_bond_master(const struct net_device *dev)
4179 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4182 static inline bool netif_is_bond_slave(const struct net_device *dev)
4184 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4187 static inline bool netif_supports_nofcs(struct net_device *dev)
4189 return dev->priv_flags & IFF_SUPP_NOFCS;
4192 static inline bool netif_is_l3_master(const struct net_device *dev)
4194 return dev->priv_flags & IFF_L3MDEV_MASTER;
4197 static inline bool netif_is_l3_slave(const struct net_device *dev)
4199 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4202 static inline bool netif_is_bridge_master(const struct net_device *dev)
4204 return dev->priv_flags & IFF_EBRIDGE;
4207 static inline bool netif_is_bridge_port(const struct net_device *dev)
4209 return dev->priv_flags & IFF_BRIDGE_PORT;
4212 static inline bool netif_is_ovs_master(const struct net_device *dev)
4214 return dev->priv_flags & IFF_OPENVSWITCH;
4217 static inline bool netif_is_ovs_port(const struct net_device *dev)
4219 return dev->priv_flags & IFF_OVS_DATAPATH;
4222 static inline bool netif_is_team_master(const struct net_device *dev)
4224 return dev->priv_flags & IFF_TEAM;
4227 static inline bool netif_is_team_port(const struct net_device *dev)
4229 return dev->priv_flags & IFF_TEAM_PORT;
4232 static inline bool netif_is_lag_master(const struct net_device *dev)
4234 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4237 static inline bool netif_is_lag_port(const struct net_device *dev)
4239 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4242 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4244 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4247 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4248 static inline void netif_keep_dst(struct net_device *dev)
4250 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4253 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4254 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4256 /* TODO: reserve and use an additional IFF bit, if we get more users */
4257 return dev->priv_flags & IFF_MACSEC;
4260 extern struct pernet_operations __net_initdata loopback_net_ops;
4262 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4264 /* netdev_printk helpers, similar to dev_printk */
4266 static inline const char *netdev_name(const struct net_device *dev)
4268 if (!dev->name[0] || strchr(dev->name, '%'))
4269 return "(unnamed net_device)";
4273 static inline bool netdev_unregistering(const struct net_device *dev)
4275 return dev->reg_state == NETREG_UNREGISTERING;
4278 static inline const char *netdev_reg_state(const struct net_device *dev)
4280 switch (dev->reg_state) {
4281 case NETREG_UNINITIALIZED: return " (uninitialized)";
4282 case NETREG_REGISTERED: return "";
4283 case NETREG_UNREGISTERING: return " (unregistering)";
4284 case NETREG_UNREGISTERED: return " (unregistered)";
4285 case NETREG_RELEASED: return " (released)";
4286 case NETREG_DUMMY: return " (dummy)";
4289 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4290 return " (unknown)";
4294 void netdev_printk(const char *level, const struct net_device *dev,
4295 const char *format, ...);
4297 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4299 void netdev_alert(const struct net_device *dev, const char *format, ...);
4301 void netdev_crit(const struct net_device *dev, const char *format, ...);
4303 void netdev_err(const struct net_device *dev, const char *format, ...);
4305 void netdev_warn(const struct net_device *dev, const char *format, ...);
4307 void netdev_notice(const struct net_device *dev, const char *format, ...);
4309 void netdev_info(const struct net_device *dev, const char *format, ...);
4311 #define MODULE_ALIAS_NETDEV(device) \
4312 MODULE_ALIAS("netdev-" device)
4314 #if defined(CONFIG_DYNAMIC_DEBUG)
4315 #define netdev_dbg(__dev, format, args...) \
4317 dynamic_netdev_dbg(__dev, format, ##args); \
4319 #elif defined(DEBUG)
4320 #define netdev_dbg(__dev, format, args...) \
4321 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4323 #define netdev_dbg(__dev, format, args...) \
4326 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4330 #if defined(VERBOSE_DEBUG)
4331 #define netdev_vdbg netdev_dbg
4334 #define netdev_vdbg(dev, format, args...) \
4337 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4343 * netdev_WARN() acts like dev_printk(), but with the key difference
4344 * of using a WARN/WARN_ON to get the message out, including the
4345 * file/line information and a backtrace.
4347 #define netdev_WARN(dev, format, args...) \
4348 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4349 netdev_reg_state(dev), ##args)
4351 /* netif printk helpers, similar to netdev_printk */
4353 #define netif_printk(priv, type, level, dev, fmt, args...) \
4355 if (netif_msg_##type(priv)) \
4356 netdev_printk(level, (dev), fmt, ##args); \
4359 #define netif_level(level, priv, type, dev, fmt, args...) \
4361 if (netif_msg_##type(priv)) \
4362 netdev_##level(dev, fmt, ##args); \
4365 #define netif_emerg(priv, type, dev, fmt, args...) \
4366 netif_level(emerg, priv, type, dev, fmt, ##args)
4367 #define netif_alert(priv, type, dev, fmt, args...) \
4368 netif_level(alert, priv, type, dev, fmt, ##args)
4369 #define netif_crit(priv, type, dev, fmt, args...) \
4370 netif_level(crit, priv, type, dev, fmt, ##args)
4371 #define netif_err(priv, type, dev, fmt, args...) \
4372 netif_level(err, priv, type, dev, fmt, ##args)
4373 #define netif_warn(priv, type, dev, fmt, args...) \
4374 netif_level(warn, priv, type, dev, fmt, ##args)
4375 #define netif_notice(priv, type, dev, fmt, args...) \
4376 netif_level(notice, priv, type, dev, fmt, ##args)
4377 #define netif_info(priv, type, dev, fmt, args...) \
4378 netif_level(info, priv, type, dev, fmt, ##args)
4380 #if defined(CONFIG_DYNAMIC_DEBUG)
4381 #define netif_dbg(priv, type, netdev, format, args...) \
4383 if (netif_msg_##type(priv)) \
4384 dynamic_netdev_dbg(netdev, format, ##args); \
4386 #elif defined(DEBUG)
4387 #define netif_dbg(priv, type, dev, format, args...) \
4388 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4390 #define netif_dbg(priv, type, dev, format, args...) \
4393 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4398 /* if @cond then downgrade to debug, else print at @level */
4399 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4402 netif_dbg(priv, type, netdev, fmt, ##args); \
4404 netif_ ## level(priv, type, netdev, fmt, ##args); \
4407 #if defined(VERBOSE_DEBUG)
4408 #define netif_vdbg netif_dbg
4410 #define netif_vdbg(priv, type, dev, format, args...) \
4413 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4419 * The list of packet types we will receive (as opposed to discard)
4420 * and the routines to invoke.
4422 * Why 16. Because with 16 the only overlap we get on a hash of the
4423 * low nibble of the protocol value is RARP/SNAP/X.25.
4425 * NOTE: That is no longer true with the addition of VLAN tags. Not
4426 * sure which should go first, but I bet it won't make much
4427 * difference if we are running VLANs. The good news is that
4428 * this protocol won't be in the list unless compiled in, so
4429 * the average user (w/out VLANs) will not be adversely affected.
4445 #define PTYPE_HASH_SIZE (16)
4446 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4448 #endif /* _LINUX_NETDEVICE_H */