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/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.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>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
63 void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
66 /* hardware address assignment types */
67 #define NET_ADDR_PERM 0 /* address is permanent (default) */
68 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
69 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
70 #define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
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(), 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_POLICED 0x03 /* skb is shot by police */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
116 typedef enum netdev_tx netdev_tx_t;
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 static inline bool dev_xmit_complete(int rc)
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 if (likely(rc < NET_XMIT_MASK))
137 * Compute the worst case header length according to the protocols
141 #if 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;
202 struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
215 struct rcu_head rcu_head;
218 struct netdev_hw_addr_list {
219 struct list_head list;
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 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 (*rebuild)(struct sk_buff *skb);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
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 at 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
320 spinlock_t poll_lock;
323 struct net_device *dev;
324 struct sk_buff *gro_list;
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_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash */
345 typedef enum gro_result gro_result_t;
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
369 * If the rx_handler consumed to skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other ways.
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
377 * If the rx_handler consider the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
379 * are registered on exact device (ptype->dev == skb->dev).
381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
388 enum rx_handler_result {
394 typedef enum rx_handler_result rx_handler_result_t;
395 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
397 void __napi_schedule(struct napi_struct *n);
399 static inline bool napi_disable_pending(struct napi_struct *n)
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
405 * napi_schedule_prep - check if napi can be scheduled
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
413 static inline bool napi_schedule_prep(struct napi_struct *n)
415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
420 * napi_schedule - schedule NAPI poll
423 * Schedule NAPI poll routine to be called if it is not already
426 static inline void napi_schedule(struct napi_struct *n)
428 if (napi_schedule_prep(n))
432 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
433 static inline bool napi_reschedule(struct napi_struct *napi)
435 if (napi_schedule_prep(napi)) {
436 __napi_schedule(napi);
443 * napi_complete - NAPI processing complete
446 * Mark NAPI processing as complete.
448 void __napi_complete(struct napi_struct *n);
449 void napi_complete(struct napi_struct *n);
452 * napi_by_id - lookup a NAPI by napi_id
453 * @napi_id: hashed napi_id
455 * lookup @napi_id in napi_hash table
456 * must be called under rcu_read_lock()
458 struct napi_struct *napi_by_id(unsigned int napi_id);
461 * napi_hash_add - add a NAPI to global hashtable
462 * @napi: napi context
464 * generate a new napi_id and store a @napi under it in napi_hash
466 void napi_hash_add(struct napi_struct *napi);
469 * napi_hash_del - remove a NAPI from global table
470 * @napi: napi context
472 * Warning: caller must observe rcu grace period
473 * before freeing memory containing @napi
475 void napi_hash_del(struct napi_struct *napi);
478 * napi_disable - prevent NAPI from scheduling
481 * Stop NAPI from being scheduled on this context.
482 * Waits till any outstanding processing completes.
484 static inline void napi_disable(struct napi_struct *n)
487 set_bit(NAPI_STATE_DISABLE, &n->state);
488 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
490 clear_bit(NAPI_STATE_DISABLE, &n->state);
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_clear_bit();
504 clear_bit(NAPI_STATE_SCHED, &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 while (test_bit(NAPI_STATE_SCHED, &n->state))
522 # define napi_synchronize(n) barrier()
525 enum netdev_queue_state_t {
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
528 __QUEUE_STATE_FROZEN,
529 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
530 (1 << __QUEUE_STATE_STACK_XOFF))
531 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
532 (1 << __QUEUE_STATE_FROZEN))
535 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
536 * netif_tx_* functions below are used to manipulate this flag. The
537 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
538 * queue independently. The netif_xmit_*stopped functions below are called
539 * to check if the queue has been stopped by the driver or stack (either
540 * of the XOFF bits are set in the state). Drivers should not need to call
541 * netif_xmit*stopped functions, they should only be using netif_tx_*.
544 struct netdev_queue {
548 struct net_device *dev;
550 struct Qdisc *qdisc_sleeping;
554 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
560 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
563 * please use this field instead of dev->trans_start
565 unsigned long trans_start;
568 * Number of TX timeouts for this queue
569 * (/sys/class/net/DEV/Q/trans_timeout)
571 unsigned long trans_timeout;
578 } ____cacheline_aligned_in_smp;
580 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
582 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
589 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
591 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 * This structure holds an RPS map which can be of variable length. The
599 * map is an array of CPUs.
606 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
609 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
610 * tail pointer for that CPU's input queue at the time of last enqueue, and
611 * a hardware filter index.
613 struct rps_dev_flow {
616 unsigned int last_qtail;
618 #define RPS_NO_FILTER 0xffff
621 * The rps_dev_flow_table structure contains a table of flow mappings.
623 struct rps_dev_flow_table {
626 struct rps_dev_flow flows[0];
628 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
629 ((_num) * sizeof(struct rps_dev_flow)))
632 * The rps_sock_flow_table contains mappings of flows to the last CPU
633 * on which they were processed by the application (set in recvmsg).
635 struct rps_sock_flow_table {
639 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
640 ((_num) * sizeof(u16)))
642 #define RPS_NO_CPU 0xffff
644 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
648 unsigned int cpu, index = hash & table->mask;
650 /* We only give a hint, preemption can change cpu under us */
651 cpu = raw_smp_processor_id();
653 if (table->ents[index] != cpu)
654 table->ents[index] = cpu;
658 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
662 table->ents[hash & table->mask] = RPS_NO_CPU;
665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667 #ifdef CONFIG_RFS_ACCEL
668 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
671 #endif /* CONFIG_RPS */
673 /* This structure contains an instance of an RX queue. */
674 struct netdev_rx_queue {
676 struct rps_map __rcu *rps_map;
677 struct rps_dev_flow_table __rcu *rps_flow_table;
680 struct net_device *dev;
681 } ____cacheline_aligned_in_smp;
684 * RX queue sysfs structures and functions.
686 struct rx_queue_attribute {
687 struct attribute attr;
688 ssize_t (*show)(struct netdev_rx_queue *queue,
689 struct rx_queue_attribute *attr, char *buf);
690 ssize_t (*store)(struct netdev_rx_queue *queue,
691 struct rx_queue_attribute *attr, const char *buf, size_t len);
696 * This structure holds an XPS map which can be of variable length. The
697 * map is an array of queues.
701 unsigned int alloc_len;
705 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
706 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
710 * This structure holds all XPS maps for device. Maps are indexed by CPU.
712 struct xps_dev_maps {
714 struct xps_map __rcu *cpu_map[0];
716 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
717 (nr_cpu_ids * sizeof(struct xps_map *)))
718 #endif /* CONFIG_XPS */
720 #define TC_MAX_QUEUE 16
721 #define TC_BITMASK 15
722 /* HW offloaded queuing disciplines txq count and offset maps */
723 struct netdev_tc_txq {
728 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
730 * This structure is to hold information about the device
731 * configured to run FCoE protocol stack.
733 struct netdev_fcoe_hbainfo {
734 char manufacturer[64];
735 char serial_number[64];
736 char hardware_version[64];
737 char driver_version[64];
738 char optionrom_version[64];
739 char firmware_version[64];
741 char model_description[256];
745 #define MAX_PHYS_PORT_ID_LEN 32
747 /* This structure holds a unique identifier to identify the
748 * physical port used by a netdevice.
750 struct netdev_phys_port_id {
751 unsigned char id[MAX_PHYS_PORT_ID_LEN];
752 unsigned char id_len;
755 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
756 struct sk_buff *skb);
759 * This structure defines the management hooks for network devices.
760 * The following hooks can be defined; unless noted otherwise, they are
761 * optional and can be filled with a null pointer.
763 * int (*ndo_init)(struct net_device *dev);
764 * This function is called once when network device is registered.
765 * The network device can use this to any late stage initializaton
766 * or semantic validattion. It can fail with an error code which will
767 * be propogated back to register_netdev
769 * void (*ndo_uninit)(struct net_device *dev);
770 * This function is called when device is unregistered or when registration
771 * fails. It is not called if init fails.
773 * int (*ndo_open)(struct net_device *dev);
774 * This function is called when network device transistions to the up
777 * int (*ndo_stop)(struct net_device *dev);
778 * This function is called when network device transistions to the down
781 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
782 * struct net_device *dev);
783 * Called when a packet needs to be transmitted.
784 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
785 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
786 * Required can not be NULL.
788 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
789 * void *accel_priv, select_queue_fallback_t fallback);
790 * Called to decide which queue to when device supports multiple
793 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
794 * This function is called to allow device receiver to make
795 * changes to configuration when multicast or promiscious is enabled.
797 * void (*ndo_set_rx_mode)(struct net_device *dev);
798 * This function is called device changes address list filtering.
799 * If driver handles unicast address filtering, it should set
800 * IFF_UNICAST_FLT to its priv_flags.
802 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
803 * This function is called when the Media Access Control address
804 * needs to be changed. If this interface is not defined, the
805 * mac address can not be changed.
807 * int (*ndo_validate_addr)(struct net_device *dev);
808 * Test if Media Access Control address is valid for the device.
810 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
811 * Called when a user request an ioctl which can't be handled by
812 * the generic interface code. If not defined ioctl's return
813 * not supported error code.
815 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
816 * Used to set network devices bus interface parameters. This interface
817 * is retained for legacy reason, new devices should use the bus
818 * interface (PCI) for low level management.
820 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
821 * Called when a user wants to change the Maximum Transfer Unit
822 * of a device. If not defined, any request to change MTU will
823 * will return an error.
825 * void (*ndo_tx_timeout)(struct net_device *dev);
826 * Callback uses when the transmitter has not made any progress
827 * for dev->watchdog ticks.
829 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
830 * struct rtnl_link_stats64 *storage);
831 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
832 * Called when a user wants to get the network device usage
833 * statistics. Drivers must do one of the following:
834 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
835 * rtnl_link_stats64 structure passed by the caller.
836 * 2. Define @ndo_get_stats to update a net_device_stats structure
837 * (which should normally be dev->stats) and return a pointer to
838 * it. The structure may be changed asynchronously only if each
839 * field is written atomically.
840 * 3. Update dev->stats asynchronously and atomically, and define
843 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
844 * If device support VLAN filtering this function is called when a
845 * VLAN id is registered.
847 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
848 * If device support VLAN filtering this function is called when a
849 * VLAN id is unregistered.
851 * void (*ndo_poll_controller)(struct net_device *dev);
853 * SR-IOV management functions.
854 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
855 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
856 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
857 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
858 * int (*ndo_get_vf_config)(struct net_device *dev,
859 * int vf, struct ifla_vf_info *ivf);
860 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
861 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
862 * struct nlattr *port[]);
863 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
864 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
865 * Called to setup 'tc' number of traffic classes in the net device. This
866 * is always called from the stack with the rtnl lock held and netif tx
867 * queues stopped. This allows the netdevice to perform queue management
870 * Fiber Channel over Ethernet (FCoE) offload functions.
871 * int (*ndo_fcoe_enable)(struct net_device *dev);
872 * Called when the FCoE protocol stack wants to start using LLD for FCoE
873 * so the underlying device can perform whatever needed configuration or
874 * initialization to support acceleration of FCoE traffic.
876 * int (*ndo_fcoe_disable)(struct net_device *dev);
877 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
878 * so the underlying device can perform whatever needed clean-ups to
879 * stop supporting acceleration of FCoE traffic.
881 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
882 * struct scatterlist *sgl, unsigned int sgc);
883 * Called when the FCoE Initiator wants to initialize an I/O that
884 * is a possible candidate for Direct Data Placement (DDP). The LLD can
885 * perform necessary setup and returns 1 to indicate the device is set up
886 * successfully to perform DDP on this I/O, otherwise this returns 0.
888 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
889 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
890 * indicated by the FC exchange id 'xid', so the underlying device can
891 * clean up and reuse resources for later DDP requests.
893 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
894 * struct scatterlist *sgl, unsigned int sgc);
895 * Called when the FCoE Target wants to initialize an I/O that
896 * is a possible candidate for Direct Data Placement (DDP). The LLD can
897 * perform necessary setup and returns 1 to indicate the device is set up
898 * successfully to perform DDP on this I/O, otherwise this returns 0.
900 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
901 * struct netdev_fcoe_hbainfo *hbainfo);
902 * Called when the FCoE Protocol stack wants information on the underlying
903 * device. This information is utilized by the FCoE protocol stack to
904 * register attributes with Fiber Channel management service as per the
905 * FC-GS Fabric Device Management Information(FDMI) specification.
907 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
908 * Called when the underlying device wants to override default World Wide
909 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
910 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
911 * protocol stack to use.
914 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
915 * u16 rxq_index, u32 flow_id);
916 * Set hardware filter for RFS. rxq_index is the target queue index;
917 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
918 * Return the filter ID on success, or a negative error code.
920 * Slave management functions (for bridge, bonding, etc).
921 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
922 * Called to make another netdev an underling.
924 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
925 * Called to release previously enslaved netdev.
927 * Feature/offload setting functions.
928 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
929 * netdev_features_t features);
930 * Adjusts the requested feature flags according to device-specific
931 * constraints, and returns the resulting flags. Must not modify
934 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
935 * Called to update device configuration to new features. Passed
936 * feature set might be less than what was returned by ndo_fix_features()).
937 * Must return >0 or -errno if it changed dev->features itself.
939 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
940 * struct net_device *dev,
941 * const unsigned char *addr, u16 flags)
942 * Adds an FDB entry to dev for addr.
943 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
944 * struct net_device *dev,
945 * const unsigned char *addr)
946 * Deletes the FDB entry from dev coresponding to addr.
947 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
948 * struct net_device *dev, int idx)
949 * Used to add FDB entries to dump requests. Implementers should add
950 * entries to skb and update idx with the number of entries.
952 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
953 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
954 * struct net_device *dev, u32 filter_mask)
956 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
957 * Called to change device carrier. Soft-devices (like dummy, team, etc)
958 * which do not represent real hardware may define this to allow their
959 * userspace components to manage their virtual carrier state. Devices
960 * that determine carrier state from physical hardware properties (eg
961 * network cables) or protocol-dependent mechanisms (eg
962 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
964 * int (*ndo_get_phys_port_id)(struct net_device *dev,
965 * struct netdev_phys_port_id *ppid);
966 * Called to get ID of physical port of this device. If driver does
967 * not implement this, it is assumed that the hw is not able to have
968 * multiple net devices on single physical port.
970 * void (*ndo_add_vxlan_port)(struct net_device *dev,
971 * sa_family_t sa_family, __be16 port);
972 * Called by vxlan to notiy a driver about the UDP port and socket
973 * address family that vxlan is listnening to. It is called only when
974 * a new port starts listening. The operation is protected by the
975 * vxlan_net->sock_lock.
977 * void (*ndo_del_vxlan_port)(struct net_device *dev,
978 * sa_family_t sa_family, __be16 port);
979 * Called by vxlan to notify the driver about a UDP port and socket
980 * address family that vxlan is not listening to anymore. The operation
981 * is protected by the vxlan_net->sock_lock.
983 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
984 * struct net_device *dev)
985 * Called by upper layer devices to accelerate switching or other
986 * station functionality into hardware. 'pdev is the lowerdev
987 * to use for the offload and 'dev' is the net device that will
988 * back the offload. Returns a pointer to the private structure
989 * the upper layer will maintain.
990 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
991 * Called by upper layer device to delete the station created
992 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
993 * the station and priv is the structure returned by the add
995 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
996 * struct net_device *dev,
998 * Callback to use for xmit over the accelerated station. This
999 * is used in place of ndo_start_xmit on accelerated net
1002 struct net_device_ops {
1003 int (*ndo_init)(struct net_device *dev);
1004 void (*ndo_uninit)(struct net_device *dev);
1005 int (*ndo_open)(struct net_device *dev);
1006 int (*ndo_stop)(struct net_device *dev);
1007 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1008 struct net_device *dev);
1009 u16 (*ndo_select_queue)(struct net_device *dev,
1010 struct sk_buff *skb,
1012 select_queue_fallback_t fallback);
1013 void (*ndo_change_rx_flags)(struct net_device *dev,
1015 void (*ndo_set_rx_mode)(struct net_device *dev);
1016 int (*ndo_set_mac_address)(struct net_device *dev,
1018 int (*ndo_validate_addr)(struct net_device *dev);
1019 int (*ndo_do_ioctl)(struct net_device *dev,
1020 struct ifreq *ifr, int cmd);
1021 int (*ndo_set_config)(struct net_device *dev,
1023 int (*ndo_change_mtu)(struct net_device *dev,
1025 int (*ndo_neigh_setup)(struct net_device *dev,
1026 struct neigh_parms *);
1027 void (*ndo_tx_timeout) (struct net_device *dev);
1029 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1030 struct rtnl_link_stats64 *storage);
1031 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1033 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1034 __be16 proto, u16 vid);
1035 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1036 __be16 proto, u16 vid);
1037 #ifdef CONFIG_NET_POLL_CONTROLLER
1038 void (*ndo_poll_controller)(struct net_device *dev);
1039 int (*ndo_netpoll_setup)(struct net_device *dev,
1040 struct netpoll_info *info);
1041 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1043 #ifdef CONFIG_NET_RX_BUSY_POLL
1044 int (*ndo_busy_poll)(struct napi_struct *dev);
1046 int (*ndo_set_vf_mac)(struct net_device *dev,
1047 int queue, u8 *mac);
1048 int (*ndo_set_vf_vlan)(struct net_device *dev,
1049 int queue, u16 vlan, u8 qos);
1050 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1052 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1053 int vf, bool setting);
1054 int (*ndo_get_vf_config)(struct net_device *dev,
1056 struct ifla_vf_info *ivf);
1057 int (*ndo_set_vf_link_state)(struct net_device *dev,
1058 int vf, int link_state);
1059 int (*ndo_set_vf_port)(struct net_device *dev,
1061 struct nlattr *port[]);
1062 int (*ndo_get_vf_port)(struct net_device *dev,
1063 int vf, struct sk_buff *skb);
1064 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1065 #if IS_ENABLED(CONFIG_FCOE)
1066 int (*ndo_fcoe_enable)(struct net_device *dev);
1067 int (*ndo_fcoe_disable)(struct net_device *dev);
1068 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1070 struct scatterlist *sgl,
1072 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1074 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1076 struct scatterlist *sgl,
1078 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1079 struct netdev_fcoe_hbainfo *hbainfo);
1082 #if IS_ENABLED(CONFIG_LIBFCOE)
1083 #define NETDEV_FCOE_WWNN 0
1084 #define NETDEV_FCOE_WWPN 1
1085 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1086 u64 *wwn, int type);
1089 #ifdef CONFIG_RFS_ACCEL
1090 int (*ndo_rx_flow_steer)(struct net_device *dev,
1091 const struct sk_buff *skb,
1095 int (*ndo_add_slave)(struct net_device *dev,
1096 struct net_device *slave_dev);
1097 int (*ndo_del_slave)(struct net_device *dev,
1098 struct net_device *slave_dev);
1099 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1100 netdev_features_t features);
1101 int (*ndo_set_features)(struct net_device *dev,
1102 netdev_features_t features);
1103 int (*ndo_neigh_construct)(struct neighbour *n);
1104 void (*ndo_neigh_destroy)(struct neighbour *n);
1106 int (*ndo_fdb_add)(struct ndmsg *ndm,
1107 struct nlattr *tb[],
1108 struct net_device *dev,
1109 const unsigned char *addr,
1111 int (*ndo_fdb_del)(struct ndmsg *ndm,
1112 struct nlattr *tb[],
1113 struct net_device *dev,
1114 const unsigned char *addr);
1115 int (*ndo_fdb_dump)(struct sk_buff *skb,
1116 struct netlink_callback *cb,
1117 struct net_device *dev,
1120 int (*ndo_bridge_setlink)(struct net_device *dev,
1121 struct nlmsghdr *nlh);
1122 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1124 struct net_device *dev,
1126 int (*ndo_bridge_dellink)(struct net_device *dev,
1127 struct nlmsghdr *nlh);
1128 int (*ndo_change_carrier)(struct net_device *dev,
1130 int (*ndo_get_phys_port_id)(struct net_device *dev,
1131 struct netdev_phys_port_id *ppid);
1132 void (*ndo_add_vxlan_port)(struct net_device *dev,
1133 sa_family_t sa_family,
1135 void (*ndo_del_vxlan_port)(struct net_device *dev,
1136 sa_family_t sa_family,
1139 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1140 struct net_device *dev);
1141 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1144 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1145 struct net_device *dev,
1150 * enum net_device_priv_flags - &struct net_device priv_flags
1152 * These are the &struct net_device, they are only set internally
1153 * by drivers and used in the kernel. These flags are invisible to
1154 * userspace, this means that the order of these flags can change
1155 * during any kernel release.
1157 * You should have a pretty good reason to be extending these flags.
1159 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1160 * @IFF_EBRIDGE: Ethernet bridging device
1161 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1162 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1163 * @IFF_MASTER_ALB: bonding master, balance-alb
1164 * @IFF_BONDING: bonding master or slave
1165 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1166 * @IFF_ISATAP: ISATAP interface (RFC4214)
1167 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1168 * @IFF_WAN_HDLC: WAN HDLC device
1169 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1171 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1172 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1173 * @IFF_MACVLAN_PORT: device used as macvlan port
1174 * @IFF_BRIDGE_PORT: device used as bridge port
1175 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1176 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1177 * @IFF_UNICAST_FLT: Supports unicast filtering
1178 * @IFF_TEAM_PORT: device used as team port
1179 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1180 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1181 * change when it's running
1182 * @IFF_MACVLAN: Macvlan device
1184 enum netdev_priv_flags {
1185 IFF_802_1Q_VLAN = 1<<0,
1187 IFF_SLAVE_INACTIVE = 1<<2,
1188 IFF_MASTER_8023AD = 1<<3,
1189 IFF_MASTER_ALB = 1<<4,
1191 IFF_SLAVE_NEEDARP = 1<<6,
1193 IFF_MASTER_ARPMON = 1<<8,
1194 IFF_WAN_HDLC = 1<<9,
1195 IFF_XMIT_DST_RELEASE = 1<<10,
1196 IFF_DONT_BRIDGE = 1<<11,
1197 IFF_DISABLE_NETPOLL = 1<<12,
1198 IFF_MACVLAN_PORT = 1<<13,
1199 IFF_BRIDGE_PORT = 1<<14,
1200 IFF_OVS_DATAPATH = 1<<15,
1201 IFF_TX_SKB_SHARING = 1<<16,
1202 IFF_UNICAST_FLT = 1<<17,
1203 IFF_TEAM_PORT = 1<<18,
1204 IFF_SUPP_NOFCS = 1<<19,
1205 IFF_LIVE_ADDR_CHANGE = 1<<20,
1206 IFF_MACVLAN = 1<<21,
1209 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1210 #define IFF_EBRIDGE IFF_EBRIDGE
1211 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1212 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1213 #define IFF_MASTER_ALB IFF_MASTER_ALB
1214 #define IFF_BONDING IFF_BONDING
1215 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1216 #define IFF_ISATAP IFF_ISATAP
1217 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1218 #define IFF_WAN_HDLC IFF_WAN_HDLC
1219 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1220 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1221 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1222 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1223 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1224 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1225 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1226 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1227 #define IFF_TEAM_PORT IFF_TEAM_PORT
1228 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1229 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1230 #define IFF_MACVLAN IFF_MACVLAN
1233 * The DEVICE structure.
1234 * Actually, this whole structure is a big mistake. It mixes I/O
1235 * data with strictly "high-level" data, and it has to know about
1236 * almost every data structure used in the INET module.
1238 * FIXME: cleanup struct net_device such that network protocol info
1245 * This is the first field of the "visible" part of this structure
1246 * (i.e. as seen by users in the "Space.c" file). It is the name
1249 char name[IFNAMSIZ];
1251 /* device name hash chain, please keep it close to name[] */
1252 struct hlist_node name_hlist;
1258 * I/O specific fields
1259 * FIXME: Merge these and struct ifmap into one
1261 unsigned long mem_end; /* shared mem end */
1262 unsigned long mem_start; /* shared mem start */
1263 unsigned long base_addr; /* device I/O address */
1264 int irq; /* device IRQ number */
1267 * Some hardware also needs these fields, but they are not
1268 * part of the usual set specified in Space.c.
1271 unsigned long state;
1273 struct list_head dev_list;
1274 struct list_head napi_list;
1275 struct list_head unreg_list;
1276 struct list_head close_list;
1278 /* directly linked devices, like slaves for bonding */
1280 struct list_head upper;
1281 struct list_head lower;
1284 /* all linked devices, *including* neighbours */
1286 struct list_head upper;
1287 struct list_head lower;
1291 /* currently active device features */
1292 netdev_features_t features;
1293 /* user-changeable features */
1294 netdev_features_t hw_features;
1295 /* user-requested features */
1296 netdev_features_t wanted_features;
1297 /* mask of features inheritable by VLAN devices */
1298 netdev_features_t vlan_features;
1299 /* mask of features inherited by encapsulating devices
1300 * This field indicates what encapsulation offloads
1301 * the hardware is capable of doing, and drivers will
1302 * need to set them appropriately.
1304 netdev_features_t hw_enc_features;
1305 /* mask of fetures inheritable by MPLS */
1306 netdev_features_t mpls_features;
1308 /* Interface index. Unique device identifier */
1312 struct net_device_stats stats;
1314 /* dropped packets by core network, Do not use this in drivers */
1315 atomic_long_t rx_dropped;
1316 atomic_long_t tx_dropped;
1318 #ifdef CONFIG_WIRELESS_EXT
1319 /* List of functions to handle Wireless Extensions (instead of ioctl).
1320 * See <net/iw_handler.h> for details. Jean II */
1321 const struct iw_handler_def * wireless_handlers;
1322 /* Instance data managed by the core of Wireless Extensions. */
1323 struct iw_public_data * wireless_data;
1325 /* Management operations */
1326 const struct net_device_ops *netdev_ops;
1327 const struct ethtool_ops *ethtool_ops;
1328 const struct forwarding_accel_ops *fwd_ops;
1330 /* Hardware header description */
1331 const struct header_ops *header_ops;
1333 unsigned int flags; /* interface flags (a la BSD) */
1334 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1335 * See if.h for definitions. */
1336 unsigned short gflags;
1337 unsigned short padded; /* How much padding added by alloc_netdev() */
1339 unsigned char operstate; /* RFC2863 operstate */
1340 unsigned char link_mode; /* mapping policy to operstate */
1342 unsigned char if_port; /* Selectable AUI, TP,..*/
1343 unsigned char dma; /* DMA channel */
1345 unsigned int mtu; /* interface MTU value */
1346 unsigned short type; /* interface hardware type */
1347 unsigned short hard_header_len; /* hardware hdr length */
1349 /* extra head- and tailroom the hardware may need, but not in all cases
1350 * can this be guaranteed, especially tailroom. Some cases also use
1351 * LL_MAX_HEADER instead to allocate the skb.
1353 unsigned short needed_headroom;
1354 unsigned short needed_tailroom;
1356 /* Interface address info. */
1357 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1358 unsigned char addr_assign_type; /* hw address assignment type */
1359 unsigned char addr_len; /* hardware address length */
1360 unsigned short neigh_priv_len;
1361 unsigned short dev_id; /* Used to differentiate devices
1362 * that share the same link
1365 unsigned short dev_port; /* Used to differentiate
1366 * devices that share the same
1369 spinlock_t addr_list_lock;
1370 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1371 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1372 struct netdev_hw_addr_list dev_addrs; /* list of device
1376 struct kset *queues_kset;
1380 unsigned int promiscuity;
1381 unsigned int allmulti;
1384 /* Protocol specific pointers */
1386 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1387 struct vlan_info __rcu *vlan_info; /* VLAN info */
1389 #if IS_ENABLED(CONFIG_NET_DSA)
1390 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1392 #if IS_ENABLED(CONFIG_TIPC)
1393 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */
1395 void *atalk_ptr; /* AppleTalk link */
1396 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1397 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1398 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1399 void *ax25_ptr; /* AX.25 specific data */
1400 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1401 assign before registering */
1404 * Cache lines mostly used on receive path (including eth_type_trans())
1406 unsigned long last_rx; /* Time of last Rx */
1408 /* Interface address info used in eth_type_trans() */
1409 unsigned char *dev_addr; /* hw address, (before bcast
1410 because most packets are
1415 struct netdev_rx_queue *_rx;
1417 /* Number of RX queues allocated at register_netdev() time */
1418 unsigned int num_rx_queues;
1420 /* Number of RX queues currently active in device */
1421 unsigned int real_num_rx_queues;
1425 rx_handler_func_t __rcu *rx_handler;
1426 void __rcu *rx_handler_data;
1428 struct netdev_queue __rcu *ingress_queue;
1429 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1433 * Cache lines mostly used on transmit path
1435 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1437 /* Number of TX queues allocated at alloc_netdev_mq() time */
1438 unsigned int num_tx_queues;
1440 /* Number of TX queues currently active in device */
1441 unsigned int real_num_tx_queues;
1443 /* root qdisc from userspace point of view */
1444 struct Qdisc *qdisc;
1446 unsigned long tx_queue_len; /* Max frames per queue allowed */
1447 spinlock_t tx_global_lock;
1450 struct xps_dev_maps __rcu *xps_maps;
1452 #ifdef CONFIG_RFS_ACCEL
1453 /* CPU reverse-mapping for RX completion interrupts, indexed
1454 * by RX queue number. Assigned by driver. This must only be
1455 * set if the ndo_rx_flow_steer operation is defined. */
1456 struct cpu_rmap *rx_cpu_rmap;
1459 /* These may be needed for future network-power-down code. */
1462 * trans_start here is expensive for high speed devices on SMP,
1463 * please use netdev_queue->trans_start instead.
1465 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1467 int watchdog_timeo; /* used by dev_watchdog() */
1468 struct timer_list watchdog_timer;
1470 /* Number of references to this device */
1471 int __percpu *pcpu_refcnt;
1473 /* delayed register/unregister */
1474 struct list_head todo_list;
1475 /* device index hash chain */
1476 struct hlist_node index_hlist;
1478 struct list_head link_watch_list;
1480 /* register/unregister state machine */
1481 enum { NETREG_UNINITIALIZED=0,
1482 NETREG_REGISTERED, /* completed register_netdevice */
1483 NETREG_UNREGISTERING, /* called unregister_netdevice */
1484 NETREG_UNREGISTERED, /* completed unregister todo */
1485 NETREG_RELEASED, /* called free_netdev */
1486 NETREG_DUMMY, /* dummy device for NAPI poll */
1489 bool dismantle; /* device is going do be freed */
1492 RTNL_LINK_INITIALIZED,
1493 RTNL_LINK_INITIALIZING,
1494 } rtnl_link_state:16;
1496 /* Called from unregister, can be used to call free_netdev */
1497 void (*destructor)(struct net_device *dev);
1499 #ifdef CONFIG_NETPOLL
1500 struct netpoll_info __rcu *npinfo;
1503 #ifdef CONFIG_NET_NS
1504 /* Network namespace this network device is inside */
1508 /* mid-layer private */
1511 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1512 struct pcpu_sw_netstats __percpu *tstats;
1513 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1514 struct pcpu_vstats __percpu *vstats; /* veth stats */
1517 struct garp_port __rcu *garp_port;
1519 struct mrp_port __rcu *mrp_port;
1521 /* class/net/name entry */
1523 /* space for optional device, statistics, and wireless sysfs groups */
1524 const struct attribute_group *sysfs_groups[4];
1525 /* space for optional per-rx queue attributes */
1526 const struct attribute_group *sysfs_rx_queue_group;
1528 /* rtnetlink link ops */
1529 const struct rtnl_link_ops *rtnl_link_ops;
1531 /* for setting kernel sock attribute on TCP connection setup */
1532 #define GSO_MAX_SIZE 65536
1533 unsigned int gso_max_size;
1534 #define GSO_MAX_SEGS 65535
1538 /* Data Center Bridging netlink ops */
1539 const struct dcbnl_rtnl_ops *dcbnl_ops;
1542 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1543 u8 prio_tc_map[TC_BITMASK + 1];
1545 #if IS_ENABLED(CONFIG_FCOE)
1546 /* max exchange id for FCoE LRO by ddp */
1547 unsigned int fcoe_ddp_xid;
1549 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1550 struct netprio_map __rcu *priomap;
1552 /* phy device may attach itself for hardware timestamping */
1553 struct phy_device *phydev;
1555 struct lock_class_key *qdisc_tx_busylock;
1557 /* group the device belongs to */
1560 struct pm_qos_request pm_qos_req;
1562 #define to_net_dev(d) container_of(d, struct net_device, dev)
1564 #define NETDEV_ALIGN 32
1567 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1569 return dev->prio_tc_map[prio & TC_BITMASK];
1573 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1575 if (tc >= dev->num_tc)
1578 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1583 void netdev_reset_tc(struct net_device *dev)
1586 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1587 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1591 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1593 if (tc >= dev->num_tc)
1596 dev->tc_to_txq[tc].count = count;
1597 dev->tc_to_txq[tc].offset = offset;
1602 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1604 if (num_tc > TC_MAX_QUEUE)
1607 dev->num_tc = num_tc;
1612 int netdev_get_num_tc(struct net_device *dev)
1618 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1621 return &dev->_tx[index];
1624 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1625 void (*f)(struct net_device *,
1626 struct netdev_queue *,
1632 for (i = 0; i < dev->num_tx_queues; i++)
1633 f(dev, &dev->_tx[i], arg);
1636 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1637 struct sk_buff *skb,
1641 * Net namespace inlines
1644 struct net *dev_net(const struct net_device *dev)
1646 return read_pnet(&dev->nd_net);
1650 void dev_net_set(struct net_device *dev, struct net *net)
1652 #ifdef CONFIG_NET_NS
1653 release_net(dev->nd_net);
1654 dev->nd_net = hold_net(net);
1658 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1660 #ifdef CONFIG_NET_DSA_TAG_DSA
1661 if (dev->dsa_ptr != NULL)
1662 return dsa_uses_dsa_tags(dev->dsa_ptr);
1668 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1670 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1671 if (dev->dsa_ptr != NULL)
1672 return dsa_uses_trailer_tags(dev->dsa_ptr);
1679 * netdev_priv - access network device private data
1680 * @dev: network device
1682 * Get network device private data
1684 static inline void *netdev_priv(const struct net_device *dev)
1686 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1689 /* Set the sysfs physical device reference for the network logical device
1690 * if set prior to registration will cause a symlink during initialization.
1692 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1694 /* Set the sysfs device type for the network logical device to allow
1695 * fine-grained identification of different network device types. For
1696 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1698 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1700 /* Default NAPI poll() weight
1701 * Device drivers are strongly advised to not use bigger value
1703 #define NAPI_POLL_WEIGHT 64
1706 * netif_napi_add - initialize a napi context
1707 * @dev: network device
1708 * @napi: napi context
1709 * @poll: polling function
1710 * @weight: default weight
1712 * netif_napi_add() must be used to initialize a napi context prior to calling
1713 * *any* of the other napi related functions.
1715 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1716 int (*poll)(struct napi_struct *, int), int weight);
1719 * netif_napi_del - remove a napi context
1720 * @napi: napi context
1722 * netif_napi_del() removes a napi context from the network device napi list
1724 void netif_napi_del(struct napi_struct *napi);
1726 struct napi_gro_cb {
1727 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1730 /* Length of frag0. */
1731 unsigned int frag0_len;
1733 /* This indicates where we are processing relative to skb->data. */
1736 /* This is non-zero if the packet cannot be merged with the new skb. */
1739 /* Save the IP ID here and check when we get to the transport layer */
1742 /* Number of segments aggregated. */
1745 /* This is non-zero if the packet may be of the same flow. */
1750 #define NAPI_GRO_FREE 1
1751 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1753 /* jiffies when first packet was created/queued */
1756 /* Used in ipv6_gro_receive() */
1759 /* Used in udp_gro_receive */
1762 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1765 /* used in skb_gro_receive() slow path */
1766 struct sk_buff *last;
1769 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1771 struct packet_type {
1772 __be16 type; /* This is really htons(ether_type). */
1773 struct net_device *dev; /* NULL is wildcarded here */
1774 int (*func) (struct sk_buff *,
1775 struct net_device *,
1776 struct packet_type *,
1777 struct net_device *);
1778 bool (*id_match)(struct packet_type *ptype,
1780 void *af_packet_priv;
1781 struct list_head list;
1784 struct offload_callbacks {
1785 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1786 netdev_features_t features);
1787 int (*gso_send_check)(struct sk_buff *skb);
1788 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1789 struct sk_buff *skb);
1790 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1793 struct packet_offload {
1794 __be16 type; /* This is really htons(ether_type). */
1795 struct offload_callbacks callbacks;
1796 struct list_head list;
1799 struct udp_offload {
1801 struct offload_callbacks callbacks;
1804 /* often modified stats are per cpu, other are shared (netdev->stats) */
1805 struct pcpu_sw_netstats {
1810 struct u64_stats_sync syncp;
1813 #define netdev_alloc_pcpu_stats(type) \
1815 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1818 for_each_possible_cpu(i) { \
1819 typeof(type) *stat; \
1820 stat = per_cpu_ptr(pcpu_stats, i); \
1821 u64_stats_init(&stat->syncp); \
1827 #include <linux/notifier.h>
1829 /* netdevice notifier chain. Please remember to update the rtnetlink
1830 * notification exclusion list in rtnetlink_event() when adding new
1833 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1834 #define NETDEV_DOWN 0x0002
1835 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1836 detected a hardware crash and restarted
1837 - we can use this eg to kick tcp sessions
1839 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1840 #define NETDEV_REGISTER 0x0005
1841 #define NETDEV_UNREGISTER 0x0006
1842 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1843 #define NETDEV_CHANGEADDR 0x0008
1844 #define NETDEV_GOING_DOWN 0x0009
1845 #define NETDEV_CHANGENAME 0x000A
1846 #define NETDEV_FEAT_CHANGE 0x000B
1847 #define NETDEV_BONDING_FAILOVER 0x000C
1848 #define NETDEV_PRE_UP 0x000D
1849 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1850 #define NETDEV_POST_TYPE_CHANGE 0x000F
1851 #define NETDEV_POST_INIT 0x0010
1852 #define NETDEV_UNREGISTER_FINAL 0x0011
1853 #define NETDEV_RELEASE 0x0012
1854 #define NETDEV_NOTIFY_PEERS 0x0013
1855 #define NETDEV_JOIN 0x0014
1856 #define NETDEV_CHANGEUPPER 0x0015
1857 #define NETDEV_RESEND_IGMP 0x0016
1858 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1860 int register_netdevice_notifier(struct notifier_block *nb);
1861 int unregister_netdevice_notifier(struct notifier_block *nb);
1863 struct netdev_notifier_info {
1864 struct net_device *dev;
1867 struct netdev_notifier_change_info {
1868 struct netdev_notifier_info info; /* must be first */
1869 unsigned int flags_changed;
1872 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1873 struct net_device *dev)
1878 static inline struct net_device *
1879 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1884 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1887 extern rwlock_t dev_base_lock; /* Device list lock */
1889 #define for_each_netdev(net, d) \
1890 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1891 #define for_each_netdev_reverse(net, d) \
1892 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1893 #define for_each_netdev_rcu(net, d) \
1894 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1895 #define for_each_netdev_safe(net, d, n) \
1896 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1897 #define for_each_netdev_continue(net, d) \
1898 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1899 #define for_each_netdev_continue_rcu(net, d) \
1900 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1901 #define for_each_netdev_in_bond_rcu(bond, slave) \
1902 for_each_netdev_rcu(&init_net, slave) \
1903 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1904 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1906 static inline struct net_device *next_net_device(struct net_device *dev)
1908 struct list_head *lh;
1912 lh = dev->dev_list.next;
1913 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1916 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1918 struct list_head *lh;
1922 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1923 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1926 static inline struct net_device *first_net_device(struct net *net)
1928 return list_empty(&net->dev_base_head) ? NULL :
1929 net_device_entry(net->dev_base_head.next);
1932 static inline struct net_device *first_net_device_rcu(struct net *net)
1934 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1936 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1939 int netdev_boot_setup_check(struct net_device *dev);
1940 unsigned long netdev_boot_base(const char *prefix, int unit);
1941 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1942 const char *hwaddr);
1943 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1944 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1945 void dev_add_pack(struct packet_type *pt);
1946 void dev_remove_pack(struct packet_type *pt);
1947 void __dev_remove_pack(struct packet_type *pt);
1948 void dev_add_offload(struct packet_offload *po);
1949 void dev_remove_offload(struct packet_offload *po);
1951 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1952 unsigned short mask);
1953 struct net_device *dev_get_by_name(struct net *net, const char *name);
1954 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1955 struct net_device *__dev_get_by_name(struct net *net, const char *name);
1956 int dev_alloc_name(struct net_device *dev, const char *name);
1957 int dev_open(struct net_device *dev);
1958 int dev_close(struct net_device *dev);
1959 void dev_disable_lro(struct net_device *dev);
1960 int dev_loopback_xmit(struct sk_buff *newskb);
1961 int dev_queue_xmit(struct sk_buff *skb);
1962 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
1963 int register_netdevice(struct net_device *dev);
1964 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1965 void unregister_netdevice_many(struct list_head *head);
1966 static inline void unregister_netdevice(struct net_device *dev)
1968 unregister_netdevice_queue(dev, NULL);
1971 int netdev_refcnt_read(const struct net_device *dev);
1972 void free_netdev(struct net_device *dev);
1973 void netdev_freemem(struct net_device *dev);
1974 void synchronize_net(void);
1975 int init_dummy_netdev(struct net_device *dev);
1977 struct net_device *dev_get_by_index(struct net *net, int ifindex);
1978 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1979 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1980 int netdev_get_name(struct net *net, char *name, int ifindex);
1981 int dev_restart(struct net_device *dev);
1982 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1984 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1986 return NAPI_GRO_CB(skb)->data_offset;
1989 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1991 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1994 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1996 NAPI_GRO_CB(skb)->data_offset += len;
1999 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2000 unsigned int offset)
2002 return NAPI_GRO_CB(skb)->frag0 + offset;
2005 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2007 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2010 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2011 unsigned int offset)
2013 if (!pskb_may_pull(skb, hlen))
2016 NAPI_GRO_CB(skb)->frag0 = NULL;
2017 NAPI_GRO_CB(skb)->frag0_len = 0;
2018 return skb->data + offset;
2021 static inline void *skb_gro_mac_header(struct sk_buff *skb)
2023 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
2026 static inline void *skb_gro_network_header(struct sk_buff *skb)
2028 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2029 skb_network_offset(skb);
2032 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2033 const void *start, unsigned int len)
2035 if (skb->ip_summed == CHECKSUM_COMPLETE)
2036 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2037 csum_partial(start, len, 0));
2040 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2041 unsigned short type,
2042 const void *daddr, const void *saddr,
2045 if (!dev->header_ops || !dev->header_ops->create)
2048 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2051 static inline int dev_parse_header(const struct sk_buff *skb,
2052 unsigned char *haddr)
2054 const struct net_device *dev = skb->dev;
2056 if (!dev->header_ops || !dev->header_ops->parse)
2058 return dev->header_ops->parse(skb, haddr);
2061 static inline int dev_rebuild_header(struct sk_buff *skb)
2063 const struct net_device *dev = skb->dev;
2065 if (!dev->header_ops || !dev->header_ops->rebuild)
2067 return dev->header_ops->rebuild(skb);
2070 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2071 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2072 static inline int unregister_gifconf(unsigned int family)
2074 return register_gifconf(family, NULL);
2077 #ifdef CONFIG_NET_FLOW_LIMIT
2078 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2079 struct sd_flow_limit {
2081 unsigned int num_buckets;
2082 unsigned int history_head;
2083 u16 history[FLOW_LIMIT_HISTORY];
2087 extern int netdev_flow_limit_table_len;
2088 #endif /* CONFIG_NET_FLOW_LIMIT */
2091 * Incoming packets are placed on per-cpu queues
2093 struct softnet_data {
2094 struct Qdisc *output_queue;
2095 struct Qdisc **output_queue_tailp;
2096 struct list_head poll_list;
2097 struct sk_buff *completion_queue;
2098 struct sk_buff_head process_queue;
2101 unsigned int processed;
2102 unsigned int time_squeeze;
2103 unsigned int cpu_collision;
2104 unsigned int received_rps;
2107 struct softnet_data *rps_ipi_list;
2109 /* Elements below can be accessed between CPUs for RPS */
2110 struct call_single_data csd ____cacheline_aligned_in_smp;
2111 struct softnet_data *rps_ipi_next;
2113 unsigned int input_queue_head;
2114 unsigned int input_queue_tail;
2116 unsigned int dropped;
2117 struct sk_buff_head input_pkt_queue;
2118 struct napi_struct backlog;
2120 #ifdef CONFIG_NET_FLOW_LIMIT
2121 struct sd_flow_limit __rcu *flow_limit;
2125 static inline void input_queue_head_incr(struct softnet_data *sd)
2128 sd->input_queue_head++;
2132 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2133 unsigned int *qtail)
2136 *qtail = ++sd->input_queue_tail;
2140 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2142 void __netif_schedule(struct Qdisc *q);
2144 static inline void netif_schedule_queue(struct netdev_queue *txq)
2146 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
2147 __netif_schedule(txq->qdisc);
2150 static inline void netif_tx_schedule_all(struct net_device *dev)
2154 for (i = 0; i < dev->num_tx_queues; i++)
2155 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2158 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2160 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2164 * netif_start_queue - allow transmit
2165 * @dev: network device
2167 * Allow upper layers to call the device hard_start_xmit routine.
2169 static inline void netif_start_queue(struct net_device *dev)
2171 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2174 static inline void netif_tx_start_all_queues(struct net_device *dev)
2178 for (i = 0; i < dev->num_tx_queues; i++) {
2179 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2180 netif_tx_start_queue(txq);
2184 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
2186 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
2187 __netif_schedule(dev_queue->qdisc);
2191 * netif_wake_queue - restart transmit
2192 * @dev: network device
2194 * Allow upper layers to call the device hard_start_xmit routine.
2195 * Used for flow control when transmit resources are available.
2197 static inline void netif_wake_queue(struct net_device *dev)
2199 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2202 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2206 for (i = 0; i < dev->num_tx_queues; i++) {
2207 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2208 netif_tx_wake_queue(txq);
2212 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2214 if (WARN_ON(!dev_queue)) {
2215 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2218 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2222 * netif_stop_queue - stop transmitted packets
2223 * @dev: network device
2225 * Stop upper layers calling the device hard_start_xmit routine.
2226 * Used for flow control when transmit resources are unavailable.
2228 static inline void netif_stop_queue(struct net_device *dev)
2230 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2233 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2237 for (i = 0; i < dev->num_tx_queues; i++) {
2238 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2239 netif_tx_stop_queue(txq);
2243 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2245 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2249 * netif_queue_stopped - test if transmit queue is flowblocked
2250 * @dev: network device
2252 * Test if transmit queue on device is currently unable to send.
2254 static inline bool netif_queue_stopped(const struct net_device *dev)
2256 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2259 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2261 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2264 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2266 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2269 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2273 dql_queued(&dev_queue->dql, bytes);
2275 if (likely(dql_avail(&dev_queue->dql) >= 0))
2278 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2281 * The XOFF flag must be set before checking the dql_avail below,
2282 * because in netdev_tx_completed_queue we update the dql_completed
2283 * before checking the XOFF flag.
2287 /* check again in case another CPU has just made room avail */
2288 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2289 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2294 * netdev_sent_queue - report the number of bytes queued to hardware
2295 * @dev: network device
2296 * @bytes: number of bytes queued to the hardware device queue
2298 * Report the number of bytes queued for sending/completion to the network
2299 * device hardware queue. @bytes should be a good approximation and should
2300 * exactly match netdev_completed_queue() @bytes
2302 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2304 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2307 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2308 unsigned int pkts, unsigned int bytes)
2311 if (unlikely(!bytes))
2314 dql_completed(&dev_queue->dql, bytes);
2317 * Without the memory barrier there is a small possiblity that
2318 * netdev_tx_sent_queue will miss the update and cause the queue to
2319 * be stopped forever
2323 if (dql_avail(&dev_queue->dql) < 0)
2326 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2327 netif_schedule_queue(dev_queue);
2332 * netdev_completed_queue - report bytes and packets completed by device
2333 * @dev: network device
2334 * @pkts: actual number of packets sent over the medium
2335 * @bytes: actual number of bytes sent over the medium
2337 * Report the number of bytes and packets transmitted by the network device
2338 * hardware queue over the physical medium, @bytes must exactly match the
2339 * @bytes amount passed to netdev_sent_queue()
2341 static inline void netdev_completed_queue(struct net_device *dev,
2342 unsigned int pkts, unsigned int bytes)
2344 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2347 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2350 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2356 * netdev_reset_queue - reset the packets and bytes count of a network device
2357 * @dev_queue: network device
2359 * Reset the bytes and packet count of a network device and clear the
2360 * software flow control OFF bit for this network device
2362 static inline void netdev_reset_queue(struct net_device *dev_queue)
2364 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2368 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2369 * @dev: network device
2370 * @queue_index: given tx queue index
2372 * Returns 0 if given tx queue index >= number of device tx queues,
2373 * otherwise returns the originally passed tx queue index.
2375 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2377 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2378 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2379 dev->name, queue_index,
2380 dev->real_num_tx_queues);
2388 * netif_running - test if up
2389 * @dev: network device
2391 * Test if the device has been brought up.
2393 static inline bool netif_running(const struct net_device *dev)
2395 return test_bit(__LINK_STATE_START, &dev->state);
2399 * Routines to manage the subqueues on a device. We only need start
2400 * stop, and a check if it's stopped. All other device management is
2401 * done at the overall netdevice level.
2402 * Also test the device if we're multiqueue.
2406 * netif_start_subqueue - allow sending packets on subqueue
2407 * @dev: network device
2408 * @queue_index: sub queue index
2410 * Start individual transmit queue of a device with multiple transmit queues.
2412 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2414 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2416 netif_tx_start_queue(txq);
2420 * netif_stop_subqueue - stop sending packets on subqueue
2421 * @dev: network device
2422 * @queue_index: sub queue index
2424 * Stop individual transmit queue of a device with multiple transmit queues.
2426 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2428 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2429 netif_tx_stop_queue(txq);
2433 * netif_subqueue_stopped - test status of subqueue
2434 * @dev: network device
2435 * @queue_index: sub queue index
2437 * Check individual transmit queue of a device with multiple transmit queues.
2439 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2442 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2444 return netif_tx_queue_stopped(txq);
2447 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2448 struct sk_buff *skb)
2450 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2454 * netif_wake_subqueue - allow sending packets on subqueue
2455 * @dev: network device
2456 * @queue_index: sub queue index
2458 * Resume individual transmit queue of a device with multiple transmit queues.
2460 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2462 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2463 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2464 __netif_schedule(txq->qdisc);
2468 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2471 static inline int netif_set_xps_queue(struct net_device *dev,
2472 const struct cpumask *mask,
2480 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2481 * as a distribution range limit for the returned value.
2483 static inline u16 skb_tx_hash(const struct net_device *dev,
2484 const struct sk_buff *skb)
2486 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2490 * netif_is_multiqueue - test if device has multiple transmit queues
2491 * @dev: network device
2493 * Check if device has multiple transmit queues
2495 static inline bool netif_is_multiqueue(const struct net_device *dev)
2497 return dev->num_tx_queues > 1;
2500 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2503 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2505 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2512 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2513 const struct net_device *from_dev)
2517 err = netif_set_real_num_tx_queues(to_dev,
2518 from_dev->real_num_tx_queues);
2522 return netif_set_real_num_rx_queues(to_dev,
2523 from_dev->real_num_rx_queues);
2530 static inline unsigned int get_netdev_rx_queue_index(
2531 struct netdev_rx_queue *queue)
2533 struct net_device *dev = queue->dev;
2534 int index = queue - dev->_rx;
2536 BUG_ON(index >= dev->num_rx_queues);
2541 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2542 int netif_get_num_default_rss_queues(void);
2544 enum skb_free_reason {
2545 SKB_REASON_CONSUMED,
2549 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2550 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2553 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2554 * interrupt context or with hardware interrupts being disabled.
2555 * (in_irq() || irqs_disabled())
2557 * We provide four helpers that can be used in following contexts :
2559 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2560 * replacing kfree_skb(skb)
2562 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2563 * Typically used in place of consume_skb(skb) in TX completion path
2565 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2566 * replacing kfree_skb(skb)
2568 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2569 * and consumed a packet. Used in place of consume_skb(skb)
2571 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2573 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2576 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2578 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2581 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2583 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2586 static inline void dev_consume_skb_any(struct sk_buff *skb)
2588 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2591 int netif_rx(struct sk_buff *skb);
2592 int netif_rx_ni(struct sk_buff *skb);
2593 int netif_receive_skb(struct sk_buff *skb);
2594 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2595 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2596 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2597 gro_result_t napi_gro_frags(struct napi_struct *napi);
2598 struct packet_offload *gro_find_receive_by_type(__be16 type);
2599 struct packet_offload *gro_find_complete_by_type(__be16 type);
2601 static inline void napi_free_frags(struct napi_struct *napi)
2603 kfree_skb(napi->skb);
2607 int netdev_rx_handler_register(struct net_device *dev,
2608 rx_handler_func_t *rx_handler,
2609 void *rx_handler_data);
2610 void netdev_rx_handler_unregister(struct net_device *dev);
2612 bool dev_valid_name(const char *name);
2613 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2614 int dev_ethtool(struct net *net, struct ifreq *);
2615 unsigned int dev_get_flags(const struct net_device *);
2616 int __dev_change_flags(struct net_device *, unsigned int flags);
2617 int dev_change_flags(struct net_device *, unsigned int);
2618 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2619 unsigned int gchanges);
2620 int dev_change_name(struct net_device *, const char *);
2621 int dev_set_alias(struct net_device *, const char *, size_t);
2622 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2623 int dev_set_mtu(struct net_device *, int);
2624 void dev_set_group(struct net_device *, int);
2625 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2626 int dev_change_carrier(struct net_device *, bool new_carrier);
2627 int dev_get_phys_port_id(struct net_device *dev,
2628 struct netdev_phys_port_id *ppid);
2629 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2630 struct netdev_queue *txq);
2631 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2633 extern int netdev_budget;
2635 /* Called by rtnetlink.c:rtnl_unlock() */
2636 void netdev_run_todo(void);
2639 * dev_put - release reference to device
2640 * @dev: network device
2642 * Release reference to device to allow it to be freed.
2644 static inline void dev_put(struct net_device *dev)
2646 this_cpu_dec(*dev->pcpu_refcnt);
2650 * dev_hold - get reference to device
2651 * @dev: network device
2653 * Hold reference to device to keep it from being freed.
2655 static inline void dev_hold(struct net_device *dev)
2657 this_cpu_inc(*dev->pcpu_refcnt);
2660 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2661 * and _off may be called from IRQ context, but it is caller
2662 * who is responsible for serialization of these calls.
2664 * The name carrier is inappropriate, these functions should really be
2665 * called netif_lowerlayer_*() because they represent the state of any
2666 * kind of lower layer not just hardware media.
2669 void linkwatch_init_dev(struct net_device *dev);
2670 void linkwatch_fire_event(struct net_device *dev);
2671 void linkwatch_forget_dev(struct net_device *dev);
2674 * netif_carrier_ok - test if carrier present
2675 * @dev: network device
2677 * Check if carrier is present on device
2679 static inline bool netif_carrier_ok(const struct net_device *dev)
2681 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2684 unsigned long dev_trans_start(struct net_device *dev);
2686 void __netdev_watchdog_up(struct net_device *dev);
2688 void netif_carrier_on(struct net_device *dev);
2690 void netif_carrier_off(struct net_device *dev);
2693 * netif_dormant_on - mark device as dormant.
2694 * @dev: network device
2696 * Mark device as dormant (as per RFC2863).
2698 * The dormant state indicates that the relevant interface is not
2699 * actually in a condition to pass packets (i.e., it is not 'up') but is
2700 * in a "pending" state, waiting for some external event. For "on-
2701 * demand" interfaces, this new state identifies the situation where the
2702 * interface is waiting for events to place it in the up state.
2705 static inline void netif_dormant_on(struct net_device *dev)
2707 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2708 linkwatch_fire_event(dev);
2712 * netif_dormant_off - set device as not dormant.
2713 * @dev: network device
2715 * Device is not in dormant state.
2717 static inline void netif_dormant_off(struct net_device *dev)
2719 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2720 linkwatch_fire_event(dev);
2724 * netif_dormant - test if carrier present
2725 * @dev: network device
2727 * Check if carrier is present on device
2729 static inline bool netif_dormant(const struct net_device *dev)
2731 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2736 * netif_oper_up - test if device is operational
2737 * @dev: network device
2739 * Check if carrier is operational
2741 static inline bool netif_oper_up(const struct net_device *dev)
2743 return (dev->operstate == IF_OPER_UP ||
2744 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2748 * netif_device_present - is device available or removed
2749 * @dev: network device
2751 * Check if device has not been removed from system.
2753 static inline bool netif_device_present(struct net_device *dev)
2755 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2758 void netif_device_detach(struct net_device *dev);
2760 void netif_device_attach(struct net_device *dev);
2763 * Network interface message level settings
2767 NETIF_MSG_DRV = 0x0001,
2768 NETIF_MSG_PROBE = 0x0002,
2769 NETIF_MSG_LINK = 0x0004,
2770 NETIF_MSG_TIMER = 0x0008,
2771 NETIF_MSG_IFDOWN = 0x0010,
2772 NETIF_MSG_IFUP = 0x0020,
2773 NETIF_MSG_RX_ERR = 0x0040,
2774 NETIF_MSG_TX_ERR = 0x0080,
2775 NETIF_MSG_TX_QUEUED = 0x0100,
2776 NETIF_MSG_INTR = 0x0200,
2777 NETIF_MSG_TX_DONE = 0x0400,
2778 NETIF_MSG_RX_STATUS = 0x0800,
2779 NETIF_MSG_PKTDATA = 0x1000,
2780 NETIF_MSG_HW = 0x2000,
2781 NETIF_MSG_WOL = 0x4000,
2784 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2785 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2786 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2787 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2788 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2789 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2790 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2791 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2792 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2793 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2794 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2795 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2796 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2797 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2798 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2800 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2803 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2804 return default_msg_enable_bits;
2805 if (debug_value == 0) /* no output */
2807 /* set low N bits */
2808 return (1 << debug_value) - 1;
2811 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2813 spin_lock(&txq->_xmit_lock);
2814 txq->xmit_lock_owner = cpu;
2817 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2819 spin_lock_bh(&txq->_xmit_lock);
2820 txq->xmit_lock_owner = smp_processor_id();
2823 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2825 bool ok = spin_trylock(&txq->_xmit_lock);
2827 txq->xmit_lock_owner = smp_processor_id();
2831 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2833 txq->xmit_lock_owner = -1;
2834 spin_unlock(&txq->_xmit_lock);
2837 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2839 txq->xmit_lock_owner = -1;
2840 spin_unlock_bh(&txq->_xmit_lock);
2843 static inline void txq_trans_update(struct netdev_queue *txq)
2845 if (txq->xmit_lock_owner != -1)
2846 txq->trans_start = jiffies;
2850 * netif_tx_lock - grab network device transmit lock
2851 * @dev: network device
2853 * Get network device transmit lock
2855 static inline void netif_tx_lock(struct net_device *dev)
2860 spin_lock(&dev->tx_global_lock);
2861 cpu = smp_processor_id();
2862 for (i = 0; i < dev->num_tx_queues; i++) {
2863 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2865 /* We are the only thread of execution doing a
2866 * freeze, but we have to grab the _xmit_lock in
2867 * order to synchronize with threads which are in
2868 * the ->hard_start_xmit() handler and already
2869 * checked the frozen bit.
2871 __netif_tx_lock(txq, cpu);
2872 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2873 __netif_tx_unlock(txq);
2877 static inline void netif_tx_lock_bh(struct net_device *dev)
2883 static inline void netif_tx_unlock(struct net_device *dev)
2887 for (i = 0; i < dev->num_tx_queues; i++) {
2888 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2890 /* No need to grab the _xmit_lock here. If the
2891 * queue is not stopped for another reason, we
2894 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2895 netif_schedule_queue(txq);
2897 spin_unlock(&dev->tx_global_lock);
2900 static inline void netif_tx_unlock_bh(struct net_device *dev)
2902 netif_tx_unlock(dev);
2906 #define HARD_TX_LOCK(dev, txq, cpu) { \
2907 if ((dev->features & NETIF_F_LLTX) == 0) { \
2908 __netif_tx_lock(txq, cpu); \
2912 #define HARD_TX_TRYLOCK(dev, txq) \
2913 (((dev->features & NETIF_F_LLTX) == 0) ? \
2914 __netif_tx_trylock(txq) : \
2917 #define HARD_TX_UNLOCK(dev, txq) { \
2918 if ((dev->features & NETIF_F_LLTX) == 0) { \
2919 __netif_tx_unlock(txq); \
2923 static inline void netif_tx_disable(struct net_device *dev)
2929 cpu = smp_processor_id();
2930 for (i = 0; i < dev->num_tx_queues; i++) {
2931 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2933 __netif_tx_lock(txq, cpu);
2934 netif_tx_stop_queue(txq);
2935 __netif_tx_unlock(txq);
2940 static inline void netif_addr_lock(struct net_device *dev)
2942 spin_lock(&dev->addr_list_lock);
2945 static inline void netif_addr_lock_nested(struct net_device *dev)
2947 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2950 static inline void netif_addr_lock_bh(struct net_device *dev)
2952 spin_lock_bh(&dev->addr_list_lock);
2955 static inline void netif_addr_unlock(struct net_device *dev)
2957 spin_unlock(&dev->addr_list_lock);
2960 static inline void netif_addr_unlock_bh(struct net_device *dev)
2962 spin_unlock_bh(&dev->addr_list_lock);
2966 * dev_addrs walker. Should be used only for read access. Call with
2967 * rcu_read_lock held.
2969 #define for_each_dev_addr(dev, ha) \
2970 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2972 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2974 void ether_setup(struct net_device *dev);
2976 /* Support for loadable net-drivers */
2977 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2978 void (*setup)(struct net_device *),
2979 unsigned int txqs, unsigned int rxqs);
2980 #define alloc_netdev(sizeof_priv, name, setup) \
2981 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2983 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2984 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2986 int register_netdev(struct net_device *dev);
2987 void unregister_netdev(struct net_device *dev);
2989 /* General hardware address lists handling functions */
2990 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2991 struct netdev_hw_addr_list *from_list, int addr_len);
2992 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2993 struct netdev_hw_addr_list *from_list, int addr_len);
2994 void __hw_addr_init(struct netdev_hw_addr_list *list);
2996 /* Functions used for device addresses handling */
2997 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2998 unsigned char addr_type);
2999 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3000 unsigned char addr_type);
3001 void dev_addr_flush(struct net_device *dev);
3002 int dev_addr_init(struct net_device *dev);
3004 /* Functions used for unicast addresses handling */
3005 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3006 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3007 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3008 int dev_uc_sync(struct net_device *to, struct net_device *from);
3009 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3010 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3011 void dev_uc_flush(struct net_device *dev);
3012 void dev_uc_init(struct net_device *dev);
3014 /* Functions used for multicast addresses handling */
3015 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3016 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3017 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3018 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3019 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3020 int dev_mc_sync(struct net_device *to, struct net_device *from);
3021 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3022 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3023 void dev_mc_flush(struct net_device *dev);
3024 void dev_mc_init(struct net_device *dev);
3026 /* Functions used for secondary unicast and multicast support */
3027 void dev_set_rx_mode(struct net_device *dev);
3028 void __dev_set_rx_mode(struct net_device *dev);
3029 int dev_set_promiscuity(struct net_device *dev, int inc);
3030 int dev_set_allmulti(struct net_device *dev, int inc);
3031 void netdev_state_change(struct net_device *dev);
3032 void netdev_notify_peers(struct net_device *dev);
3033 void netdev_features_change(struct net_device *dev);
3034 /* Load a device via the kmod */
3035 void dev_load(struct net *net, const char *name);
3036 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3037 struct rtnl_link_stats64 *storage);
3038 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3039 const struct net_device_stats *netdev_stats);
3041 extern int netdev_max_backlog;
3042 extern int netdev_tstamp_prequeue;
3043 extern int weight_p;
3044 extern int bpf_jit_enable;
3046 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3047 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3048 struct list_head **iter);
3050 /* iterate through upper list, must be called under RCU read lock */
3051 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3052 for (iter = &(dev)->all_adj_list.upper, \
3053 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3055 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3057 void *netdev_lower_get_next_private(struct net_device *dev,
3058 struct list_head **iter);
3059 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3060 struct list_head **iter);
3062 #define netdev_for_each_lower_private(dev, priv, iter) \
3063 for (iter = (dev)->adj_list.lower.next, \
3064 priv = netdev_lower_get_next_private(dev, &(iter)); \
3066 priv = netdev_lower_get_next_private(dev, &(iter)))
3068 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3069 for (iter = &(dev)->adj_list.lower, \
3070 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3072 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3074 void *netdev_adjacent_get_private(struct list_head *adj_list);
3075 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3076 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3077 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3078 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3079 int netdev_master_upper_dev_link(struct net_device *dev,
3080 struct net_device *upper_dev);
3081 int netdev_master_upper_dev_link_private(struct net_device *dev,
3082 struct net_device *upper_dev,
3084 void netdev_upper_dev_unlink(struct net_device *dev,
3085 struct net_device *upper_dev);
3086 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3087 void *netdev_lower_dev_get_private(struct net_device *dev,
3088 struct net_device *lower_dev);
3089 int skb_checksum_help(struct sk_buff *skb);
3090 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3091 netdev_features_t features, bool tx_path);
3092 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3093 netdev_features_t features);
3096 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3098 return __skb_gso_segment(skb, features, true);
3100 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3102 static inline bool can_checksum_protocol(netdev_features_t features,
3105 return ((features & NETIF_F_GEN_CSUM) ||
3106 ((features & NETIF_F_V4_CSUM) &&
3107 protocol == htons(ETH_P_IP)) ||
3108 ((features & NETIF_F_V6_CSUM) &&
3109 protocol == htons(ETH_P_IPV6)) ||
3110 ((features & NETIF_F_FCOE_CRC) &&
3111 protocol == htons(ETH_P_FCOE)));
3115 void netdev_rx_csum_fault(struct net_device *dev);
3117 static inline void netdev_rx_csum_fault(struct net_device *dev)
3121 /* rx skb timestamps */
3122 void net_enable_timestamp(void);
3123 void net_disable_timestamp(void);
3125 #ifdef CONFIG_PROC_FS
3126 int __init dev_proc_init(void);
3128 #define dev_proc_init() 0
3131 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3133 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3136 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3138 return netdev_class_create_file_ns(class_attr, NULL);
3141 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3143 netdev_class_remove_file_ns(class_attr, NULL);
3146 extern struct kobj_ns_type_operations net_ns_type_operations;
3148 const char *netdev_drivername(const struct net_device *dev);
3150 void linkwatch_run_queue(void);
3152 static inline netdev_features_t netdev_get_wanted_features(
3153 struct net_device *dev)
3155 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3157 netdev_features_t netdev_increment_features(netdev_features_t all,
3158 netdev_features_t one, netdev_features_t mask);
3160 /* Allow TSO being used on stacked device :
3161 * Performing the GSO segmentation before last device
3162 * is a performance improvement.
3164 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3165 netdev_features_t mask)
3167 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3170 int __netdev_update_features(struct net_device *dev);
3171 void netdev_update_features(struct net_device *dev);
3172 void netdev_change_features(struct net_device *dev);
3174 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3175 struct net_device *dev);
3177 netdev_features_t netif_skb_dev_features(struct sk_buff *skb,
3178 const struct net_device *dev);
3179 static inline netdev_features_t netif_skb_features(struct sk_buff *skb)
3181 return netif_skb_dev_features(skb, skb->dev);
3184 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3186 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3188 /* check flags correspondence */
3189 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3190 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3191 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3192 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3193 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3194 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3196 return (features & feature) == feature;
3199 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3201 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3202 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3205 static inline bool netif_needs_gso(struct sk_buff *skb,
3206 netdev_features_t features)
3208 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3209 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3210 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3213 static inline void netif_set_gso_max_size(struct net_device *dev,
3216 dev->gso_max_size = size;
3219 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3220 int pulled_hlen, u16 mac_offset,
3223 skb->protocol = protocol;
3224 skb->encapsulation = 1;
3225 skb_push(skb, pulled_hlen);
3226 skb_reset_transport_header(skb);
3227 skb->mac_header = mac_offset;
3228 skb->network_header = skb->mac_header + mac_len;
3229 skb->mac_len = mac_len;
3232 static inline bool netif_is_macvlan(struct net_device *dev)
3234 return dev->priv_flags & IFF_MACVLAN;
3237 static inline bool netif_is_bond_master(struct net_device *dev)
3239 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3242 static inline bool netif_is_bond_slave(struct net_device *dev)
3244 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3247 static inline bool netif_supports_nofcs(struct net_device *dev)
3249 return dev->priv_flags & IFF_SUPP_NOFCS;
3252 extern struct pernet_operations __net_initdata loopback_net_ops;
3254 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3256 /* netdev_printk helpers, similar to dev_printk */
3258 static inline const char *netdev_name(const struct net_device *dev)
3260 if (dev->reg_state != NETREG_REGISTERED)
3261 return "(unregistered net_device)";
3266 int netdev_printk(const char *level, const struct net_device *dev,
3267 const char *format, ...);
3269 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3271 int netdev_alert(const struct net_device *dev, const char *format, ...);
3273 int netdev_crit(const struct net_device *dev, const char *format, ...);
3275 int netdev_err(const struct net_device *dev, const char *format, ...);
3277 int netdev_warn(const struct net_device *dev, const char *format, ...);
3279 int netdev_notice(const struct net_device *dev, const char *format, ...);
3281 int netdev_info(const struct net_device *dev, const char *format, ...);
3283 #define MODULE_ALIAS_NETDEV(device) \
3284 MODULE_ALIAS("netdev-" device)
3286 #if defined(CONFIG_DYNAMIC_DEBUG)
3287 #define netdev_dbg(__dev, format, args...) \
3289 dynamic_netdev_dbg(__dev, format, ##args); \
3291 #elif defined(DEBUG)
3292 #define netdev_dbg(__dev, format, args...) \
3293 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3295 #define netdev_dbg(__dev, format, args...) \
3298 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3303 #if defined(VERBOSE_DEBUG)
3304 #define netdev_vdbg netdev_dbg
3307 #define netdev_vdbg(dev, format, args...) \
3310 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3316 * netdev_WARN() acts like dev_printk(), but with the key difference
3317 * of using a WARN/WARN_ON to get the message out, including the
3318 * file/line information and a backtrace.
3320 #define netdev_WARN(dev, format, args...) \
3321 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3323 /* netif printk helpers, similar to netdev_printk */
3325 #define netif_printk(priv, type, level, dev, fmt, args...) \
3327 if (netif_msg_##type(priv)) \
3328 netdev_printk(level, (dev), fmt, ##args); \
3331 #define netif_level(level, priv, type, dev, fmt, args...) \
3333 if (netif_msg_##type(priv)) \
3334 netdev_##level(dev, fmt, ##args); \
3337 #define netif_emerg(priv, type, dev, fmt, args...) \
3338 netif_level(emerg, priv, type, dev, fmt, ##args)
3339 #define netif_alert(priv, type, dev, fmt, args...) \
3340 netif_level(alert, priv, type, dev, fmt, ##args)
3341 #define netif_crit(priv, type, dev, fmt, args...) \
3342 netif_level(crit, priv, type, dev, fmt, ##args)
3343 #define netif_err(priv, type, dev, fmt, args...) \
3344 netif_level(err, priv, type, dev, fmt, ##args)
3345 #define netif_warn(priv, type, dev, fmt, args...) \
3346 netif_level(warn, priv, type, dev, fmt, ##args)
3347 #define netif_notice(priv, type, dev, fmt, args...) \
3348 netif_level(notice, priv, type, dev, fmt, ##args)
3349 #define netif_info(priv, type, dev, fmt, args...) \
3350 netif_level(info, priv, type, dev, fmt, ##args)
3352 #if defined(CONFIG_DYNAMIC_DEBUG)
3353 #define netif_dbg(priv, type, netdev, format, args...) \
3355 if (netif_msg_##type(priv)) \
3356 dynamic_netdev_dbg(netdev, format, ##args); \
3358 #elif defined(DEBUG)
3359 #define netif_dbg(priv, type, dev, format, args...) \
3360 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3362 #define netif_dbg(priv, type, dev, format, args...) \
3365 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3370 #if defined(VERBOSE_DEBUG)
3371 #define netif_vdbg netif_dbg
3373 #define netif_vdbg(priv, type, dev, format, args...) \
3376 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3382 * The list of packet types we will receive (as opposed to discard)
3383 * and the routines to invoke.
3385 * Why 16. Because with 16 the only overlap we get on a hash of the
3386 * low nibble of the protocol value is RARP/SNAP/X.25.
3388 * NOTE: That is no longer true with the addition of VLAN tags. Not
3389 * sure which should go first, but I bet it won't make much
3390 * difference if we are running VLANs. The good news is that
3391 * this protocol won't be in the list unless compiled in, so
3392 * the average user (w/out VLANs) will not be adversely affected.
3408 #define PTYPE_HASH_SIZE (16)
3409 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3411 #endif /* _LINUX_NETDEVICE_H */