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 <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
47 #include <net/dcbnl.h>
49 #include <net/netprio_cgroup.h>
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
61 void netdev_set_default_ethtool_ops(struct net_device *dev,
62 const struct ethtool_ops *ops);
64 /* Backlog congestion levels */
65 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
66 #define NET_RX_DROP 1 /* packet dropped */
69 * Transmit return codes: transmit return codes originate from three different
72 * - qdisc return codes
73 * - driver transmit return codes
76 * Drivers are allowed to return any one of those in their hard_start_xmit()
77 * function. Real network devices commonly used with qdiscs should only return
78 * the driver transmit return codes though - when qdiscs are used, the actual
79 * transmission happens asynchronously, so the value is not propagated to
80 * higher layers. Virtual network devices transmit synchronously, in this case
81 * the driver transmit return codes are consumed by dev_queue_xmit(), all
82 * others are propagated to higher layers.
85 /* qdisc ->enqueue() return codes. */
86 #define NET_XMIT_SUCCESS 0x00
87 #define NET_XMIT_DROP 0x01 /* skb dropped */
88 #define NET_XMIT_CN 0x02 /* congestion notification */
89 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
90 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
92 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
93 * indicates that the device will soon be dropping packets, or already drops
94 * some packets of the same priority; prompting us to send less aggressively. */
95 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
96 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
98 /* Driver transmit return codes */
99 #define NETDEV_TX_MASK 0xf0
102 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
103 NETDEV_TX_OK = 0x00, /* driver took care of packet */
104 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
105 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
107 typedef enum netdev_tx netdev_tx_t;
110 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
111 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
113 static inline bool dev_xmit_complete(int rc)
116 * Positive cases with an skb consumed by a driver:
117 * - successful transmission (rc == NETDEV_TX_OK)
118 * - error while transmitting (rc < 0)
119 * - error while queueing to a different device (rc & NET_XMIT_MASK)
121 if (likely(rc < NET_XMIT_MASK))
128 * Compute the worst case header length according to the protocols
132 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
133 # if defined(CONFIG_MAC80211_MESH)
134 # define LL_MAX_HEADER 128
136 # define LL_MAX_HEADER 96
139 # define LL_MAX_HEADER 32
142 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
143 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
144 #define MAX_HEADER LL_MAX_HEADER
146 #define MAX_HEADER (LL_MAX_HEADER + 48)
150 * Old network device statistics. Fields are native words
151 * (unsigned long) so they can be read and written atomically.
154 struct net_device_stats {
155 unsigned long rx_packets;
156 unsigned long tx_packets;
157 unsigned long rx_bytes;
158 unsigned long tx_bytes;
159 unsigned long rx_errors;
160 unsigned long tx_errors;
161 unsigned long rx_dropped;
162 unsigned long tx_dropped;
163 unsigned long multicast;
164 unsigned long collisions;
165 unsigned long rx_length_errors;
166 unsigned long rx_over_errors;
167 unsigned long rx_crc_errors;
168 unsigned long rx_frame_errors;
169 unsigned long rx_fifo_errors;
170 unsigned long rx_missed_errors;
171 unsigned long tx_aborted_errors;
172 unsigned long tx_carrier_errors;
173 unsigned long tx_fifo_errors;
174 unsigned long tx_heartbeat_errors;
175 unsigned long tx_window_errors;
176 unsigned long rx_compressed;
177 unsigned long tx_compressed;
181 #include <linux/cache.h>
182 #include <linux/skbuff.h>
185 #include <linux/static_key.h>
186 extern struct static_key rps_needed;
193 struct netdev_hw_addr {
194 struct list_head list;
195 unsigned char addr[MAX_ADDR_LEN];
197 #define NETDEV_HW_ADDR_T_LAN 1
198 #define NETDEV_HW_ADDR_T_SAN 2
199 #define NETDEV_HW_ADDR_T_SLAVE 3
200 #define NETDEV_HW_ADDR_T_UNICAST 4
201 #define NETDEV_HW_ADDR_T_MULTICAST 5
206 struct rcu_head rcu_head;
209 struct netdev_hw_addr_list {
210 struct list_head list;
214 #define netdev_hw_addr_list_count(l) ((l)->count)
215 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
216 #define netdev_hw_addr_list_for_each(ha, l) \
217 list_for_each_entry(ha, &(l)->list, list)
219 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
220 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
221 #define netdev_for_each_uc_addr(ha, dev) \
222 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
224 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
225 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
226 #define netdev_for_each_mc_addr(ha, dev) \
227 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
234 /* cached hardware header; allow for machine alignment needs. */
235 #define HH_DATA_MOD 16
236 #define HH_DATA_OFF(__len) \
237 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
238 #define HH_DATA_ALIGN(__len) \
239 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
240 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
243 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
245 * dev->hard_header_len ? (dev->hard_header_len +
246 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
248 * We could use other alignment values, but we must maintain the
249 * relationship HH alignment <= LL alignment.
251 #define LL_RESERVED_SPACE(dev) \
252 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
253 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
254 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
257 int (*create) (struct sk_buff *skb, struct net_device *dev,
258 unsigned short type, const void *daddr,
259 const void *saddr, unsigned int len);
260 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
261 int (*rebuild)(struct sk_buff *skb);
262 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
263 void (*cache_update)(struct hh_cache *hh,
264 const struct net_device *dev,
265 const unsigned char *haddr);
268 /* These flag bits are private to the generic network queueing
269 * layer, they may not be explicitly referenced by any other
273 enum netdev_state_t {
275 __LINK_STATE_PRESENT,
276 __LINK_STATE_NOCARRIER,
277 __LINK_STATE_LINKWATCH_PENDING,
278 __LINK_STATE_DORMANT,
283 * This structure holds at boot time configured netdevice settings. They
284 * are then used in the device probing.
286 struct netdev_boot_setup {
290 #define NETDEV_BOOT_SETUP_MAX 8
292 int __init netdev_boot_setup(char *str);
295 * Structure for NAPI scheduling similar to tasklet but with weighting
298 /* The poll_list must only be managed by the entity which
299 * changes the state of the NAPI_STATE_SCHED bit. This means
300 * whoever atomically sets that bit can add this napi_struct
301 * to the per-cpu poll_list, and whoever clears that bit
302 * can remove from the list right before clearing the bit.
304 struct list_head poll_list;
308 unsigned int gro_count;
309 int (*poll)(struct napi_struct *, int);
310 #ifdef CONFIG_NETPOLL
311 spinlock_t poll_lock;
314 struct net_device *dev;
315 struct sk_buff *gro_list;
317 struct list_head dev_list;
318 struct hlist_node napi_hash_node;
319 unsigned int napi_id;
323 NAPI_STATE_SCHED, /* Poll is scheduled */
324 NAPI_STATE_DISABLE, /* Disable pending */
325 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
326 NAPI_STATE_HASHED, /* In NAPI hash */
336 typedef enum gro_result gro_result_t;
339 * enum rx_handler_result - Possible return values for rx_handlers.
340 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
342 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
343 * case skb->dev was changed by rx_handler.
344 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
345 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
347 * rx_handlers are functions called from inside __netif_receive_skb(), to do
348 * special processing of the skb, prior to delivery to protocol handlers.
350 * Currently, a net_device can only have a single rx_handler registered. Trying
351 * to register a second rx_handler will return -EBUSY.
353 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
354 * To unregister a rx_handler on a net_device, use
355 * netdev_rx_handler_unregister().
357 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
360 * If the rx_handler consumed to skb in some way, it should return
361 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
362 * the skb to be delivered in some other ways.
364 * If the rx_handler changed skb->dev, to divert the skb to another
365 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
366 * new device will be called if it exists.
368 * If the rx_handler consider the skb should be ignored, it should return
369 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
370 * are registered on exact device (ptype->dev == skb->dev).
372 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
373 * delivered, it should return RX_HANDLER_PASS.
375 * A device without a registered rx_handler will behave as if rx_handler
376 * returned RX_HANDLER_PASS.
379 enum rx_handler_result {
385 typedef enum rx_handler_result rx_handler_result_t;
386 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
388 void __napi_schedule(struct napi_struct *n);
390 static inline bool napi_disable_pending(struct napi_struct *n)
392 return test_bit(NAPI_STATE_DISABLE, &n->state);
396 * napi_schedule_prep - check if napi can be scheduled
399 * Test if NAPI routine is already running, and if not mark
400 * it as running. This is used as a condition variable
401 * insure only one NAPI poll instance runs. We also make
402 * sure there is no pending NAPI disable.
404 static inline bool napi_schedule_prep(struct napi_struct *n)
406 return !napi_disable_pending(n) &&
407 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
411 * napi_schedule - schedule NAPI poll
414 * Schedule NAPI poll routine to be called if it is not already
417 static inline void napi_schedule(struct napi_struct *n)
419 if (napi_schedule_prep(n))
423 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
424 static inline bool napi_reschedule(struct napi_struct *napi)
426 if (napi_schedule_prep(napi)) {
427 __napi_schedule(napi);
434 * napi_complete - NAPI processing complete
437 * Mark NAPI processing as complete.
439 void __napi_complete(struct napi_struct *n);
440 void napi_complete(struct napi_struct *n);
443 * napi_by_id - lookup a NAPI by napi_id
444 * @napi_id: hashed napi_id
446 * lookup @napi_id in napi_hash table
447 * must be called under rcu_read_lock()
449 struct napi_struct *napi_by_id(unsigned int napi_id);
452 * napi_hash_add - add a NAPI to global hashtable
453 * @napi: napi context
455 * generate a new napi_id and store a @napi under it in napi_hash
457 void napi_hash_add(struct napi_struct *napi);
460 * napi_hash_del - remove a NAPI from global table
461 * @napi: napi context
463 * Warning: caller must observe rcu grace period
464 * before freeing memory containing @napi
466 void napi_hash_del(struct napi_struct *napi);
469 * napi_disable - prevent NAPI from scheduling
472 * Stop NAPI from being scheduled on this context.
473 * Waits till any outstanding processing completes.
475 static inline void napi_disable(struct napi_struct *n)
478 set_bit(NAPI_STATE_DISABLE, &n->state);
479 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
481 clear_bit(NAPI_STATE_DISABLE, &n->state);
485 * napi_enable - enable NAPI scheduling
488 * Resume NAPI from being scheduled on this context.
489 * Must be paired with napi_disable.
491 static inline void napi_enable(struct napi_struct *n)
493 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
494 smp_mb__before_atomic();
495 clear_bit(NAPI_STATE_SCHED, &n->state);
500 * napi_synchronize - wait until NAPI is not running
503 * Wait until NAPI is done being scheduled on this context.
504 * Waits till any outstanding processing completes but
505 * does not disable future activations.
507 static inline void napi_synchronize(const struct napi_struct *n)
509 while (test_bit(NAPI_STATE_SCHED, &n->state))
513 # define napi_synchronize(n) barrier()
516 enum netdev_queue_state_t {
517 __QUEUE_STATE_DRV_XOFF,
518 __QUEUE_STATE_STACK_XOFF,
519 __QUEUE_STATE_FROZEN,
522 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
523 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
524 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
526 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
527 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
529 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
533 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
534 * netif_tx_* functions below are used to manipulate this flag. The
535 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
536 * queue independently. The netif_xmit_*stopped functions below are called
537 * to check if the queue has been stopped by the driver or stack (either
538 * of the XOFF bits are set in the state). Drivers should not need to call
539 * netif_xmit*stopped functions, they should only be using netif_tx_*.
542 struct netdev_queue {
546 struct net_device *dev;
547 struct Qdisc __rcu *qdisc;
548 struct Qdisc *qdisc_sleeping;
552 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
558 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
561 * please use this field instead of dev->trans_start
563 unsigned long trans_start;
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
569 unsigned long trans_timeout;
576 } ____cacheline_aligned_in_smp;
578 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
580 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
587 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 * This structure holds an RPS map which can be of variable length. The
597 * map is an array of CPUs.
604 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
607 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
608 * tail pointer for that CPU's input queue at the time of last enqueue, and
609 * a hardware filter index.
611 struct rps_dev_flow {
614 unsigned int last_qtail;
616 #define RPS_NO_FILTER 0xffff
619 * The rps_dev_flow_table structure contains a table of flow mappings.
621 struct rps_dev_flow_table {
624 struct rps_dev_flow flows[0];
626 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
627 ((_num) * sizeof(struct rps_dev_flow)))
630 * The rps_sock_flow_table contains mappings of flows to the last CPU
631 * on which they were processed by the application (set in recvmsg).
633 struct rps_sock_flow_table {
637 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
638 ((_num) * sizeof(u16)))
640 #define RPS_NO_CPU 0xffff
642 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
646 unsigned int cpu, index = hash & table->mask;
648 /* We only give a hint, preemption can change cpu under us */
649 cpu = raw_smp_processor_id();
651 if (table->ents[index] != cpu)
652 table->ents[index] = cpu;
656 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
660 table->ents[hash & table->mask] = RPS_NO_CPU;
663 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665 #ifdef CONFIG_RFS_ACCEL
666 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
669 #endif /* CONFIG_RPS */
671 /* This structure contains an instance of an RX queue. */
672 struct netdev_rx_queue {
674 struct rps_map __rcu *rps_map;
675 struct rps_dev_flow_table __rcu *rps_flow_table;
678 struct net_device *dev;
679 } ____cacheline_aligned_in_smp;
682 * RX queue sysfs structures and functions.
684 struct rx_queue_attribute {
685 struct attribute attr;
686 ssize_t (*show)(struct netdev_rx_queue *queue,
687 struct rx_queue_attribute *attr, char *buf);
688 ssize_t (*store)(struct netdev_rx_queue *queue,
689 struct rx_queue_attribute *attr, const char *buf, size_t len);
694 * This structure holds an XPS map which can be of variable length. The
695 * map is an array of queues.
699 unsigned int alloc_len;
703 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
704 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
708 * This structure holds all XPS maps for device. Maps are indexed by CPU.
710 struct xps_dev_maps {
712 struct xps_map __rcu *cpu_map[0];
714 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
715 (nr_cpu_ids * sizeof(struct xps_map *)))
716 #endif /* CONFIG_XPS */
718 #define TC_MAX_QUEUE 16
719 #define TC_BITMASK 15
720 /* HW offloaded queuing disciplines txq count and offset maps */
721 struct netdev_tc_txq {
726 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
728 * This structure is to hold information about the device
729 * configured to run FCoE protocol stack.
731 struct netdev_fcoe_hbainfo {
732 char manufacturer[64];
733 char serial_number[64];
734 char hardware_version[64];
735 char driver_version[64];
736 char optionrom_version[64];
737 char firmware_version[64];
739 char model_description[256];
743 #define MAX_PHYS_PORT_ID_LEN 32
745 /* This structure holds a unique identifier to identify the
746 * physical port used by a netdevice.
748 struct netdev_phys_port_id {
749 unsigned char id[MAX_PHYS_PORT_ID_LEN];
750 unsigned char id_len;
753 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
754 struct sk_buff *skb);
757 * This structure defines the management hooks for network devices.
758 * The following hooks can be defined; unless noted otherwise, they are
759 * optional and can be filled with a null pointer.
761 * int (*ndo_init)(struct net_device *dev);
762 * This function is called once when network device is registered.
763 * The network device can use this to any late stage initializaton
764 * or semantic validattion. It can fail with an error code which will
765 * be propogated back to register_netdev
767 * void (*ndo_uninit)(struct net_device *dev);
768 * This function is called when device is unregistered or when registration
769 * fails. It is not called if init fails.
771 * int (*ndo_open)(struct net_device *dev);
772 * This function is called when network device transistions to the up
775 * int (*ndo_stop)(struct net_device *dev);
776 * This function is called when network device transistions to the down
779 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
780 * struct net_device *dev);
781 * Called when a packet needs to be transmitted.
782 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
783 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
784 * Required can not be NULL.
786 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
787 * void *accel_priv, select_queue_fallback_t fallback);
788 * Called to decide which queue to when device supports multiple
791 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
792 * This function is called to allow device receiver to make
793 * changes to configuration when multicast or promiscious is enabled.
795 * void (*ndo_set_rx_mode)(struct net_device *dev);
796 * This function is called device changes address list filtering.
797 * If driver handles unicast address filtering, it should set
798 * IFF_UNICAST_FLT to its priv_flags.
800 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
801 * This function is called when the Media Access Control address
802 * needs to be changed. If this interface is not defined, the
803 * mac address can not be changed.
805 * int (*ndo_validate_addr)(struct net_device *dev);
806 * Test if Media Access Control address is valid for the device.
808 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
809 * Called when a user request an ioctl which can't be handled by
810 * the generic interface code. If not defined ioctl's return
811 * not supported error code.
813 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
814 * Used to set network devices bus interface parameters. This interface
815 * is retained for legacy reason, new devices should use the bus
816 * interface (PCI) for low level management.
818 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
819 * Called when a user wants to change the Maximum Transfer Unit
820 * of a device. If not defined, any request to change MTU will
821 * will return an error.
823 * void (*ndo_tx_timeout)(struct net_device *dev);
824 * Callback uses when the transmitter has not made any progress
825 * for dev->watchdog ticks.
827 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
828 * struct rtnl_link_stats64 *storage);
829 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
830 * Called when a user wants to get the network device usage
831 * statistics. Drivers must do one of the following:
832 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
833 * rtnl_link_stats64 structure passed by the caller.
834 * 2. Define @ndo_get_stats to update a net_device_stats structure
835 * (which should normally be dev->stats) and return a pointer to
836 * it. The structure may be changed asynchronously only if each
837 * field is written atomically.
838 * 3. Update dev->stats asynchronously and atomically, and define
841 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
842 * If device support VLAN filtering this function is called when a
843 * VLAN id is registered.
845 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
846 * If device support VLAN filtering this function is called when a
847 * VLAN id is unregistered.
849 * void (*ndo_poll_controller)(struct net_device *dev);
851 * SR-IOV management functions.
852 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
853 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
854 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
856 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
857 * int (*ndo_get_vf_config)(struct net_device *dev,
858 * int vf, struct ifla_vf_info *ivf);
859 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
860 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
861 * struct nlattr *port[]);
862 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
863 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
864 * Called to setup 'tc' number of traffic classes in the net device. This
865 * is always called from the stack with the rtnl lock held and netif tx
866 * queues stopped. This allows the netdevice to perform queue management
869 * Fiber Channel over Ethernet (FCoE) offload functions.
870 * int (*ndo_fcoe_enable)(struct net_device *dev);
871 * Called when the FCoE protocol stack wants to start using LLD for FCoE
872 * so the underlying device can perform whatever needed configuration or
873 * initialization to support acceleration of FCoE traffic.
875 * int (*ndo_fcoe_disable)(struct net_device *dev);
876 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
877 * so the underlying device can perform whatever needed clean-ups to
878 * stop supporting acceleration of FCoE traffic.
880 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
881 * struct scatterlist *sgl, unsigned int sgc);
882 * Called when the FCoE Initiator wants to initialize an I/O that
883 * is a possible candidate for Direct Data Placement (DDP). The LLD can
884 * perform necessary setup and returns 1 to indicate the device is set up
885 * successfully to perform DDP on this I/O, otherwise this returns 0.
887 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
888 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
889 * indicated by the FC exchange id 'xid', so the underlying device can
890 * clean up and reuse resources for later DDP requests.
892 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
893 * struct scatterlist *sgl, unsigned int sgc);
894 * Called when the FCoE Target wants to initialize an I/O that
895 * is a possible candidate for Direct Data Placement (DDP). The LLD can
896 * perform necessary setup and returns 1 to indicate the device is set up
897 * successfully to perform DDP on this I/O, otherwise this returns 0.
899 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
900 * struct netdev_fcoe_hbainfo *hbainfo);
901 * Called when the FCoE Protocol stack wants information on the underlying
902 * device. This information is utilized by the FCoE protocol stack to
903 * register attributes with Fiber Channel management service as per the
904 * FC-GS Fabric Device Management Information(FDMI) specification.
906 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
907 * Called when the underlying device wants to override default World Wide
908 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
909 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
910 * protocol stack to use.
913 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
914 * u16 rxq_index, u32 flow_id);
915 * Set hardware filter for RFS. rxq_index is the target queue index;
916 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
917 * Return the filter ID on success, or a negative error code.
919 * Slave management functions (for bridge, bonding, etc).
920 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
921 * Called to make another netdev an underling.
923 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
924 * Called to release previously enslaved netdev.
926 * Feature/offload setting functions.
927 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
928 * netdev_features_t features);
929 * Adjusts the requested feature flags according to device-specific
930 * constraints, and returns the resulting flags. Must not modify
933 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
934 * Called to update device configuration to new features. Passed
935 * feature set might be less than what was returned by ndo_fix_features()).
936 * Must return >0 or -errno if it changed dev->features itself.
938 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
939 * struct net_device *dev,
940 * const unsigned char *addr, u16 flags)
941 * Adds an FDB entry to dev for addr.
942 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
943 * struct net_device *dev,
944 * const unsigned char *addr)
945 * Deletes the FDB entry from dev coresponding to addr.
946 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
947 * struct net_device *dev, struct net_device *filter_dev,
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_rate)(struct net_device *dev,
1051 int vf, int min_tx_rate,
1053 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1054 int vf, bool setting);
1055 int (*ndo_get_vf_config)(struct net_device *dev,
1057 struct ifla_vf_info *ivf);
1058 int (*ndo_set_vf_link_state)(struct net_device *dev,
1059 int vf, int link_state);
1060 int (*ndo_set_vf_port)(struct net_device *dev,
1062 struct nlattr *port[]);
1063 int (*ndo_get_vf_port)(struct net_device *dev,
1064 int vf, struct sk_buff *skb);
1065 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1066 #if IS_ENABLED(CONFIG_FCOE)
1067 int (*ndo_fcoe_enable)(struct net_device *dev);
1068 int (*ndo_fcoe_disable)(struct net_device *dev);
1069 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1071 struct scatterlist *sgl,
1073 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1075 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1077 struct scatterlist *sgl,
1079 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1080 struct netdev_fcoe_hbainfo *hbainfo);
1083 #if IS_ENABLED(CONFIG_LIBFCOE)
1084 #define NETDEV_FCOE_WWNN 0
1085 #define NETDEV_FCOE_WWPN 1
1086 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1087 u64 *wwn, int type);
1090 #ifdef CONFIG_RFS_ACCEL
1091 int (*ndo_rx_flow_steer)(struct net_device *dev,
1092 const struct sk_buff *skb,
1096 int (*ndo_add_slave)(struct net_device *dev,
1097 struct net_device *slave_dev);
1098 int (*ndo_del_slave)(struct net_device *dev,
1099 struct net_device *slave_dev);
1100 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1101 netdev_features_t features);
1102 int (*ndo_set_features)(struct net_device *dev,
1103 netdev_features_t features);
1104 int (*ndo_neigh_construct)(struct neighbour *n);
1105 void (*ndo_neigh_destroy)(struct neighbour *n);
1107 int (*ndo_fdb_add)(struct ndmsg *ndm,
1108 struct nlattr *tb[],
1109 struct net_device *dev,
1110 const unsigned char *addr,
1112 int (*ndo_fdb_del)(struct ndmsg *ndm,
1113 struct nlattr *tb[],
1114 struct net_device *dev,
1115 const unsigned char *addr);
1116 int (*ndo_fdb_dump)(struct sk_buff *skb,
1117 struct netlink_callback *cb,
1118 struct net_device *dev,
1119 struct net_device *filter_dev,
1122 int (*ndo_bridge_setlink)(struct net_device *dev,
1123 struct nlmsghdr *nlh);
1124 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1126 struct net_device *dev,
1128 int (*ndo_bridge_dellink)(struct net_device *dev,
1129 struct nlmsghdr *nlh);
1130 int (*ndo_change_carrier)(struct net_device *dev,
1132 int (*ndo_get_phys_port_id)(struct net_device *dev,
1133 struct netdev_phys_port_id *ppid);
1134 void (*ndo_add_vxlan_port)(struct net_device *dev,
1135 sa_family_t sa_family,
1137 void (*ndo_del_vxlan_port)(struct net_device *dev,
1138 sa_family_t sa_family,
1141 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1142 struct net_device *dev);
1143 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1146 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1147 struct net_device *dev,
1149 int (*ndo_get_lock_subclass)(struct net_device *dev);
1153 * enum net_device_priv_flags - &struct net_device priv_flags
1155 * These are the &struct net_device, they are only set internally
1156 * by drivers and used in the kernel. These flags are invisible to
1157 * userspace, this means that the order of these flags can change
1158 * during any kernel release.
1160 * You should have a pretty good reason to be extending these flags.
1162 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1163 * @IFF_EBRIDGE: Ethernet bridging device
1164 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1165 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1166 * @IFF_MASTER_ALB: bonding master, balance-alb
1167 * @IFF_BONDING: bonding master or slave
1168 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1169 * @IFF_ISATAP: ISATAP interface (RFC4214)
1170 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1171 * @IFF_WAN_HDLC: WAN HDLC device
1172 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1174 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1175 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1176 * @IFF_MACVLAN_PORT: device used as macvlan port
1177 * @IFF_BRIDGE_PORT: device used as bridge port
1178 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1179 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1180 * @IFF_UNICAST_FLT: Supports unicast filtering
1181 * @IFF_TEAM_PORT: device used as team port
1182 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1183 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1184 * change when it's running
1185 * @IFF_MACVLAN: Macvlan device
1187 enum netdev_priv_flags {
1188 IFF_802_1Q_VLAN = 1<<0,
1190 IFF_SLAVE_INACTIVE = 1<<2,
1191 IFF_MASTER_8023AD = 1<<3,
1192 IFF_MASTER_ALB = 1<<4,
1194 IFF_SLAVE_NEEDARP = 1<<6,
1196 IFF_MASTER_ARPMON = 1<<8,
1197 IFF_WAN_HDLC = 1<<9,
1198 IFF_XMIT_DST_RELEASE = 1<<10,
1199 IFF_DONT_BRIDGE = 1<<11,
1200 IFF_DISABLE_NETPOLL = 1<<12,
1201 IFF_MACVLAN_PORT = 1<<13,
1202 IFF_BRIDGE_PORT = 1<<14,
1203 IFF_OVS_DATAPATH = 1<<15,
1204 IFF_TX_SKB_SHARING = 1<<16,
1205 IFF_UNICAST_FLT = 1<<17,
1206 IFF_TEAM_PORT = 1<<18,
1207 IFF_SUPP_NOFCS = 1<<19,
1208 IFF_LIVE_ADDR_CHANGE = 1<<20,
1209 IFF_MACVLAN = 1<<21,
1210 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1213 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1214 #define IFF_EBRIDGE IFF_EBRIDGE
1215 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1216 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1217 #define IFF_MASTER_ALB IFF_MASTER_ALB
1218 #define IFF_BONDING IFF_BONDING
1219 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1220 #define IFF_ISATAP IFF_ISATAP
1221 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1222 #define IFF_WAN_HDLC IFF_WAN_HDLC
1223 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1224 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1225 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1226 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1227 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1228 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1229 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1230 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1231 #define IFF_TEAM_PORT IFF_TEAM_PORT
1232 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1233 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1234 #define IFF_MACVLAN IFF_MACVLAN
1235 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1238 * struct net_device - The DEVICE structure.
1239 * Actually, this whole structure is a big mistake. It mixes I/O
1240 * data with strictly "high-level" data, and it has to know about
1241 * almost every data structure used in the INET module.
1243 * @name: This is the first field of the "visible" part of this structure
1244 * (i.e. as seen by users in the "Space.c" file). It is the name
1247 * @name_hlist: Device name hash chain, please keep it close to name[]
1248 * @ifalias: SNMP alias
1249 * @mem_end: Shared memory end
1250 * @mem_start: Shared memory start
1251 * @base_addr: Device I/O address
1252 * @irq: Device IRQ number
1254 * @state: Generic network queuing layer state, see netdev_state_t
1255 * @dev_list: The global list of network devices
1256 * @napi_list: List entry, that is used for polling napi devices
1257 * @unreg_list: List entry, that is used, when we are unregistering the
1258 * device, see the function unregister_netdev
1259 * @close_list: List entry, that is used, when we are closing the device
1261 * @adj_list: Directly linked devices, like slaves for bonding
1262 * @all_adj_list: All linked devices, *including* neighbours
1263 * @features: Currently active device features
1264 * @hw_features: User-changeable features
1266 * @wanted_features: User-requested features
1267 * @vlan_features: Mask of features inheritable by VLAN devices
1269 * @hw_enc_features: Mask of features inherited by encapsulating devices
1270 * This field indicates what encapsulation
1271 * offloads the hardware is capable of doing,
1272 * and drivers will need to set them appropriately.
1274 * @mpls_features: Mask of features inheritable by MPLS
1276 * @ifindex: interface index
1277 * @iflink: unique device identifier
1279 * @stats: Statistics struct, which was left as a legacy, use
1280 * rtnl_link_stats64 instead
1282 * @rx_dropped: Dropped packets by core network,
1283 * do not use this in drivers
1284 * @tx_dropped: Dropped packets by core network,
1285 * do not use this in drivers
1287 * @carrier_changes: Stats to monitor carrier on<->off transitions
1289 * @wireless_handlers: List of functions to handle Wireless Extensions,
1291 * see <net/iw_handler.h> for details.
1292 * @wireless_data: Instance data managed by the core of wireless extensions
1294 * @netdev_ops: Includes several pointers to callbacks,
1295 * if one wants to override the ndo_*() functions
1296 * @ethtool_ops: Management operations
1297 * @fwd_ops: Management operations
1298 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1299 * of Layer 2 headers.
1301 * @flags: Interface flags (a la BSD)
1302 * @priv_flags: Like 'flags' but invisible to userspace,
1303 * see if.h for the definitions
1304 * @gflags: Global flags ( kept as legacy )
1305 * @padded: How much padding added by alloc_netdev()
1306 * @operstate: RFC2863 operstate
1307 * @link_mode: Mapping policy to operstate
1308 * @if_port: Selectable AUI, TP, ...
1310 * @mtu: Interface MTU value
1311 * @type: Interface hardware type
1312 * @hard_header_len: Hardware header length
1314 * @needed_headroom: Extra headroom the hardware may need, but not in all
1315 * cases can this be guaranteed
1316 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1317 * cases can this be guaranteed. Some cases also use
1318 * LL_MAX_HEADER instead to allocate the skb
1320 * interface address info:
1322 * @perm_addr: Permanent hw address
1323 * @addr_assign_type: Hw address assignment type
1324 * @addr_len: Hardware address length
1325 * @neigh_priv_len; Used in neigh_alloc(),
1326 * initialized only in atm/clip.c
1327 * @dev_id: Used to differentiate devices that share
1328 * the same link layer address
1329 * @dev_port: Used to differentiate devices that share
1331 * @addr_list_lock: XXX: need comments on this one
1332 * @uc: unicast mac addresses
1333 * @mc: multicast mac addresses
1334 * @dev_addrs: list of device hw addresses
1335 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1336 * @uc_promisc: Counter, that indicates, that promiscuous mode
1337 * has been enabled due to the need to listen to
1338 * additional unicast addresses in a device that
1339 * does not implement ndo_set_rx_mode()
1340 * @promiscuity: Number of times, the NIC is told to work in
1341 * Promiscuous mode, if it becomes 0 the NIC will
1342 * exit from working in Promiscuous mode
1343 * @allmulti: Counter, enables or disables allmulticast mode
1345 * @vlan_info: VLAN info
1346 * @dsa_ptr: dsa specific data
1347 * @tipc_ptr: TIPC specific data
1348 * @atalk_ptr: AppleTalk link
1349 * @ip_ptr: IPv4 specific data
1350 * @dn_ptr: DECnet specific data
1351 * @ip6_ptr: IPv6 specific data
1352 * @ax25_ptr: AX.25 specific data
1353 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1355 * @last_rx: Time of last Rx
1356 * @dev_addr: Hw address (before bcast,
1357 * because most packets are unicast)
1359 * @_rx: Array of RX queues
1360 * @num_rx_queues: Number of RX queues
1361 * allocated at register_netdev() time
1362 * @real_num_rx_queues: Number of RX queues currently active in device
1364 * @rx_handler: handler for received packets
1365 * @rx_handler_data: XXX: need comments on this one
1366 * @ingress_queue: XXX: need comments on this one
1367 * @broadcast: hw bcast address
1369 * @_tx: Array of TX queues
1370 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1371 * @real_num_tx_queues: Number of TX queues currently active in device
1372 * @qdisc: Root qdisc from userspace point of view
1373 * @tx_queue_len: Max frames per queue allowed
1374 * @tx_global_lock: XXX: need comments on this one
1376 * @xps_maps: XXX: need comments on this one
1378 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1379 * indexed by RX queue number. Assigned by driver.
1380 * This must only be set if the ndo_rx_flow_steer
1381 * operation is defined
1383 * @trans_start: Time (in jiffies) of last Tx
1384 * @watchdog_timeo: Represents the timeout that is used by
1385 * the watchdog ( see dev_watchdog() )
1386 * @watchdog_timer: List of timers
1388 * @pcpu_refcnt: Number of references to this device
1389 * @todo_list: Delayed register/unregister
1390 * @index_hlist: Device index hash chain
1391 * @link_watch_list: XXX: need comments on this one
1393 * @reg_state: Register/unregister state machine
1394 * @dismantle: Device is going to be freed
1395 * @rtnl_link_state: This enum represents the phases of creating
1398 * @destructor: Called from unregister,
1399 * can be used to call free_netdev
1400 * @npinfo: XXX: need comments on this one
1401 * @nd_net: Network namespace this network device is inside
1403 * @ml_priv: Mid-layer private
1404 * @lstats: Loopback statistics
1405 * @tstats: Tunnel statistics
1406 * @dstats: Dummy statistics
1407 * @vstats: Virtual ethernet statistics
1412 * @dev: Class/net/name entry
1413 * @sysfs_groups: Space for optional device, statistics and wireless
1416 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1417 * @rtnl_link_ops: Rtnl_link_ops
1419 * @gso_max_size: Maximum size of generic segmentation offload
1420 * @gso_max_segs: Maximum number of segments that can be passed to the
1422 * @gso_min_segs: Minimum number of segments that can be passed to the
1425 * @dcbnl_ops: Data Center Bridging netlink ops
1426 * @num_tc: Number of traffic classes in the net device
1427 * @tc_to_txq: XXX: need comments on this one
1428 * @prio_tc_map XXX: need comments on this one
1430 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1432 * @priomap: XXX: need comments on this one
1433 * @phydev: Physical device may attach itself
1434 * for hardware timestamping
1436 * @qdisc_tx_busylock: XXX: need comments on this one
1438 * @group: The group, that the device belongs to
1439 * @pm_qos_req: Power Management QoS object
1441 * FIXME: cleanup struct net_device such that network protocol info
1446 char name[IFNAMSIZ];
1447 struct hlist_node name_hlist;
1450 * I/O specific fields
1451 * FIXME: Merge these and struct ifmap into one
1453 unsigned long mem_end;
1454 unsigned long mem_start;
1455 unsigned long base_addr;
1459 * Some hardware also needs these fields (state,dev_list,
1460 * napi_list,unreg_list,close_list) but they are not
1461 * part of the usual set specified in Space.c.
1464 unsigned long state;
1466 struct list_head dev_list;
1467 struct list_head napi_list;
1468 struct list_head unreg_list;
1469 struct list_head close_list;
1472 struct list_head upper;
1473 struct list_head lower;
1477 struct list_head upper;
1478 struct list_head lower;
1481 netdev_features_t features;
1482 netdev_features_t hw_features;
1483 netdev_features_t wanted_features;
1484 netdev_features_t vlan_features;
1485 netdev_features_t hw_enc_features;
1486 netdev_features_t mpls_features;
1491 struct net_device_stats stats;
1493 atomic_long_t rx_dropped;
1494 atomic_long_t tx_dropped;
1496 atomic_t carrier_changes;
1498 #ifdef CONFIG_WIRELESS_EXT
1499 const struct iw_handler_def * wireless_handlers;
1500 struct iw_public_data * wireless_data;
1502 const struct net_device_ops *netdev_ops;
1503 const struct ethtool_ops *ethtool_ops;
1504 const struct forwarding_accel_ops *fwd_ops;
1506 const struct header_ops *header_ops;
1509 unsigned int priv_flags;
1511 unsigned short gflags;
1512 unsigned short padded;
1514 unsigned char operstate;
1515 unsigned char link_mode;
1517 unsigned char if_port;
1521 unsigned short type;
1522 unsigned short hard_header_len;
1524 unsigned short needed_headroom;
1525 unsigned short needed_tailroom;
1527 /* Interface address info. */
1528 unsigned char perm_addr[MAX_ADDR_LEN];
1529 unsigned char addr_assign_type;
1530 unsigned char addr_len;
1531 unsigned short neigh_priv_len;
1532 unsigned short dev_id;
1533 unsigned short dev_port;
1534 spinlock_t addr_list_lock;
1535 struct netdev_hw_addr_list uc;
1536 struct netdev_hw_addr_list mc;
1537 struct netdev_hw_addr_list dev_addrs;
1540 struct kset *queues_kset;
1543 unsigned char name_assign_type;
1546 unsigned int promiscuity;
1547 unsigned int allmulti;
1550 /* Protocol specific pointers */
1552 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1553 struct vlan_info __rcu *vlan_info;
1555 #if IS_ENABLED(CONFIG_NET_DSA)
1556 struct dsa_switch_tree *dsa_ptr;
1558 #if IS_ENABLED(CONFIG_TIPC)
1559 struct tipc_bearer __rcu *tipc_ptr;
1562 struct in_device __rcu *ip_ptr;
1563 struct dn_dev __rcu *dn_ptr;
1564 struct inet6_dev __rcu *ip6_ptr;
1566 struct wireless_dev *ieee80211_ptr;
1569 * Cache lines mostly used on receive path (including eth_type_trans())
1571 unsigned long last_rx;
1573 /* Interface address info used in eth_type_trans() */
1574 unsigned char *dev_addr;
1578 struct netdev_rx_queue *_rx;
1580 unsigned int num_rx_queues;
1581 unsigned int real_num_rx_queues;
1585 rx_handler_func_t __rcu *rx_handler;
1586 void __rcu *rx_handler_data;
1588 struct netdev_queue __rcu *ingress_queue;
1589 unsigned char broadcast[MAX_ADDR_LEN];
1593 * Cache lines mostly used on transmit path
1595 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1596 unsigned int num_tx_queues;
1597 unsigned int real_num_tx_queues;
1598 struct Qdisc *qdisc;
1599 unsigned long tx_queue_len;
1600 spinlock_t tx_global_lock;
1603 struct xps_dev_maps __rcu *xps_maps;
1605 #ifdef CONFIG_RFS_ACCEL
1606 struct cpu_rmap *rx_cpu_rmap;
1609 /* These may be needed for future network-power-down code. */
1612 * trans_start here is expensive for high speed devices on SMP,
1613 * please use netdev_queue->trans_start instead.
1615 unsigned long trans_start;
1618 struct timer_list watchdog_timer;
1620 int __percpu *pcpu_refcnt;
1621 struct list_head todo_list;
1623 struct hlist_node index_hlist;
1624 struct list_head link_watch_list;
1626 enum { NETREG_UNINITIALIZED=0,
1627 NETREG_REGISTERED, /* completed register_netdevice */
1628 NETREG_UNREGISTERING, /* called unregister_netdevice */
1629 NETREG_UNREGISTERED, /* completed unregister todo */
1630 NETREG_RELEASED, /* called free_netdev */
1631 NETREG_DUMMY, /* dummy device for NAPI poll */
1637 RTNL_LINK_INITIALIZED,
1638 RTNL_LINK_INITIALIZING,
1639 } rtnl_link_state:16;
1641 void (*destructor)(struct net_device *dev);
1643 #ifdef CONFIG_NETPOLL
1644 struct netpoll_info __rcu *npinfo;
1647 #ifdef CONFIG_NET_NS
1651 /* mid-layer private */
1654 struct pcpu_lstats __percpu *lstats;
1655 struct pcpu_sw_netstats __percpu *tstats;
1656 struct pcpu_dstats __percpu *dstats;
1657 struct pcpu_vstats __percpu *vstats;
1660 struct garp_port __rcu *garp_port;
1661 struct mrp_port __rcu *mrp_port;
1664 const struct attribute_group *sysfs_groups[4];
1665 const struct attribute_group *sysfs_rx_queue_group;
1667 const struct rtnl_link_ops *rtnl_link_ops;
1669 /* for setting kernel sock attribute on TCP connection setup */
1670 #define GSO_MAX_SIZE 65536
1671 unsigned int gso_max_size;
1672 #define GSO_MAX_SEGS 65535
1676 const struct dcbnl_rtnl_ops *dcbnl_ops;
1679 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1680 u8 prio_tc_map[TC_BITMASK + 1];
1682 #if IS_ENABLED(CONFIG_FCOE)
1683 unsigned int fcoe_ddp_xid;
1685 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1686 struct netprio_map __rcu *priomap;
1688 struct phy_device *phydev;
1689 struct lock_class_key *qdisc_tx_busylock;
1691 struct pm_qos_request pm_qos_req;
1693 #define to_net_dev(d) container_of(d, struct net_device, dev)
1695 #define NETDEV_ALIGN 32
1698 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1700 return dev->prio_tc_map[prio & TC_BITMASK];
1704 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1706 if (tc >= dev->num_tc)
1709 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1714 void netdev_reset_tc(struct net_device *dev)
1717 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1718 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1722 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1724 if (tc >= dev->num_tc)
1727 dev->tc_to_txq[tc].count = count;
1728 dev->tc_to_txq[tc].offset = offset;
1733 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1735 if (num_tc > TC_MAX_QUEUE)
1738 dev->num_tc = num_tc;
1743 int netdev_get_num_tc(struct net_device *dev)
1749 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1752 return &dev->_tx[index];
1755 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1756 const struct sk_buff *skb)
1758 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1761 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1762 void (*f)(struct net_device *,
1763 struct netdev_queue *,
1769 for (i = 0; i < dev->num_tx_queues; i++)
1770 f(dev, &dev->_tx[i], arg);
1773 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1774 struct sk_buff *skb,
1778 * Net namespace inlines
1781 struct net *dev_net(const struct net_device *dev)
1783 return read_pnet(&dev->nd_net);
1787 void dev_net_set(struct net_device *dev, struct net *net)
1789 #ifdef CONFIG_NET_NS
1790 release_net(dev->nd_net);
1791 dev->nd_net = hold_net(net);
1795 static inline bool netdev_uses_dsa(struct net_device *dev)
1797 #if IS_ENABLED(CONFIG_NET_DSA)
1798 if (dev->dsa_ptr != NULL)
1799 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1805 * netdev_priv - access network device private data
1806 * @dev: network device
1808 * Get network device private data
1810 static inline void *netdev_priv(const struct net_device *dev)
1812 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1815 /* Set the sysfs physical device reference for the network logical device
1816 * if set prior to registration will cause a symlink during initialization.
1818 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1820 /* Set the sysfs device type for the network logical device to allow
1821 * fine-grained identification of different network device types. For
1822 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1824 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1826 /* Default NAPI poll() weight
1827 * Device drivers are strongly advised to not use bigger value
1829 #define NAPI_POLL_WEIGHT 64
1832 * netif_napi_add - initialize a napi context
1833 * @dev: network device
1834 * @napi: napi context
1835 * @poll: polling function
1836 * @weight: default weight
1838 * netif_napi_add() must be used to initialize a napi context prior to calling
1839 * *any* of the other napi related functions.
1841 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1842 int (*poll)(struct napi_struct *, int), int weight);
1845 * netif_napi_del - remove a napi context
1846 * @napi: napi context
1848 * netif_napi_del() removes a napi context from the network device napi list
1850 void netif_napi_del(struct napi_struct *napi);
1852 struct napi_gro_cb {
1853 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1856 /* Length of frag0. */
1857 unsigned int frag0_len;
1859 /* This indicates where we are processing relative to skb->data. */
1862 /* This is non-zero if the packet cannot be merged with the new skb. */
1865 /* Save the IP ID here and check when we get to the transport layer */
1868 /* Number of segments aggregated. */
1871 /* This is non-zero if the packet may be of the same flow. */
1876 #define NAPI_GRO_FREE 1
1877 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1879 /* jiffies when first packet was created/queued */
1882 /* Used in ipv6_gro_receive() and foo-over-udp */
1885 /* Used in udp_gro_receive */
1888 /* GRO checksum is valid */
1891 /* Number of checksums via CHECKSUM_UNNECESSARY */
1894 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1897 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1900 /* used in skb_gro_receive() slow path */
1901 struct sk_buff *last;
1904 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1906 struct packet_type {
1907 __be16 type; /* This is really htons(ether_type). */
1908 struct net_device *dev; /* NULL is wildcarded here */
1909 int (*func) (struct sk_buff *,
1910 struct net_device *,
1911 struct packet_type *,
1912 struct net_device *);
1913 bool (*id_match)(struct packet_type *ptype,
1915 void *af_packet_priv;
1916 struct list_head list;
1919 struct offload_callbacks {
1920 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1921 netdev_features_t features);
1922 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1923 struct sk_buff *skb);
1924 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1927 struct packet_offload {
1928 __be16 type; /* This is really htons(ether_type). */
1929 struct offload_callbacks callbacks;
1930 struct list_head list;
1933 struct udp_offload {
1936 struct offload_callbacks callbacks;
1939 /* often modified stats are per cpu, other are shared (netdev->stats) */
1940 struct pcpu_sw_netstats {
1945 struct u64_stats_sync syncp;
1948 #define netdev_alloc_pcpu_stats(type) \
1950 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1953 for_each_possible_cpu(i) { \
1954 typeof(type) *stat; \
1955 stat = per_cpu_ptr(pcpu_stats, i); \
1956 u64_stats_init(&stat->syncp); \
1962 #include <linux/notifier.h>
1964 /* netdevice notifier chain. Please remember to update the rtnetlink
1965 * notification exclusion list in rtnetlink_event() when adding new
1968 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1969 #define NETDEV_DOWN 0x0002
1970 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1971 detected a hardware crash and restarted
1972 - we can use this eg to kick tcp sessions
1974 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1975 #define NETDEV_REGISTER 0x0005
1976 #define NETDEV_UNREGISTER 0x0006
1977 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1978 #define NETDEV_CHANGEADDR 0x0008
1979 #define NETDEV_GOING_DOWN 0x0009
1980 #define NETDEV_CHANGENAME 0x000A
1981 #define NETDEV_FEAT_CHANGE 0x000B
1982 #define NETDEV_BONDING_FAILOVER 0x000C
1983 #define NETDEV_PRE_UP 0x000D
1984 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1985 #define NETDEV_POST_TYPE_CHANGE 0x000F
1986 #define NETDEV_POST_INIT 0x0010
1987 #define NETDEV_UNREGISTER_FINAL 0x0011
1988 #define NETDEV_RELEASE 0x0012
1989 #define NETDEV_NOTIFY_PEERS 0x0013
1990 #define NETDEV_JOIN 0x0014
1991 #define NETDEV_CHANGEUPPER 0x0015
1992 #define NETDEV_RESEND_IGMP 0x0016
1993 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1994 #define NETDEV_CHANGEINFODATA 0x0018
1996 int register_netdevice_notifier(struct notifier_block *nb);
1997 int unregister_netdevice_notifier(struct notifier_block *nb);
1999 struct netdev_notifier_info {
2000 struct net_device *dev;
2003 struct netdev_notifier_change_info {
2004 struct netdev_notifier_info info; /* must be first */
2005 unsigned int flags_changed;
2008 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2009 struct net_device *dev)
2014 static inline struct net_device *
2015 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2020 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2023 extern rwlock_t dev_base_lock; /* Device list lock */
2025 #define for_each_netdev(net, d) \
2026 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2027 #define for_each_netdev_reverse(net, d) \
2028 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2029 #define for_each_netdev_rcu(net, d) \
2030 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2031 #define for_each_netdev_safe(net, d, n) \
2032 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2033 #define for_each_netdev_continue(net, d) \
2034 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2035 #define for_each_netdev_continue_rcu(net, d) \
2036 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2037 #define for_each_netdev_in_bond_rcu(bond, slave) \
2038 for_each_netdev_rcu(&init_net, slave) \
2039 if (netdev_master_upper_dev_get_rcu(slave) == bond)
2040 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2042 static inline struct net_device *next_net_device(struct net_device *dev)
2044 struct list_head *lh;
2048 lh = dev->dev_list.next;
2049 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2052 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2054 struct list_head *lh;
2058 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2059 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2062 static inline struct net_device *first_net_device(struct net *net)
2064 return list_empty(&net->dev_base_head) ? NULL :
2065 net_device_entry(net->dev_base_head.next);
2068 static inline struct net_device *first_net_device_rcu(struct net *net)
2070 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2072 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2075 int netdev_boot_setup_check(struct net_device *dev);
2076 unsigned long netdev_boot_base(const char *prefix, int unit);
2077 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2078 const char *hwaddr);
2079 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2080 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2081 void dev_add_pack(struct packet_type *pt);
2082 void dev_remove_pack(struct packet_type *pt);
2083 void __dev_remove_pack(struct packet_type *pt);
2084 void dev_add_offload(struct packet_offload *po);
2085 void dev_remove_offload(struct packet_offload *po);
2087 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2088 unsigned short mask);
2089 struct net_device *dev_get_by_name(struct net *net, const char *name);
2090 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2091 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2092 int dev_alloc_name(struct net_device *dev, const char *name);
2093 int dev_open(struct net_device *dev);
2094 int dev_close(struct net_device *dev);
2095 void dev_disable_lro(struct net_device *dev);
2096 int dev_loopback_xmit(struct sk_buff *newskb);
2097 int dev_queue_xmit(struct sk_buff *skb);
2098 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2099 int register_netdevice(struct net_device *dev);
2100 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2101 void unregister_netdevice_many(struct list_head *head);
2102 static inline void unregister_netdevice(struct net_device *dev)
2104 unregister_netdevice_queue(dev, NULL);
2107 int netdev_refcnt_read(const struct net_device *dev);
2108 void free_netdev(struct net_device *dev);
2109 void netdev_freemem(struct net_device *dev);
2110 void synchronize_net(void);
2111 int init_dummy_netdev(struct net_device *dev);
2113 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2114 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2115 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2116 int netdev_get_name(struct net *net, char *name, int ifindex);
2117 int dev_restart(struct net_device *dev);
2118 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2120 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2122 return NAPI_GRO_CB(skb)->data_offset;
2125 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2127 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2130 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2132 NAPI_GRO_CB(skb)->data_offset += len;
2135 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2136 unsigned int offset)
2138 return NAPI_GRO_CB(skb)->frag0 + offset;
2141 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2143 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2146 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2147 unsigned int offset)
2149 if (!pskb_may_pull(skb, hlen))
2152 NAPI_GRO_CB(skb)->frag0 = NULL;
2153 NAPI_GRO_CB(skb)->frag0_len = 0;
2154 return skb->data + offset;
2157 static inline void *skb_gro_network_header(struct sk_buff *skb)
2159 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2160 skb_network_offset(skb);
2163 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2164 const void *start, unsigned int len)
2166 if (NAPI_GRO_CB(skb)->csum_valid)
2167 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2168 csum_partial(start, len, 0));
2171 /* GRO checksum functions. These are logical equivalents of the normal
2172 * checksum functions (in skbuff.h) except that they operate on the GRO
2173 * offsets and fields in sk_buff.
2176 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2178 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2182 return (skb->ip_summed != CHECKSUM_PARTIAL &&
2183 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2184 (!zero_okay || check));
2187 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2190 if (NAPI_GRO_CB(skb)->csum_valid &&
2191 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2194 NAPI_GRO_CB(skb)->csum = psum;
2196 return __skb_gro_checksum_complete(skb);
2199 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2201 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2202 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2203 NAPI_GRO_CB(skb)->csum_cnt--;
2205 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2206 * verified a new top level checksum or an encapsulated one
2207 * during GRO. This saves work if we fallback to normal path.
2209 __skb_incr_checksum_unnecessary(skb);
2213 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2216 __sum16 __ret = 0; \
2217 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2218 __ret = __skb_gro_checksum_validate_complete(skb, \
2219 compute_pseudo(skb, proto)); \
2221 __skb_mark_checksum_bad(skb); \
2223 skb_gro_incr_csum_unnecessary(skb); \
2227 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2228 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2230 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2232 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2234 #define skb_gro_checksum_simple_validate(skb) \
2235 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2237 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2239 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2240 !NAPI_GRO_CB(skb)->csum_valid);
2243 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2244 __sum16 check, __wsum pseudo)
2246 NAPI_GRO_CB(skb)->csum = ~pseudo;
2247 NAPI_GRO_CB(skb)->csum_valid = 1;
2250 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2252 if (__skb_gro_checksum_convert_check(skb)) \
2253 __skb_gro_checksum_convert(skb, check, \
2254 compute_pseudo(skb, proto)); \
2257 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2258 unsigned short type,
2259 const void *daddr, const void *saddr,
2262 if (!dev->header_ops || !dev->header_ops->create)
2265 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2268 static inline int dev_parse_header(const struct sk_buff *skb,
2269 unsigned char *haddr)
2271 const struct net_device *dev = skb->dev;
2273 if (!dev->header_ops || !dev->header_ops->parse)
2275 return dev->header_ops->parse(skb, haddr);
2278 static inline int dev_rebuild_header(struct sk_buff *skb)
2280 const struct net_device *dev = skb->dev;
2282 if (!dev->header_ops || !dev->header_ops->rebuild)
2284 return dev->header_ops->rebuild(skb);
2287 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2288 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2289 static inline int unregister_gifconf(unsigned int family)
2291 return register_gifconf(family, NULL);
2294 #ifdef CONFIG_NET_FLOW_LIMIT
2295 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2296 struct sd_flow_limit {
2298 unsigned int num_buckets;
2299 unsigned int history_head;
2300 u16 history[FLOW_LIMIT_HISTORY];
2304 extern int netdev_flow_limit_table_len;
2305 #endif /* CONFIG_NET_FLOW_LIMIT */
2308 * Incoming packets are placed on per-cpu queues
2310 struct softnet_data {
2311 struct Qdisc *output_queue;
2312 struct Qdisc **output_queue_tailp;
2313 struct list_head poll_list;
2314 struct sk_buff *completion_queue;
2315 struct sk_buff_head process_queue;
2318 unsigned int processed;
2319 unsigned int time_squeeze;
2320 unsigned int cpu_collision;
2321 unsigned int received_rps;
2324 struct softnet_data *rps_ipi_list;
2326 /* Elements below can be accessed between CPUs for RPS */
2327 struct call_single_data csd ____cacheline_aligned_in_smp;
2328 struct softnet_data *rps_ipi_next;
2330 unsigned int input_queue_head;
2331 unsigned int input_queue_tail;
2333 unsigned int dropped;
2334 struct sk_buff_head input_pkt_queue;
2335 struct napi_struct backlog;
2337 #ifdef CONFIG_NET_FLOW_LIMIT
2338 struct sd_flow_limit __rcu *flow_limit;
2342 static inline void input_queue_head_incr(struct softnet_data *sd)
2345 sd->input_queue_head++;
2349 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2350 unsigned int *qtail)
2353 *qtail = ++sd->input_queue_tail;
2357 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2359 void __netif_schedule(struct Qdisc *q);
2360 void netif_schedule_queue(struct netdev_queue *txq);
2362 static inline void netif_tx_schedule_all(struct net_device *dev)
2366 for (i = 0; i < dev->num_tx_queues; i++)
2367 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2370 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2372 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2376 * netif_start_queue - allow transmit
2377 * @dev: network device
2379 * Allow upper layers to call the device hard_start_xmit routine.
2381 static inline void netif_start_queue(struct net_device *dev)
2383 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2386 static inline void netif_tx_start_all_queues(struct net_device *dev)
2390 for (i = 0; i < dev->num_tx_queues; i++) {
2391 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2392 netif_tx_start_queue(txq);
2396 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2399 * netif_wake_queue - restart transmit
2400 * @dev: network device
2402 * Allow upper layers to call the device hard_start_xmit routine.
2403 * Used for flow control when transmit resources are available.
2405 static inline void netif_wake_queue(struct net_device *dev)
2407 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2410 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2414 for (i = 0; i < dev->num_tx_queues; i++) {
2415 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2416 netif_tx_wake_queue(txq);
2420 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2422 if (WARN_ON(!dev_queue)) {
2423 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2426 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2430 * netif_stop_queue - stop transmitted packets
2431 * @dev: network device
2433 * Stop upper layers calling the device hard_start_xmit routine.
2434 * Used for flow control when transmit resources are unavailable.
2436 static inline void netif_stop_queue(struct net_device *dev)
2438 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2441 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2445 for (i = 0; i < dev->num_tx_queues; i++) {
2446 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2447 netif_tx_stop_queue(txq);
2451 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2453 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2457 * netif_queue_stopped - test if transmit queue is flowblocked
2458 * @dev: network device
2460 * Test if transmit queue on device is currently unable to send.
2462 static inline bool netif_queue_stopped(const struct net_device *dev)
2464 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2467 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2469 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2473 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2475 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2479 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2481 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2485 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2486 * @dev_queue: pointer to transmit queue
2488 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2489 * to give appropriate hint to the cpu.
2491 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2494 prefetchw(&dev_queue->dql.num_queued);
2499 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2500 * @dev_queue: pointer to transmit queue
2502 * BQL enabled drivers might use this helper in their TX completion path,
2503 * to give appropriate hint to the cpu.
2505 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2508 prefetchw(&dev_queue->dql.limit);
2512 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2516 dql_queued(&dev_queue->dql, bytes);
2518 if (likely(dql_avail(&dev_queue->dql) >= 0))
2521 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2524 * The XOFF flag must be set before checking the dql_avail below,
2525 * because in netdev_tx_completed_queue we update the dql_completed
2526 * before checking the XOFF flag.
2530 /* check again in case another CPU has just made room avail */
2531 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2532 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2537 * netdev_sent_queue - report the number of bytes queued to hardware
2538 * @dev: network device
2539 * @bytes: number of bytes queued to the hardware device queue
2541 * Report the number of bytes queued for sending/completion to the network
2542 * device hardware queue. @bytes should be a good approximation and should
2543 * exactly match netdev_completed_queue() @bytes
2545 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2547 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2550 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2551 unsigned int pkts, unsigned int bytes)
2554 if (unlikely(!bytes))
2557 dql_completed(&dev_queue->dql, bytes);
2560 * Without the memory barrier there is a small possiblity that
2561 * netdev_tx_sent_queue will miss the update and cause the queue to
2562 * be stopped forever
2566 if (dql_avail(&dev_queue->dql) < 0)
2569 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2570 netif_schedule_queue(dev_queue);
2575 * netdev_completed_queue - report bytes and packets completed by device
2576 * @dev: network device
2577 * @pkts: actual number of packets sent over the medium
2578 * @bytes: actual number of bytes sent over the medium
2580 * Report the number of bytes and packets transmitted by the network device
2581 * hardware queue over the physical medium, @bytes must exactly match the
2582 * @bytes amount passed to netdev_sent_queue()
2584 static inline void netdev_completed_queue(struct net_device *dev,
2585 unsigned int pkts, unsigned int bytes)
2587 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2590 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2593 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2599 * netdev_reset_queue - reset the packets and bytes count of a network device
2600 * @dev_queue: network device
2602 * Reset the bytes and packet count of a network device and clear the
2603 * software flow control OFF bit for this network device
2605 static inline void netdev_reset_queue(struct net_device *dev_queue)
2607 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2611 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2612 * @dev: network device
2613 * @queue_index: given tx queue index
2615 * Returns 0 if given tx queue index >= number of device tx queues,
2616 * otherwise returns the originally passed tx queue index.
2618 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2620 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2621 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2622 dev->name, queue_index,
2623 dev->real_num_tx_queues);
2631 * netif_running - test if up
2632 * @dev: network device
2634 * Test if the device has been brought up.
2636 static inline bool netif_running(const struct net_device *dev)
2638 return test_bit(__LINK_STATE_START, &dev->state);
2642 * Routines to manage the subqueues on a device. We only need start
2643 * stop, and a check if it's stopped. All other device management is
2644 * done at the overall netdevice level.
2645 * Also test the device if we're multiqueue.
2649 * netif_start_subqueue - allow sending packets on subqueue
2650 * @dev: network device
2651 * @queue_index: sub queue index
2653 * Start individual transmit queue of a device with multiple transmit queues.
2655 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2657 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2659 netif_tx_start_queue(txq);
2663 * netif_stop_subqueue - stop sending packets on subqueue
2664 * @dev: network device
2665 * @queue_index: sub queue index
2667 * Stop individual transmit queue of a device with multiple transmit queues.
2669 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2671 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2672 netif_tx_stop_queue(txq);
2676 * netif_subqueue_stopped - test status of subqueue
2677 * @dev: network device
2678 * @queue_index: sub queue index
2680 * Check individual transmit queue of a device with multiple transmit queues.
2682 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2685 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2687 return netif_tx_queue_stopped(txq);
2690 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2691 struct sk_buff *skb)
2693 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2696 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2699 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2702 static inline int netif_set_xps_queue(struct net_device *dev,
2703 const struct cpumask *mask,
2711 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2712 * as a distribution range limit for the returned value.
2714 static inline u16 skb_tx_hash(const struct net_device *dev,
2715 struct sk_buff *skb)
2717 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2721 * netif_is_multiqueue - test if device has multiple transmit queues
2722 * @dev: network device
2724 * Check if device has multiple transmit queues
2726 static inline bool netif_is_multiqueue(const struct net_device *dev)
2728 return dev->num_tx_queues > 1;
2731 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2734 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2736 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2743 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2744 const struct net_device *from_dev)
2748 err = netif_set_real_num_tx_queues(to_dev,
2749 from_dev->real_num_tx_queues);
2753 return netif_set_real_num_rx_queues(to_dev,
2754 from_dev->real_num_rx_queues);
2761 static inline unsigned int get_netdev_rx_queue_index(
2762 struct netdev_rx_queue *queue)
2764 struct net_device *dev = queue->dev;
2765 int index = queue - dev->_rx;
2767 BUG_ON(index >= dev->num_rx_queues);
2772 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2773 int netif_get_num_default_rss_queues(void);
2775 enum skb_free_reason {
2776 SKB_REASON_CONSUMED,
2780 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2781 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2784 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2785 * interrupt context or with hardware interrupts being disabled.
2786 * (in_irq() || irqs_disabled())
2788 * We provide four helpers that can be used in following contexts :
2790 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2791 * replacing kfree_skb(skb)
2793 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2794 * Typically used in place of consume_skb(skb) in TX completion path
2796 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2797 * replacing kfree_skb(skb)
2799 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2800 * and consumed a packet. Used in place of consume_skb(skb)
2802 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2804 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2807 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2809 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2812 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2814 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2817 static inline void dev_consume_skb_any(struct sk_buff *skb)
2819 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2822 int netif_rx(struct sk_buff *skb);
2823 int netif_rx_ni(struct sk_buff *skb);
2824 int netif_receive_skb(struct sk_buff *skb);
2825 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2826 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2827 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2828 gro_result_t napi_gro_frags(struct napi_struct *napi);
2829 struct packet_offload *gro_find_receive_by_type(__be16 type);
2830 struct packet_offload *gro_find_complete_by_type(__be16 type);
2832 static inline void napi_free_frags(struct napi_struct *napi)
2834 kfree_skb(napi->skb);
2838 int netdev_rx_handler_register(struct net_device *dev,
2839 rx_handler_func_t *rx_handler,
2840 void *rx_handler_data);
2841 void netdev_rx_handler_unregister(struct net_device *dev);
2843 bool dev_valid_name(const char *name);
2844 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2845 int dev_ethtool(struct net *net, struct ifreq *);
2846 unsigned int dev_get_flags(const struct net_device *);
2847 int __dev_change_flags(struct net_device *, unsigned int flags);
2848 int dev_change_flags(struct net_device *, unsigned int);
2849 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2850 unsigned int gchanges);
2851 int dev_change_name(struct net_device *, const char *);
2852 int dev_set_alias(struct net_device *, const char *, size_t);
2853 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2854 int dev_set_mtu(struct net_device *, int);
2855 void dev_set_group(struct net_device *, int);
2856 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2857 int dev_change_carrier(struct net_device *, bool new_carrier);
2858 int dev_get_phys_port_id(struct net_device *dev,
2859 struct netdev_phys_port_id *ppid);
2860 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2861 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2862 struct netdev_queue *txq, int *ret);
2863 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2864 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2865 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2867 extern int netdev_budget;
2869 /* Called by rtnetlink.c:rtnl_unlock() */
2870 void netdev_run_todo(void);
2873 * dev_put - release reference to device
2874 * @dev: network device
2876 * Release reference to device to allow it to be freed.
2878 static inline void dev_put(struct net_device *dev)
2880 this_cpu_dec(*dev->pcpu_refcnt);
2884 * dev_hold - get reference to device
2885 * @dev: network device
2887 * Hold reference to device to keep it from being freed.
2889 static inline void dev_hold(struct net_device *dev)
2891 this_cpu_inc(*dev->pcpu_refcnt);
2894 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2895 * and _off may be called from IRQ context, but it is caller
2896 * who is responsible for serialization of these calls.
2898 * The name carrier is inappropriate, these functions should really be
2899 * called netif_lowerlayer_*() because they represent the state of any
2900 * kind of lower layer not just hardware media.
2903 void linkwatch_init_dev(struct net_device *dev);
2904 void linkwatch_fire_event(struct net_device *dev);
2905 void linkwatch_forget_dev(struct net_device *dev);
2908 * netif_carrier_ok - test if carrier present
2909 * @dev: network device
2911 * Check if carrier is present on device
2913 static inline bool netif_carrier_ok(const struct net_device *dev)
2915 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2918 unsigned long dev_trans_start(struct net_device *dev);
2920 void __netdev_watchdog_up(struct net_device *dev);
2922 void netif_carrier_on(struct net_device *dev);
2924 void netif_carrier_off(struct net_device *dev);
2927 * netif_dormant_on - mark device as dormant.
2928 * @dev: network device
2930 * Mark device as dormant (as per RFC2863).
2932 * The dormant state indicates that the relevant interface is not
2933 * actually in a condition to pass packets (i.e., it is not 'up') but is
2934 * in a "pending" state, waiting for some external event. For "on-
2935 * demand" interfaces, this new state identifies the situation where the
2936 * interface is waiting for events to place it in the up state.
2939 static inline void netif_dormant_on(struct net_device *dev)
2941 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2942 linkwatch_fire_event(dev);
2946 * netif_dormant_off - set device as not dormant.
2947 * @dev: network device
2949 * Device is not in dormant state.
2951 static inline void netif_dormant_off(struct net_device *dev)
2953 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2954 linkwatch_fire_event(dev);
2958 * netif_dormant - test if carrier present
2959 * @dev: network device
2961 * Check if carrier is present on device
2963 static inline bool netif_dormant(const struct net_device *dev)
2965 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2970 * netif_oper_up - test if device is operational
2971 * @dev: network device
2973 * Check if carrier is operational
2975 static inline bool netif_oper_up(const struct net_device *dev)
2977 return (dev->operstate == IF_OPER_UP ||
2978 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2982 * netif_device_present - is device available or removed
2983 * @dev: network device
2985 * Check if device has not been removed from system.
2987 static inline bool netif_device_present(struct net_device *dev)
2989 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2992 void netif_device_detach(struct net_device *dev);
2994 void netif_device_attach(struct net_device *dev);
2997 * Network interface message level settings
3001 NETIF_MSG_DRV = 0x0001,
3002 NETIF_MSG_PROBE = 0x0002,
3003 NETIF_MSG_LINK = 0x0004,
3004 NETIF_MSG_TIMER = 0x0008,
3005 NETIF_MSG_IFDOWN = 0x0010,
3006 NETIF_MSG_IFUP = 0x0020,
3007 NETIF_MSG_RX_ERR = 0x0040,
3008 NETIF_MSG_TX_ERR = 0x0080,
3009 NETIF_MSG_TX_QUEUED = 0x0100,
3010 NETIF_MSG_INTR = 0x0200,
3011 NETIF_MSG_TX_DONE = 0x0400,
3012 NETIF_MSG_RX_STATUS = 0x0800,
3013 NETIF_MSG_PKTDATA = 0x1000,
3014 NETIF_MSG_HW = 0x2000,
3015 NETIF_MSG_WOL = 0x4000,
3018 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3019 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3020 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3021 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3022 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3023 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3024 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3025 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3026 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3027 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3028 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3029 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3030 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3031 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3032 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3034 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3037 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3038 return default_msg_enable_bits;
3039 if (debug_value == 0) /* no output */
3041 /* set low N bits */
3042 return (1 << debug_value) - 1;
3045 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3047 spin_lock(&txq->_xmit_lock);
3048 txq->xmit_lock_owner = cpu;
3051 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3053 spin_lock_bh(&txq->_xmit_lock);
3054 txq->xmit_lock_owner = smp_processor_id();
3057 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3059 bool ok = spin_trylock(&txq->_xmit_lock);
3061 txq->xmit_lock_owner = smp_processor_id();
3065 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3067 txq->xmit_lock_owner = -1;
3068 spin_unlock(&txq->_xmit_lock);
3071 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3073 txq->xmit_lock_owner = -1;
3074 spin_unlock_bh(&txq->_xmit_lock);
3077 static inline void txq_trans_update(struct netdev_queue *txq)
3079 if (txq->xmit_lock_owner != -1)
3080 txq->trans_start = jiffies;
3084 * netif_tx_lock - grab network device transmit lock
3085 * @dev: network device
3087 * Get network device transmit lock
3089 static inline void netif_tx_lock(struct net_device *dev)
3094 spin_lock(&dev->tx_global_lock);
3095 cpu = smp_processor_id();
3096 for (i = 0; i < dev->num_tx_queues; i++) {
3097 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3099 /* We are the only thread of execution doing a
3100 * freeze, but we have to grab the _xmit_lock in
3101 * order to synchronize with threads which are in
3102 * the ->hard_start_xmit() handler and already
3103 * checked the frozen bit.
3105 __netif_tx_lock(txq, cpu);
3106 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3107 __netif_tx_unlock(txq);
3111 static inline void netif_tx_lock_bh(struct net_device *dev)
3117 static inline void netif_tx_unlock(struct net_device *dev)
3121 for (i = 0; i < dev->num_tx_queues; i++) {
3122 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3124 /* No need to grab the _xmit_lock here. If the
3125 * queue is not stopped for another reason, we
3128 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3129 netif_schedule_queue(txq);
3131 spin_unlock(&dev->tx_global_lock);
3134 static inline void netif_tx_unlock_bh(struct net_device *dev)
3136 netif_tx_unlock(dev);
3140 #define HARD_TX_LOCK(dev, txq, cpu) { \
3141 if ((dev->features & NETIF_F_LLTX) == 0) { \
3142 __netif_tx_lock(txq, cpu); \
3146 #define HARD_TX_TRYLOCK(dev, txq) \
3147 (((dev->features & NETIF_F_LLTX) == 0) ? \
3148 __netif_tx_trylock(txq) : \
3151 #define HARD_TX_UNLOCK(dev, txq) { \
3152 if ((dev->features & NETIF_F_LLTX) == 0) { \
3153 __netif_tx_unlock(txq); \
3157 static inline void netif_tx_disable(struct net_device *dev)
3163 cpu = smp_processor_id();
3164 for (i = 0; i < dev->num_tx_queues; i++) {
3165 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3167 __netif_tx_lock(txq, cpu);
3168 netif_tx_stop_queue(txq);
3169 __netif_tx_unlock(txq);
3174 static inline void netif_addr_lock(struct net_device *dev)
3176 spin_lock(&dev->addr_list_lock);
3179 static inline void netif_addr_lock_nested(struct net_device *dev)
3181 int subclass = SINGLE_DEPTH_NESTING;
3183 if (dev->netdev_ops->ndo_get_lock_subclass)
3184 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3186 spin_lock_nested(&dev->addr_list_lock, subclass);
3189 static inline void netif_addr_lock_bh(struct net_device *dev)
3191 spin_lock_bh(&dev->addr_list_lock);
3194 static inline void netif_addr_unlock(struct net_device *dev)
3196 spin_unlock(&dev->addr_list_lock);
3199 static inline void netif_addr_unlock_bh(struct net_device *dev)
3201 spin_unlock_bh(&dev->addr_list_lock);
3205 * dev_addrs walker. Should be used only for read access. Call with
3206 * rcu_read_lock held.
3208 #define for_each_dev_addr(dev, ha) \
3209 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3211 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3213 void ether_setup(struct net_device *dev);
3215 /* Support for loadable net-drivers */
3216 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3217 unsigned char name_assign_type,
3218 void (*setup)(struct net_device *),
3219 unsigned int txqs, unsigned int rxqs);
3220 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3221 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3223 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3224 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3227 int register_netdev(struct net_device *dev);
3228 void unregister_netdev(struct net_device *dev);
3230 /* General hardware address lists handling functions */
3231 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3232 struct netdev_hw_addr_list *from_list, int addr_len);
3233 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3234 struct netdev_hw_addr_list *from_list, int addr_len);
3235 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3236 struct net_device *dev,
3237 int (*sync)(struct net_device *, const unsigned char *),
3238 int (*unsync)(struct net_device *,
3239 const unsigned char *));
3240 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3241 struct net_device *dev,
3242 int (*unsync)(struct net_device *,
3243 const unsigned char *));
3244 void __hw_addr_init(struct netdev_hw_addr_list *list);
3246 /* Functions used for device addresses handling */
3247 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3248 unsigned char addr_type);
3249 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3250 unsigned char addr_type);
3251 void dev_addr_flush(struct net_device *dev);
3252 int dev_addr_init(struct net_device *dev);
3254 /* Functions used for unicast addresses handling */
3255 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3256 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3257 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3258 int dev_uc_sync(struct net_device *to, struct net_device *from);
3259 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3260 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3261 void dev_uc_flush(struct net_device *dev);
3262 void dev_uc_init(struct net_device *dev);
3265 * __dev_uc_sync - Synchonize device's unicast list
3266 * @dev: device to sync
3267 * @sync: function to call if address should be added
3268 * @unsync: function to call if address should be removed
3270 * Add newly added addresses to the interface, and release
3271 * addresses that have been deleted.
3273 static inline int __dev_uc_sync(struct net_device *dev,
3274 int (*sync)(struct net_device *,
3275 const unsigned char *),
3276 int (*unsync)(struct net_device *,
3277 const unsigned char *))
3279 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3283 * __dev_uc_unsync - Remove synchronized addresses from device
3284 * @dev: device to sync
3285 * @unsync: function to call if address should be removed
3287 * Remove all addresses that were added to the device by dev_uc_sync().
3289 static inline void __dev_uc_unsync(struct net_device *dev,
3290 int (*unsync)(struct net_device *,
3291 const unsigned char *))
3293 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3296 /* Functions used for multicast addresses handling */
3297 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3298 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3299 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3300 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3301 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3302 int dev_mc_sync(struct net_device *to, struct net_device *from);
3303 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3304 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3305 void dev_mc_flush(struct net_device *dev);
3306 void dev_mc_init(struct net_device *dev);
3309 * __dev_mc_sync - Synchonize device's multicast list
3310 * @dev: device to sync
3311 * @sync: function to call if address should be added
3312 * @unsync: function to call if address should be removed
3314 * Add newly added addresses to the interface, and release
3315 * addresses that have been deleted.
3317 static inline int __dev_mc_sync(struct net_device *dev,
3318 int (*sync)(struct net_device *,
3319 const unsigned char *),
3320 int (*unsync)(struct net_device *,
3321 const unsigned char *))
3323 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3327 * __dev_mc_unsync - Remove synchronized addresses from device
3328 * @dev: device to sync
3329 * @unsync: function to call if address should be removed
3331 * Remove all addresses that were added to the device by dev_mc_sync().
3333 static inline void __dev_mc_unsync(struct net_device *dev,
3334 int (*unsync)(struct net_device *,
3335 const unsigned char *))
3337 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3340 /* Functions used for secondary unicast and multicast support */
3341 void dev_set_rx_mode(struct net_device *dev);
3342 void __dev_set_rx_mode(struct net_device *dev);
3343 int dev_set_promiscuity(struct net_device *dev, int inc);
3344 int dev_set_allmulti(struct net_device *dev, int inc);
3345 void netdev_state_change(struct net_device *dev);
3346 void netdev_notify_peers(struct net_device *dev);
3347 void netdev_features_change(struct net_device *dev);
3348 /* Load a device via the kmod */
3349 void dev_load(struct net *net, const char *name);
3350 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3351 struct rtnl_link_stats64 *storage);
3352 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3353 const struct net_device_stats *netdev_stats);
3355 extern int netdev_max_backlog;
3356 extern int netdev_tstamp_prequeue;
3357 extern int weight_p;
3358 extern int bpf_jit_enable;
3360 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3361 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3362 struct list_head **iter);
3363 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3364 struct list_head **iter);
3366 /* iterate through upper list, must be called under RCU read lock */
3367 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3368 for (iter = &(dev)->adj_list.upper, \
3369 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3371 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3373 /* iterate through upper list, must be called under RCU read lock */
3374 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3375 for (iter = &(dev)->all_adj_list.upper, \
3376 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3378 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3380 void *netdev_lower_get_next_private(struct net_device *dev,
3381 struct list_head **iter);
3382 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3383 struct list_head **iter);
3385 #define netdev_for_each_lower_private(dev, priv, iter) \
3386 for (iter = (dev)->adj_list.lower.next, \
3387 priv = netdev_lower_get_next_private(dev, &(iter)); \
3389 priv = netdev_lower_get_next_private(dev, &(iter)))
3391 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3392 for (iter = &(dev)->adj_list.lower, \
3393 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3395 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3397 void *netdev_lower_get_next(struct net_device *dev,
3398 struct list_head **iter);
3399 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3400 for (iter = &(dev)->adj_list.lower, \
3401 ldev = netdev_lower_get_next(dev, &(iter)); \
3403 ldev = netdev_lower_get_next(dev, &(iter)))
3405 void *netdev_adjacent_get_private(struct list_head *adj_list);
3406 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3407 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3408 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3409 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3410 int netdev_master_upper_dev_link(struct net_device *dev,
3411 struct net_device *upper_dev);
3412 int netdev_master_upper_dev_link_private(struct net_device *dev,
3413 struct net_device *upper_dev,
3415 void netdev_upper_dev_unlink(struct net_device *dev,
3416 struct net_device *upper_dev);
3417 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3418 void *netdev_lower_dev_get_private(struct net_device *dev,
3419 struct net_device *lower_dev);
3420 int dev_get_nest_level(struct net_device *dev,
3421 bool (*type_check)(struct net_device *dev));
3422 int skb_checksum_help(struct sk_buff *skb);
3423 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3424 netdev_features_t features, bool tx_path);
3425 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3426 netdev_features_t features);
3429 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3431 return __skb_gso_segment(skb, features, true);
3433 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3435 static inline bool can_checksum_protocol(netdev_features_t features,
3438 return ((features & NETIF_F_GEN_CSUM) ||
3439 ((features & NETIF_F_V4_CSUM) &&
3440 protocol == htons(ETH_P_IP)) ||
3441 ((features & NETIF_F_V6_CSUM) &&
3442 protocol == htons(ETH_P_IPV6)) ||
3443 ((features & NETIF_F_FCOE_CRC) &&
3444 protocol == htons(ETH_P_FCOE)));
3448 void netdev_rx_csum_fault(struct net_device *dev);
3450 static inline void netdev_rx_csum_fault(struct net_device *dev)
3454 /* rx skb timestamps */
3455 void net_enable_timestamp(void);
3456 void net_disable_timestamp(void);
3458 #ifdef CONFIG_PROC_FS
3459 int __init dev_proc_init(void);
3461 #define dev_proc_init() 0
3464 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3465 struct sk_buff *skb, struct net_device *dev,
3468 skb->xmit_more = more ? 1 : 0;
3469 return ops->ndo_start_xmit(skb, dev);
3472 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3473 struct netdev_queue *txq, bool more)
3475 const struct net_device_ops *ops = dev->netdev_ops;
3478 rc = __netdev_start_xmit(ops, skb, dev, more);
3479 if (rc == NETDEV_TX_OK)
3480 txq_trans_update(txq);
3485 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3487 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3490 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3492 return netdev_class_create_file_ns(class_attr, NULL);
3495 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3497 netdev_class_remove_file_ns(class_attr, NULL);
3500 extern struct kobj_ns_type_operations net_ns_type_operations;
3502 const char *netdev_drivername(const struct net_device *dev);
3504 void linkwatch_run_queue(void);
3506 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3507 netdev_features_t f2)
3509 if (f1 & NETIF_F_GEN_CSUM)
3510 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3511 if (f2 & NETIF_F_GEN_CSUM)
3512 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3514 if (f1 & NETIF_F_GEN_CSUM)
3515 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3520 static inline netdev_features_t netdev_get_wanted_features(
3521 struct net_device *dev)
3523 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3525 netdev_features_t netdev_increment_features(netdev_features_t all,
3526 netdev_features_t one, netdev_features_t mask);
3528 /* Allow TSO being used on stacked device :
3529 * Performing the GSO segmentation before last device
3530 * is a performance improvement.
3532 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3533 netdev_features_t mask)
3535 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3538 int __netdev_update_features(struct net_device *dev);
3539 void netdev_update_features(struct net_device *dev);
3540 void netdev_change_features(struct net_device *dev);
3542 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3543 struct net_device *dev);
3545 netdev_features_t netif_skb_features(struct sk_buff *skb);
3547 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3549 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3551 /* check flags correspondence */
3552 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3553 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3554 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3555 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3556 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3557 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3558 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3559 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3560 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3561 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3562 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3563 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3564 BUILD_BUG_ON(SKB_GSO_MPLS != (NETIF_F_GSO_MPLS >> NETIF_F_GSO_SHIFT));
3566 return (features & feature) == feature;
3569 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3571 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3572 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3575 static inline bool netif_needs_gso(struct sk_buff *skb,
3576 netdev_features_t features)
3578 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3579 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3580 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3583 static inline void netif_set_gso_max_size(struct net_device *dev,
3586 dev->gso_max_size = size;
3589 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3590 int pulled_hlen, u16 mac_offset,
3593 skb->protocol = protocol;
3594 skb->encapsulation = 1;
3595 skb_push(skb, pulled_hlen);
3596 skb_reset_transport_header(skb);
3597 skb->mac_header = mac_offset;
3598 skb->network_header = skb->mac_header + mac_len;
3599 skb->mac_len = mac_len;
3602 static inline bool netif_is_macvlan(struct net_device *dev)
3604 return dev->priv_flags & IFF_MACVLAN;
3607 static inline bool netif_is_bond_master(struct net_device *dev)
3609 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3612 static inline bool netif_is_bond_slave(struct net_device *dev)
3614 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3617 static inline bool netif_supports_nofcs(struct net_device *dev)
3619 return dev->priv_flags & IFF_SUPP_NOFCS;
3622 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3623 static inline void netif_keep_dst(struct net_device *dev)
3625 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3628 extern struct pernet_operations __net_initdata loopback_net_ops;
3630 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3632 /* netdev_printk helpers, similar to dev_printk */
3634 static inline const char *netdev_name(const struct net_device *dev)
3636 if (!dev->name[0] || strchr(dev->name, '%'))
3637 return "(unnamed net_device)";
3641 static inline const char *netdev_reg_state(const struct net_device *dev)
3643 switch (dev->reg_state) {
3644 case NETREG_UNINITIALIZED: return " (uninitialized)";
3645 case NETREG_REGISTERED: return "";
3646 case NETREG_UNREGISTERING: return " (unregistering)";
3647 case NETREG_UNREGISTERED: return " (unregistered)";
3648 case NETREG_RELEASED: return " (released)";
3649 case NETREG_DUMMY: return " (dummy)";
3652 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3653 return " (unknown)";
3657 void netdev_printk(const char *level, const struct net_device *dev,
3658 const char *format, ...);
3660 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3662 void netdev_alert(const struct net_device *dev, const char *format, ...);
3664 void netdev_crit(const struct net_device *dev, const char *format, ...);
3666 void netdev_err(const struct net_device *dev, const char *format, ...);
3668 void netdev_warn(const struct net_device *dev, const char *format, ...);
3670 void netdev_notice(const struct net_device *dev, const char *format, ...);
3672 void netdev_info(const struct net_device *dev, const char *format, ...);
3674 #define MODULE_ALIAS_NETDEV(device) \
3675 MODULE_ALIAS("netdev-" device)
3677 #if defined(CONFIG_DYNAMIC_DEBUG)
3678 #define netdev_dbg(__dev, format, args...) \
3680 dynamic_netdev_dbg(__dev, format, ##args); \
3682 #elif defined(DEBUG)
3683 #define netdev_dbg(__dev, format, args...) \
3684 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3686 #define netdev_dbg(__dev, format, args...) \
3689 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3693 #if defined(VERBOSE_DEBUG)
3694 #define netdev_vdbg netdev_dbg
3697 #define netdev_vdbg(dev, format, args...) \
3700 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3706 * netdev_WARN() acts like dev_printk(), but with the key difference
3707 * of using a WARN/WARN_ON to get the message out, including the
3708 * file/line information and a backtrace.
3710 #define netdev_WARN(dev, format, args...) \
3711 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3712 netdev_reg_state(dev), ##args)
3714 /* netif printk helpers, similar to netdev_printk */
3716 #define netif_printk(priv, type, level, dev, fmt, args...) \
3718 if (netif_msg_##type(priv)) \
3719 netdev_printk(level, (dev), fmt, ##args); \
3722 #define netif_level(level, priv, type, dev, fmt, args...) \
3724 if (netif_msg_##type(priv)) \
3725 netdev_##level(dev, fmt, ##args); \
3728 #define netif_emerg(priv, type, dev, fmt, args...) \
3729 netif_level(emerg, priv, type, dev, fmt, ##args)
3730 #define netif_alert(priv, type, dev, fmt, args...) \
3731 netif_level(alert, priv, type, dev, fmt, ##args)
3732 #define netif_crit(priv, type, dev, fmt, args...) \
3733 netif_level(crit, priv, type, dev, fmt, ##args)
3734 #define netif_err(priv, type, dev, fmt, args...) \
3735 netif_level(err, priv, type, dev, fmt, ##args)
3736 #define netif_warn(priv, type, dev, fmt, args...) \
3737 netif_level(warn, priv, type, dev, fmt, ##args)
3738 #define netif_notice(priv, type, dev, fmt, args...) \
3739 netif_level(notice, priv, type, dev, fmt, ##args)
3740 #define netif_info(priv, type, dev, fmt, args...) \
3741 netif_level(info, priv, type, dev, fmt, ##args)
3743 #if defined(CONFIG_DYNAMIC_DEBUG)
3744 #define netif_dbg(priv, type, netdev, format, args...) \
3746 if (netif_msg_##type(priv)) \
3747 dynamic_netdev_dbg(netdev, format, ##args); \
3749 #elif defined(DEBUG)
3750 #define netif_dbg(priv, type, dev, format, args...) \
3751 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3753 #define netif_dbg(priv, type, dev, format, args...) \
3756 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3761 #if defined(VERBOSE_DEBUG)
3762 #define netif_vdbg netif_dbg
3764 #define netif_vdbg(priv, type, dev, format, args...) \
3767 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3773 * The list of packet types we will receive (as opposed to discard)
3774 * and the routines to invoke.
3776 * Why 16. Because with 16 the only overlap we get on a hash of the
3777 * low nibble of the protocol value is RARP/SNAP/X.25.
3779 * NOTE: That is no longer true with the addition of VLAN tags. Not
3780 * sure which should go first, but I bet it won't make much
3781 * difference if we are running VLANs. The good news is that
3782 * this protocol won't be in the list unless compiled in, so
3783 * the average user (w/out VLANs) will not be adversely affected.
3799 #define PTYPE_HASH_SIZE (16)
3800 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3802 #endif /* _LINUX_NETDEVICE_H */