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 hrtimer timer;
318 struct list_head dev_list;
319 struct hlist_node napi_hash_node;
320 unsigned int napi_id;
324 NAPI_STATE_SCHED, /* Poll is scheduled */
325 NAPI_STATE_DISABLE, /* Disable pending */
326 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
327 NAPI_STATE_HASHED, /* In NAPI hash */
337 typedef enum gro_result gro_result_t;
340 * enum rx_handler_result - Possible return values for rx_handlers.
341 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
343 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
344 * case skb->dev was changed by rx_handler.
345 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
346 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
348 * rx_handlers are functions called from inside __netif_receive_skb(), to do
349 * special processing of the skb, prior to delivery to protocol handlers.
351 * Currently, a net_device can only have a single rx_handler registered. Trying
352 * to register a second rx_handler will return -EBUSY.
354 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
355 * To unregister a rx_handler on a net_device, use
356 * netdev_rx_handler_unregister().
358 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * If the rx_handler consumed to skb in some way, it should return
362 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
363 * the skb to be delivered in some other ways.
365 * If the rx_handler changed skb->dev, to divert the skb to another
366 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
367 * new device will be called if it exists.
369 * If the rx_handler consider the skb should be ignored, it should return
370 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
371 * are registered on exact device (ptype->dev == skb->dev).
373 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
374 * delivered, it should return RX_HANDLER_PASS.
376 * A device without a registered rx_handler will behave as if rx_handler
377 * returned RX_HANDLER_PASS.
380 enum rx_handler_result {
386 typedef enum rx_handler_result rx_handler_result_t;
387 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
389 void __napi_schedule(struct napi_struct *n);
390 void __napi_schedule_irqoff(struct napi_struct *n);
392 static inline bool napi_disable_pending(struct napi_struct *n)
394 return test_bit(NAPI_STATE_DISABLE, &n->state);
398 * napi_schedule_prep - check if napi can be scheduled
401 * Test if NAPI routine is already running, and if not mark
402 * it as running. This is used as a condition variable
403 * insure only one NAPI poll instance runs. We also make
404 * sure there is no pending NAPI disable.
406 static inline bool napi_schedule_prep(struct napi_struct *n)
408 return !napi_disable_pending(n) &&
409 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
413 * napi_schedule - schedule NAPI poll
416 * Schedule NAPI poll routine to be called if it is not already
419 static inline void napi_schedule(struct napi_struct *n)
421 if (napi_schedule_prep(n))
426 * napi_schedule_irqoff - schedule NAPI poll
429 * Variant of napi_schedule(), assuming hard irqs are masked.
431 static inline void napi_schedule_irqoff(struct napi_struct *n)
433 if (napi_schedule_prep(n))
434 __napi_schedule_irqoff(n);
437 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
438 static inline bool napi_reschedule(struct napi_struct *napi)
440 if (napi_schedule_prep(napi)) {
441 __napi_schedule(napi);
447 void __napi_complete(struct napi_struct *n);
448 void napi_complete_done(struct napi_struct *n, int work_done);
450 * napi_complete - NAPI processing complete
453 * Mark NAPI processing as complete.
454 * Consider using napi_complete_done() instead.
456 static inline void napi_complete(struct napi_struct *n)
458 return napi_complete_done(n, 0);
462 * napi_by_id - lookup a NAPI by napi_id
463 * @napi_id: hashed napi_id
465 * lookup @napi_id in napi_hash table
466 * must be called under rcu_read_lock()
468 struct napi_struct *napi_by_id(unsigned int napi_id);
471 * napi_hash_add - add a NAPI to global hashtable
472 * @napi: napi context
474 * generate a new napi_id and store a @napi under it in napi_hash
476 void napi_hash_add(struct napi_struct *napi);
479 * napi_hash_del - remove a NAPI from global table
480 * @napi: napi context
482 * Warning: caller must observe rcu grace period
483 * before freeing memory containing @napi
485 void napi_hash_del(struct napi_struct *napi);
488 * napi_disable - prevent NAPI from scheduling
491 * Stop NAPI from being scheduled on this context.
492 * Waits till any outstanding processing completes.
494 void napi_disable(struct napi_struct *n);
497 * napi_enable - enable NAPI scheduling
500 * Resume NAPI from being scheduled on this context.
501 * Must be paired with napi_disable.
503 static inline void napi_enable(struct napi_struct *n)
505 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
506 smp_mb__before_atomic();
507 clear_bit(NAPI_STATE_SCHED, &n->state);
512 * napi_synchronize - wait until NAPI is not running
515 * Wait until NAPI is done being scheduled on this context.
516 * Waits till any outstanding processing completes but
517 * does not disable future activations.
519 static inline void napi_synchronize(const struct napi_struct *n)
521 while (test_bit(NAPI_STATE_SCHED, &n->state))
525 # define napi_synchronize(n) barrier()
528 enum netdev_queue_state_t {
529 __QUEUE_STATE_DRV_XOFF,
530 __QUEUE_STATE_STACK_XOFF,
531 __QUEUE_STATE_FROZEN,
534 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
535 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
538 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
539 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
541 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
545 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
546 * netif_tx_* functions below are used to manipulate this flag. The
547 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
548 * queue independently. The netif_xmit_*stopped functions below are called
549 * to check if the queue has been stopped by the driver or stack (either
550 * of the XOFF bits are set in the state). Drivers should not need to call
551 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 struct netdev_queue {
558 struct net_device *dev;
559 struct Qdisc __rcu *qdisc;
560 struct Qdisc *qdisc_sleeping;
564 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
570 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 * please use this field instead of dev->trans_start
575 unsigned long trans_start;
578 * Number of TX timeouts for this queue
579 * (/sys/class/net/DEV/Q/trans_timeout)
581 unsigned long trans_timeout;
588 } ____cacheline_aligned_in_smp;
590 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
592 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 * This structure holds an RPS map which can be of variable length. The
609 * map is an array of CPUs.
616 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
619 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
620 * tail pointer for that CPU's input queue at the time of last enqueue, and
621 * a hardware filter index.
623 struct rps_dev_flow {
626 unsigned int last_qtail;
628 #define RPS_NO_FILTER 0xffff
631 * The rps_dev_flow_table structure contains a table of flow mappings.
633 struct rps_dev_flow_table {
636 struct rps_dev_flow flows[0];
638 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
639 ((_num) * sizeof(struct rps_dev_flow)))
642 * The rps_sock_flow_table contains mappings of flows to the last CPU
643 * on which they were processed by the application (set in recvmsg).
645 struct rps_sock_flow_table {
649 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
650 ((_num) * sizeof(u16)))
652 #define RPS_NO_CPU 0xffff
654 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
658 unsigned int cpu, index = hash & table->mask;
660 /* We only give a hint, preemption can change cpu under us */
661 cpu = raw_smp_processor_id();
663 if (table->ents[index] != cpu)
664 table->ents[index] = cpu;
668 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
672 table->ents[hash & table->mask] = RPS_NO_CPU;
675 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
677 #ifdef CONFIG_RFS_ACCEL
678 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
681 #endif /* CONFIG_RPS */
683 /* This structure contains an instance of an RX queue. */
684 struct netdev_rx_queue {
686 struct rps_map __rcu *rps_map;
687 struct rps_dev_flow_table __rcu *rps_flow_table;
690 struct net_device *dev;
691 } ____cacheline_aligned_in_smp;
694 * RX queue sysfs structures and functions.
696 struct rx_queue_attribute {
697 struct attribute attr;
698 ssize_t (*show)(struct netdev_rx_queue *queue,
699 struct rx_queue_attribute *attr, char *buf);
700 ssize_t (*store)(struct netdev_rx_queue *queue,
701 struct rx_queue_attribute *attr, const char *buf, size_t len);
706 * This structure holds an XPS map which can be of variable length. The
707 * map is an array of queues.
711 unsigned int alloc_len;
715 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
716 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
720 * This structure holds all XPS maps for device. Maps are indexed by CPU.
722 struct xps_dev_maps {
724 struct xps_map __rcu *cpu_map[0];
726 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
727 (nr_cpu_ids * sizeof(struct xps_map *)))
728 #endif /* CONFIG_XPS */
730 #define TC_MAX_QUEUE 16
731 #define TC_BITMASK 15
732 /* HW offloaded queuing disciplines txq count and offset maps */
733 struct netdev_tc_txq {
738 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
740 * This structure is to hold information about the device
741 * configured to run FCoE protocol stack.
743 struct netdev_fcoe_hbainfo {
744 char manufacturer[64];
745 char serial_number[64];
746 char hardware_version[64];
747 char driver_version[64];
748 char optionrom_version[64];
749 char firmware_version[64];
751 char model_description[256];
755 #define MAX_PHYS_PORT_ID_LEN 32
757 /* This structure holds a unique identifier to identify the
758 * physical port used by a netdevice.
760 struct netdev_phys_port_id {
761 unsigned char id[MAX_PHYS_PORT_ID_LEN];
762 unsigned char id_len;
765 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
766 struct sk_buff *skb);
769 * This structure defines the management hooks for network devices.
770 * The following hooks can be defined; unless noted otherwise, they are
771 * optional and can be filled with a null pointer.
773 * int (*ndo_init)(struct net_device *dev);
774 * This function is called once when network device is registered.
775 * The network device can use this to any late stage initializaton
776 * or semantic validattion. It can fail with an error code which will
777 * be propogated back to register_netdev
779 * void (*ndo_uninit)(struct net_device *dev);
780 * This function is called when device is unregistered or when registration
781 * fails. It is not called if init fails.
783 * int (*ndo_open)(struct net_device *dev);
784 * This function is called when network device transistions to the up
787 * int (*ndo_stop)(struct net_device *dev);
788 * This function is called when network device transistions to the down
791 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
792 * struct net_device *dev);
793 * Called when a packet needs to be transmitted.
794 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
795 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
796 * Required can not be NULL.
798 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
799 * void *accel_priv, select_queue_fallback_t fallback);
800 * Called to decide which queue to when device supports multiple
803 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
804 * This function is called to allow device receiver to make
805 * changes to configuration when multicast or promiscious is enabled.
807 * void (*ndo_set_rx_mode)(struct net_device *dev);
808 * This function is called device changes address list filtering.
809 * If driver handles unicast address filtering, it should set
810 * IFF_UNICAST_FLT to its priv_flags.
812 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
813 * This function is called when the Media Access Control address
814 * needs to be changed. If this interface is not defined, the
815 * mac address can not be changed.
817 * int (*ndo_validate_addr)(struct net_device *dev);
818 * Test if Media Access Control address is valid for the device.
820 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
821 * Called when a user request an ioctl which can't be handled by
822 * the generic interface code. If not defined ioctl's return
823 * not supported error code.
825 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
826 * Used to set network devices bus interface parameters. This interface
827 * is retained for legacy reason, new devices should use the bus
828 * interface (PCI) for low level management.
830 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
831 * Called when a user wants to change the Maximum Transfer Unit
832 * of a device. If not defined, any request to change MTU will
833 * will return an error.
835 * void (*ndo_tx_timeout)(struct net_device *dev);
836 * Callback uses when the transmitter has not made any progress
837 * for dev->watchdog ticks.
839 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
840 * struct rtnl_link_stats64 *storage);
841 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
842 * Called when a user wants to get the network device usage
843 * statistics. Drivers must do one of the following:
844 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
845 * rtnl_link_stats64 structure passed by the caller.
846 * 2. Define @ndo_get_stats to update a net_device_stats structure
847 * (which should normally be dev->stats) and return a pointer to
848 * it. The structure may be changed asynchronously only if each
849 * field is written atomically.
850 * 3. Update dev->stats asynchronously and atomically, and define
853 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
854 * If device support VLAN filtering this function is called when a
855 * VLAN id is registered.
857 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
858 * If device support VLAN filtering this function is called when a
859 * VLAN id is unregistered.
861 * void (*ndo_poll_controller)(struct net_device *dev);
863 * SR-IOV management functions.
864 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
865 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
866 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
868 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
869 * int (*ndo_get_vf_config)(struct net_device *dev,
870 * int vf, struct ifla_vf_info *ivf);
871 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
872 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
873 * struct nlattr *port[]);
874 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
875 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
876 * Called to setup 'tc' number of traffic classes in the net device. This
877 * is always called from the stack with the rtnl lock held and netif tx
878 * queues stopped. This allows the netdevice to perform queue management
881 * Fiber Channel over Ethernet (FCoE) offload functions.
882 * int (*ndo_fcoe_enable)(struct net_device *dev);
883 * Called when the FCoE protocol stack wants to start using LLD for FCoE
884 * so the underlying device can perform whatever needed configuration or
885 * initialization to support acceleration of FCoE traffic.
887 * int (*ndo_fcoe_disable)(struct net_device *dev);
888 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
889 * so the underlying device can perform whatever needed clean-ups to
890 * stop supporting acceleration of FCoE traffic.
892 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
893 * struct scatterlist *sgl, unsigned int sgc);
894 * Called when the FCoE Initiator 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_ddp_done)(struct net_device *dev, u16 xid);
900 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
901 * indicated by the FC exchange id 'xid', so the underlying device can
902 * clean up and reuse resources for later DDP requests.
904 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
905 * struct scatterlist *sgl, unsigned int sgc);
906 * Called when the FCoE Target wants to initialize an I/O that
907 * is a possible candidate for Direct Data Placement (DDP). The LLD can
908 * perform necessary setup and returns 1 to indicate the device is set up
909 * successfully to perform DDP on this I/O, otherwise this returns 0.
911 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
912 * struct netdev_fcoe_hbainfo *hbainfo);
913 * Called when the FCoE Protocol stack wants information on the underlying
914 * device. This information is utilized by the FCoE protocol stack to
915 * register attributes with Fiber Channel management service as per the
916 * FC-GS Fabric Device Management Information(FDMI) specification.
918 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
919 * Called when the underlying device wants to override default World Wide
920 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
921 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
922 * protocol stack to use.
925 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
926 * u16 rxq_index, u32 flow_id);
927 * Set hardware filter for RFS. rxq_index is the target queue index;
928 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
929 * Return the filter ID on success, or a negative error code.
931 * Slave management functions (for bridge, bonding, etc).
932 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
933 * Called to make another netdev an underling.
935 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
936 * Called to release previously enslaved netdev.
938 * Feature/offload setting functions.
939 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
940 * netdev_features_t features);
941 * Adjusts the requested feature flags according to device-specific
942 * constraints, and returns the resulting flags. Must not modify
945 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
946 * Called to update device configuration to new features. Passed
947 * feature set might be less than what was returned by ndo_fix_features()).
948 * Must return >0 or -errno if it changed dev->features itself.
950 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
951 * struct net_device *dev,
952 * const unsigned char *addr, u16 flags)
953 * Adds an FDB entry to dev for addr.
954 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
955 * struct net_device *dev,
956 * const unsigned char *addr)
957 * Deletes the FDB entry from dev coresponding to addr.
958 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
959 * struct net_device *dev, struct net_device *filter_dev,
961 * Used to add FDB entries to dump requests. Implementers should add
962 * entries to skb and update idx with the number of entries.
964 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
965 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
966 * struct net_device *dev, u32 filter_mask)
968 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
969 * Called to change device carrier. Soft-devices (like dummy, team, etc)
970 * which do not represent real hardware may define this to allow their
971 * userspace components to manage their virtual carrier state. Devices
972 * that determine carrier state from physical hardware properties (eg
973 * network cables) or protocol-dependent mechanisms (eg
974 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
976 * int (*ndo_get_phys_port_id)(struct net_device *dev,
977 * struct netdev_phys_port_id *ppid);
978 * Called to get ID of physical port of this device. If driver does
979 * not implement this, it is assumed that the hw is not able to have
980 * multiple net devices on single physical port.
982 * void (*ndo_add_vxlan_port)(struct net_device *dev,
983 * sa_family_t sa_family, __be16 port);
984 * Called by vxlan to notiy a driver about the UDP port and socket
985 * address family that vxlan is listnening to. It is called only when
986 * a new port starts listening. The operation is protected by the
987 * vxlan_net->sock_lock.
989 * void (*ndo_del_vxlan_port)(struct net_device *dev,
990 * sa_family_t sa_family, __be16 port);
991 * Called by vxlan to notify the driver about a UDP port and socket
992 * address family that vxlan is not listening to anymore. The operation
993 * is protected by the vxlan_net->sock_lock.
995 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
996 * struct net_device *dev)
997 * Called by upper layer devices to accelerate switching or other
998 * station functionality into hardware. 'pdev is the lowerdev
999 * to use for the offload and 'dev' is the net device that will
1000 * back the offload. Returns a pointer to the private structure
1001 * the upper layer will maintain.
1002 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1003 * Called by upper layer device to delete the station created
1004 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1005 * the station and priv is the structure returned by the add
1007 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1008 * struct net_device *dev,
1010 * Callback to use for xmit over the accelerated station. This
1011 * is used in place of ndo_start_xmit on accelerated net
1013 * bool (*ndo_gso_check) (struct sk_buff *skb,
1014 * struct net_device *dev);
1015 * Called by core transmit path to determine if device is capable of
1016 * performing GSO on a packet. The device returns true if it is
1017 * able to GSO the packet, false otherwise. If the return value is
1018 * false the stack will do software GSO.
1020 struct net_device_ops {
1021 int (*ndo_init)(struct net_device *dev);
1022 void (*ndo_uninit)(struct net_device *dev);
1023 int (*ndo_open)(struct net_device *dev);
1024 int (*ndo_stop)(struct net_device *dev);
1025 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1026 struct net_device *dev);
1027 u16 (*ndo_select_queue)(struct net_device *dev,
1028 struct sk_buff *skb,
1030 select_queue_fallback_t fallback);
1031 void (*ndo_change_rx_flags)(struct net_device *dev,
1033 void (*ndo_set_rx_mode)(struct net_device *dev);
1034 int (*ndo_set_mac_address)(struct net_device *dev,
1036 int (*ndo_validate_addr)(struct net_device *dev);
1037 int (*ndo_do_ioctl)(struct net_device *dev,
1038 struct ifreq *ifr, int cmd);
1039 int (*ndo_set_config)(struct net_device *dev,
1041 int (*ndo_change_mtu)(struct net_device *dev,
1043 int (*ndo_neigh_setup)(struct net_device *dev,
1044 struct neigh_parms *);
1045 void (*ndo_tx_timeout) (struct net_device *dev);
1047 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1048 struct rtnl_link_stats64 *storage);
1049 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1051 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1052 __be16 proto, u16 vid);
1053 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1054 __be16 proto, u16 vid);
1055 #ifdef CONFIG_NET_POLL_CONTROLLER
1056 void (*ndo_poll_controller)(struct net_device *dev);
1057 int (*ndo_netpoll_setup)(struct net_device *dev,
1058 struct netpoll_info *info);
1059 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1061 #ifdef CONFIG_NET_RX_BUSY_POLL
1062 int (*ndo_busy_poll)(struct napi_struct *dev);
1064 int (*ndo_set_vf_mac)(struct net_device *dev,
1065 int queue, u8 *mac);
1066 int (*ndo_set_vf_vlan)(struct net_device *dev,
1067 int queue, u16 vlan, u8 qos);
1068 int (*ndo_set_vf_rate)(struct net_device *dev,
1069 int vf, int min_tx_rate,
1071 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1072 int vf, bool setting);
1073 int (*ndo_get_vf_config)(struct net_device *dev,
1075 struct ifla_vf_info *ivf);
1076 int (*ndo_set_vf_link_state)(struct net_device *dev,
1077 int vf, int link_state);
1078 int (*ndo_set_vf_port)(struct net_device *dev,
1080 struct nlattr *port[]);
1081 int (*ndo_get_vf_port)(struct net_device *dev,
1082 int vf, struct sk_buff *skb);
1083 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1084 #if IS_ENABLED(CONFIG_FCOE)
1085 int (*ndo_fcoe_enable)(struct net_device *dev);
1086 int (*ndo_fcoe_disable)(struct net_device *dev);
1087 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1089 struct scatterlist *sgl,
1091 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1093 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1095 struct scatterlist *sgl,
1097 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1098 struct netdev_fcoe_hbainfo *hbainfo);
1101 #if IS_ENABLED(CONFIG_LIBFCOE)
1102 #define NETDEV_FCOE_WWNN 0
1103 #define NETDEV_FCOE_WWPN 1
1104 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1105 u64 *wwn, int type);
1108 #ifdef CONFIG_RFS_ACCEL
1109 int (*ndo_rx_flow_steer)(struct net_device *dev,
1110 const struct sk_buff *skb,
1114 int (*ndo_add_slave)(struct net_device *dev,
1115 struct net_device *slave_dev);
1116 int (*ndo_del_slave)(struct net_device *dev,
1117 struct net_device *slave_dev);
1118 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1119 netdev_features_t features);
1120 int (*ndo_set_features)(struct net_device *dev,
1121 netdev_features_t features);
1122 int (*ndo_neigh_construct)(struct neighbour *n);
1123 void (*ndo_neigh_destroy)(struct neighbour *n);
1125 int (*ndo_fdb_add)(struct ndmsg *ndm,
1126 struct nlattr *tb[],
1127 struct net_device *dev,
1128 const unsigned char *addr,
1130 int (*ndo_fdb_del)(struct ndmsg *ndm,
1131 struct nlattr *tb[],
1132 struct net_device *dev,
1133 const unsigned char *addr);
1134 int (*ndo_fdb_dump)(struct sk_buff *skb,
1135 struct netlink_callback *cb,
1136 struct net_device *dev,
1137 struct net_device *filter_dev,
1140 int (*ndo_bridge_setlink)(struct net_device *dev,
1141 struct nlmsghdr *nlh);
1142 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1144 struct net_device *dev,
1146 int (*ndo_bridge_dellink)(struct net_device *dev,
1147 struct nlmsghdr *nlh);
1148 int (*ndo_change_carrier)(struct net_device *dev,
1150 int (*ndo_get_phys_port_id)(struct net_device *dev,
1151 struct netdev_phys_port_id *ppid);
1152 void (*ndo_add_vxlan_port)(struct net_device *dev,
1153 sa_family_t sa_family,
1155 void (*ndo_del_vxlan_port)(struct net_device *dev,
1156 sa_family_t sa_family,
1159 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1160 struct net_device *dev);
1161 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1164 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1165 struct net_device *dev,
1167 int (*ndo_get_lock_subclass)(struct net_device *dev);
1168 bool (*ndo_gso_check) (struct sk_buff *skb,
1169 struct net_device *dev);
1173 * enum net_device_priv_flags - &struct net_device priv_flags
1175 * These are the &struct net_device, they are only set internally
1176 * by drivers and used in the kernel. These flags are invisible to
1177 * userspace, this means that the order of these flags can change
1178 * during any kernel release.
1180 * You should have a pretty good reason to be extending these flags.
1182 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1183 * @IFF_EBRIDGE: Ethernet bridging device
1184 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1185 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1186 * @IFF_MASTER_ALB: bonding master, balance-alb
1187 * @IFF_BONDING: bonding master or slave
1188 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1189 * @IFF_ISATAP: ISATAP interface (RFC4214)
1190 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1191 * @IFF_WAN_HDLC: WAN HDLC device
1192 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1194 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1195 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1196 * @IFF_MACVLAN_PORT: device used as macvlan port
1197 * @IFF_BRIDGE_PORT: device used as bridge port
1198 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1199 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1200 * @IFF_UNICAST_FLT: Supports unicast filtering
1201 * @IFF_TEAM_PORT: device used as team port
1202 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1203 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1204 * change when it's running
1205 * @IFF_MACVLAN: Macvlan device
1207 enum netdev_priv_flags {
1208 IFF_802_1Q_VLAN = 1<<0,
1210 IFF_SLAVE_INACTIVE = 1<<2,
1211 IFF_MASTER_8023AD = 1<<3,
1212 IFF_MASTER_ALB = 1<<4,
1214 IFF_SLAVE_NEEDARP = 1<<6,
1216 IFF_MASTER_ARPMON = 1<<8,
1217 IFF_WAN_HDLC = 1<<9,
1218 IFF_XMIT_DST_RELEASE = 1<<10,
1219 IFF_DONT_BRIDGE = 1<<11,
1220 IFF_DISABLE_NETPOLL = 1<<12,
1221 IFF_MACVLAN_PORT = 1<<13,
1222 IFF_BRIDGE_PORT = 1<<14,
1223 IFF_OVS_DATAPATH = 1<<15,
1224 IFF_TX_SKB_SHARING = 1<<16,
1225 IFF_UNICAST_FLT = 1<<17,
1226 IFF_TEAM_PORT = 1<<18,
1227 IFF_SUPP_NOFCS = 1<<19,
1228 IFF_LIVE_ADDR_CHANGE = 1<<20,
1229 IFF_MACVLAN = 1<<21,
1230 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1233 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1234 #define IFF_EBRIDGE IFF_EBRIDGE
1235 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1236 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1237 #define IFF_MASTER_ALB IFF_MASTER_ALB
1238 #define IFF_BONDING IFF_BONDING
1239 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1240 #define IFF_ISATAP IFF_ISATAP
1241 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1242 #define IFF_WAN_HDLC IFF_WAN_HDLC
1243 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1244 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1245 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1246 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1247 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1248 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1249 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1250 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1251 #define IFF_TEAM_PORT IFF_TEAM_PORT
1252 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1253 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1254 #define IFF_MACVLAN IFF_MACVLAN
1255 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1258 * struct net_device - The DEVICE structure.
1259 * Actually, this whole structure is a big mistake. It mixes I/O
1260 * data with strictly "high-level" data, and it has to know about
1261 * almost every data structure used in the INET module.
1263 * @name: This is the first field of the "visible" part of this structure
1264 * (i.e. as seen by users in the "Space.c" file). It is the name
1267 * @name_hlist: Device name hash chain, please keep it close to name[]
1268 * @ifalias: SNMP alias
1269 * @mem_end: Shared memory end
1270 * @mem_start: Shared memory start
1271 * @base_addr: Device I/O address
1272 * @irq: Device IRQ number
1274 * @state: Generic network queuing layer state, see netdev_state_t
1275 * @dev_list: The global list of network devices
1276 * @napi_list: List entry, that is used for polling napi devices
1277 * @unreg_list: List entry, that is used, when we are unregistering the
1278 * device, see the function unregister_netdev
1279 * @close_list: List entry, that is used, when we are closing the device
1281 * @adj_list: Directly linked devices, like slaves for bonding
1282 * @all_adj_list: All linked devices, *including* neighbours
1283 * @features: Currently active device features
1284 * @hw_features: User-changeable features
1286 * @wanted_features: User-requested features
1287 * @vlan_features: Mask of features inheritable by VLAN devices
1289 * @hw_enc_features: Mask of features inherited by encapsulating devices
1290 * This field indicates what encapsulation
1291 * offloads the hardware is capable of doing,
1292 * and drivers will need to set them appropriately.
1294 * @mpls_features: Mask of features inheritable by MPLS
1296 * @ifindex: interface index
1297 * @iflink: unique device identifier
1299 * @stats: Statistics struct, which was left as a legacy, use
1300 * rtnl_link_stats64 instead
1302 * @rx_dropped: Dropped packets by core network,
1303 * do not use this in drivers
1304 * @tx_dropped: Dropped packets by core network,
1305 * do not use this in drivers
1307 * @carrier_changes: Stats to monitor carrier on<->off transitions
1309 * @wireless_handlers: List of functions to handle Wireless Extensions,
1311 * see <net/iw_handler.h> for details.
1312 * @wireless_data: Instance data managed by the core of wireless extensions
1314 * @netdev_ops: Includes several pointers to callbacks,
1315 * if one wants to override the ndo_*() functions
1316 * @ethtool_ops: Management operations
1317 * @fwd_ops: Management operations
1318 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1319 * of Layer 2 headers.
1321 * @flags: Interface flags (a la BSD)
1322 * @priv_flags: Like 'flags' but invisible to userspace,
1323 * see if.h for the definitions
1324 * @gflags: Global flags ( kept as legacy )
1325 * @padded: How much padding added by alloc_netdev()
1326 * @operstate: RFC2863 operstate
1327 * @link_mode: Mapping policy to operstate
1328 * @if_port: Selectable AUI, TP, ...
1330 * @mtu: Interface MTU value
1331 * @type: Interface hardware type
1332 * @hard_header_len: Hardware header length
1334 * @needed_headroom: Extra headroom the hardware may need, but not in all
1335 * cases can this be guaranteed
1336 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1337 * cases can this be guaranteed. Some cases also use
1338 * LL_MAX_HEADER instead to allocate the skb
1340 * interface address info:
1342 * @perm_addr: Permanent hw address
1343 * @addr_assign_type: Hw address assignment type
1344 * @addr_len: Hardware address length
1345 * @neigh_priv_len; Used in neigh_alloc(),
1346 * initialized only in atm/clip.c
1347 * @dev_id: Used to differentiate devices that share
1348 * the same link layer address
1349 * @dev_port: Used to differentiate devices that share
1351 * @addr_list_lock: XXX: need comments on this one
1352 * @uc: unicast mac addresses
1353 * @mc: multicast mac addresses
1354 * @dev_addrs: list of device hw addresses
1355 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1356 * @uc_promisc: Counter, that indicates, that promiscuous mode
1357 * has been enabled due to the need to listen to
1358 * additional unicast addresses in a device that
1359 * does not implement ndo_set_rx_mode()
1360 * @promiscuity: Number of times, the NIC is told to work in
1361 * Promiscuous mode, if it becomes 0 the NIC will
1362 * exit from working in Promiscuous mode
1363 * @allmulti: Counter, enables or disables allmulticast mode
1365 * @vlan_info: VLAN info
1366 * @dsa_ptr: dsa specific data
1367 * @tipc_ptr: TIPC specific data
1368 * @atalk_ptr: AppleTalk link
1369 * @ip_ptr: IPv4 specific data
1370 * @dn_ptr: DECnet specific data
1371 * @ip6_ptr: IPv6 specific data
1372 * @ax25_ptr: AX.25 specific data
1373 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1375 * @last_rx: Time of last Rx
1376 * @dev_addr: Hw address (before bcast,
1377 * because most packets are unicast)
1379 * @_rx: Array of RX queues
1380 * @num_rx_queues: Number of RX queues
1381 * allocated at register_netdev() time
1382 * @real_num_rx_queues: Number of RX queues currently active in device
1384 * @rx_handler: handler for received packets
1385 * @rx_handler_data: XXX: need comments on this one
1386 * @ingress_queue: XXX: need comments on this one
1387 * @broadcast: hw bcast address
1389 * @_tx: Array of TX queues
1390 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1391 * @real_num_tx_queues: Number of TX queues currently active in device
1392 * @qdisc: Root qdisc from userspace point of view
1393 * @tx_queue_len: Max frames per queue allowed
1394 * @tx_global_lock: XXX: need comments on this one
1396 * @xps_maps: XXX: need comments on this one
1398 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1399 * indexed by RX queue number. Assigned by driver.
1400 * This must only be set if the ndo_rx_flow_steer
1401 * operation is defined
1403 * @trans_start: Time (in jiffies) of last Tx
1404 * @watchdog_timeo: Represents the timeout that is used by
1405 * the watchdog ( see dev_watchdog() )
1406 * @watchdog_timer: List of timers
1408 * @pcpu_refcnt: Number of references to this device
1409 * @todo_list: Delayed register/unregister
1410 * @index_hlist: Device index hash chain
1411 * @link_watch_list: XXX: need comments on this one
1413 * @reg_state: Register/unregister state machine
1414 * @dismantle: Device is going to be freed
1415 * @rtnl_link_state: This enum represents the phases of creating
1418 * @destructor: Called from unregister,
1419 * can be used to call free_netdev
1420 * @npinfo: XXX: need comments on this one
1421 * @nd_net: Network namespace this network device is inside
1423 * @ml_priv: Mid-layer private
1424 * @lstats: Loopback statistics
1425 * @tstats: Tunnel statistics
1426 * @dstats: Dummy statistics
1427 * @vstats: Virtual ethernet statistics
1432 * @dev: Class/net/name entry
1433 * @sysfs_groups: Space for optional device, statistics and wireless
1436 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1437 * @rtnl_link_ops: Rtnl_link_ops
1439 * @gso_max_size: Maximum size of generic segmentation offload
1440 * @gso_max_segs: Maximum number of segments that can be passed to the
1442 * @gso_min_segs: Minimum number of segments that can be passed to the
1445 * @dcbnl_ops: Data Center Bridging netlink ops
1446 * @num_tc: Number of traffic classes in the net device
1447 * @tc_to_txq: XXX: need comments on this one
1448 * @prio_tc_map XXX: need comments on this one
1450 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1452 * @priomap: XXX: need comments on this one
1453 * @phydev: Physical device may attach itself
1454 * for hardware timestamping
1456 * @qdisc_tx_busylock: XXX: need comments on this one
1458 * @group: The group, that the device belongs to
1459 * @pm_qos_req: Power Management QoS object
1461 * FIXME: cleanup struct net_device such that network protocol info
1466 char name[IFNAMSIZ];
1467 struct hlist_node name_hlist;
1470 * I/O specific fields
1471 * FIXME: Merge these and struct ifmap into one
1473 unsigned long mem_end;
1474 unsigned long mem_start;
1475 unsigned long base_addr;
1479 * Some hardware also needs these fields (state,dev_list,
1480 * napi_list,unreg_list,close_list) but they are not
1481 * part of the usual set specified in Space.c.
1484 unsigned long state;
1486 struct list_head dev_list;
1487 struct list_head napi_list;
1488 struct list_head unreg_list;
1489 struct list_head close_list;
1492 struct list_head upper;
1493 struct list_head lower;
1497 struct list_head upper;
1498 struct list_head lower;
1501 netdev_features_t features;
1502 netdev_features_t hw_features;
1503 netdev_features_t wanted_features;
1504 netdev_features_t vlan_features;
1505 netdev_features_t hw_enc_features;
1506 netdev_features_t mpls_features;
1511 struct net_device_stats stats;
1513 atomic_long_t rx_dropped;
1514 atomic_long_t tx_dropped;
1516 atomic_t carrier_changes;
1518 #ifdef CONFIG_WIRELESS_EXT
1519 const struct iw_handler_def * wireless_handlers;
1520 struct iw_public_data * wireless_data;
1522 const struct net_device_ops *netdev_ops;
1523 const struct ethtool_ops *ethtool_ops;
1524 const struct forwarding_accel_ops *fwd_ops;
1526 const struct header_ops *header_ops;
1529 unsigned int priv_flags;
1531 unsigned short gflags;
1532 unsigned short padded;
1534 unsigned char operstate;
1535 unsigned char link_mode;
1537 unsigned char if_port;
1541 unsigned short type;
1542 unsigned short hard_header_len;
1544 unsigned short needed_headroom;
1545 unsigned short needed_tailroom;
1547 /* Interface address info. */
1548 unsigned char perm_addr[MAX_ADDR_LEN];
1549 unsigned char addr_assign_type;
1550 unsigned char addr_len;
1551 unsigned short neigh_priv_len;
1552 unsigned short dev_id;
1553 unsigned short dev_port;
1554 spinlock_t addr_list_lock;
1555 struct netdev_hw_addr_list uc;
1556 struct netdev_hw_addr_list mc;
1557 struct netdev_hw_addr_list dev_addrs;
1560 struct kset *queues_kset;
1563 unsigned char name_assign_type;
1566 unsigned int promiscuity;
1567 unsigned int allmulti;
1570 /* Protocol specific pointers */
1572 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1573 struct vlan_info __rcu *vlan_info;
1575 #if IS_ENABLED(CONFIG_NET_DSA)
1576 struct dsa_switch_tree *dsa_ptr;
1578 #if IS_ENABLED(CONFIG_TIPC)
1579 struct tipc_bearer __rcu *tipc_ptr;
1582 struct in_device __rcu *ip_ptr;
1583 struct dn_dev __rcu *dn_ptr;
1584 struct inet6_dev __rcu *ip6_ptr;
1586 struct wireless_dev *ieee80211_ptr;
1589 * Cache lines mostly used on receive path (including eth_type_trans())
1591 unsigned long last_rx;
1593 /* Interface address info used in eth_type_trans() */
1594 unsigned char *dev_addr;
1598 struct netdev_rx_queue *_rx;
1600 unsigned int num_rx_queues;
1601 unsigned int real_num_rx_queues;
1605 unsigned long gro_flush_timeout;
1606 rx_handler_func_t __rcu *rx_handler;
1607 void __rcu *rx_handler_data;
1609 struct netdev_queue __rcu *ingress_queue;
1610 unsigned char broadcast[MAX_ADDR_LEN];
1614 * Cache lines mostly used on transmit path
1616 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1617 unsigned int num_tx_queues;
1618 unsigned int real_num_tx_queues;
1619 struct Qdisc *qdisc;
1620 unsigned long tx_queue_len;
1621 spinlock_t tx_global_lock;
1624 struct xps_dev_maps __rcu *xps_maps;
1626 #ifdef CONFIG_RFS_ACCEL
1627 struct cpu_rmap *rx_cpu_rmap;
1630 /* These may be needed for future network-power-down code. */
1633 * trans_start here is expensive for high speed devices on SMP,
1634 * please use netdev_queue->trans_start instead.
1636 unsigned long trans_start;
1639 struct timer_list watchdog_timer;
1641 int __percpu *pcpu_refcnt;
1642 struct list_head todo_list;
1644 struct hlist_node index_hlist;
1645 struct list_head link_watch_list;
1647 enum { NETREG_UNINITIALIZED=0,
1648 NETREG_REGISTERED, /* completed register_netdevice */
1649 NETREG_UNREGISTERING, /* called unregister_netdevice */
1650 NETREG_UNREGISTERED, /* completed unregister todo */
1651 NETREG_RELEASED, /* called free_netdev */
1652 NETREG_DUMMY, /* dummy device for NAPI poll */
1658 RTNL_LINK_INITIALIZED,
1659 RTNL_LINK_INITIALIZING,
1660 } rtnl_link_state:16;
1662 void (*destructor)(struct net_device *dev);
1664 #ifdef CONFIG_NETPOLL
1665 struct netpoll_info __rcu *npinfo;
1668 #ifdef CONFIG_NET_NS
1672 /* mid-layer private */
1675 struct pcpu_lstats __percpu *lstats;
1676 struct pcpu_sw_netstats __percpu *tstats;
1677 struct pcpu_dstats __percpu *dstats;
1678 struct pcpu_vstats __percpu *vstats;
1681 struct garp_port __rcu *garp_port;
1682 struct mrp_port __rcu *mrp_port;
1685 const struct attribute_group *sysfs_groups[4];
1686 const struct attribute_group *sysfs_rx_queue_group;
1688 const struct rtnl_link_ops *rtnl_link_ops;
1690 /* for setting kernel sock attribute on TCP connection setup */
1691 #define GSO_MAX_SIZE 65536
1692 unsigned int gso_max_size;
1693 #define GSO_MAX_SEGS 65535
1697 const struct dcbnl_rtnl_ops *dcbnl_ops;
1700 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1701 u8 prio_tc_map[TC_BITMASK + 1];
1703 #if IS_ENABLED(CONFIG_FCOE)
1704 unsigned int fcoe_ddp_xid;
1706 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1707 struct netprio_map __rcu *priomap;
1709 struct phy_device *phydev;
1710 struct lock_class_key *qdisc_tx_busylock;
1712 struct pm_qos_request pm_qos_req;
1714 #define to_net_dev(d) container_of(d, struct net_device, dev)
1716 #define NETDEV_ALIGN 32
1719 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1721 return dev->prio_tc_map[prio & TC_BITMASK];
1725 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1727 if (tc >= dev->num_tc)
1730 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1735 void netdev_reset_tc(struct net_device *dev)
1738 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1739 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1743 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1745 if (tc >= dev->num_tc)
1748 dev->tc_to_txq[tc].count = count;
1749 dev->tc_to_txq[tc].offset = offset;
1754 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1756 if (num_tc > TC_MAX_QUEUE)
1759 dev->num_tc = num_tc;
1764 int netdev_get_num_tc(struct net_device *dev)
1770 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1773 return &dev->_tx[index];
1776 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1777 const struct sk_buff *skb)
1779 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1782 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1783 void (*f)(struct net_device *,
1784 struct netdev_queue *,
1790 for (i = 0; i < dev->num_tx_queues; i++)
1791 f(dev, &dev->_tx[i], arg);
1794 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1795 struct sk_buff *skb,
1799 * Net namespace inlines
1802 struct net *dev_net(const struct net_device *dev)
1804 return read_pnet(&dev->nd_net);
1808 void dev_net_set(struct net_device *dev, struct net *net)
1810 #ifdef CONFIG_NET_NS
1811 release_net(dev->nd_net);
1812 dev->nd_net = hold_net(net);
1816 static inline bool netdev_uses_dsa(struct net_device *dev)
1818 #if IS_ENABLED(CONFIG_NET_DSA)
1819 if (dev->dsa_ptr != NULL)
1820 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1826 * netdev_priv - access network device private data
1827 * @dev: network device
1829 * Get network device private data
1831 static inline void *netdev_priv(const struct net_device *dev)
1833 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1836 /* Set the sysfs physical device reference for the network logical device
1837 * if set prior to registration will cause a symlink during initialization.
1839 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1841 /* Set the sysfs device type for the network logical device to allow
1842 * fine-grained identification of different network device types. For
1843 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1845 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1847 /* Default NAPI poll() weight
1848 * Device drivers are strongly advised to not use bigger value
1850 #define NAPI_POLL_WEIGHT 64
1853 * netif_napi_add - initialize a napi context
1854 * @dev: network device
1855 * @napi: napi context
1856 * @poll: polling function
1857 * @weight: default weight
1859 * netif_napi_add() must be used to initialize a napi context prior to calling
1860 * *any* of the other napi related functions.
1862 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1863 int (*poll)(struct napi_struct *, int), int weight);
1866 * netif_napi_del - remove a napi context
1867 * @napi: napi context
1869 * netif_napi_del() removes a napi context from the network device napi list
1871 void netif_napi_del(struct napi_struct *napi);
1873 struct napi_gro_cb {
1874 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1877 /* Length of frag0. */
1878 unsigned int frag0_len;
1880 /* This indicates where we are processing relative to skb->data. */
1883 /* This is non-zero if the packet cannot be merged with the new skb. */
1886 /* Save the IP ID here and check when we get to the transport layer */
1889 /* Number of segments aggregated. */
1892 /* This is non-zero if the packet may be of the same flow. */
1897 #define NAPI_GRO_FREE 1
1898 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1900 /* jiffies when first packet was created/queued */
1903 /* Used in ipv6_gro_receive() and foo-over-udp */
1906 /* Used in udp_gro_receive */
1909 /* GRO checksum is valid */
1912 /* Number of checksums via CHECKSUM_UNNECESSARY */
1915 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1918 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1921 /* used in skb_gro_receive() slow path */
1922 struct sk_buff *last;
1925 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1927 struct packet_type {
1928 __be16 type; /* This is really htons(ether_type). */
1929 struct net_device *dev; /* NULL is wildcarded here */
1930 int (*func) (struct sk_buff *,
1931 struct net_device *,
1932 struct packet_type *,
1933 struct net_device *);
1934 bool (*id_match)(struct packet_type *ptype,
1936 void *af_packet_priv;
1937 struct list_head list;
1940 struct offload_callbacks {
1941 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1942 netdev_features_t features);
1943 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1944 struct sk_buff *skb);
1945 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1948 struct packet_offload {
1949 __be16 type; /* This is really htons(ether_type). */
1950 struct offload_callbacks callbacks;
1951 struct list_head list;
1954 struct udp_offload {
1957 struct offload_callbacks callbacks;
1960 /* often modified stats are per cpu, other are shared (netdev->stats) */
1961 struct pcpu_sw_netstats {
1966 struct u64_stats_sync syncp;
1969 #define netdev_alloc_pcpu_stats(type) \
1971 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1974 for_each_possible_cpu(i) { \
1975 typeof(type) *stat; \
1976 stat = per_cpu_ptr(pcpu_stats, i); \
1977 u64_stats_init(&stat->syncp); \
1983 #include <linux/notifier.h>
1985 /* netdevice notifier chain. Please remember to update the rtnetlink
1986 * notification exclusion list in rtnetlink_event() when adding new
1989 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1990 #define NETDEV_DOWN 0x0002
1991 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1992 detected a hardware crash and restarted
1993 - we can use this eg to kick tcp sessions
1995 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1996 #define NETDEV_REGISTER 0x0005
1997 #define NETDEV_UNREGISTER 0x0006
1998 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1999 #define NETDEV_CHANGEADDR 0x0008
2000 #define NETDEV_GOING_DOWN 0x0009
2001 #define NETDEV_CHANGENAME 0x000A
2002 #define NETDEV_FEAT_CHANGE 0x000B
2003 #define NETDEV_BONDING_FAILOVER 0x000C
2004 #define NETDEV_PRE_UP 0x000D
2005 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2006 #define NETDEV_POST_TYPE_CHANGE 0x000F
2007 #define NETDEV_POST_INIT 0x0010
2008 #define NETDEV_UNREGISTER_FINAL 0x0011
2009 #define NETDEV_RELEASE 0x0012
2010 #define NETDEV_NOTIFY_PEERS 0x0013
2011 #define NETDEV_JOIN 0x0014
2012 #define NETDEV_CHANGEUPPER 0x0015
2013 #define NETDEV_RESEND_IGMP 0x0016
2014 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2015 #define NETDEV_CHANGEINFODATA 0x0018
2017 int register_netdevice_notifier(struct notifier_block *nb);
2018 int unregister_netdevice_notifier(struct notifier_block *nb);
2020 struct netdev_notifier_info {
2021 struct net_device *dev;
2024 struct netdev_notifier_change_info {
2025 struct netdev_notifier_info info; /* must be first */
2026 unsigned int flags_changed;
2029 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2030 struct net_device *dev)
2035 static inline struct net_device *
2036 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2041 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2044 extern rwlock_t dev_base_lock; /* Device list lock */
2046 #define for_each_netdev(net, d) \
2047 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2048 #define for_each_netdev_reverse(net, d) \
2049 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2050 #define for_each_netdev_rcu(net, d) \
2051 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2052 #define for_each_netdev_safe(net, d, n) \
2053 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2054 #define for_each_netdev_continue(net, d) \
2055 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2056 #define for_each_netdev_continue_rcu(net, d) \
2057 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2058 #define for_each_netdev_in_bond_rcu(bond, slave) \
2059 for_each_netdev_rcu(&init_net, slave) \
2060 if (netdev_master_upper_dev_get_rcu(slave) == bond)
2061 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2063 static inline struct net_device *next_net_device(struct net_device *dev)
2065 struct list_head *lh;
2069 lh = dev->dev_list.next;
2070 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2073 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2075 struct list_head *lh;
2079 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2080 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2083 static inline struct net_device *first_net_device(struct net *net)
2085 return list_empty(&net->dev_base_head) ? NULL :
2086 net_device_entry(net->dev_base_head.next);
2089 static inline struct net_device *first_net_device_rcu(struct net *net)
2091 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2093 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2096 int netdev_boot_setup_check(struct net_device *dev);
2097 unsigned long netdev_boot_base(const char *prefix, int unit);
2098 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2099 const char *hwaddr);
2100 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2101 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2102 void dev_add_pack(struct packet_type *pt);
2103 void dev_remove_pack(struct packet_type *pt);
2104 void __dev_remove_pack(struct packet_type *pt);
2105 void dev_add_offload(struct packet_offload *po);
2106 void dev_remove_offload(struct packet_offload *po);
2108 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2109 unsigned short mask);
2110 struct net_device *dev_get_by_name(struct net *net, const char *name);
2111 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2112 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2113 int dev_alloc_name(struct net_device *dev, const char *name);
2114 int dev_open(struct net_device *dev);
2115 int dev_close(struct net_device *dev);
2116 void dev_disable_lro(struct net_device *dev);
2117 int dev_loopback_xmit(struct sk_buff *newskb);
2118 int dev_queue_xmit(struct sk_buff *skb);
2119 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2120 int register_netdevice(struct net_device *dev);
2121 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2122 void unregister_netdevice_many(struct list_head *head);
2123 static inline void unregister_netdevice(struct net_device *dev)
2125 unregister_netdevice_queue(dev, NULL);
2128 int netdev_refcnt_read(const struct net_device *dev);
2129 void free_netdev(struct net_device *dev);
2130 void netdev_freemem(struct net_device *dev);
2131 void synchronize_net(void);
2132 int init_dummy_netdev(struct net_device *dev);
2134 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2135 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2136 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2137 int netdev_get_name(struct net *net, char *name, int ifindex);
2138 int dev_restart(struct net_device *dev);
2139 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2141 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2143 return NAPI_GRO_CB(skb)->data_offset;
2146 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2148 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2151 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2153 NAPI_GRO_CB(skb)->data_offset += len;
2156 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2157 unsigned int offset)
2159 return NAPI_GRO_CB(skb)->frag0 + offset;
2162 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2164 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2167 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2168 unsigned int offset)
2170 if (!pskb_may_pull(skb, hlen))
2173 NAPI_GRO_CB(skb)->frag0 = NULL;
2174 NAPI_GRO_CB(skb)->frag0_len = 0;
2175 return skb->data + offset;
2178 static inline void *skb_gro_network_header(struct sk_buff *skb)
2180 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2181 skb_network_offset(skb);
2184 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2185 const void *start, unsigned int len)
2187 if (NAPI_GRO_CB(skb)->csum_valid)
2188 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2189 csum_partial(start, len, 0));
2192 /* GRO checksum functions. These are logical equivalents of the normal
2193 * checksum functions (in skbuff.h) except that they operate on the GRO
2194 * offsets and fields in sk_buff.
2197 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2199 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2203 return (skb->ip_summed != CHECKSUM_PARTIAL &&
2204 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2205 (!zero_okay || check));
2208 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2211 if (NAPI_GRO_CB(skb)->csum_valid &&
2212 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2215 NAPI_GRO_CB(skb)->csum = psum;
2217 return __skb_gro_checksum_complete(skb);
2220 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2222 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2223 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2224 NAPI_GRO_CB(skb)->csum_cnt--;
2226 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2227 * verified a new top level checksum or an encapsulated one
2228 * during GRO. This saves work if we fallback to normal path.
2230 __skb_incr_checksum_unnecessary(skb);
2234 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2237 __sum16 __ret = 0; \
2238 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2239 __ret = __skb_gro_checksum_validate_complete(skb, \
2240 compute_pseudo(skb, proto)); \
2242 __skb_mark_checksum_bad(skb); \
2244 skb_gro_incr_csum_unnecessary(skb); \
2248 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2249 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2251 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2253 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2255 #define skb_gro_checksum_simple_validate(skb) \
2256 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2258 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2260 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2261 !NAPI_GRO_CB(skb)->csum_valid);
2264 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2265 __sum16 check, __wsum pseudo)
2267 NAPI_GRO_CB(skb)->csum = ~pseudo;
2268 NAPI_GRO_CB(skb)->csum_valid = 1;
2271 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2273 if (__skb_gro_checksum_convert_check(skb)) \
2274 __skb_gro_checksum_convert(skb, check, \
2275 compute_pseudo(skb, proto)); \
2278 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2279 unsigned short type,
2280 const void *daddr, const void *saddr,
2283 if (!dev->header_ops || !dev->header_ops->create)
2286 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2289 static inline int dev_parse_header(const struct sk_buff *skb,
2290 unsigned char *haddr)
2292 const struct net_device *dev = skb->dev;
2294 if (!dev->header_ops || !dev->header_ops->parse)
2296 return dev->header_ops->parse(skb, haddr);
2299 static inline int dev_rebuild_header(struct sk_buff *skb)
2301 const struct net_device *dev = skb->dev;
2303 if (!dev->header_ops || !dev->header_ops->rebuild)
2305 return dev->header_ops->rebuild(skb);
2308 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2309 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2310 static inline int unregister_gifconf(unsigned int family)
2312 return register_gifconf(family, NULL);
2315 #ifdef CONFIG_NET_FLOW_LIMIT
2316 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2317 struct sd_flow_limit {
2319 unsigned int num_buckets;
2320 unsigned int history_head;
2321 u16 history[FLOW_LIMIT_HISTORY];
2325 extern int netdev_flow_limit_table_len;
2326 #endif /* CONFIG_NET_FLOW_LIMIT */
2329 * Incoming packets are placed on per-cpu queues
2331 struct softnet_data {
2332 struct list_head poll_list;
2333 struct sk_buff_head process_queue;
2336 unsigned int processed;
2337 unsigned int time_squeeze;
2338 unsigned int cpu_collision;
2339 unsigned int received_rps;
2341 struct softnet_data *rps_ipi_list;
2343 #ifdef CONFIG_NET_FLOW_LIMIT
2344 struct sd_flow_limit __rcu *flow_limit;
2346 struct Qdisc *output_queue;
2347 struct Qdisc **output_queue_tailp;
2348 struct sk_buff *completion_queue;
2351 /* Elements below can be accessed between CPUs for RPS */
2352 struct call_single_data csd ____cacheline_aligned_in_smp;
2353 struct softnet_data *rps_ipi_next;
2355 unsigned int input_queue_head;
2356 unsigned int input_queue_tail;
2358 unsigned int dropped;
2359 struct sk_buff_head input_pkt_queue;
2360 struct napi_struct backlog;
2364 static inline void input_queue_head_incr(struct softnet_data *sd)
2367 sd->input_queue_head++;
2371 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2372 unsigned int *qtail)
2375 *qtail = ++sd->input_queue_tail;
2379 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2381 void __netif_schedule(struct Qdisc *q);
2382 void netif_schedule_queue(struct netdev_queue *txq);
2384 static inline void netif_tx_schedule_all(struct net_device *dev)
2388 for (i = 0; i < dev->num_tx_queues; i++)
2389 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2392 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2394 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2398 * netif_start_queue - allow transmit
2399 * @dev: network device
2401 * Allow upper layers to call the device hard_start_xmit routine.
2403 static inline void netif_start_queue(struct net_device *dev)
2405 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2408 static inline void netif_tx_start_all_queues(struct net_device *dev)
2412 for (i = 0; i < dev->num_tx_queues; i++) {
2413 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2414 netif_tx_start_queue(txq);
2418 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2421 * netif_wake_queue - restart transmit
2422 * @dev: network device
2424 * Allow upper layers to call the device hard_start_xmit routine.
2425 * Used for flow control when transmit resources are available.
2427 static inline void netif_wake_queue(struct net_device *dev)
2429 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2432 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2436 for (i = 0; i < dev->num_tx_queues; i++) {
2437 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2438 netif_tx_wake_queue(txq);
2442 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2444 if (WARN_ON(!dev_queue)) {
2445 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2448 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2452 * netif_stop_queue - stop transmitted packets
2453 * @dev: network device
2455 * Stop upper layers calling the device hard_start_xmit routine.
2456 * Used for flow control when transmit resources are unavailable.
2458 static inline void netif_stop_queue(struct net_device *dev)
2460 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2463 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2467 for (i = 0; i < dev->num_tx_queues; i++) {
2468 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2469 netif_tx_stop_queue(txq);
2473 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2475 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2479 * netif_queue_stopped - test if transmit queue is flowblocked
2480 * @dev: network device
2482 * Test if transmit queue on device is currently unable to send.
2484 static inline bool netif_queue_stopped(const struct net_device *dev)
2486 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2489 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2491 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2495 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2497 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2501 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2503 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2507 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2508 * @dev_queue: pointer to transmit queue
2510 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2511 * to give appropriate hint to the cpu.
2513 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2516 prefetchw(&dev_queue->dql.num_queued);
2521 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2522 * @dev_queue: pointer to transmit queue
2524 * BQL enabled drivers might use this helper in their TX completion path,
2525 * to give appropriate hint to the cpu.
2527 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2530 prefetchw(&dev_queue->dql.limit);
2534 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2538 dql_queued(&dev_queue->dql, bytes);
2540 if (likely(dql_avail(&dev_queue->dql) >= 0))
2543 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2546 * The XOFF flag must be set before checking the dql_avail below,
2547 * because in netdev_tx_completed_queue we update the dql_completed
2548 * before checking the XOFF flag.
2552 /* check again in case another CPU has just made room avail */
2553 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2554 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2559 * netdev_sent_queue - report the number of bytes queued to hardware
2560 * @dev: network device
2561 * @bytes: number of bytes queued to the hardware device queue
2563 * Report the number of bytes queued for sending/completion to the network
2564 * device hardware queue. @bytes should be a good approximation and should
2565 * exactly match netdev_completed_queue() @bytes
2567 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2569 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2572 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2573 unsigned int pkts, unsigned int bytes)
2576 if (unlikely(!bytes))
2579 dql_completed(&dev_queue->dql, bytes);
2582 * Without the memory barrier there is a small possiblity that
2583 * netdev_tx_sent_queue will miss the update and cause the queue to
2584 * be stopped forever
2588 if (dql_avail(&dev_queue->dql) < 0)
2591 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2592 netif_schedule_queue(dev_queue);
2597 * netdev_completed_queue - report bytes and packets completed by device
2598 * @dev: network device
2599 * @pkts: actual number of packets sent over the medium
2600 * @bytes: actual number of bytes sent over the medium
2602 * Report the number of bytes and packets transmitted by the network device
2603 * hardware queue over the physical medium, @bytes must exactly match the
2604 * @bytes amount passed to netdev_sent_queue()
2606 static inline void netdev_completed_queue(struct net_device *dev,
2607 unsigned int pkts, unsigned int bytes)
2609 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2612 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2615 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2621 * netdev_reset_queue - reset the packets and bytes count of a network device
2622 * @dev_queue: network device
2624 * Reset the bytes and packet count of a network device and clear the
2625 * software flow control OFF bit for this network device
2627 static inline void netdev_reset_queue(struct net_device *dev_queue)
2629 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2633 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2634 * @dev: network device
2635 * @queue_index: given tx queue index
2637 * Returns 0 if given tx queue index >= number of device tx queues,
2638 * otherwise returns the originally passed tx queue index.
2640 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2642 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2643 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2644 dev->name, queue_index,
2645 dev->real_num_tx_queues);
2653 * netif_running - test if up
2654 * @dev: network device
2656 * Test if the device has been brought up.
2658 static inline bool netif_running(const struct net_device *dev)
2660 return test_bit(__LINK_STATE_START, &dev->state);
2664 * Routines to manage the subqueues on a device. We only need start
2665 * stop, and a check if it's stopped. All other device management is
2666 * done at the overall netdevice level.
2667 * Also test the device if we're multiqueue.
2671 * netif_start_subqueue - allow sending packets on subqueue
2672 * @dev: network device
2673 * @queue_index: sub queue index
2675 * Start individual transmit queue of a device with multiple transmit queues.
2677 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2679 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2681 netif_tx_start_queue(txq);
2685 * netif_stop_subqueue - stop sending packets on subqueue
2686 * @dev: network device
2687 * @queue_index: sub queue index
2689 * Stop individual transmit queue of a device with multiple transmit queues.
2691 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2693 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2694 netif_tx_stop_queue(txq);
2698 * netif_subqueue_stopped - test status of subqueue
2699 * @dev: network device
2700 * @queue_index: sub queue index
2702 * Check individual transmit queue of a device with multiple transmit queues.
2704 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2707 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2709 return netif_tx_queue_stopped(txq);
2712 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2713 struct sk_buff *skb)
2715 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2718 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2721 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2724 static inline int netif_set_xps_queue(struct net_device *dev,
2725 const struct cpumask *mask,
2733 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2734 * as a distribution range limit for the returned value.
2736 static inline u16 skb_tx_hash(const struct net_device *dev,
2737 struct sk_buff *skb)
2739 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2743 * netif_is_multiqueue - test if device has multiple transmit queues
2744 * @dev: network device
2746 * Check if device has multiple transmit queues
2748 static inline bool netif_is_multiqueue(const struct net_device *dev)
2750 return dev->num_tx_queues > 1;
2753 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2756 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2758 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2766 static inline unsigned int get_netdev_rx_queue_index(
2767 struct netdev_rx_queue *queue)
2769 struct net_device *dev = queue->dev;
2770 int index = queue - dev->_rx;
2772 BUG_ON(index >= dev->num_rx_queues);
2777 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2778 int netif_get_num_default_rss_queues(void);
2780 enum skb_free_reason {
2781 SKB_REASON_CONSUMED,
2785 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2786 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2789 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2790 * interrupt context or with hardware interrupts being disabled.
2791 * (in_irq() || irqs_disabled())
2793 * We provide four helpers that can be used in following contexts :
2795 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2796 * replacing kfree_skb(skb)
2798 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2799 * Typically used in place of consume_skb(skb) in TX completion path
2801 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2802 * replacing kfree_skb(skb)
2804 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2805 * and consumed a packet. Used in place of consume_skb(skb)
2807 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2809 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2812 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2814 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2817 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2819 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2822 static inline void dev_consume_skb_any(struct sk_buff *skb)
2824 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2827 int netif_rx(struct sk_buff *skb);
2828 int netif_rx_ni(struct sk_buff *skb);
2829 int netif_receive_skb(struct sk_buff *skb);
2830 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2831 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2832 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2833 gro_result_t napi_gro_frags(struct napi_struct *napi);
2834 struct packet_offload *gro_find_receive_by_type(__be16 type);
2835 struct packet_offload *gro_find_complete_by_type(__be16 type);
2837 static inline void napi_free_frags(struct napi_struct *napi)
2839 kfree_skb(napi->skb);
2843 int netdev_rx_handler_register(struct net_device *dev,
2844 rx_handler_func_t *rx_handler,
2845 void *rx_handler_data);
2846 void netdev_rx_handler_unregister(struct net_device *dev);
2848 bool dev_valid_name(const char *name);
2849 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2850 int dev_ethtool(struct net *net, struct ifreq *);
2851 unsigned int dev_get_flags(const struct net_device *);
2852 int __dev_change_flags(struct net_device *, unsigned int flags);
2853 int dev_change_flags(struct net_device *, unsigned int);
2854 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2855 unsigned int gchanges);
2856 int dev_change_name(struct net_device *, const char *);
2857 int dev_set_alias(struct net_device *, const char *, size_t);
2858 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2859 int dev_set_mtu(struct net_device *, int);
2860 void dev_set_group(struct net_device *, int);
2861 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2862 int dev_change_carrier(struct net_device *, bool new_carrier);
2863 int dev_get_phys_port_id(struct net_device *dev,
2864 struct netdev_phys_port_id *ppid);
2865 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2866 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2867 struct netdev_queue *txq, int *ret);
2868 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2869 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2870 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2872 extern int netdev_budget;
2874 /* Called by rtnetlink.c:rtnl_unlock() */
2875 void netdev_run_todo(void);
2878 * dev_put - release reference to device
2879 * @dev: network device
2881 * Release reference to device to allow it to be freed.
2883 static inline void dev_put(struct net_device *dev)
2885 this_cpu_dec(*dev->pcpu_refcnt);
2889 * dev_hold - get reference to device
2890 * @dev: network device
2892 * Hold reference to device to keep it from being freed.
2894 static inline void dev_hold(struct net_device *dev)
2896 this_cpu_inc(*dev->pcpu_refcnt);
2899 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2900 * and _off may be called from IRQ context, but it is caller
2901 * who is responsible for serialization of these calls.
2903 * The name carrier is inappropriate, these functions should really be
2904 * called netif_lowerlayer_*() because they represent the state of any
2905 * kind of lower layer not just hardware media.
2908 void linkwatch_init_dev(struct net_device *dev);
2909 void linkwatch_fire_event(struct net_device *dev);
2910 void linkwatch_forget_dev(struct net_device *dev);
2913 * netif_carrier_ok - test if carrier present
2914 * @dev: network device
2916 * Check if carrier is present on device
2918 static inline bool netif_carrier_ok(const struct net_device *dev)
2920 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2923 unsigned long dev_trans_start(struct net_device *dev);
2925 void __netdev_watchdog_up(struct net_device *dev);
2927 void netif_carrier_on(struct net_device *dev);
2929 void netif_carrier_off(struct net_device *dev);
2932 * netif_dormant_on - mark device as dormant.
2933 * @dev: network device
2935 * Mark device as dormant (as per RFC2863).
2937 * The dormant state indicates that the relevant interface is not
2938 * actually in a condition to pass packets (i.e., it is not 'up') but is
2939 * in a "pending" state, waiting for some external event. For "on-
2940 * demand" interfaces, this new state identifies the situation where the
2941 * interface is waiting for events to place it in the up state.
2944 static inline void netif_dormant_on(struct net_device *dev)
2946 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2947 linkwatch_fire_event(dev);
2951 * netif_dormant_off - set device as not dormant.
2952 * @dev: network device
2954 * Device is not in dormant state.
2956 static inline void netif_dormant_off(struct net_device *dev)
2958 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2959 linkwatch_fire_event(dev);
2963 * netif_dormant - test if carrier present
2964 * @dev: network device
2966 * Check if carrier is present on device
2968 static inline bool netif_dormant(const struct net_device *dev)
2970 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2975 * netif_oper_up - test if device is operational
2976 * @dev: network device
2978 * Check if carrier is operational
2980 static inline bool netif_oper_up(const struct net_device *dev)
2982 return (dev->operstate == IF_OPER_UP ||
2983 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2987 * netif_device_present - is device available or removed
2988 * @dev: network device
2990 * Check if device has not been removed from system.
2992 static inline bool netif_device_present(struct net_device *dev)
2994 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2997 void netif_device_detach(struct net_device *dev);
2999 void netif_device_attach(struct net_device *dev);
3002 * Network interface message level settings
3006 NETIF_MSG_DRV = 0x0001,
3007 NETIF_MSG_PROBE = 0x0002,
3008 NETIF_MSG_LINK = 0x0004,
3009 NETIF_MSG_TIMER = 0x0008,
3010 NETIF_MSG_IFDOWN = 0x0010,
3011 NETIF_MSG_IFUP = 0x0020,
3012 NETIF_MSG_RX_ERR = 0x0040,
3013 NETIF_MSG_TX_ERR = 0x0080,
3014 NETIF_MSG_TX_QUEUED = 0x0100,
3015 NETIF_MSG_INTR = 0x0200,
3016 NETIF_MSG_TX_DONE = 0x0400,
3017 NETIF_MSG_RX_STATUS = 0x0800,
3018 NETIF_MSG_PKTDATA = 0x1000,
3019 NETIF_MSG_HW = 0x2000,
3020 NETIF_MSG_WOL = 0x4000,
3023 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3024 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3025 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3026 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3027 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3028 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3029 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3030 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3031 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3032 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3033 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3034 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3035 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3036 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3037 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3039 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3042 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3043 return default_msg_enable_bits;
3044 if (debug_value == 0) /* no output */
3046 /* set low N bits */
3047 return (1 << debug_value) - 1;
3050 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3052 spin_lock(&txq->_xmit_lock);
3053 txq->xmit_lock_owner = cpu;
3056 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3058 spin_lock_bh(&txq->_xmit_lock);
3059 txq->xmit_lock_owner = smp_processor_id();
3062 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3064 bool ok = spin_trylock(&txq->_xmit_lock);
3066 txq->xmit_lock_owner = smp_processor_id();
3070 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3072 txq->xmit_lock_owner = -1;
3073 spin_unlock(&txq->_xmit_lock);
3076 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3078 txq->xmit_lock_owner = -1;
3079 spin_unlock_bh(&txq->_xmit_lock);
3082 static inline void txq_trans_update(struct netdev_queue *txq)
3084 if (txq->xmit_lock_owner != -1)
3085 txq->trans_start = jiffies;
3089 * netif_tx_lock - grab network device transmit lock
3090 * @dev: network device
3092 * Get network device transmit lock
3094 static inline void netif_tx_lock(struct net_device *dev)
3099 spin_lock(&dev->tx_global_lock);
3100 cpu = smp_processor_id();
3101 for (i = 0; i < dev->num_tx_queues; i++) {
3102 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3104 /* We are the only thread of execution doing a
3105 * freeze, but we have to grab the _xmit_lock in
3106 * order to synchronize with threads which are in
3107 * the ->hard_start_xmit() handler and already
3108 * checked the frozen bit.
3110 __netif_tx_lock(txq, cpu);
3111 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3112 __netif_tx_unlock(txq);
3116 static inline void netif_tx_lock_bh(struct net_device *dev)
3122 static inline void netif_tx_unlock(struct net_device *dev)
3126 for (i = 0; i < dev->num_tx_queues; i++) {
3127 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3129 /* No need to grab the _xmit_lock here. If the
3130 * queue is not stopped for another reason, we
3133 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3134 netif_schedule_queue(txq);
3136 spin_unlock(&dev->tx_global_lock);
3139 static inline void netif_tx_unlock_bh(struct net_device *dev)
3141 netif_tx_unlock(dev);
3145 #define HARD_TX_LOCK(dev, txq, cpu) { \
3146 if ((dev->features & NETIF_F_LLTX) == 0) { \
3147 __netif_tx_lock(txq, cpu); \
3151 #define HARD_TX_TRYLOCK(dev, txq) \
3152 (((dev->features & NETIF_F_LLTX) == 0) ? \
3153 __netif_tx_trylock(txq) : \
3156 #define HARD_TX_UNLOCK(dev, txq) { \
3157 if ((dev->features & NETIF_F_LLTX) == 0) { \
3158 __netif_tx_unlock(txq); \
3162 static inline void netif_tx_disable(struct net_device *dev)
3168 cpu = smp_processor_id();
3169 for (i = 0; i < dev->num_tx_queues; i++) {
3170 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3172 __netif_tx_lock(txq, cpu);
3173 netif_tx_stop_queue(txq);
3174 __netif_tx_unlock(txq);
3179 static inline void netif_addr_lock(struct net_device *dev)
3181 spin_lock(&dev->addr_list_lock);
3184 static inline void netif_addr_lock_nested(struct net_device *dev)
3186 int subclass = SINGLE_DEPTH_NESTING;
3188 if (dev->netdev_ops->ndo_get_lock_subclass)
3189 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3191 spin_lock_nested(&dev->addr_list_lock, subclass);
3194 static inline void netif_addr_lock_bh(struct net_device *dev)
3196 spin_lock_bh(&dev->addr_list_lock);
3199 static inline void netif_addr_unlock(struct net_device *dev)
3201 spin_unlock(&dev->addr_list_lock);
3204 static inline void netif_addr_unlock_bh(struct net_device *dev)
3206 spin_unlock_bh(&dev->addr_list_lock);
3210 * dev_addrs walker. Should be used only for read access. Call with
3211 * rcu_read_lock held.
3213 #define for_each_dev_addr(dev, ha) \
3214 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3216 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3218 void ether_setup(struct net_device *dev);
3220 /* Support for loadable net-drivers */
3221 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3222 unsigned char name_assign_type,
3223 void (*setup)(struct net_device *),
3224 unsigned int txqs, unsigned int rxqs);
3225 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3226 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3228 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3229 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3232 int register_netdev(struct net_device *dev);
3233 void unregister_netdev(struct net_device *dev);
3235 /* General hardware address lists handling functions */
3236 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3237 struct netdev_hw_addr_list *from_list, int addr_len);
3238 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3239 struct netdev_hw_addr_list *from_list, int addr_len);
3240 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3241 struct net_device *dev,
3242 int (*sync)(struct net_device *, const unsigned char *),
3243 int (*unsync)(struct net_device *,
3244 const unsigned char *));
3245 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3246 struct net_device *dev,
3247 int (*unsync)(struct net_device *,
3248 const unsigned char *));
3249 void __hw_addr_init(struct netdev_hw_addr_list *list);
3251 /* Functions used for device addresses handling */
3252 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3253 unsigned char addr_type);
3254 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3255 unsigned char addr_type);
3256 void dev_addr_flush(struct net_device *dev);
3257 int dev_addr_init(struct net_device *dev);
3259 /* Functions used for unicast addresses handling */
3260 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3261 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3262 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3263 int dev_uc_sync(struct net_device *to, struct net_device *from);
3264 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3265 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3266 void dev_uc_flush(struct net_device *dev);
3267 void dev_uc_init(struct net_device *dev);
3270 * __dev_uc_sync - Synchonize device's unicast list
3271 * @dev: device to sync
3272 * @sync: function to call if address should be added
3273 * @unsync: function to call if address should be removed
3275 * Add newly added addresses to the interface, and release
3276 * addresses that have been deleted.
3278 static inline int __dev_uc_sync(struct net_device *dev,
3279 int (*sync)(struct net_device *,
3280 const unsigned char *),
3281 int (*unsync)(struct net_device *,
3282 const unsigned char *))
3284 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3288 * __dev_uc_unsync - Remove synchronized addresses from device
3289 * @dev: device to sync
3290 * @unsync: function to call if address should be removed
3292 * Remove all addresses that were added to the device by dev_uc_sync().
3294 static inline void __dev_uc_unsync(struct net_device *dev,
3295 int (*unsync)(struct net_device *,
3296 const unsigned char *))
3298 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3301 /* Functions used for multicast addresses handling */
3302 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3303 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3304 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3305 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3306 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3307 int dev_mc_sync(struct net_device *to, struct net_device *from);
3308 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3309 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3310 void dev_mc_flush(struct net_device *dev);
3311 void dev_mc_init(struct net_device *dev);
3314 * __dev_mc_sync - Synchonize device's multicast list
3315 * @dev: device to sync
3316 * @sync: function to call if address should be added
3317 * @unsync: function to call if address should be removed
3319 * Add newly added addresses to the interface, and release
3320 * addresses that have been deleted.
3322 static inline int __dev_mc_sync(struct net_device *dev,
3323 int (*sync)(struct net_device *,
3324 const unsigned char *),
3325 int (*unsync)(struct net_device *,
3326 const unsigned char *))
3328 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3332 * __dev_mc_unsync - Remove synchronized addresses from device
3333 * @dev: device to sync
3334 * @unsync: function to call if address should be removed
3336 * Remove all addresses that were added to the device by dev_mc_sync().
3338 static inline void __dev_mc_unsync(struct net_device *dev,
3339 int (*unsync)(struct net_device *,
3340 const unsigned char *))
3342 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3345 /* Functions used for secondary unicast and multicast support */
3346 void dev_set_rx_mode(struct net_device *dev);
3347 void __dev_set_rx_mode(struct net_device *dev);
3348 int dev_set_promiscuity(struct net_device *dev, int inc);
3349 int dev_set_allmulti(struct net_device *dev, int inc);
3350 void netdev_state_change(struct net_device *dev);
3351 void netdev_notify_peers(struct net_device *dev);
3352 void netdev_features_change(struct net_device *dev);
3353 /* Load a device via the kmod */
3354 void dev_load(struct net *net, const char *name);
3355 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3356 struct rtnl_link_stats64 *storage);
3357 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3358 const struct net_device_stats *netdev_stats);
3360 extern int netdev_max_backlog;
3361 extern int netdev_tstamp_prequeue;
3362 extern int weight_p;
3363 extern int bpf_jit_enable;
3365 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3366 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3367 struct list_head **iter);
3368 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3369 struct list_head **iter);
3371 /* iterate through upper list, must be called under RCU read lock */
3372 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3373 for (iter = &(dev)->adj_list.upper, \
3374 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3376 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3378 /* iterate through upper list, must be called under RCU read lock */
3379 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3380 for (iter = &(dev)->all_adj_list.upper, \
3381 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3383 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3385 void *netdev_lower_get_next_private(struct net_device *dev,
3386 struct list_head **iter);
3387 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3388 struct list_head **iter);
3390 #define netdev_for_each_lower_private(dev, priv, iter) \
3391 for (iter = (dev)->adj_list.lower.next, \
3392 priv = netdev_lower_get_next_private(dev, &(iter)); \
3394 priv = netdev_lower_get_next_private(dev, &(iter)))
3396 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3397 for (iter = &(dev)->adj_list.lower, \
3398 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3400 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3402 void *netdev_lower_get_next(struct net_device *dev,
3403 struct list_head **iter);
3404 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3405 for (iter = &(dev)->adj_list.lower, \
3406 ldev = netdev_lower_get_next(dev, &(iter)); \
3408 ldev = netdev_lower_get_next(dev, &(iter)))
3410 void *netdev_adjacent_get_private(struct list_head *adj_list);
3411 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3412 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3413 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3414 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3415 int netdev_master_upper_dev_link(struct net_device *dev,
3416 struct net_device *upper_dev);
3417 int netdev_master_upper_dev_link_private(struct net_device *dev,
3418 struct net_device *upper_dev,
3420 void netdev_upper_dev_unlink(struct net_device *dev,
3421 struct net_device *upper_dev);
3422 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3423 void *netdev_lower_dev_get_private(struct net_device *dev,
3424 struct net_device *lower_dev);
3425 int dev_get_nest_level(struct net_device *dev,
3426 bool (*type_check)(struct net_device *dev));
3427 int skb_checksum_help(struct sk_buff *skb);
3428 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3429 netdev_features_t features, bool tx_path);
3430 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3431 netdev_features_t features);
3434 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3436 return __skb_gso_segment(skb, features, true);
3438 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3440 static inline bool can_checksum_protocol(netdev_features_t features,
3443 return ((features & NETIF_F_GEN_CSUM) ||
3444 ((features & NETIF_F_V4_CSUM) &&
3445 protocol == htons(ETH_P_IP)) ||
3446 ((features & NETIF_F_V6_CSUM) &&
3447 protocol == htons(ETH_P_IPV6)) ||
3448 ((features & NETIF_F_FCOE_CRC) &&
3449 protocol == htons(ETH_P_FCOE)));
3453 void netdev_rx_csum_fault(struct net_device *dev);
3455 static inline void netdev_rx_csum_fault(struct net_device *dev)
3459 /* rx skb timestamps */
3460 void net_enable_timestamp(void);
3461 void net_disable_timestamp(void);
3463 #ifdef CONFIG_PROC_FS
3464 int __init dev_proc_init(void);
3466 #define dev_proc_init() 0
3469 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3470 struct sk_buff *skb, struct net_device *dev,
3473 skb->xmit_more = more ? 1 : 0;
3474 return ops->ndo_start_xmit(skb, dev);
3477 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3478 struct netdev_queue *txq, bool more)
3480 const struct net_device_ops *ops = dev->netdev_ops;
3483 rc = __netdev_start_xmit(ops, skb, dev, more);
3484 if (rc == NETDEV_TX_OK)
3485 txq_trans_update(txq);
3490 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3492 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3495 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3497 return netdev_class_create_file_ns(class_attr, NULL);
3500 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3502 netdev_class_remove_file_ns(class_attr, NULL);
3505 extern struct kobj_ns_type_operations net_ns_type_operations;
3507 const char *netdev_drivername(const struct net_device *dev);
3509 void linkwatch_run_queue(void);
3511 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3512 netdev_features_t f2)
3514 if (f1 & NETIF_F_GEN_CSUM)
3515 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3516 if (f2 & NETIF_F_GEN_CSUM)
3517 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3519 if (f1 & NETIF_F_GEN_CSUM)
3520 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3525 static inline netdev_features_t netdev_get_wanted_features(
3526 struct net_device *dev)
3528 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3530 netdev_features_t netdev_increment_features(netdev_features_t all,
3531 netdev_features_t one, netdev_features_t mask);
3533 /* Allow TSO being used on stacked device :
3534 * Performing the GSO segmentation before last device
3535 * is a performance improvement.
3537 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3538 netdev_features_t mask)
3540 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3543 int __netdev_update_features(struct net_device *dev);
3544 void netdev_update_features(struct net_device *dev);
3545 void netdev_change_features(struct net_device *dev);
3547 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3548 struct net_device *dev);
3550 netdev_features_t netif_skb_features(struct sk_buff *skb);
3552 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3554 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3556 /* check flags correspondence */
3557 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3558 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3559 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3560 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3561 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3562 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3563 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3564 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3565 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3566 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3567 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3568 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3569 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3571 return (features & feature) == feature;
3574 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3576 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3577 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3580 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3581 netdev_features_t features)
3583 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3584 (dev->netdev_ops->ndo_gso_check &&
3585 !dev->netdev_ops->ndo_gso_check(skb, dev)) ||
3586 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3587 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3590 static inline void netif_set_gso_max_size(struct net_device *dev,
3593 dev->gso_max_size = size;
3596 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3597 int pulled_hlen, u16 mac_offset,
3600 skb->protocol = protocol;
3601 skb->encapsulation = 1;
3602 skb_push(skb, pulled_hlen);
3603 skb_reset_transport_header(skb);
3604 skb->mac_header = mac_offset;
3605 skb->network_header = skb->mac_header + mac_len;
3606 skb->mac_len = mac_len;
3609 static inline bool netif_is_macvlan(struct net_device *dev)
3611 return dev->priv_flags & IFF_MACVLAN;
3614 static inline bool netif_is_bond_master(struct net_device *dev)
3616 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3619 static inline bool netif_is_bond_slave(struct net_device *dev)
3621 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3624 static inline bool netif_supports_nofcs(struct net_device *dev)
3626 return dev->priv_flags & IFF_SUPP_NOFCS;
3629 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3630 static inline void netif_keep_dst(struct net_device *dev)
3632 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3635 extern struct pernet_operations __net_initdata loopback_net_ops;
3637 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3639 /* netdev_printk helpers, similar to dev_printk */
3641 static inline const char *netdev_name(const struct net_device *dev)
3643 if (!dev->name[0] || strchr(dev->name, '%'))
3644 return "(unnamed net_device)";
3648 static inline const char *netdev_reg_state(const struct net_device *dev)
3650 switch (dev->reg_state) {
3651 case NETREG_UNINITIALIZED: return " (uninitialized)";
3652 case NETREG_REGISTERED: return "";
3653 case NETREG_UNREGISTERING: return " (unregistering)";
3654 case NETREG_UNREGISTERED: return " (unregistered)";
3655 case NETREG_RELEASED: return " (released)";
3656 case NETREG_DUMMY: return " (dummy)";
3659 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3660 return " (unknown)";
3664 void netdev_printk(const char *level, const struct net_device *dev,
3665 const char *format, ...);
3667 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3669 void netdev_alert(const struct net_device *dev, const char *format, ...);
3671 void netdev_crit(const struct net_device *dev, const char *format, ...);
3673 void netdev_err(const struct net_device *dev, const char *format, ...);
3675 void netdev_warn(const struct net_device *dev, const char *format, ...);
3677 void netdev_notice(const struct net_device *dev, const char *format, ...);
3679 void netdev_info(const struct net_device *dev, const char *format, ...);
3681 #define MODULE_ALIAS_NETDEV(device) \
3682 MODULE_ALIAS("netdev-" device)
3684 #if defined(CONFIG_DYNAMIC_DEBUG)
3685 #define netdev_dbg(__dev, format, args...) \
3687 dynamic_netdev_dbg(__dev, format, ##args); \
3689 #elif defined(DEBUG)
3690 #define netdev_dbg(__dev, format, args...) \
3691 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3693 #define netdev_dbg(__dev, format, args...) \
3696 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3700 #if defined(VERBOSE_DEBUG)
3701 #define netdev_vdbg netdev_dbg
3704 #define netdev_vdbg(dev, format, args...) \
3707 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3713 * netdev_WARN() acts like dev_printk(), but with the key difference
3714 * of using a WARN/WARN_ON to get the message out, including the
3715 * file/line information and a backtrace.
3717 #define netdev_WARN(dev, format, args...) \
3718 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3719 netdev_reg_state(dev), ##args)
3721 /* netif printk helpers, similar to netdev_printk */
3723 #define netif_printk(priv, type, level, dev, fmt, args...) \
3725 if (netif_msg_##type(priv)) \
3726 netdev_printk(level, (dev), fmt, ##args); \
3729 #define netif_level(level, priv, type, dev, fmt, args...) \
3731 if (netif_msg_##type(priv)) \
3732 netdev_##level(dev, fmt, ##args); \
3735 #define netif_emerg(priv, type, dev, fmt, args...) \
3736 netif_level(emerg, priv, type, dev, fmt, ##args)
3737 #define netif_alert(priv, type, dev, fmt, args...) \
3738 netif_level(alert, priv, type, dev, fmt, ##args)
3739 #define netif_crit(priv, type, dev, fmt, args...) \
3740 netif_level(crit, priv, type, dev, fmt, ##args)
3741 #define netif_err(priv, type, dev, fmt, args...) \
3742 netif_level(err, priv, type, dev, fmt, ##args)
3743 #define netif_warn(priv, type, dev, fmt, args...) \
3744 netif_level(warn, priv, type, dev, fmt, ##args)
3745 #define netif_notice(priv, type, dev, fmt, args...) \
3746 netif_level(notice, priv, type, dev, fmt, ##args)
3747 #define netif_info(priv, type, dev, fmt, args...) \
3748 netif_level(info, priv, type, dev, fmt, ##args)
3750 #if defined(CONFIG_DYNAMIC_DEBUG)
3751 #define netif_dbg(priv, type, netdev, format, args...) \
3753 if (netif_msg_##type(priv)) \
3754 dynamic_netdev_dbg(netdev, format, ##args); \
3756 #elif defined(DEBUG)
3757 #define netif_dbg(priv, type, dev, format, args...) \
3758 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3760 #define netif_dbg(priv, type, dev, format, args...) \
3763 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3768 #if defined(VERBOSE_DEBUG)
3769 #define netif_vdbg netif_dbg
3771 #define netif_vdbg(priv, type, dev, format, args...) \
3774 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3780 * The list of packet types we will receive (as opposed to discard)
3781 * and the routines to invoke.
3783 * Why 16. Because with 16 the only overlap we get on a hash of the
3784 * low nibble of the protocol value is RARP/SNAP/X.25.
3786 * NOTE: That is no longer true with the addition of VLAN tags. Not
3787 * sure which should go first, but I bet it won't make much
3788 * difference if we are running VLANs. The good news is that
3789 * this protocol won't be in the list unless compiled in, so
3790 * the average user (w/out VLANs) will not be adversely affected.
3806 #define PTYPE_HASH_SIZE (16)
3807 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3809 #endif /* _LINUX_NETDEVICE_H */