2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
63 void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
66 /* Backlog congestion levels */
67 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
68 #define NET_RX_DROP 1 /* packet dropped */
71 * Transmit return codes: transmit return codes originate from three different
74 * - qdisc return codes
75 * - driver transmit return codes
78 * Drivers are allowed to return any one of those in their hard_start_xmit()
79 * function. Real network devices commonly used with qdiscs should only return
80 * the driver transmit return codes though - when qdiscs are used, the actual
81 * transmission happens asynchronously, so the value is not propagated to
82 * higher layers. Virtual network devices transmit synchronously, in this case
83 * the driver transmit return codes are consumed by dev_queue_xmit(), all
84 * others are propagated to higher layers.
87 /* qdisc ->enqueue() return codes. */
88 #define NET_XMIT_SUCCESS 0x00
89 #define NET_XMIT_DROP 0x01 /* skb dropped */
90 #define NET_XMIT_CN 0x02 /* congestion notification */
91 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
92 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
95 * indicates that the device will soon be dropping packets, or already drops
96 * some packets of the same priority; prompting us to send less aggressively. */
97 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
98 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100 /* Driver transmit return codes */
101 #define NETDEV_TX_MASK 0xf0
104 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
105 NETDEV_TX_OK = 0x00, /* driver took care of packet */
106 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
107 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109 typedef enum netdev_tx netdev_tx_t;
112 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
113 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 static inline bool dev_xmit_complete(int rc)
118 * Positive cases with an skb consumed by a driver:
119 * - successful transmission (rc == NETDEV_TX_OK)
120 * - error while transmitting (rc < 0)
121 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 if (likely(rc < NET_XMIT_MASK))
130 * Compute the worst case header length according to the protocols
134 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
135 # if defined(CONFIG_MAC80211_MESH)
136 # define LL_MAX_HEADER 128
138 # define LL_MAX_HEADER 96
141 # define LL_MAX_HEADER 32
144 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
145 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
146 #define MAX_HEADER LL_MAX_HEADER
148 #define MAX_HEADER (LL_MAX_HEADER + 48)
152 * Old network device statistics. Fields are native words
153 * (unsigned long) so they can be read and written atomically.
156 struct net_device_stats {
157 unsigned long rx_packets;
158 unsigned long tx_packets;
159 unsigned long rx_bytes;
160 unsigned long tx_bytes;
161 unsigned long rx_errors;
162 unsigned long tx_errors;
163 unsigned long rx_dropped;
164 unsigned long tx_dropped;
165 unsigned long multicast;
166 unsigned long collisions;
167 unsigned long rx_length_errors;
168 unsigned long rx_over_errors;
169 unsigned long rx_crc_errors;
170 unsigned long rx_frame_errors;
171 unsigned long rx_fifo_errors;
172 unsigned long rx_missed_errors;
173 unsigned long tx_aborted_errors;
174 unsigned long tx_carrier_errors;
175 unsigned long tx_fifo_errors;
176 unsigned long tx_heartbeat_errors;
177 unsigned long tx_window_errors;
178 unsigned long rx_compressed;
179 unsigned long tx_compressed;
183 #include <linux/cache.h>
184 #include <linux/skbuff.h>
187 #include <linux/static_key.h>
188 extern struct static_key rps_needed;
195 struct netdev_hw_addr {
196 struct list_head list;
197 unsigned char addr[MAX_ADDR_LEN];
199 #define NETDEV_HW_ADDR_T_LAN 1
200 #define NETDEV_HW_ADDR_T_SAN 2
201 #define NETDEV_HW_ADDR_T_SLAVE 3
202 #define NETDEV_HW_ADDR_T_UNICAST 4
203 #define NETDEV_HW_ADDR_T_MULTICAST 5
208 struct rcu_head rcu_head;
211 struct netdev_hw_addr_list {
212 struct list_head list;
216 #define netdev_hw_addr_list_count(l) ((l)->count)
217 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
218 #define netdev_hw_addr_list_for_each(ha, l) \
219 list_for_each_entry(ha, &(l)->list, list)
221 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
222 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
223 #define netdev_for_each_uc_addr(ha, dev) \
224 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
226 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
227 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
228 #define netdev_for_each_mc_addr(ha, dev) \
229 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236 /* cached hardware header; allow for machine alignment needs. */
237 #define HH_DATA_MOD 16
238 #define HH_DATA_OFF(__len) \
239 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
240 #define HH_DATA_ALIGN(__len) \
241 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
242 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
245 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * dev->hard_header_len ? (dev->hard_header_len +
248 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 * We could use other alignment values, but we must maintain the
251 * relationship HH alignment <= LL alignment.
253 #define LL_RESERVED_SPACE(dev) \
254 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
255 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
256 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
259 int (*create) (struct sk_buff *skb, struct net_device *dev,
260 unsigned short type, const void *daddr,
261 const void *saddr, unsigned int len);
262 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
263 int (*rebuild)(struct sk_buff *skb);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
275 enum netdev_state_t {
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
288 struct netdev_boot_setup {
292 #define NETDEV_BOOT_SETUP_MAX 8
294 int __init netdev_boot_setup(char *str);
297 * Structure for NAPI scheduling similar to tasklet but with weighting
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
306 struct list_head poll_list;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
316 struct net_device *dev;
317 struct sk_buff *gro_list;
319 struct list_head dev_list;
320 struct hlist_node napi_hash_node;
321 unsigned int napi_id;
325 NAPI_STATE_SCHED, /* Poll is scheduled */
326 NAPI_STATE_DISABLE, /* Disable pending */
327 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
328 NAPI_STATE_HASHED, /* In NAPI hash */
338 typedef enum gro_result gro_result_t;
341 * enum rx_handler_result - Possible return values for rx_handlers.
342 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
345 * case skb->dev was changed by rx_handler.
346 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
347 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 * rx_handlers are functions called from inside __netif_receive_skb(), to do
350 * special processing of the skb, prior to delivery to protocol handlers.
352 * Currently, a net_device can only have a single rx_handler registered. Trying
353 * to register a second rx_handler will return -EBUSY.
355 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
356 * To unregister a rx_handler on a net_device, use
357 * netdev_rx_handler_unregister().
359 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
362 * If the rx_handler consumed to skb in some way, it should return
363 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
364 * the skb to be delivered in some other ways.
366 * If the rx_handler changed skb->dev, to divert the skb to another
367 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
368 * new device will be called if it exists.
370 * If the rx_handler consider the skb should be ignored, it should return
371 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
372 * are registered on exact device (ptype->dev == skb->dev).
374 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
375 * delivered, it should return RX_HANDLER_PASS.
377 * A device without a registered rx_handler will behave as if rx_handler
378 * returned RX_HANDLER_PASS.
381 enum rx_handler_result {
387 typedef enum rx_handler_result rx_handler_result_t;
388 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
390 void __napi_schedule(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))
425 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
426 static inline bool napi_reschedule(struct napi_struct *napi)
428 if (napi_schedule_prep(napi)) {
429 __napi_schedule(napi);
436 * napi_complete - NAPI processing complete
439 * Mark NAPI processing as complete.
441 void __napi_complete(struct napi_struct *n);
442 void napi_complete(struct napi_struct *n);
445 * napi_by_id - lookup a NAPI by napi_id
446 * @napi_id: hashed napi_id
448 * lookup @napi_id in napi_hash table
449 * must be called under rcu_read_lock()
451 struct napi_struct *napi_by_id(unsigned int napi_id);
454 * napi_hash_add - add a NAPI to global hashtable
455 * @napi: napi context
457 * generate a new napi_id and store a @napi under it in napi_hash
459 void napi_hash_add(struct napi_struct *napi);
462 * napi_hash_del - remove a NAPI from global table
463 * @napi: napi context
465 * Warning: caller must observe rcu grace period
466 * before freeing memory containing @napi
468 void napi_hash_del(struct napi_struct *napi);
471 * napi_disable - prevent NAPI from scheduling
474 * Stop NAPI from being scheduled on this context.
475 * Waits till any outstanding processing completes.
477 static inline void napi_disable(struct napi_struct *n)
480 set_bit(NAPI_STATE_DISABLE, &n->state);
481 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
483 clear_bit(NAPI_STATE_DISABLE, &n->state);
487 * napi_enable - enable NAPI scheduling
490 * Resume NAPI from being scheduled on this context.
491 * Must be paired with napi_disable.
493 static inline void napi_enable(struct napi_struct *n)
495 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
496 smp_mb__before_clear_bit();
497 clear_bit(NAPI_STATE_SCHED, &n->state);
502 * napi_synchronize - wait until NAPI is not running
505 * Wait until NAPI is done being scheduled on this context.
506 * Waits till any outstanding processing completes but
507 * does not disable future activations.
509 static inline void napi_synchronize(const struct napi_struct *n)
511 while (test_bit(NAPI_STATE_SCHED, &n->state))
515 # define napi_synchronize(n) barrier()
518 enum netdev_queue_state_t {
519 __QUEUE_STATE_DRV_XOFF,
520 __QUEUE_STATE_STACK_XOFF,
521 __QUEUE_STATE_FROZEN,
522 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
523 (1 << __QUEUE_STATE_STACK_XOFF))
524 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
525 (1 << __QUEUE_STATE_FROZEN))
528 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
529 * netif_tx_* functions below are used to manipulate this flag. The
530 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
531 * queue independently. The netif_xmit_*stopped functions below are called
532 * to check if the queue has been stopped by the driver or stack (either
533 * of the XOFF bits are set in the state). Drivers should not need to call
534 * netif_xmit*stopped functions, they should only be using netif_tx_*.
537 struct netdev_queue {
541 struct net_device *dev;
543 struct Qdisc *qdisc_sleeping;
547 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
553 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
556 * please use this field instead of dev->trans_start
558 unsigned long trans_start;
561 * Number of TX timeouts for this queue
562 * (/sys/class/net/DEV/Q/trans_timeout)
564 unsigned long trans_timeout;
571 } ____cacheline_aligned_in_smp;
573 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
575 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
582 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
584 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 * This structure holds an RPS map which can be of variable length. The
592 * map is an array of CPUs.
599 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
602 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
603 * tail pointer for that CPU's input queue at the time of last enqueue, and
604 * a hardware filter index.
606 struct rps_dev_flow {
609 unsigned int last_qtail;
611 #define RPS_NO_FILTER 0xffff
614 * The rps_dev_flow_table structure contains a table of flow mappings.
616 struct rps_dev_flow_table {
619 struct rps_dev_flow flows[0];
621 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
622 ((_num) * sizeof(struct rps_dev_flow)))
625 * The rps_sock_flow_table contains mappings of flows to the last CPU
626 * on which they were processed by the application (set in recvmsg).
628 struct rps_sock_flow_table {
632 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
633 ((_num) * sizeof(u16)))
635 #define RPS_NO_CPU 0xffff
637 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
641 unsigned int cpu, index = hash & table->mask;
643 /* We only give a hint, preemption can change cpu under us */
644 cpu = raw_smp_processor_id();
646 if (table->ents[index] != cpu)
647 table->ents[index] = cpu;
651 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
655 table->ents[hash & table->mask] = RPS_NO_CPU;
658 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660 #ifdef CONFIG_RFS_ACCEL
661 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
664 #endif /* CONFIG_RPS */
666 /* This structure contains an instance of an RX queue. */
667 struct netdev_rx_queue {
669 struct rps_map __rcu *rps_map;
670 struct rps_dev_flow_table __rcu *rps_flow_table;
673 struct net_device *dev;
674 } ____cacheline_aligned_in_smp;
677 * RX queue sysfs structures and functions.
679 struct rx_queue_attribute {
680 struct attribute attr;
681 ssize_t (*show)(struct netdev_rx_queue *queue,
682 struct rx_queue_attribute *attr, char *buf);
683 ssize_t (*store)(struct netdev_rx_queue *queue,
684 struct rx_queue_attribute *attr, const char *buf, size_t len);
689 * This structure holds an XPS map which can be of variable length. The
690 * map is an array of queues.
694 unsigned int alloc_len;
698 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
699 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
703 * This structure holds all XPS maps for device. Maps are indexed by CPU.
705 struct xps_dev_maps {
707 struct xps_map __rcu *cpu_map[0];
709 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
710 (nr_cpu_ids * sizeof(struct xps_map *)))
711 #endif /* CONFIG_XPS */
713 #define TC_MAX_QUEUE 16
714 #define TC_BITMASK 15
715 /* HW offloaded queuing disciplines txq count and offset maps */
716 struct netdev_tc_txq {
721 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
723 * This structure is to hold information about the device
724 * configured to run FCoE protocol stack.
726 struct netdev_fcoe_hbainfo {
727 char manufacturer[64];
728 char serial_number[64];
729 char hardware_version[64];
730 char driver_version[64];
731 char optionrom_version[64];
732 char firmware_version[64];
734 char model_description[256];
738 #define MAX_PHYS_PORT_ID_LEN 32
740 /* This structure holds a unique identifier to identify the
741 * physical port used by a netdevice.
743 struct netdev_phys_port_id {
744 unsigned char id[MAX_PHYS_PORT_ID_LEN];
745 unsigned char id_len;
748 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
749 struct sk_buff *skb);
752 * This structure defines the management hooks for network devices.
753 * The following hooks can be defined; unless noted otherwise, they are
754 * optional and can be filled with a null pointer.
756 * int (*ndo_init)(struct net_device *dev);
757 * This function is called once when network device is registered.
758 * The network device can use this to any late stage initializaton
759 * or semantic validattion. It can fail with an error code which will
760 * be propogated back to register_netdev
762 * void (*ndo_uninit)(struct net_device *dev);
763 * This function is called when device is unregistered or when registration
764 * fails. It is not called if init fails.
766 * int (*ndo_open)(struct net_device *dev);
767 * This function is called when network device transistions to the up
770 * int (*ndo_stop)(struct net_device *dev);
771 * This function is called when network device transistions to the down
774 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
775 * struct net_device *dev);
776 * Called when a packet needs to be transmitted.
777 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
778 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
779 * Required can not be NULL.
781 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
782 * void *accel_priv, select_queue_fallback_t fallback);
783 * Called to decide which queue to when device supports multiple
786 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
787 * This function is called to allow device receiver to make
788 * changes to configuration when multicast or promiscious is enabled.
790 * void (*ndo_set_rx_mode)(struct net_device *dev);
791 * This function is called device changes address list filtering.
792 * If driver handles unicast address filtering, it should set
793 * IFF_UNICAST_FLT to its priv_flags.
795 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
796 * This function is called when the Media Access Control address
797 * needs to be changed. If this interface is not defined, the
798 * mac address can not be changed.
800 * int (*ndo_validate_addr)(struct net_device *dev);
801 * Test if Media Access Control address is valid for the device.
803 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
804 * Called when a user request an ioctl which can't be handled by
805 * the generic interface code. If not defined ioctl's return
806 * not supported error code.
808 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
809 * Used to set network devices bus interface parameters. This interface
810 * is retained for legacy reason, new devices should use the bus
811 * interface (PCI) for low level management.
813 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
814 * Called when a user wants to change the Maximum Transfer Unit
815 * of a device. If not defined, any request to change MTU will
816 * will return an error.
818 * void (*ndo_tx_timeout)(struct net_device *dev);
819 * Callback uses when the transmitter has not made any progress
820 * for dev->watchdog ticks.
822 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
823 * struct rtnl_link_stats64 *storage);
824 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
825 * Called when a user wants to get the network device usage
826 * statistics. Drivers must do one of the following:
827 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
828 * rtnl_link_stats64 structure passed by the caller.
829 * 2. Define @ndo_get_stats to update a net_device_stats structure
830 * (which should normally be dev->stats) and return a pointer to
831 * it. The structure may be changed asynchronously only if each
832 * field is written atomically.
833 * 3. Update dev->stats asynchronously and atomically, and define
836 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
837 * If device support VLAN filtering this function is called when a
838 * VLAN id is registered.
840 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
841 * If device support VLAN filtering this function is called when a
842 * VLAN id is unregistered.
844 * void (*ndo_poll_controller)(struct net_device *dev);
846 * SR-IOV management functions.
847 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
848 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
849 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
850 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
851 * int (*ndo_get_vf_config)(struct net_device *dev,
852 * int vf, struct ifla_vf_info *ivf);
853 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
854 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
855 * struct nlattr *port[]);
856 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
857 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
858 * Called to setup 'tc' number of traffic classes in the net device. This
859 * is always called from the stack with the rtnl lock held and netif tx
860 * queues stopped. This allows the netdevice to perform queue management
863 * Fiber Channel over Ethernet (FCoE) offload functions.
864 * int (*ndo_fcoe_enable)(struct net_device *dev);
865 * Called when the FCoE protocol stack wants to start using LLD for FCoE
866 * so the underlying device can perform whatever needed configuration or
867 * initialization to support acceleration of FCoE traffic.
869 * int (*ndo_fcoe_disable)(struct net_device *dev);
870 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
871 * so the underlying device can perform whatever needed clean-ups to
872 * stop supporting acceleration of FCoE traffic.
874 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
875 * struct scatterlist *sgl, unsigned int sgc);
876 * Called when the FCoE Initiator wants to initialize an I/O that
877 * is a possible candidate for Direct Data Placement (DDP). The LLD can
878 * perform necessary setup and returns 1 to indicate the device is set up
879 * successfully to perform DDP on this I/O, otherwise this returns 0.
881 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
882 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
883 * indicated by the FC exchange id 'xid', so the underlying device can
884 * clean up and reuse resources for later DDP requests.
886 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
887 * struct scatterlist *sgl, unsigned int sgc);
888 * Called when the FCoE Target wants to initialize an I/O that
889 * is a possible candidate for Direct Data Placement (DDP). The LLD can
890 * perform necessary setup and returns 1 to indicate the device is set up
891 * successfully to perform DDP on this I/O, otherwise this returns 0.
893 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
894 * struct netdev_fcoe_hbainfo *hbainfo);
895 * Called when the FCoE Protocol stack wants information on the underlying
896 * device. This information is utilized by the FCoE protocol stack to
897 * register attributes with Fiber Channel management service as per the
898 * FC-GS Fabric Device Management Information(FDMI) specification.
900 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
901 * Called when the underlying device wants to override default World Wide
902 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
903 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
904 * protocol stack to use.
907 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
908 * u16 rxq_index, u32 flow_id);
909 * Set hardware filter for RFS. rxq_index is the target queue index;
910 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
911 * Return the filter ID on success, or a negative error code.
913 * Slave management functions (for bridge, bonding, etc).
914 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
915 * Called to make another netdev an underling.
917 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
918 * Called to release previously enslaved netdev.
920 * Feature/offload setting functions.
921 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
922 * netdev_features_t features);
923 * Adjusts the requested feature flags according to device-specific
924 * constraints, and returns the resulting flags. Must not modify
927 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
928 * Called to update device configuration to new features. Passed
929 * feature set might be less than what was returned by ndo_fix_features()).
930 * Must return >0 or -errno if it changed dev->features itself.
932 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
933 * struct net_device *dev,
934 * const unsigned char *addr, u16 flags)
935 * Adds an FDB entry to dev for addr.
936 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
937 * struct net_device *dev,
938 * const unsigned char *addr)
939 * Deletes the FDB entry from dev coresponding to addr.
940 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
941 * struct net_device *dev, int idx)
942 * Used to add FDB entries to dump requests. Implementers should add
943 * entries to skb and update idx with the number of entries.
945 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
946 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
947 * struct net_device *dev, u32 filter_mask)
949 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
950 * Called to change device carrier. Soft-devices (like dummy, team, etc)
951 * which do not represent real hardware may define this to allow their
952 * userspace components to manage their virtual carrier state. Devices
953 * that determine carrier state from physical hardware properties (eg
954 * network cables) or protocol-dependent mechanisms (eg
955 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
957 * int (*ndo_get_phys_port_id)(struct net_device *dev,
958 * struct netdev_phys_port_id *ppid);
959 * Called to get ID of physical port of this device. If driver does
960 * not implement this, it is assumed that the hw is not able to have
961 * multiple net devices on single physical port.
963 * void (*ndo_add_vxlan_port)(struct net_device *dev,
964 * sa_family_t sa_family, __be16 port);
965 * Called by vxlan to notiy a driver about the UDP port and socket
966 * address family that vxlan is listnening to. It is called only when
967 * a new port starts listening. The operation is protected by the
968 * vxlan_net->sock_lock.
970 * void (*ndo_del_vxlan_port)(struct net_device *dev,
971 * sa_family_t sa_family, __be16 port);
972 * Called by vxlan to notify the driver about a UDP port and socket
973 * address family that vxlan is not listening to anymore. The operation
974 * is protected by the vxlan_net->sock_lock.
976 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
977 * struct net_device *dev)
978 * Called by upper layer devices to accelerate switching or other
979 * station functionality into hardware. 'pdev is the lowerdev
980 * to use for the offload and 'dev' is the net device that will
981 * back the offload. Returns a pointer to the private structure
982 * the upper layer will maintain.
983 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
984 * Called by upper layer device to delete the station created
985 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
986 * the station and priv is the structure returned by the add
988 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
989 * struct net_device *dev,
991 * Callback to use for xmit over the accelerated station. This
992 * is used in place of ndo_start_xmit on accelerated net
995 struct net_device_ops {
996 int (*ndo_init)(struct net_device *dev);
997 void (*ndo_uninit)(struct net_device *dev);
998 int (*ndo_open)(struct net_device *dev);
999 int (*ndo_stop)(struct net_device *dev);
1000 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1001 struct net_device *dev);
1002 u16 (*ndo_select_queue)(struct net_device *dev,
1003 struct sk_buff *skb,
1005 select_queue_fallback_t fallback);
1006 void (*ndo_change_rx_flags)(struct net_device *dev,
1008 void (*ndo_set_rx_mode)(struct net_device *dev);
1009 int (*ndo_set_mac_address)(struct net_device *dev,
1011 int (*ndo_validate_addr)(struct net_device *dev);
1012 int (*ndo_do_ioctl)(struct net_device *dev,
1013 struct ifreq *ifr, int cmd);
1014 int (*ndo_set_config)(struct net_device *dev,
1016 int (*ndo_change_mtu)(struct net_device *dev,
1018 int (*ndo_neigh_setup)(struct net_device *dev,
1019 struct neigh_parms *);
1020 void (*ndo_tx_timeout) (struct net_device *dev);
1022 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1023 struct rtnl_link_stats64 *storage);
1024 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1026 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1027 __be16 proto, u16 vid);
1028 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1029 __be16 proto, u16 vid);
1030 #ifdef CONFIG_NET_POLL_CONTROLLER
1031 void (*ndo_poll_controller)(struct net_device *dev);
1032 int (*ndo_netpoll_setup)(struct net_device *dev,
1033 struct netpoll_info *info);
1034 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1036 #ifdef CONFIG_NET_RX_BUSY_POLL
1037 int (*ndo_busy_poll)(struct napi_struct *dev);
1039 int (*ndo_set_vf_mac)(struct net_device *dev,
1040 int queue, u8 *mac);
1041 int (*ndo_set_vf_vlan)(struct net_device *dev,
1042 int queue, u16 vlan, u8 qos);
1043 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1045 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1046 int vf, bool setting);
1047 int (*ndo_get_vf_config)(struct net_device *dev,
1049 struct ifla_vf_info *ivf);
1050 int (*ndo_set_vf_link_state)(struct net_device *dev,
1051 int vf, int link_state);
1052 int (*ndo_set_vf_port)(struct net_device *dev,
1054 struct nlattr *port[]);
1055 int (*ndo_get_vf_port)(struct net_device *dev,
1056 int vf, struct sk_buff *skb);
1057 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1058 #if IS_ENABLED(CONFIG_FCOE)
1059 int (*ndo_fcoe_enable)(struct net_device *dev);
1060 int (*ndo_fcoe_disable)(struct net_device *dev);
1061 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1063 struct scatterlist *sgl,
1065 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1067 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1069 struct scatterlist *sgl,
1071 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1072 struct netdev_fcoe_hbainfo *hbainfo);
1075 #if IS_ENABLED(CONFIG_LIBFCOE)
1076 #define NETDEV_FCOE_WWNN 0
1077 #define NETDEV_FCOE_WWPN 1
1078 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1079 u64 *wwn, int type);
1082 #ifdef CONFIG_RFS_ACCEL
1083 int (*ndo_rx_flow_steer)(struct net_device *dev,
1084 const struct sk_buff *skb,
1088 int (*ndo_add_slave)(struct net_device *dev,
1089 struct net_device *slave_dev);
1090 int (*ndo_del_slave)(struct net_device *dev,
1091 struct net_device *slave_dev);
1092 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1093 netdev_features_t features);
1094 int (*ndo_set_features)(struct net_device *dev,
1095 netdev_features_t features);
1096 int (*ndo_neigh_construct)(struct neighbour *n);
1097 void (*ndo_neigh_destroy)(struct neighbour *n);
1099 int (*ndo_fdb_add)(struct ndmsg *ndm,
1100 struct nlattr *tb[],
1101 struct net_device *dev,
1102 const unsigned char *addr,
1104 int (*ndo_fdb_del)(struct ndmsg *ndm,
1105 struct nlattr *tb[],
1106 struct net_device *dev,
1107 const unsigned char *addr);
1108 int (*ndo_fdb_dump)(struct sk_buff *skb,
1109 struct netlink_callback *cb,
1110 struct net_device *dev,
1113 int (*ndo_bridge_setlink)(struct net_device *dev,
1114 struct nlmsghdr *nlh);
1115 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1117 struct net_device *dev,
1119 int (*ndo_bridge_dellink)(struct net_device *dev,
1120 struct nlmsghdr *nlh);
1121 int (*ndo_change_carrier)(struct net_device *dev,
1123 int (*ndo_get_phys_port_id)(struct net_device *dev,
1124 struct netdev_phys_port_id *ppid);
1125 void (*ndo_add_vxlan_port)(struct net_device *dev,
1126 sa_family_t sa_family,
1128 void (*ndo_del_vxlan_port)(struct net_device *dev,
1129 sa_family_t sa_family,
1132 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1133 struct net_device *dev);
1134 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1137 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1138 struct net_device *dev,
1143 * enum net_device_priv_flags - &struct net_device priv_flags
1145 * These are the &struct net_device, they are only set internally
1146 * by drivers and used in the kernel. These flags are invisible to
1147 * userspace, this means that the order of these flags can change
1148 * during any kernel release.
1150 * You should have a pretty good reason to be extending these flags.
1152 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1153 * @IFF_EBRIDGE: Ethernet bridging device
1154 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1155 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1156 * @IFF_MASTER_ALB: bonding master, balance-alb
1157 * @IFF_BONDING: bonding master or slave
1158 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1159 * @IFF_ISATAP: ISATAP interface (RFC4214)
1160 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1161 * @IFF_WAN_HDLC: WAN HDLC device
1162 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1164 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1165 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1166 * @IFF_MACVLAN_PORT: device used as macvlan port
1167 * @IFF_BRIDGE_PORT: device used as bridge port
1168 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1169 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1170 * @IFF_UNICAST_FLT: Supports unicast filtering
1171 * @IFF_TEAM_PORT: device used as team port
1172 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1173 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1174 * change when it's running
1175 * @IFF_MACVLAN: Macvlan device
1177 enum netdev_priv_flags {
1178 IFF_802_1Q_VLAN = 1<<0,
1180 IFF_SLAVE_INACTIVE = 1<<2,
1181 IFF_MASTER_8023AD = 1<<3,
1182 IFF_MASTER_ALB = 1<<4,
1184 IFF_SLAVE_NEEDARP = 1<<6,
1186 IFF_MASTER_ARPMON = 1<<8,
1187 IFF_WAN_HDLC = 1<<9,
1188 IFF_XMIT_DST_RELEASE = 1<<10,
1189 IFF_DONT_BRIDGE = 1<<11,
1190 IFF_DISABLE_NETPOLL = 1<<12,
1191 IFF_MACVLAN_PORT = 1<<13,
1192 IFF_BRIDGE_PORT = 1<<14,
1193 IFF_OVS_DATAPATH = 1<<15,
1194 IFF_TX_SKB_SHARING = 1<<16,
1195 IFF_UNICAST_FLT = 1<<17,
1196 IFF_TEAM_PORT = 1<<18,
1197 IFF_SUPP_NOFCS = 1<<19,
1198 IFF_LIVE_ADDR_CHANGE = 1<<20,
1199 IFF_MACVLAN = 1<<21,
1202 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1203 #define IFF_EBRIDGE IFF_EBRIDGE
1204 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1205 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1206 #define IFF_MASTER_ALB IFF_MASTER_ALB
1207 #define IFF_BONDING IFF_BONDING
1208 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1209 #define IFF_ISATAP IFF_ISATAP
1210 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1211 #define IFF_WAN_HDLC IFF_WAN_HDLC
1212 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1213 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1214 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1215 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1216 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1217 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1218 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1219 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1220 #define IFF_TEAM_PORT IFF_TEAM_PORT
1221 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1222 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1223 #define IFF_MACVLAN IFF_MACVLAN
1226 * The DEVICE structure.
1227 * Actually, this whole structure is a big mistake. It mixes I/O
1228 * data with strictly "high-level" data, and it has to know about
1229 * almost every data structure used in the INET module.
1231 * FIXME: cleanup struct net_device such that network protocol info
1238 * This is the first field of the "visible" part of this structure
1239 * (i.e. as seen by users in the "Space.c" file). It is the name
1242 char name[IFNAMSIZ];
1244 /* device name hash chain, please keep it close to name[] */
1245 struct hlist_node name_hlist;
1251 * I/O specific fields
1252 * FIXME: Merge these and struct ifmap into one
1254 unsigned long mem_end; /* shared mem end */
1255 unsigned long mem_start; /* shared mem start */
1256 unsigned long base_addr; /* device I/O address */
1257 int irq; /* device IRQ number */
1260 * Some hardware also needs these fields, but they are not
1261 * part of the usual set specified in Space.c.
1264 unsigned long state;
1266 struct list_head dev_list;
1267 struct list_head napi_list;
1268 struct list_head unreg_list;
1269 struct list_head close_list;
1271 /* directly linked devices, like slaves for bonding */
1273 struct list_head upper;
1274 struct list_head lower;
1277 /* all linked devices, *including* neighbours */
1279 struct list_head upper;
1280 struct list_head lower;
1284 /* currently active device features */
1285 netdev_features_t features;
1286 /* user-changeable features */
1287 netdev_features_t hw_features;
1288 /* user-requested features */
1289 netdev_features_t wanted_features;
1290 /* mask of features inheritable by VLAN devices */
1291 netdev_features_t vlan_features;
1292 /* mask of features inherited by encapsulating devices
1293 * This field indicates what encapsulation offloads
1294 * the hardware is capable of doing, and drivers will
1295 * need to set them appropriately.
1297 netdev_features_t hw_enc_features;
1298 /* mask of fetures inheritable by MPLS */
1299 netdev_features_t mpls_features;
1301 /* Interface index. Unique device identifier */
1305 struct net_device_stats stats;
1307 /* dropped packets by core network, Do not use this in drivers */
1308 atomic_long_t rx_dropped;
1309 atomic_long_t tx_dropped;
1311 /* Stats to monitor carrier on<->off transitions */
1312 atomic_t carrier_changes;
1314 #ifdef CONFIG_WIRELESS_EXT
1315 /* List of functions to handle Wireless Extensions (instead of ioctl).
1316 * See <net/iw_handler.h> for details. Jean II */
1317 const struct iw_handler_def * wireless_handlers;
1318 /* Instance data managed by the core of Wireless Extensions. */
1319 struct iw_public_data * wireless_data;
1321 /* Management operations */
1322 const struct net_device_ops *netdev_ops;
1323 const struct ethtool_ops *ethtool_ops;
1324 const struct forwarding_accel_ops *fwd_ops;
1326 /* Hardware header description */
1327 const struct header_ops *header_ops;
1329 unsigned int flags; /* interface flags (a la BSD) */
1330 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1331 * See if.h for definitions. */
1332 unsigned short gflags;
1333 unsigned short padded; /* How much padding added by alloc_netdev() */
1335 unsigned char operstate; /* RFC2863 operstate */
1336 unsigned char link_mode; /* mapping policy to operstate */
1338 unsigned char if_port; /* Selectable AUI, TP,..*/
1339 unsigned char dma; /* DMA channel */
1341 unsigned int mtu; /* interface MTU value */
1342 unsigned short type; /* interface hardware type */
1343 unsigned short hard_header_len; /* hardware hdr length */
1345 /* extra head- and tailroom the hardware may need, but not in all cases
1346 * can this be guaranteed, especially tailroom. Some cases also use
1347 * LL_MAX_HEADER instead to allocate the skb.
1349 unsigned short needed_headroom;
1350 unsigned short needed_tailroom;
1352 /* Interface address info. */
1353 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1354 unsigned char addr_assign_type; /* hw address assignment type */
1355 unsigned char addr_len; /* hardware address length */
1356 unsigned short neigh_priv_len;
1357 unsigned short dev_id; /* Used to differentiate devices
1358 * that share the same link
1361 unsigned short dev_port; /* Used to differentiate
1362 * devices that share the same
1365 spinlock_t addr_list_lock;
1366 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1367 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1368 struct netdev_hw_addr_list dev_addrs; /* list of device
1372 struct kset *queues_kset;
1376 unsigned int promiscuity;
1377 unsigned int allmulti;
1380 /* Protocol specific pointers */
1382 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1383 struct vlan_info __rcu *vlan_info; /* VLAN info */
1385 #if IS_ENABLED(CONFIG_NET_DSA)
1386 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1388 #if IS_ENABLED(CONFIG_TIPC)
1389 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */
1391 void *atalk_ptr; /* AppleTalk link */
1392 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1393 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1394 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1395 void *ax25_ptr; /* AX.25 specific data */
1396 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1397 assign before registering */
1400 * Cache lines mostly used on receive path (including eth_type_trans())
1402 unsigned long last_rx; /* Time of last Rx */
1404 /* Interface address info used in eth_type_trans() */
1405 unsigned char *dev_addr; /* hw address, (before bcast
1406 because most packets are
1411 struct netdev_rx_queue *_rx;
1413 /* Number of RX queues allocated at register_netdev() time */
1414 unsigned int num_rx_queues;
1416 /* Number of RX queues currently active in device */
1417 unsigned int real_num_rx_queues;
1421 rx_handler_func_t __rcu *rx_handler;
1422 void __rcu *rx_handler_data;
1424 struct netdev_queue __rcu *ingress_queue;
1425 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1429 * Cache lines mostly used on transmit path
1431 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1433 /* Number of TX queues allocated at alloc_netdev_mq() time */
1434 unsigned int num_tx_queues;
1436 /* Number of TX queues currently active in device */
1437 unsigned int real_num_tx_queues;
1439 /* root qdisc from userspace point of view */
1440 struct Qdisc *qdisc;
1442 unsigned long tx_queue_len; /* Max frames per queue allowed */
1443 spinlock_t tx_global_lock;
1446 struct xps_dev_maps __rcu *xps_maps;
1448 #ifdef CONFIG_RFS_ACCEL
1449 /* CPU reverse-mapping for RX completion interrupts, indexed
1450 * by RX queue number. Assigned by driver. This must only be
1451 * set if the ndo_rx_flow_steer operation is defined. */
1452 struct cpu_rmap *rx_cpu_rmap;
1455 /* These may be needed for future network-power-down code. */
1458 * trans_start here is expensive for high speed devices on SMP,
1459 * please use netdev_queue->trans_start instead.
1461 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1463 int watchdog_timeo; /* used by dev_watchdog() */
1464 struct timer_list watchdog_timer;
1466 /* Number of references to this device */
1467 int __percpu *pcpu_refcnt;
1469 /* delayed register/unregister */
1470 struct list_head todo_list;
1471 /* device index hash chain */
1472 struct hlist_node index_hlist;
1474 struct list_head link_watch_list;
1476 /* register/unregister state machine */
1477 enum { NETREG_UNINITIALIZED=0,
1478 NETREG_REGISTERED, /* completed register_netdevice */
1479 NETREG_UNREGISTERING, /* called unregister_netdevice */
1480 NETREG_UNREGISTERED, /* completed unregister todo */
1481 NETREG_RELEASED, /* called free_netdev */
1482 NETREG_DUMMY, /* dummy device for NAPI poll */
1485 bool dismantle; /* device is going do be freed */
1488 RTNL_LINK_INITIALIZED,
1489 RTNL_LINK_INITIALIZING,
1490 } rtnl_link_state:16;
1492 /* Called from unregister, can be used to call free_netdev */
1493 void (*destructor)(struct net_device *dev);
1495 #ifdef CONFIG_NETPOLL
1496 struct netpoll_info __rcu *npinfo;
1499 #ifdef CONFIG_NET_NS
1500 /* Network namespace this network device is inside */
1504 /* mid-layer private */
1507 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1508 struct pcpu_sw_netstats __percpu *tstats;
1509 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1510 struct pcpu_vstats __percpu *vstats; /* veth stats */
1513 struct garp_port __rcu *garp_port;
1515 struct mrp_port __rcu *mrp_port;
1517 /* class/net/name entry */
1519 /* space for optional device, statistics, and wireless sysfs groups */
1520 const struct attribute_group *sysfs_groups[4];
1521 /* space for optional per-rx queue attributes */
1522 const struct attribute_group *sysfs_rx_queue_group;
1524 /* rtnetlink link ops */
1525 const struct rtnl_link_ops *rtnl_link_ops;
1527 /* for setting kernel sock attribute on TCP connection setup */
1528 #define GSO_MAX_SIZE 65536
1529 unsigned int gso_max_size;
1530 #define GSO_MAX_SEGS 65535
1534 /* Data Center Bridging netlink ops */
1535 const struct dcbnl_rtnl_ops *dcbnl_ops;
1538 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1539 u8 prio_tc_map[TC_BITMASK + 1];
1541 #if IS_ENABLED(CONFIG_FCOE)
1542 /* max exchange id for FCoE LRO by ddp */
1543 unsigned int fcoe_ddp_xid;
1545 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1546 struct netprio_map __rcu *priomap;
1548 /* phy device may attach itself for hardware timestamping */
1549 struct phy_device *phydev;
1551 struct lock_class_key *qdisc_tx_busylock;
1553 /* group the device belongs to */
1556 struct pm_qos_request pm_qos_req;
1558 #define to_net_dev(d) container_of(d, struct net_device, dev)
1560 #define NETDEV_ALIGN 32
1563 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1565 return dev->prio_tc_map[prio & TC_BITMASK];
1569 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1571 if (tc >= dev->num_tc)
1574 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1579 void netdev_reset_tc(struct net_device *dev)
1582 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1583 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1587 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1589 if (tc >= dev->num_tc)
1592 dev->tc_to_txq[tc].count = count;
1593 dev->tc_to_txq[tc].offset = offset;
1598 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1600 if (num_tc > TC_MAX_QUEUE)
1603 dev->num_tc = num_tc;
1608 int netdev_get_num_tc(struct net_device *dev)
1614 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1617 return &dev->_tx[index];
1620 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1621 void (*f)(struct net_device *,
1622 struct netdev_queue *,
1628 for (i = 0; i < dev->num_tx_queues; i++)
1629 f(dev, &dev->_tx[i], arg);
1632 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1633 struct sk_buff *skb,
1637 * Net namespace inlines
1640 struct net *dev_net(const struct net_device *dev)
1642 return read_pnet(&dev->nd_net);
1646 void dev_net_set(struct net_device *dev, struct net *net)
1648 #ifdef CONFIG_NET_NS
1649 release_net(dev->nd_net);
1650 dev->nd_net = hold_net(net);
1654 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1656 #ifdef CONFIG_NET_DSA_TAG_DSA
1657 if (dev->dsa_ptr != NULL)
1658 return dsa_uses_dsa_tags(dev->dsa_ptr);
1664 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1666 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1667 if (dev->dsa_ptr != NULL)
1668 return dsa_uses_trailer_tags(dev->dsa_ptr);
1675 * netdev_priv - access network device private data
1676 * @dev: network device
1678 * Get network device private data
1680 static inline void *netdev_priv(const struct net_device *dev)
1682 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1685 /* Set the sysfs physical device reference for the network logical device
1686 * if set prior to registration will cause a symlink during initialization.
1688 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1690 /* Set the sysfs device type for the network logical device to allow
1691 * fine-grained identification of different network device types. For
1692 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1694 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1696 /* Default NAPI poll() weight
1697 * Device drivers are strongly advised to not use bigger value
1699 #define NAPI_POLL_WEIGHT 64
1702 * netif_napi_add - initialize a napi context
1703 * @dev: network device
1704 * @napi: napi context
1705 * @poll: polling function
1706 * @weight: default weight
1708 * netif_napi_add() must be used to initialize a napi context prior to calling
1709 * *any* of the other napi related functions.
1711 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1712 int (*poll)(struct napi_struct *, int), int weight);
1715 * netif_napi_del - remove a napi context
1716 * @napi: napi context
1718 * netif_napi_del() removes a napi context from the network device napi list
1720 void netif_napi_del(struct napi_struct *napi);
1722 struct napi_gro_cb {
1723 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1726 /* Length of frag0. */
1727 unsigned int frag0_len;
1729 /* This indicates where we are processing relative to skb->data. */
1732 /* This is non-zero if the packet cannot be merged with the new skb. */
1735 /* Save the IP ID here and check when we get to the transport layer */
1738 /* Number of segments aggregated. */
1741 /* This is non-zero if the packet may be of the same flow. */
1746 #define NAPI_GRO_FREE 1
1747 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1749 /* jiffies when first packet was created/queued */
1752 /* Used in ipv6_gro_receive() */
1755 /* Used in udp_gro_receive */
1758 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1761 /* used in skb_gro_receive() slow path */
1762 struct sk_buff *last;
1765 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1767 struct packet_type {
1768 __be16 type; /* This is really htons(ether_type). */
1769 struct net_device *dev; /* NULL is wildcarded here */
1770 int (*func) (struct sk_buff *,
1771 struct net_device *,
1772 struct packet_type *,
1773 struct net_device *);
1774 bool (*id_match)(struct packet_type *ptype,
1776 void *af_packet_priv;
1777 struct list_head list;
1780 struct offload_callbacks {
1781 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1782 netdev_features_t features);
1783 int (*gso_send_check)(struct sk_buff *skb);
1784 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1785 struct sk_buff *skb);
1786 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1789 struct packet_offload {
1790 __be16 type; /* This is really htons(ether_type). */
1791 struct offload_callbacks callbacks;
1792 struct list_head list;
1795 struct udp_offload {
1797 struct offload_callbacks callbacks;
1800 /* often modified stats are per cpu, other are shared (netdev->stats) */
1801 struct pcpu_sw_netstats {
1806 struct u64_stats_sync syncp;
1809 #define netdev_alloc_pcpu_stats(type) \
1811 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1814 for_each_possible_cpu(i) { \
1815 typeof(type) *stat; \
1816 stat = per_cpu_ptr(pcpu_stats, i); \
1817 u64_stats_init(&stat->syncp); \
1823 #include <linux/notifier.h>
1825 /* netdevice notifier chain. Please remember to update the rtnetlink
1826 * notification exclusion list in rtnetlink_event() when adding new
1829 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1830 #define NETDEV_DOWN 0x0002
1831 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1832 detected a hardware crash and restarted
1833 - we can use this eg to kick tcp sessions
1835 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1836 #define NETDEV_REGISTER 0x0005
1837 #define NETDEV_UNREGISTER 0x0006
1838 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1839 #define NETDEV_CHANGEADDR 0x0008
1840 #define NETDEV_GOING_DOWN 0x0009
1841 #define NETDEV_CHANGENAME 0x000A
1842 #define NETDEV_FEAT_CHANGE 0x000B
1843 #define NETDEV_BONDING_FAILOVER 0x000C
1844 #define NETDEV_PRE_UP 0x000D
1845 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1846 #define NETDEV_POST_TYPE_CHANGE 0x000F
1847 #define NETDEV_POST_INIT 0x0010
1848 #define NETDEV_UNREGISTER_FINAL 0x0011
1849 #define NETDEV_RELEASE 0x0012
1850 #define NETDEV_NOTIFY_PEERS 0x0013
1851 #define NETDEV_JOIN 0x0014
1852 #define NETDEV_CHANGEUPPER 0x0015
1853 #define NETDEV_RESEND_IGMP 0x0016
1854 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1856 int register_netdevice_notifier(struct notifier_block *nb);
1857 int unregister_netdevice_notifier(struct notifier_block *nb);
1859 struct netdev_notifier_info {
1860 struct net_device *dev;
1863 struct netdev_notifier_change_info {
1864 struct netdev_notifier_info info; /* must be first */
1865 unsigned int flags_changed;
1868 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1869 struct net_device *dev)
1874 static inline struct net_device *
1875 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1880 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1883 extern rwlock_t dev_base_lock; /* Device list lock */
1885 #define for_each_netdev(net, d) \
1886 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1887 #define for_each_netdev_reverse(net, d) \
1888 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1889 #define for_each_netdev_rcu(net, d) \
1890 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1891 #define for_each_netdev_safe(net, d, n) \
1892 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1893 #define for_each_netdev_continue(net, d) \
1894 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1895 #define for_each_netdev_continue_rcu(net, d) \
1896 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1897 #define for_each_netdev_in_bond_rcu(bond, slave) \
1898 for_each_netdev_rcu(&init_net, slave) \
1899 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1900 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1902 static inline struct net_device *next_net_device(struct net_device *dev)
1904 struct list_head *lh;
1908 lh = dev->dev_list.next;
1909 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1912 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1914 struct list_head *lh;
1918 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1919 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1922 static inline struct net_device *first_net_device(struct net *net)
1924 return list_empty(&net->dev_base_head) ? NULL :
1925 net_device_entry(net->dev_base_head.next);
1928 static inline struct net_device *first_net_device_rcu(struct net *net)
1930 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1932 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1935 int netdev_boot_setup_check(struct net_device *dev);
1936 unsigned long netdev_boot_base(const char *prefix, int unit);
1937 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1938 const char *hwaddr);
1939 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1940 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1941 void dev_add_pack(struct packet_type *pt);
1942 void dev_remove_pack(struct packet_type *pt);
1943 void __dev_remove_pack(struct packet_type *pt);
1944 void dev_add_offload(struct packet_offload *po);
1945 void dev_remove_offload(struct packet_offload *po);
1947 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1948 unsigned short mask);
1949 struct net_device *dev_get_by_name(struct net *net, const char *name);
1950 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1951 struct net_device *__dev_get_by_name(struct net *net, const char *name);
1952 int dev_alloc_name(struct net_device *dev, const char *name);
1953 int dev_open(struct net_device *dev);
1954 int dev_close(struct net_device *dev);
1955 void dev_disable_lro(struct net_device *dev);
1956 int dev_loopback_xmit(struct sk_buff *newskb);
1957 int dev_queue_xmit(struct sk_buff *skb);
1958 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
1959 int register_netdevice(struct net_device *dev);
1960 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1961 void unregister_netdevice_many(struct list_head *head);
1962 static inline void unregister_netdevice(struct net_device *dev)
1964 unregister_netdevice_queue(dev, NULL);
1967 int netdev_refcnt_read(const struct net_device *dev);
1968 void free_netdev(struct net_device *dev);
1969 void netdev_freemem(struct net_device *dev);
1970 void synchronize_net(void);
1971 int init_dummy_netdev(struct net_device *dev);
1973 struct net_device *dev_get_by_index(struct net *net, int ifindex);
1974 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1975 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1976 int netdev_get_name(struct net *net, char *name, int ifindex);
1977 int dev_restart(struct net_device *dev);
1978 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1980 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1982 return NAPI_GRO_CB(skb)->data_offset;
1985 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1987 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1990 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1992 NAPI_GRO_CB(skb)->data_offset += len;
1995 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1996 unsigned int offset)
1998 return NAPI_GRO_CB(skb)->frag0 + offset;
2001 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2003 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2006 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2007 unsigned int offset)
2009 if (!pskb_may_pull(skb, hlen))
2012 NAPI_GRO_CB(skb)->frag0 = NULL;
2013 NAPI_GRO_CB(skb)->frag0_len = 0;
2014 return skb->data + offset;
2017 static inline void *skb_gro_network_header(struct sk_buff *skb)
2019 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2020 skb_network_offset(skb);
2023 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2024 const void *start, unsigned int len)
2026 if (skb->ip_summed == CHECKSUM_COMPLETE)
2027 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2028 csum_partial(start, len, 0));
2031 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2032 unsigned short type,
2033 const void *daddr, const void *saddr,
2036 if (!dev->header_ops || !dev->header_ops->create)
2039 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2042 static inline int dev_parse_header(const struct sk_buff *skb,
2043 unsigned char *haddr)
2045 const struct net_device *dev = skb->dev;
2047 if (!dev->header_ops || !dev->header_ops->parse)
2049 return dev->header_ops->parse(skb, haddr);
2052 static inline int dev_rebuild_header(struct sk_buff *skb)
2054 const struct net_device *dev = skb->dev;
2056 if (!dev->header_ops || !dev->header_ops->rebuild)
2058 return dev->header_ops->rebuild(skb);
2061 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2062 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2063 static inline int unregister_gifconf(unsigned int family)
2065 return register_gifconf(family, NULL);
2068 #ifdef CONFIG_NET_FLOW_LIMIT
2069 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2070 struct sd_flow_limit {
2072 unsigned int num_buckets;
2073 unsigned int history_head;
2074 u16 history[FLOW_LIMIT_HISTORY];
2078 extern int netdev_flow_limit_table_len;
2079 #endif /* CONFIG_NET_FLOW_LIMIT */
2082 * Incoming packets are placed on per-cpu queues
2084 struct softnet_data {
2085 struct Qdisc *output_queue;
2086 struct Qdisc **output_queue_tailp;
2087 struct list_head poll_list;
2088 struct sk_buff *completion_queue;
2089 struct sk_buff_head process_queue;
2092 unsigned int processed;
2093 unsigned int time_squeeze;
2094 unsigned int cpu_collision;
2095 unsigned int received_rps;
2098 struct softnet_data *rps_ipi_list;
2100 /* Elements below can be accessed between CPUs for RPS */
2101 struct call_single_data csd ____cacheline_aligned_in_smp;
2102 struct softnet_data *rps_ipi_next;
2104 unsigned int input_queue_head;
2105 unsigned int input_queue_tail;
2107 unsigned int dropped;
2108 struct sk_buff_head input_pkt_queue;
2109 struct napi_struct backlog;
2111 #ifdef CONFIG_NET_FLOW_LIMIT
2112 struct sd_flow_limit __rcu *flow_limit;
2116 static inline void input_queue_head_incr(struct softnet_data *sd)
2119 sd->input_queue_head++;
2123 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2124 unsigned int *qtail)
2127 *qtail = ++sd->input_queue_tail;
2131 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2133 void __netif_schedule(struct Qdisc *q);
2135 static inline void netif_schedule_queue(struct netdev_queue *txq)
2137 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
2138 __netif_schedule(txq->qdisc);
2141 static inline void netif_tx_schedule_all(struct net_device *dev)
2145 for (i = 0; i < dev->num_tx_queues; i++)
2146 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2149 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2151 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2155 * netif_start_queue - allow transmit
2156 * @dev: network device
2158 * Allow upper layers to call the device hard_start_xmit routine.
2160 static inline void netif_start_queue(struct net_device *dev)
2162 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2165 static inline void netif_tx_start_all_queues(struct net_device *dev)
2169 for (i = 0; i < dev->num_tx_queues; i++) {
2170 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2171 netif_tx_start_queue(txq);
2175 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
2177 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
2178 __netif_schedule(dev_queue->qdisc);
2182 * netif_wake_queue - restart transmit
2183 * @dev: network device
2185 * Allow upper layers to call the device hard_start_xmit routine.
2186 * Used for flow control when transmit resources are available.
2188 static inline void netif_wake_queue(struct net_device *dev)
2190 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2193 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2197 for (i = 0; i < dev->num_tx_queues; i++) {
2198 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2199 netif_tx_wake_queue(txq);
2203 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2205 if (WARN_ON(!dev_queue)) {
2206 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2209 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2213 * netif_stop_queue - stop transmitted packets
2214 * @dev: network device
2216 * Stop upper layers calling the device hard_start_xmit routine.
2217 * Used for flow control when transmit resources are unavailable.
2219 static inline void netif_stop_queue(struct net_device *dev)
2221 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2224 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2228 for (i = 0; i < dev->num_tx_queues; i++) {
2229 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2230 netif_tx_stop_queue(txq);
2234 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2236 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2240 * netif_queue_stopped - test if transmit queue is flowblocked
2241 * @dev: network device
2243 * Test if transmit queue on device is currently unable to send.
2245 static inline bool netif_queue_stopped(const struct net_device *dev)
2247 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2250 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2252 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2255 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2257 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2260 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2264 dql_queued(&dev_queue->dql, bytes);
2266 if (likely(dql_avail(&dev_queue->dql) >= 0))
2269 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2272 * The XOFF flag must be set before checking the dql_avail below,
2273 * because in netdev_tx_completed_queue we update the dql_completed
2274 * before checking the XOFF flag.
2278 /* check again in case another CPU has just made room avail */
2279 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2280 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2285 * netdev_sent_queue - report the number of bytes queued to hardware
2286 * @dev: network device
2287 * @bytes: number of bytes queued to the hardware device queue
2289 * Report the number of bytes queued for sending/completion to the network
2290 * device hardware queue. @bytes should be a good approximation and should
2291 * exactly match netdev_completed_queue() @bytes
2293 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2295 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2298 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2299 unsigned int pkts, unsigned int bytes)
2302 if (unlikely(!bytes))
2305 dql_completed(&dev_queue->dql, bytes);
2308 * Without the memory barrier there is a small possiblity that
2309 * netdev_tx_sent_queue will miss the update and cause the queue to
2310 * be stopped forever
2314 if (dql_avail(&dev_queue->dql) < 0)
2317 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2318 netif_schedule_queue(dev_queue);
2323 * netdev_completed_queue - report bytes and packets completed by device
2324 * @dev: network device
2325 * @pkts: actual number of packets sent over the medium
2326 * @bytes: actual number of bytes sent over the medium
2328 * Report the number of bytes and packets transmitted by the network device
2329 * hardware queue over the physical medium, @bytes must exactly match the
2330 * @bytes amount passed to netdev_sent_queue()
2332 static inline void netdev_completed_queue(struct net_device *dev,
2333 unsigned int pkts, unsigned int bytes)
2335 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2338 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2341 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2347 * netdev_reset_queue - reset the packets and bytes count of a network device
2348 * @dev_queue: network device
2350 * Reset the bytes and packet count of a network device and clear the
2351 * software flow control OFF bit for this network device
2353 static inline void netdev_reset_queue(struct net_device *dev_queue)
2355 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2359 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2360 * @dev: network device
2361 * @queue_index: given tx queue index
2363 * Returns 0 if given tx queue index >= number of device tx queues,
2364 * otherwise returns the originally passed tx queue index.
2366 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2368 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2369 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2370 dev->name, queue_index,
2371 dev->real_num_tx_queues);
2379 * netif_running - test if up
2380 * @dev: network device
2382 * Test if the device has been brought up.
2384 static inline bool netif_running(const struct net_device *dev)
2386 return test_bit(__LINK_STATE_START, &dev->state);
2390 * Routines to manage the subqueues on a device. We only need start
2391 * stop, and a check if it's stopped. All other device management is
2392 * done at the overall netdevice level.
2393 * Also test the device if we're multiqueue.
2397 * netif_start_subqueue - allow sending packets on subqueue
2398 * @dev: network device
2399 * @queue_index: sub queue index
2401 * Start individual transmit queue of a device with multiple transmit queues.
2403 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2405 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2407 netif_tx_start_queue(txq);
2411 * netif_stop_subqueue - stop sending packets on subqueue
2412 * @dev: network device
2413 * @queue_index: sub queue index
2415 * Stop individual transmit queue of a device with multiple transmit queues.
2417 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2419 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2420 netif_tx_stop_queue(txq);
2424 * netif_subqueue_stopped - test status of subqueue
2425 * @dev: network device
2426 * @queue_index: sub queue index
2428 * Check individual transmit queue of a device with multiple transmit queues.
2430 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2433 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2435 return netif_tx_queue_stopped(txq);
2438 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2439 struct sk_buff *skb)
2441 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2445 * netif_wake_subqueue - allow sending packets on subqueue
2446 * @dev: network device
2447 * @queue_index: sub queue index
2449 * Resume individual transmit queue of a device with multiple transmit queues.
2451 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2453 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2454 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2455 __netif_schedule(txq->qdisc);
2459 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2462 static inline int netif_set_xps_queue(struct net_device *dev,
2463 const struct cpumask *mask,
2471 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2472 * as a distribution range limit for the returned value.
2474 static inline u16 skb_tx_hash(const struct net_device *dev,
2475 const struct sk_buff *skb)
2477 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2481 * netif_is_multiqueue - test if device has multiple transmit queues
2482 * @dev: network device
2484 * Check if device has multiple transmit queues
2486 static inline bool netif_is_multiqueue(const struct net_device *dev)
2488 return dev->num_tx_queues > 1;
2491 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2494 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2496 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2503 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2504 const struct net_device *from_dev)
2508 err = netif_set_real_num_tx_queues(to_dev,
2509 from_dev->real_num_tx_queues);
2513 return netif_set_real_num_rx_queues(to_dev,
2514 from_dev->real_num_rx_queues);
2521 static inline unsigned int get_netdev_rx_queue_index(
2522 struct netdev_rx_queue *queue)
2524 struct net_device *dev = queue->dev;
2525 int index = queue - dev->_rx;
2527 BUG_ON(index >= dev->num_rx_queues);
2532 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2533 int netif_get_num_default_rss_queues(void);
2535 enum skb_free_reason {
2536 SKB_REASON_CONSUMED,
2540 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2541 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2544 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2545 * interrupt context or with hardware interrupts being disabled.
2546 * (in_irq() || irqs_disabled())
2548 * We provide four helpers that can be used in following contexts :
2550 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2551 * replacing kfree_skb(skb)
2553 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2554 * Typically used in place of consume_skb(skb) in TX completion path
2556 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2557 * replacing kfree_skb(skb)
2559 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2560 * and consumed a packet. Used in place of consume_skb(skb)
2562 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2564 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2567 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2569 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2572 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2574 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2577 static inline void dev_consume_skb_any(struct sk_buff *skb)
2579 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2582 int netif_rx(struct sk_buff *skb);
2583 int netif_rx_ni(struct sk_buff *skb);
2584 int netif_receive_skb(struct sk_buff *skb);
2585 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2586 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2587 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2588 gro_result_t napi_gro_frags(struct napi_struct *napi);
2589 struct packet_offload *gro_find_receive_by_type(__be16 type);
2590 struct packet_offload *gro_find_complete_by_type(__be16 type);
2592 static inline void napi_free_frags(struct napi_struct *napi)
2594 kfree_skb(napi->skb);
2598 int netdev_rx_handler_register(struct net_device *dev,
2599 rx_handler_func_t *rx_handler,
2600 void *rx_handler_data);
2601 void netdev_rx_handler_unregister(struct net_device *dev);
2603 bool dev_valid_name(const char *name);
2604 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2605 int dev_ethtool(struct net *net, struct ifreq *);
2606 unsigned int dev_get_flags(const struct net_device *);
2607 int __dev_change_flags(struct net_device *, unsigned int flags);
2608 int dev_change_flags(struct net_device *, unsigned int);
2609 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2610 unsigned int gchanges);
2611 int dev_change_name(struct net_device *, const char *);
2612 int dev_set_alias(struct net_device *, const char *, size_t);
2613 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2614 int dev_set_mtu(struct net_device *, int);
2615 void dev_set_group(struct net_device *, int);
2616 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2617 int dev_change_carrier(struct net_device *, bool new_carrier);
2618 int dev_get_phys_port_id(struct net_device *dev,
2619 struct netdev_phys_port_id *ppid);
2620 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2621 struct netdev_queue *txq);
2622 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2623 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2625 extern int netdev_budget;
2627 /* Called by rtnetlink.c:rtnl_unlock() */
2628 void netdev_run_todo(void);
2631 * dev_put - release reference to device
2632 * @dev: network device
2634 * Release reference to device to allow it to be freed.
2636 static inline void dev_put(struct net_device *dev)
2638 this_cpu_dec(*dev->pcpu_refcnt);
2642 * dev_hold - get reference to device
2643 * @dev: network device
2645 * Hold reference to device to keep it from being freed.
2647 static inline void dev_hold(struct net_device *dev)
2649 this_cpu_inc(*dev->pcpu_refcnt);
2652 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2653 * and _off may be called from IRQ context, but it is caller
2654 * who is responsible for serialization of these calls.
2656 * The name carrier is inappropriate, these functions should really be
2657 * called netif_lowerlayer_*() because they represent the state of any
2658 * kind of lower layer not just hardware media.
2661 void linkwatch_init_dev(struct net_device *dev);
2662 void linkwatch_fire_event(struct net_device *dev);
2663 void linkwatch_forget_dev(struct net_device *dev);
2666 * netif_carrier_ok - test if carrier present
2667 * @dev: network device
2669 * Check if carrier is present on device
2671 static inline bool netif_carrier_ok(const struct net_device *dev)
2673 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2676 unsigned long dev_trans_start(struct net_device *dev);
2678 void __netdev_watchdog_up(struct net_device *dev);
2680 void netif_carrier_on(struct net_device *dev);
2682 void netif_carrier_off(struct net_device *dev);
2685 * netif_dormant_on - mark device as dormant.
2686 * @dev: network device
2688 * Mark device as dormant (as per RFC2863).
2690 * The dormant state indicates that the relevant interface is not
2691 * actually in a condition to pass packets (i.e., it is not 'up') but is
2692 * in a "pending" state, waiting for some external event. For "on-
2693 * demand" interfaces, this new state identifies the situation where the
2694 * interface is waiting for events to place it in the up state.
2697 static inline void netif_dormant_on(struct net_device *dev)
2699 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2700 linkwatch_fire_event(dev);
2704 * netif_dormant_off - set device as not dormant.
2705 * @dev: network device
2707 * Device is not in dormant state.
2709 static inline void netif_dormant_off(struct net_device *dev)
2711 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2712 linkwatch_fire_event(dev);
2716 * netif_dormant - test if carrier present
2717 * @dev: network device
2719 * Check if carrier is present on device
2721 static inline bool netif_dormant(const struct net_device *dev)
2723 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2728 * netif_oper_up - test if device is operational
2729 * @dev: network device
2731 * Check if carrier is operational
2733 static inline bool netif_oper_up(const struct net_device *dev)
2735 return (dev->operstate == IF_OPER_UP ||
2736 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2740 * netif_device_present - is device available or removed
2741 * @dev: network device
2743 * Check if device has not been removed from system.
2745 static inline bool netif_device_present(struct net_device *dev)
2747 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2750 void netif_device_detach(struct net_device *dev);
2752 void netif_device_attach(struct net_device *dev);
2755 * Network interface message level settings
2759 NETIF_MSG_DRV = 0x0001,
2760 NETIF_MSG_PROBE = 0x0002,
2761 NETIF_MSG_LINK = 0x0004,
2762 NETIF_MSG_TIMER = 0x0008,
2763 NETIF_MSG_IFDOWN = 0x0010,
2764 NETIF_MSG_IFUP = 0x0020,
2765 NETIF_MSG_RX_ERR = 0x0040,
2766 NETIF_MSG_TX_ERR = 0x0080,
2767 NETIF_MSG_TX_QUEUED = 0x0100,
2768 NETIF_MSG_INTR = 0x0200,
2769 NETIF_MSG_TX_DONE = 0x0400,
2770 NETIF_MSG_RX_STATUS = 0x0800,
2771 NETIF_MSG_PKTDATA = 0x1000,
2772 NETIF_MSG_HW = 0x2000,
2773 NETIF_MSG_WOL = 0x4000,
2776 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2777 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2778 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2779 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2780 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2781 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2782 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2783 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2784 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2785 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2786 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2787 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2788 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2789 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2790 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2792 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2795 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2796 return default_msg_enable_bits;
2797 if (debug_value == 0) /* no output */
2799 /* set low N bits */
2800 return (1 << debug_value) - 1;
2803 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2805 spin_lock(&txq->_xmit_lock);
2806 txq->xmit_lock_owner = cpu;
2809 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2811 spin_lock_bh(&txq->_xmit_lock);
2812 txq->xmit_lock_owner = smp_processor_id();
2815 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2817 bool ok = spin_trylock(&txq->_xmit_lock);
2819 txq->xmit_lock_owner = smp_processor_id();
2823 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2825 txq->xmit_lock_owner = -1;
2826 spin_unlock(&txq->_xmit_lock);
2829 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2831 txq->xmit_lock_owner = -1;
2832 spin_unlock_bh(&txq->_xmit_lock);
2835 static inline void txq_trans_update(struct netdev_queue *txq)
2837 if (txq->xmit_lock_owner != -1)
2838 txq->trans_start = jiffies;
2842 * netif_tx_lock - grab network device transmit lock
2843 * @dev: network device
2845 * Get network device transmit lock
2847 static inline void netif_tx_lock(struct net_device *dev)
2852 spin_lock(&dev->tx_global_lock);
2853 cpu = smp_processor_id();
2854 for (i = 0; i < dev->num_tx_queues; i++) {
2855 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2857 /* We are the only thread of execution doing a
2858 * freeze, but we have to grab the _xmit_lock in
2859 * order to synchronize with threads which are in
2860 * the ->hard_start_xmit() handler and already
2861 * checked the frozen bit.
2863 __netif_tx_lock(txq, cpu);
2864 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2865 __netif_tx_unlock(txq);
2869 static inline void netif_tx_lock_bh(struct net_device *dev)
2875 static inline void netif_tx_unlock(struct net_device *dev)
2879 for (i = 0; i < dev->num_tx_queues; i++) {
2880 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2882 /* No need to grab the _xmit_lock here. If the
2883 * queue is not stopped for another reason, we
2886 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2887 netif_schedule_queue(txq);
2889 spin_unlock(&dev->tx_global_lock);
2892 static inline void netif_tx_unlock_bh(struct net_device *dev)
2894 netif_tx_unlock(dev);
2898 #define HARD_TX_LOCK(dev, txq, cpu) { \
2899 if ((dev->features & NETIF_F_LLTX) == 0) { \
2900 __netif_tx_lock(txq, cpu); \
2904 #define HARD_TX_TRYLOCK(dev, txq) \
2905 (((dev->features & NETIF_F_LLTX) == 0) ? \
2906 __netif_tx_trylock(txq) : \
2909 #define HARD_TX_UNLOCK(dev, txq) { \
2910 if ((dev->features & NETIF_F_LLTX) == 0) { \
2911 __netif_tx_unlock(txq); \
2915 static inline void netif_tx_disable(struct net_device *dev)
2921 cpu = smp_processor_id();
2922 for (i = 0; i < dev->num_tx_queues; i++) {
2923 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2925 __netif_tx_lock(txq, cpu);
2926 netif_tx_stop_queue(txq);
2927 __netif_tx_unlock(txq);
2932 static inline void netif_addr_lock(struct net_device *dev)
2934 spin_lock(&dev->addr_list_lock);
2937 static inline void netif_addr_lock_nested(struct net_device *dev)
2939 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2942 static inline void netif_addr_lock_bh(struct net_device *dev)
2944 spin_lock_bh(&dev->addr_list_lock);
2947 static inline void netif_addr_unlock(struct net_device *dev)
2949 spin_unlock(&dev->addr_list_lock);
2952 static inline void netif_addr_unlock_bh(struct net_device *dev)
2954 spin_unlock_bh(&dev->addr_list_lock);
2958 * dev_addrs walker. Should be used only for read access. Call with
2959 * rcu_read_lock held.
2961 #define for_each_dev_addr(dev, ha) \
2962 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2964 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2966 void ether_setup(struct net_device *dev);
2968 /* Support for loadable net-drivers */
2969 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2970 void (*setup)(struct net_device *),
2971 unsigned int txqs, unsigned int rxqs);
2972 #define alloc_netdev(sizeof_priv, name, setup) \
2973 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2975 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2976 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2978 int register_netdev(struct net_device *dev);
2979 void unregister_netdev(struct net_device *dev);
2981 /* General hardware address lists handling functions */
2982 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2983 struct netdev_hw_addr_list *from_list, int addr_len);
2984 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2985 struct netdev_hw_addr_list *from_list, int addr_len);
2986 void __hw_addr_init(struct netdev_hw_addr_list *list);
2988 /* Functions used for device addresses handling */
2989 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2990 unsigned char addr_type);
2991 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2992 unsigned char addr_type);
2993 void dev_addr_flush(struct net_device *dev);
2994 int dev_addr_init(struct net_device *dev);
2996 /* Functions used for unicast addresses handling */
2997 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2998 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2999 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3000 int dev_uc_sync(struct net_device *to, struct net_device *from);
3001 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3002 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3003 void dev_uc_flush(struct net_device *dev);
3004 void dev_uc_init(struct net_device *dev);
3006 /* Functions used for multicast addresses handling */
3007 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3008 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3009 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3010 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3011 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3012 int dev_mc_sync(struct net_device *to, struct net_device *from);
3013 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3014 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3015 void dev_mc_flush(struct net_device *dev);
3016 void dev_mc_init(struct net_device *dev);
3018 /* Functions used for secondary unicast and multicast support */
3019 void dev_set_rx_mode(struct net_device *dev);
3020 void __dev_set_rx_mode(struct net_device *dev);
3021 int dev_set_promiscuity(struct net_device *dev, int inc);
3022 int dev_set_allmulti(struct net_device *dev, int inc);
3023 void netdev_state_change(struct net_device *dev);
3024 void netdev_notify_peers(struct net_device *dev);
3025 void netdev_features_change(struct net_device *dev);
3026 /* Load a device via the kmod */
3027 void dev_load(struct net *net, const char *name);
3028 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3029 struct rtnl_link_stats64 *storage);
3030 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3031 const struct net_device_stats *netdev_stats);
3033 extern int netdev_max_backlog;
3034 extern int netdev_tstamp_prequeue;
3035 extern int weight_p;
3036 extern int bpf_jit_enable;
3038 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3039 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3040 struct list_head **iter);
3042 /* iterate through upper list, must be called under RCU read lock */
3043 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3044 for (iter = &(dev)->all_adj_list.upper, \
3045 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3047 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3049 void *netdev_lower_get_next_private(struct net_device *dev,
3050 struct list_head **iter);
3051 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3052 struct list_head **iter);
3054 #define netdev_for_each_lower_private(dev, priv, iter) \
3055 for (iter = (dev)->adj_list.lower.next, \
3056 priv = netdev_lower_get_next_private(dev, &(iter)); \
3058 priv = netdev_lower_get_next_private(dev, &(iter)))
3060 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3061 for (iter = &(dev)->adj_list.lower, \
3062 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3064 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3066 void *netdev_adjacent_get_private(struct list_head *adj_list);
3067 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3068 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3069 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3070 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3071 int netdev_master_upper_dev_link(struct net_device *dev,
3072 struct net_device *upper_dev);
3073 int netdev_master_upper_dev_link_private(struct net_device *dev,
3074 struct net_device *upper_dev,
3076 void netdev_upper_dev_unlink(struct net_device *dev,
3077 struct net_device *upper_dev);
3078 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3079 void *netdev_lower_dev_get_private(struct net_device *dev,
3080 struct net_device *lower_dev);
3081 int skb_checksum_help(struct sk_buff *skb);
3082 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3083 netdev_features_t features, bool tx_path);
3084 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3085 netdev_features_t features);
3088 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3090 return __skb_gso_segment(skb, features, true);
3092 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3094 static inline bool can_checksum_protocol(netdev_features_t features,
3097 return ((features & NETIF_F_GEN_CSUM) ||
3098 ((features & NETIF_F_V4_CSUM) &&
3099 protocol == htons(ETH_P_IP)) ||
3100 ((features & NETIF_F_V6_CSUM) &&
3101 protocol == htons(ETH_P_IPV6)) ||
3102 ((features & NETIF_F_FCOE_CRC) &&
3103 protocol == htons(ETH_P_FCOE)));
3107 void netdev_rx_csum_fault(struct net_device *dev);
3109 static inline void netdev_rx_csum_fault(struct net_device *dev)
3113 /* rx skb timestamps */
3114 void net_enable_timestamp(void);
3115 void net_disable_timestamp(void);
3117 #ifdef CONFIG_PROC_FS
3118 int __init dev_proc_init(void);
3120 #define dev_proc_init() 0
3123 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3125 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3128 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3130 return netdev_class_create_file_ns(class_attr, NULL);
3133 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3135 netdev_class_remove_file_ns(class_attr, NULL);
3138 extern struct kobj_ns_type_operations net_ns_type_operations;
3140 const char *netdev_drivername(const struct net_device *dev);
3142 void linkwatch_run_queue(void);
3144 static inline netdev_features_t netdev_get_wanted_features(
3145 struct net_device *dev)
3147 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3149 netdev_features_t netdev_increment_features(netdev_features_t all,
3150 netdev_features_t one, netdev_features_t mask);
3152 /* Allow TSO being used on stacked device :
3153 * Performing the GSO segmentation before last device
3154 * is a performance improvement.
3156 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3157 netdev_features_t mask)
3159 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3162 int __netdev_update_features(struct net_device *dev);
3163 void netdev_update_features(struct net_device *dev);
3164 void netdev_change_features(struct net_device *dev);
3166 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3167 struct net_device *dev);
3169 netdev_features_t netif_skb_dev_features(struct sk_buff *skb,
3170 const struct net_device *dev);
3171 static inline netdev_features_t netif_skb_features(struct sk_buff *skb)
3173 return netif_skb_dev_features(skb, skb->dev);
3176 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3178 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3180 /* check flags correspondence */
3181 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3182 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3183 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3184 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3185 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3186 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3188 return (features & feature) == feature;
3191 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3193 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3194 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3197 static inline bool netif_needs_gso(struct sk_buff *skb,
3198 netdev_features_t features)
3200 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3201 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3202 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3205 static inline void netif_set_gso_max_size(struct net_device *dev,
3208 dev->gso_max_size = size;
3211 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3212 int pulled_hlen, u16 mac_offset,
3215 skb->protocol = protocol;
3216 skb->encapsulation = 1;
3217 skb_push(skb, pulled_hlen);
3218 skb_reset_transport_header(skb);
3219 skb->mac_header = mac_offset;
3220 skb->network_header = skb->mac_header + mac_len;
3221 skb->mac_len = mac_len;
3224 static inline bool netif_is_macvlan(struct net_device *dev)
3226 return dev->priv_flags & IFF_MACVLAN;
3229 static inline bool netif_is_bond_master(struct net_device *dev)
3231 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3234 static inline bool netif_is_bond_slave(struct net_device *dev)
3236 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3239 static inline bool netif_supports_nofcs(struct net_device *dev)
3241 return dev->priv_flags & IFF_SUPP_NOFCS;
3244 extern struct pernet_operations __net_initdata loopback_net_ops;
3246 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3248 /* netdev_printk helpers, similar to dev_printk */
3250 static inline const char *netdev_name(const struct net_device *dev)
3252 if (dev->reg_state != NETREG_REGISTERED)
3253 return "(unregistered net_device)";
3258 int netdev_printk(const char *level, const struct net_device *dev,
3259 const char *format, ...);
3261 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3263 int netdev_alert(const struct net_device *dev, const char *format, ...);
3265 int netdev_crit(const struct net_device *dev, const char *format, ...);
3267 int netdev_err(const struct net_device *dev, const char *format, ...);
3269 int netdev_warn(const struct net_device *dev, const char *format, ...);
3271 int netdev_notice(const struct net_device *dev, const char *format, ...);
3273 int netdev_info(const struct net_device *dev, const char *format, ...);
3275 #define MODULE_ALIAS_NETDEV(device) \
3276 MODULE_ALIAS("netdev-" device)
3278 #if defined(CONFIG_DYNAMIC_DEBUG)
3279 #define netdev_dbg(__dev, format, args...) \
3281 dynamic_netdev_dbg(__dev, format, ##args); \
3283 #elif defined(DEBUG)
3284 #define netdev_dbg(__dev, format, args...) \
3285 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3287 #define netdev_dbg(__dev, format, args...) \
3290 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3295 #if defined(VERBOSE_DEBUG)
3296 #define netdev_vdbg netdev_dbg
3299 #define netdev_vdbg(dev, format, args...) \
3302 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3308 * netdev_WARN() acts like dev_printk(), but with the key difference
3309 * of using a WARN/WARN_ON to get the message out, including the
3310 * file/line information and a backtrace.
3312 #define netdev_WARN(dev, format, args...) \
3313 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3315 /* netif printk helpers, similar to netdev_printk */
3317 #define netif_printk(priv, type, level, dev, fmt, args...) \
3319 if (netif_msg_##type(priv)) \
3320 netdev_printk(level, (dev), fmt, ##args); \
3323 #define netif_level(level, priv, type, dev, fmt, args...) \
3325 if (netif_msg_##type(priv)) \
3326 netdev_##level(dev, fmt, ##args); \
3329 #define netif_emerg(priv, type, dev, fmt, args...) \
3330 netif_level(emerg, priv, type, dev, fmt, ##args)
3331 #define netif_alert(priv, type, dev, fmt, args...) \
3332 netif_level(alert, priv, type, dev, fmt, ##args)
3333 #define netif_crit(priv, type, dev, fmt, args...) \
3334 netif_level(crit, priv, type, dev, fmt, ##args)
3335 #define netif_err(priv, type, dev, fmt, args...) \
3336 netif_level(err, priv, type, dev, fmt, ##args)
3337 #define netif_warn(priv, type, dev, fmt, args...) \
3338 netif_level(warn, priv, type, dev, fmt, ##args)
3339 #define netif_notice(priv, type, dev, fmt, args...) \
3340 netif_level(notice, priv, type, dev, fmt, ##args)
3341 #define netif_info(priv, type, dev, fmt, args...) \
3342 netif_level(info, priv, type, dev, fmt, ##args)
3344 #if defined(CONFIG_DYNAMIC_DEBUG)
3345 #define netif_dbg(priv, type, netdev, format, args...) \
3347 if (netif_msg_##type(priv)) \
3348 dynamic_netdev_dbg(netdev, format, ##args); \
3350 #elif defined(DEBUG)
3351 #define netif_dbg(priv, type, dev, format, args...) \
3352 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3354 #define netif_dbg(priv, type, dev, format, args...) \
3357 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3362 #if defined(VERBOSE_DEBUG)
3363 #define netif_vdbg netif_dbg
3365 #define netif_vdbg(priv, type, dev, format, args...) \
3368 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3374 * The list of packet types we will receive (as opposed to discard)
3375 * and the routines to invoke.
3377 * Why 16. Because with 16 the only overlap we get on a hash of the
3378 * low nibble of the protocol value is RARP/SNAP/X.25.
3380 * NOTE: That is no longer true with the addition of VLAN tags. Not
3381 * sure which should go first, but I bet it won't make much
3382 * difference if we are running VLANs. The good news is that
3383 * this protocol won't be in the list unless compiled in, so
3384 * the average user (w/out VLANs) will not be adversely affected.
3400 #define PTYPE_HASH_SIZE (16)
3401 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3403 #endif /* _LINUX_NETDEVICE_H */