2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
47 #include <net/dcbnl.h>
49 #include <net/netprio_cgroup.h>
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54 #include <uapi/linux/if_bonding.h>
61 /* 802.15.4 specific */
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
72 * Transmit return codes: transmit return codes originate from three different
75 * - qdisc return codes
76 * - driver transmit return codes
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
110 typedef enum netdev_tx netdev_tx_t;
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 static inline bool dev_xmit_complete(int rc)
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 if (likely(rc < NET_XMIT_MASK))
131 * Compute the worst case header length according to the protocols
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
139 # define LL_MAX_HEADER 96
142 # define LL_MAX_HEADER 32
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
157 struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed;
196 struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
209 struct rcu_head rcu_head;
212 struct netdev_hw_addr_list {
213 struct list_head list;
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
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 hrtimer timer;
320 struct list_head dev_list;
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
326 NAPI_STATE_SCHED, /* Poll is scheduled */
327 NAPI_STATE_DISABLE, /* Disable pending */
328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED, /* In NAPI hash */
339 typedef enum gro_result gro_result_t;
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
382 enum rx_handler_result {
388 typedef enum rx_handler_result rx_handler_result_t;
389 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
391 void __napi_schedule(struct napi_struct *n);
392 void __napi_schedule_irqoff(struct napi_struct *n);
394 static inline bool napi_disable_pending(struct napi_struct *n)
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
400 * napi_schedule_prep - check if napi can be scheduled
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
408 static inline bool napi_schedule_prep(struct napi_struct *n)
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
415 * napi_schedule - schedule NAPI poll
418 * Schedule NAPI poll routine to be called if it is not already
421 static inline void napi_schedule(struct napi_struct *n)
423 if (napi_schedule_prep(n))
428 * napi_schedule_irqoff - schedule NAPI poll
431 * Variant of napi_schedule(), assuming hard irqs are masked.
433 static inline void napi_schedule_irqoff(struct napi_struct *n)
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440 static inline bool napi_reschedule(struct napi_struct *napi)
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
449 void __napi_complete(struct napi_struct *n);
450 void napi_complete_done(struct napi_struct *n, int work_done);
452 * napi_complete - NAPI processing complete
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
458 static inline void napi_complete(struct napi_struct *n)
460 return napi_complete_done(n, 0);
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
470 struct napi_struct *napi_by_id(unsigned int napi_id);
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
476 * generate a new napi_id and store a @napi under it in napi_hash
478 void napi_hash_add(struct napi_struct *napi);
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
487 void napi_hash_del(struct napi_struct *napi);
490 * napi_disable - prevent NAPI from scheduling
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
496 void napi_disable(struct napi_struct *n);
499 * napi_enable - enable NAPI scheduling
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
505 static inline void napi_enable(struct napi_struct *n)
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED, &n->state);
514 * napi_synchronize - wait until NAPI is not running
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
521 static inline void napi_synchronize(const struct napi_struct *n)
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
527 # define napi_synchronize(n) barrier()
530 enum netdev_queue_state_t {
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
533 __QUEUE_STATE_FROZEN,
536 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
540 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
543 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
556 struct netdev_queue {
560 struct net_device *dev;
561 struct Qdisc __rcu *qdisc;
562 struct Qdisc *qdisc_sleeping;
566 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 * please use this field instead of dev->trans_start
577 unsigned long trans_start;
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
583 unsigned long trans_timeout;
590 } ____cacheline_aligned_in_smp;
592 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 * This structure holds an RPS map which can be of variable length. The
611 * map is an array of CPUs.
618 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
621 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
622 * tail pointer for that CPU's input queue at the time of last enqueue, and
623 * a hardware filter index.
625 struct rps_dev_flow {
628 unsigned int last_qtail;
630 #define RPS_NO_FILTER 0xffff
633 * The rps_dev_flow_table structure contains a table of flow mappings.
635 struct rps_dev_flow_table {
638 struct rps_dev_flow flows[0];
640 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
641 ((_num) * sizeof(struct rps_dev_flow)))
644 * The rps_sock_flow_table contains mappings of flows to the last CPU
645 * on which they were processed by the application (set in recvmsg).
646 * Each entry is a 32bit value. Upper part is the high order bits
647 * of flow hash, lower part is cpu number.
648 * rps_cpu_mask is used to partition the space, depending on number of
649 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
650 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
651 * meaning we use 32-6=26 bits for the hash.
653 struct rps_sock_flow_table {
656 u32 ents[0] ____cacheline_aligned_in_smp;
658 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
660 #define RPS_NO_CPU 0xffff
662 extern u32 rps_cpu_mask;
663 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
669 unsigned int index = hash & table->mask;
670 u32 val = hash & ~rps_cpu_mask;
672 /* We only give a hint, preemption can change cpu under us */
673 val |= raw_smp_processor_id();
675 if (table->ents[index] != val)
676 table->ents[index] = val;
680 #ifdef CONFIG_RFS_ACCEL
681 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
684 #endif /* CONFIG_RPS */
686 /* This structure contains an instance of an RX queue. */
687 struct netdev_rx_queue {
689 struct rps_map __rcu *rps_map;
690 struct rps_dev_flow_table __rcu *rps_flow_table;
693 struct net_device *dev;
694 } ____cacheline_aligned_in_smp;
697 * RX queue sysfs structures and functions.
699 struct rx_queue_attribute {
700 struct attribute attr;
701 ssize_t (*show)(struct netdev_rx_queue *queue,
702 struct rx_queue_attribute *attr, char *buf);
703 ssize_t (*store)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, const char *buf, size_t len);
709 * This structure holds an XPS map which can be of variable length. The
710 * map is an array of queues.
714 unsigned int alloc_len;
718 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
719 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
723 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 struct xps_dev_maps {
727 struct xps_map __rcu *cpu_map[0];
729 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
730 (nr_cpu_ids * sizeof(struct xps_map *)))
731 #endif /* CONFIG_XPS */
733 #define TC_MAX_QUEUE 16
734 #define TC_BITMASK 15
735 /* HW offloaded queuing disciplines txq count and offset maps */
736 struct netdev_tc_txq {
741 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743 * This structure is to hold information about the device
744 * configured to run FCoE protocol stack.
746 struct netdev_fcoe_hbainfo {
747 char manufacturer[64];
748 char serial_number[64];
749 char hardware_version[64];
750 char driver_version[64];
751 char optionrom_version[64];
752 char firmware_version[64];
754 char model_description[256];
758 #define MAX_PHYS_ITEM_ID_LEN 32
760 /* This structure holds a unique identifier to identify some
761 * physical item (port for example) used by a netdevice.
763 struct netdev_phys_item_id {
764 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
765 unsigned char id_len;
768 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
769 struct sk_buff *skb);
774 * This structure defines the management hooks for network devices.
775 * The following hooks can be defined; unless noted otherwise, they are
776 * optional and can be filled with a null pointer.
778 * int (*ndo_init)(struct net_device *dev);
779 * This function is called once when network device is registered.
780 * The network device can use this to any late stage initializaton
781 * or semantic validattion. It can fail with an error code which will
782 * be propogated back to register_netdev
784 * void (*ndo_uninit)(struct net_device *dev);
785 * This function is called when device is unregistered or when registration
786 * fails. It is not called if init fails.
788 * int (*ndo_open)(struct net_device *dev);
789 * This function is called when network device transistions to the up
792 * int (*ndo_stop)(struct net_device *dev);
793 * This function is called when network device transistions to the down
796 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
797 * struct net_device *dev);
798 * Called when a packet needs to be transmitted.
799 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
800 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
801 * Required can not be NULL.
803 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
804 * void *accel_priv, select_queue_fallback_t fallback);
805 * Called to decide which queue to when device supports multiple
808 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
809 * This function is called to allow device receiver to make
810 * changes to configuration when multicast or promiscious is enabled.
812 * void (*ndo_set_rx_mode)(struct net_device *dev);
813 * This function is called device changes address list filtering.
814 * If driver handles unicast address filtering, it should set
815 * IFF_UNICAST_FLT to its priv_flags.
817 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
818 * This function is called when the Media Access Control address
819 * needs to be changed. If this interface is not defined, the
820 * mac address can not be changed.
822 * int (*ndo_validate_addr)(struct net_device *dev);
823 * Test if Media Access Control address is valid for the device.
825 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
826 * Called when a user request an ioctl which can't be handled by
827 * the generic interface code. If not defined ioctl's return
828 * not supported error code.
830 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
831 * Used to set network devices bus interface parameters. This interface
832 * is retained for legacy reason, new devices should use the bus
833 * interface (PCI) for low level management.
835 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
836 * Called when a user wants to change the Maximum Transfer Unit
837 * of a device. If not defined, any request to change MTU will
838 * will return an error.
840 * void (*ndo_tx_timeout)(struct net_device *dev);
841 * Callback uses when the transmitter has not made any progress
842 * for dev->watchdog ticks.
844 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
845 * struct rtnl_link_stats64 *storage);
846 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
847 * Called when a user wants to get the network device usage
848 * statistics. Drivers must do one of the following:
849 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
850 * rtnl_link_stats64 structure passed by the caller.
851 * 2. Define @ndo_get_stats to update a net_device_stats structure
852 * (which should normally be dev->stats) and return a pointer to
853 * it. The structure may be changed asynchronously only if each
854 * field is written atomically.
855 * 3. Update dev->stats asynchronously and atomically, and define
858 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
859 * If device support VLAN filtering this function is called when a
860 * VLAN id is registered.
862 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
863 * If device support VLAN filtering this function is called when a
864 * VLAN id is unregistered.
866 * void (*ndo_poll_controller)(struct net_device *dev);
868 * SR-IOV management functions.
869 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
870 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
871 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
873 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
874 * int (*ndo_get_vf_config)(struct net_device *dev,
875 * int vf, struct ifla_vf_info *ivf);
876 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
877 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
878 * struct nlattr *port[]);
879 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
880 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
881 * Called to setup 'tc' number of traffic classes in the net device. This
882 * is always called from the stack with the rtnl lock held and netif tx
883 * queues stopped. This allows the netdevice to perform queue management
886 * Fiber Channel over Ethernet (FCoE) offload functions.
887 * int (*ndo_fcoe_enable)(struct net_device *dev);
888 * Called when the FCoE protocol stack wants to start using LLD for FCoE
889 * so the underlying device can perform whatever needed configuration or
890 * initialization to support acceleration of FCoE traffic.
892 * int (*ndo_fcoe_disable)(struct net_device *dev);
893 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
894 * so the underlying device can perform whatever needed clean-ups to
895 * stop supporting acceleration of FCoE traffic.
897 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
898 * struct scatterlist *sgl, unsigned int sgc);
899 * Called when the FCoE Initiator wants to initialize an I/O that
900 * is a possible candidate for Direct Data Placement (DDP). The LLD can
901 * perform necessary setup and returns 1 to indicate the device is set up
902 * successfully to perform DDP on this I/O, otherwise this returns 0.
904 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
905 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
906 * indicated by the FC exchange id 'xid', so the underlying device can
907 * clean up and reuse resources for later DDP requests.
909 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
910 * struct scatterlist *sgl, unsigned int sgc);
911 * Called when the FCoE Target wants to initialize an I/O that
912 * is a possible candidate for Direct Data Placement (DDP). The LLD can
913 * perform necessary setup and returns 1 to indicate the device is set up
914 * successfully to perform DDP on this I/O, otherwise this returns 0.
916 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
917 * struct netdev_fcoe_hbainfo *hbainfo);
918 * Called when the FCoE Protocol stack wants information on the underlying
919 * device. This information is utilized by the FCoE protocol stack to
920 * register attributes with Fiber Channel management service as per the
921 * FC-GS Fabric Device Management Information(FDMI) specification.
923 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
924 * Called when the underlying device wants to override default World Wide
925 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
926 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
927 * protocol stack to use.
930 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
931 * u16 rxq_index, u32 flow_id);
932 * Set hardware filter for RFS. rxq_index is the target queue index;
933 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
934 * Return the filter ID on success, or a negative error code.
936 * Slave management functions (for bridge, bonding, etc).
937 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
938 * Called to make another netdev an underling.
940 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
941 * Called to release previously enslaved netdev.
943 * Feature/offload setting functions.
944 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
945 * netdev_features_t features);
946 * Adjusts the requested feature flags according to device-specific
947 * constraints, and returns the resulting flags. Must not modify
950 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
951 * Called to update device configuration to new features. Passed
952 * feature set might be less than what was returned by ndo_fix_features()).
953 * Must return >0 or -errno if it changed dev->features itself.
955 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
956 * struct net_device *dev,
957 * const unsigned char *addr, u16 vid, u16 flags)
958 * Adds an FDB entry to dev for addr.
959 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
960 * struct net_device *dev,
961 * const unsigned char *addr, u16 vid)
962 * Deletes the FDB entry from dev coresponding to addr.
963 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
964 * struct net_device *dev, struct net_device *filter_dev,
966 * Used to add FDB entries to dump requests. Implementers should add
967 * entries to skb and update idx with the number of entries.
969 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
970 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
971 * struct net_device *dev, u32 filter_mask)
973 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
974 * Called to change device carrier. Soft-devices (like dummy, team, etc)
975 * which do not represent real hardware may define this to allow their
976 * userspace components to manage their virtual carrier state. Devices
977 * that determine carrier state from physical hardware properties (eg
978 * network cables) or protocol-dependent mechanisms (eg
979 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
981 * int (*ndo_get_phys_port_id)(struct net_device *dev,
982 * struct netdev_phys_item_id *ppid);
983 * Called to get ID of physical port of this device. If driver does
984 * not implement this, it is assumed that the hw is not able to have
985 * multiple net devices on single physical port.
987 * void (*ndo_add_vxlan_port)(struct net_device *dev,
988 * sa_family_t sa_family, __be16 port);
989 * Called by vxlan to notiy a driver about the UDP port and socket
990 * address family that vxlan is listnening to. It is called only when
991 * a new port starts listening. The operation is protected by the
992 * vxlan_net->sock_lock.
994 * void (*ndo_del_vxlan_port)(struct net_device *dev,
995 * sa_family_t sa_family, __be16 port);
996 * Called by vxlan to notify the driver about a UDP port and socket
997 * address family that vxlan is not listening to anymore. The operation
998 * is protected by the vxlan_net->sock_lock.
1000 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1001 * struct net_device *dev)
1002 * Called by upper layer devices to accelerate switching or other
1003 * station functionality into hardware. 'pdev is the lowerdev
1004 * to use for the offload and 'dev' is the net device that will
1005 * back the offload. Returns a pointer to the private structure
1006 * the upper layer will maintain.
1007 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1008 * Called by upper layer device to delete the station created
1009 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1010 * the station and priv is the structure returned by the add
1012 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1013 * struct net_device *dev,
1015 * Callback to use for xmit over the accelerated station. This
1016 * is used in place of ndo_start_xmit on accelerated net
1018 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1019 * struct net_device *dev
1020 * netdev_features_t features);
1021 * Called by core transmit path to determine if device is capable of
1022 * performing offload operations on a given packet. This is to give
1023 * the device an opportunity to implement any restrictions that cannot
1024 * be otherwise expressed by feature flags. The check is called with
1025 * the set of features that the stack has calculated and it returns
1026 * those the driver believes to be appropriate.
1028 * int (*ndo_switch_parent_id_get)(struct net_device *dev,
1029 * struct netdev_phys_item_id *psid);
1030 * Called to get an ID of the switch chip this port is part of.
1031 * If driver implements this, it indicates that it represents a port
1033 * int (*ndo_switch_port_stp_update)(struct net_device *dev, u8 state);
1034 * Called to notify switch device port of bridge port STP
1036 * int (*ndo_sw_parent_fib_ipv4_add)(struct net_device *dev, __be32 dst,
1037 * int dst_len, struct fib_info *fi,
1038 * u8 tos, u8 type, u32 nlflags, u32 tb_id);
1039 * Called to add/modify IPv4 route to switch device.
1040 * int (*ndo_sw_parent_fib_ipv4_del)(struct net_device *dev, __be32 dst,
1041 * int dst_len, struct fib_info *fi,
1042 * u8 tos, u8 type, u32 tb_id);
1043 * Called to delete IPv4 route from switch device.
1045 struct net_device_ops {
1046 int (*ndo_init)(struct net_device *dev);
1047 void (*ndo_uninit)(struct net_device *dev);
1048 int (*ndo_open)(struct net_device *dev);
1049 int (*ndo_stop)(struct net_device *dev);
1050 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1051 struct net_device *dev);
1052 u16 (*ndo_select_queue)(struct net_device *dev,
1053 struct sk_buff *skb,
1055 select_queue_fallback_t fallback);
1056 void (*ndo_change_rx_flags)(struct net_device *dev,
1058 void (*ndo_set_rx_mode)(struct net_device *dev);
1059 int (*ndo_set_mac_address)(struct net_device *dev,
1061 int (*ndo_validate_addr)(struct net_device *dev);
1062 int (*ndo_do_ioctl)(struct net_device *dev,
1063 struct ifreq *ifr, int cmd);
1064 int (*ndo_set_config)(struct net_device *dev,
1066 int (*ndo_change_mtu)(struct net_device *dev,
1068 int (*ndo_neigh_setup)(struct net_device *dev,
1069 struct neigh_parms *);
1070 void (*ndo_tx_timeout) (struct net_device *dev);
1072 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1073 struct rtnl_link_stats64 *storage);
1074 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1076 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1077 __be16 proto, u16 vid);
1078 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1079 __be16 proto, u16 vid);
1080 #ifdef CONFIG_NET_POLL_CONTROLLER
1081 void (*ndo_poll_controller)(struct net_device *dev);
1082 int (*ndo_netpoll_setup)(struct net_device *dev,
1083 struct netpoll_info *info);
1084 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1086 #ifdef CONFIG_NET_RX_BUSY_POLL
1087 int (*ndo_busy_poll)(struct napi_struct *dev);
1089 int (*ndo_set_vf_mac)(struct net_device *dev,
1090 int queue, u8 *mac);
1091 int (*ndo_set_vf_vlan)(struct net_device *dev,
1092 int queue, u16 vlan, u8 qos);
1093 int (*ndo_set_vf_rate)(struct net_device *dev,
1094 int vf, int min_tx_rate,
1096 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1097 int vf, bool setting);
1098 int (*ndo_get_vf_config)(struct net_device *dev,
1100 struct ifla_vf_info *ivf);
1101 int (*ndo_set_vf_link_state)(struct net_device *dev,
1102 int vf, int link_state);
1103 int (*ndo_set_vf_port)(struct net_device *dev,
1105 struct nlattr *port[]);
1106 int (*ndo_get_vf_port)(struct net_device *dev,
1107 int vf, struct sk_buff *skb);
1108 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1109 #if IS_ENABLED(CONFIG_FCOE)
1110 int (*ndo_fcoe_enable)(struct net_device *dev);
1111 int (*ndo_fcoe_disable)(struct net_device *dev);
1112 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1114 struct scatterlist *sgl,
1116 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1118 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1120 struct scatterlist *sgl,
1122 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1123 struct netdev_fcoe_hbainfo *hbainfo);
1126 #if IS_ENABLED(CONFIG_LIBFCOE)
1127 #define NETDEV_FCOE_WWNN 0
1128 #define NETDEV_FCOE_WWPN 1
1129 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1130 u64 *wwn, int type);
1133 #ifdef CONFIG_RFS_ACCEL
1134 int (*ndo_rx_flow_steer)(struct net_device *dev,
1135 const struct sk_buff *skb,
1139 int (*ndo_add_slave)(struct net_device *dev,
1140 struct net_device *slave_dev);
1141 int (*ndo_del_slave)(struct net_device *dev,
1142 struct net_device *slave_dev);
1143 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1144 netdev_features_t features);
1145 int (*ndo_set_features)(struct net_device *dev,
1146 netdev_features_t features);
1147 int (*ndo_neigh_construct)(struct neighbour *n);
1148 void (*ndo_neigh_destroy)(struct neighbour *n);
1150 int (*ndo_fdb_add)(struct ndmsg *ndm,
1151 struct nlattr *tb[],
1152 struct net_device *dev,
1153 const unsigned char *addr,
1156 int (*ndo_fdb_del)(struct ndmsg *ndm,
1157 struct nlattr *tb[],
1158 struct net_device *dev,
1159 const unsigned char *addr,
1161 int (*ndo_fdb_dump)(struct sk_buff *skb,
1162 struct netlink_callback *cb,
1163 struct net_device *dev,
1164 struct net_device *filter_dev,
1167 int (*ndo_bridge_setlink)(struct net_device *dev,
1168 struct nlmsghdr *nlh,
1170 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1172 struct net_device *dev,
1174 int (*ndo_bridge_dellink)(struct net_device *dev,
1175 struct nlmsghdr *nlh,
1177 int (*ndo_change_carrier)(struct net_device *dev,
1179 int (*ndo_get_phys_port_id)(struct net_device *dev,
1180 struct netdev_phys_item_id *ppid);
1181 void (*ndo_add_vxlan_port)(struct net_device *dev,
1182 sa_family_t sa_family,
1184 void (*ndo_del_vxlan_port)(struct net_device *dev,
1185 sa_family_t sa_family,
1188 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1189 struct net_device *dev);
1190 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1193 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1194 struct net_device *dev,
1196 int (*ndo_get_lock_subclass)(struct net_device *dev);
1197 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1198 struct net_device *dev,
1199 netdev_features_t features);
1200 #ifdef CONFIG_NET_SWITCHDEV
1201 int (*ndo_switch_parent_id_get)(struct net_device *dev,
1202 struct netdev_phys_item_id *psid);
1203 int (*ndo_switch_port_stp_update)(struct net_device *dev,
1205 int (*ndo_switch_fib_ipv4_add)(struct net_device *dev,
1208 struct fib_info *fi,
1212 int (*ndo_switch_fib_ipv4_del)(struct net_device *dev,
1215 struct fib_info *fi,
1222 * enum net_device_priv_flags - &struct net_device priv_flags
1224 * These are the &struct net_device, they are only set internally
1225 * by drivers and used in the kernel. These flags are invisible to
1226 * userspace, this means that the order of these flags can change
1227 * during any kernel release.
1229 * You should have a pretty good reason to be extending these flags.
1231 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1232 * @IFF_EBRIDGE: Ethernet bridging device
1233 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1234 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1235 * @IFF_MASTER_ALB: bonding master, balance-alb
1236 * @IFF_BONDING: bonding master or slave
1237 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1238 * @IFF_ISATAP: ISATAP interface (RFC4214)
1239 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1240 * @IFF_WAN_HDLC: WAN HDLC device
1241 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1243 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1244 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1245 * @IFF_MACVLAN_PORT: device used as macvlan port
1246 * @IFF_BRIDGE_PORT: device used as bridge port
1247 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1248 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1249 * @IFF_UNICAST_FLT: Supports unicast filtering
1250 * @IFF_TEAM_PORT: device used as team port
1251 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1252 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1253 * change when it's running
1254 * @IFF_MACVLAN: Macvlan device
1256 enum netdev_priv_flags {
1257 IFF_802_1Q_VLAN = 1<<0,
1259 IFF_SLAVE_INACTIVE = 1<<2,
1260 IFF_MASTER_8023AD = 1<<3,
1261 IFF_MASTER_ALB = 1<<4,
1263 IFF_SLAVE_NEEDARP = 1<<6,
1265 IFF_MASTER_ARPMON = 1<<8,
1266 IFF_WAN_HDLC = 1<<9,
1267 IFF_XMIT_DST_RELEASE = 1<<10,
1268 IFF_DONT_BRIDGE = 1<<11,
1269 IFF_DISABLE_NETPOLL = 1<<12,
1270 IFF_MACVLAN_PORT = 1<<13,
1271 IFF_BRIDGE_PORT = 1<<14,
1272 IFF_OVS_DATAPATH = 1<<15,
1273 IFF_TX_SKB_SHARING = 1<<16,
1274 IFF_UNICAST_FLT = 1<<17,
1275 IFF_TEAM_PORT = 1<<18,
1276 IFF_SUPP_NOFCS = 1<<19,
1277 IFF_LIVE_ADDR_CHANGE = 1<<20,
1278 IFF_MACVLAN = 1<<21,
1279 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1280 IFF_IPVLAN_MASTER = 1<<23,
1281 IFF_IPVLAN_SLAVE = 1<<24,
1284 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1285 #define IFF_EBRIDGE IFF_EBRIDGE
1286 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1287 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1288 #define IFF_MASTER_ALB IFF_MASTER_ALB
1289 #define IFF_BONDING IFF_BONDING
1290 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1291 #define IFF_ISATAP IFF_ISATAP
1292 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1293 #define IFF_WAN_HDLC IFF_WAN_HDLC
1294 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1295 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1296 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1297 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1298 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1299 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1300 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1301 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1302 #define IFF_TEAM_PORT IFF_TEAM_PORT
1303 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1304 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1305 #define IFF_MACVLAN IFF_MACVLAN
1306 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1307 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1308 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1311 * struct net_device - The DEVICE structure.
1312 * Actually, this whole structure is a big mistake. It mixes I/O
1313 * data with strictly "high-level" data, and it has to know about
1314 * almost every data structure used in the INET module.
1316 * @name: This is the first field of the "visible" part of this structure
1317 * (i.e. as seen by users in the "Space.c" file). It is the name
1320 * @name_hlist: Device name hash chain, please keep it close to name[]
1321 * @ifalias: SNMP alias
1322 * @mem_end: Shared memory end
1323 * @mem_start: Shared memory start
1324 * @base_addr: Device I/O address
1325 * @irq: Device IRQ number
1327 * @state: Generic network queuing layer state, see netdev_state_t
1328 * @dev_list: The global list of network devices
1329 * @napi_list: List entry, that is used for polling napi devices
1330 * @unreg_list: List entry, that is used, when we are unregistering the
1331 * device, see the function unregister_netdev
1332 * @close_list: List entry, that is used, when we are closing the device
1334 * @adj_list: Directly linked devices, like slaves for bonding
1335 * @all_adj_list: All linked devices, *including* neighbours
1336 * @features: Currently active device features
1337 * @hw_features: User-changeable features
1339 * @wanted_features: User-requested features
1340 * @vlan_features: Mask of features inheritable by VLAN devices
1342 * @hw_enc_features: Mask of features inherited by encapsulating devices
1343 * This field indicates what encapsulation
1344 * offloads the hardware is capable of doing,
1345 * and drivers will need to set them appropriately.
1347 * @mpls_features: Mask of features inheritable by MPLS
1349 * @ifindex: interface index
1350 * @iflink: unique device identifier
1352 * @stats: Statistics struct, which was left as a legacy, use
1353 * rtnl_link_stats64 instead
1355 * @rx_dropped: Dropped packets by core network,
1356 * do not use this in drivers
1357 * @tx_dropped: Dropped packets by core network,
1358 * do not use this in drivers
1360 * @carrier_changes: Stats to monitor carrier on<->off transitions
1362 * @wireless_handlers: List of functions to handle Wireless Extensions,
1364 * see <net/iw_handler.h> for details.
1365 * @wireless_data: Instance data managed by the core of wireless extensions
1367 * @netdev_ops: Includes several pointers to callbacks,
1368 * if one wants to override the ndo_*() functions
1369 * @ethtool_ops: Management operations
1370 * @fwd_ops: Management operations
1371 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1372 * of Layer 2 headers.
1374 * @flags: Interface flags (a la BSD)
1375 * @priv_flags: Like 'flags' but invisible to userspace,
1376 * see if.h for the definitions
1377 * @gflags: Global flags ( kept as legacy )
1378 * @padded: How much padding added by alloc_netdev()
1379 * @operstate: RFC2863 operstate
1380 * @link_mode: Mapping policy to operstate
1381 * @if_port: Selectable AUI, TP, ...
1383 * @mtu: Interface MTU value
1384 * @type: Interface hardware type
1385 * @hard_header_len: Hardware header length
1387 * @needed_headroom: Extra headroom the hardware may need, but not in all
1388 * cases can this be guaranteed
1389 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1390 * cases can this be guaranteed. Some cases also use
1391 * LL_MAX_HEADER instead to allocate the skb
1393 * interface address info:
1395 * @perm_addr: Permanent hw address
1396 * @addr_assign_type: Hw address assignment type
1397 * @addr_len: Hardware address length
1398 * @neigh_priv_len; Used in neigh_alloc(),
1399 * initialized only in atm/clip.c
1400 * @dev_id: Used to differentiate devices that share
1401 * the same link layer address
1402 * @dev_port: Used to differentiate devices that share
1404 * @addr_list_lock: XXX: need comments on this one
1405 * @uc: unicast mac addresses
1406 * @mc: multicast mac addresses
1407 * @dev_addrs: list of device hw addresses
1408 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1409 * @uc_promisc: Counter, that indicates, that promiscuous mode
1410 * has been enabled due to the need to listen to
1411 * additional unicast addresses in a device that
1412 * does not implement ndo_set_rx_mode()
1413 * @promiscuity: Number of times, the NIC is told to work in
1414 * Promiscuous mode, if it becomes 0 the NIC will
1415 * exit from working in Promiscuous mode
1416 * @allmulti: Counter, enables or disables allmulticast mode
1418 * @vlan_info: VLAN info
1419 * @dsa_ptr: dsa specific data
1420 * @tipc_ptr: TIPC specific data
1421 * @atalk_ptr: AppleTalk link
1422 * @ip_ptr: IPv4 specific data
1423 * @dn_ptr: DECnet specific data
1424 * @ip6_ptr: IPv6 specific data
1425 * @ax25_ptr: AX.25 specific data
1426 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1428 * @last_rx: Time of last Rx
1429 * @dev_addr: Hw address (before bcast,
1430 * because most packets are unicast)
1432 * @_rx: Array of RX queues
1433 * @num_rx_queues: Number of RX queues
1434 * allocated at register_netdev() time
1435 * @real_num_rx_queues: Number of RX queues currently active in device
1437 * @rx_handler: handler for received packets
1438 * @rx_handler_data: XXX: need comments on this one
1439 * @ingress_queue: XXX: need comments on this one
1440 * @broadcast: hw bcast address
1442 * @_tx: Array of TX queues
1443 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1444 * @real_num_tx_queues: Number of TX queues currently active in device
1445 * @qdisc: Root qdisc from userspace point of view
1446 * @tx_queue_len: Max frames per queue allowed
1447 * @tx_global_lock: XXX: need comments on this one
1449 * @xps_maps: XXX: need comments on this one
1451 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1452 * indexed by RX queue number. Assigned by driver.
1453 * This must only be set if the ndo_rx_flow_steer
1454 * operation is defined
1456 * @trans_start: Time (in jiffies) of last Tx
1457 * @watchdog_timeo: Represents the timeout that is used by
1458 * the watchdog ( see dev_watchdog() )
1459 * @watchdog_timer: List of timers
1461 * @pcpu_refcnt: Number of references to this device
1462 * @todo_list: Delayed register/unregister
1463 * @index_hlist: Device index hash chain
1464 * @link_watch_list: XXX: need comments on this one
1466 * @reg_state: Register/unregister state machine
1467 * @dismantle: Device is going to be freed
1468 * @rtnl_link_state: This enum represents the phases of creating
1471 * @destructor: Called from unregister,
1472 * can be used to call free_netdev
1473 * @npinfo: XXX: need comments on this one
1474 * @nd_net: Network namespace this network device is inside
1476 * @ml_priv: Mid-layer private
1477 * @lstats: Loopback statistics
1478 * @tstats: Tunnel statistics
1479 * @dstats: Dummy statistics
1480 * @vstats: Virtual ethernet statistics
1485 * @dev: Class/net/name entry
1486 * @sysfs_groups: Space for optional device, statistics and wireless
1489 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1490 * @rtnl_link_ops: Rtnl_link_ops
1492 * @gso_max_size: Maximum size of generic segmentation offload
1493 * @gso_max_segs: Maximum number of segments that can be passed to the
1495 * @gso_min_segs: Minimum number of segments that can be passed to the
1498 * @dcbnl_ops: Data Center Bridging netlink ops
1499 * @num_tc: Number of traffic classes in the net device
1500 * @tc_to_txq: XXX: need comments on this one
1501 * @prio_tc_map XXX: need comments on this one
1503 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1505 * @priomap: XXX: need comments on this one
1506 * @phydev: Physical device may attach itself
1507 * for hardware timestamping
1509 * @qdisc_tx_busylock: XXX: need comments on this one
1511 * @group: The group, that the device belongs to
1512 * @pm_qos_req: Power Management QoS object
1514 * FIXME: cleanup struct net_device such that network protocol info
1519 char name[IFNAMSIZ];
1520 struct hlist_node name_hlist;
1523 * I/O specific fields
1524 * FIXME: Merge these and struct ifmap into one
1526 unsigned long mem_end;
1527 unsigned long mem_start;
1528 unsigned long base_addr;
1532 * Some hardware also needs these fields (state,dev_list,
1533 * napi_list,unreg_list,close_list) but they are not
1534 * part of the usual set specified in Space.c.
1537 unsigned long state;
1539 struct list_head dev_list;
1540 struct list_head napi_list;
1541 struct list_head unreg_list;
1542 struct list_head close_list;
1543 struct list_head ptype_all;
1544 struct list_head ptype_specific;
1547 struct list_head upper;
1548 struct list_head lower;
1552 struct list_head upper;
1553 struct list_head lower;
1556 netdev_features_t features;
1557 netdev_features_t hw_features;
1558 netdev_features_t wanted_features;
1559 netdev_features_t vlan_features;
1560 netdev_features_t hw_enc_features;
1561 netdev_features_t mpls_features;
1566 struct net_device_stats stats;
1568 atomic_long_t rx_dropped;
1569 atomic_long_t tx_dropped;
1571 atomic_t carrier_changes;
1573 #ifdef CONFIG_WIRELESS_EXT
1574 const struct iw_handler_def * wireless_handlers;
1575 struct iw_public_data * wireless_data;
1577 const struct net_device_ops *netdev_ops;
1578 const struct ethtool_ops *ethtool_ops;
1579 const struct forwarding_accel_ops *fwd_ops;
1580 #ifdef CONFIG_NET_SWITCHDEV
1581 const struct swdev_ops *swdev_ops;
1584 const struct header_ops *header_ops;
1587 unsigned int priv_flags;
1589 unsigned short gflags;
1590 unsigned short padded;
1592 unsigned char operstate;
1593 unsigned char link_mode;
1595 unsigned char if_port;
1599 unsigned short type;
1600 unsigned short hard_header_len;
1602 unsigned short needed_headroom;
1603 unsigned short needed_tailroom;
1605 /* Interface address info. */
1606 unsigned char perm_addr[MAX_ADDR_LEN];
1607 unsigned char addr_assign_type;
1608 unsigned char addr_len;
1609 unsigned short neigh_priv_len;
1610 unsigned short dev_id;
1611 unsigned short dev_port;
1612 spinlock_t addr_list_lock;
1613 struct netdev_hw_addr_list uc;
1614 struct netdev_hw_addr_list mc;
1615 struct netdev_hw_addr_list dev_addrs;
1618 struct kset *queues_kset;
1621 unsigned char name_assign_type;
1624 unsigned int promiscuity;
1625 unsigned int allmulti;
1628 /* Protocol specific pointers */
1630 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1631 struct vlan_info __rcu *vlan_info;
1633 #if IS_ENABLED(CONFIG_NET_DSA)
1634 struct dsa_switch_tree *dsa_ptr;
1636 #if IS_ENABLED(CONFIG_TIPC)
1637 struct tipc_bearer __rcu *tipc_ptr;
1640 struct in_device __rcu *ip_ptr;
1641 struct dn_dev __rcu *dn_ptr;
1642 struct inet6_dev __rcu *ip6_ptr;
1644 struct wireless_dev *ieee80211_ptr;
1645 struct wpan_dev *ieee802154_ptr;
1648 * Cache lines mostly used on receive path (including eth_type_trans())
1650 unsigned long last_rx;
1652 /* Interface address info used in eth_type_trans() */
1653 unsigned char *dev_addr;
1657 struct netdev_rx_queue *_rx;
1659 unsigned int num_rx_queues;
1660 unsigned int real_num_rx_queues;
1664 unsigned long gro_flush_timeout;
1665 rx_handler_func_t __rcu *rx_handler;
1666 void __rcu *rx_handler_data;
1668 struct netdev_queue __rcu *ingress_queue;
1669 unsigned char broadcast[MAX_ADDR_LEN];
1673 * Cache lines mostly used on transmit path
1675 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1676 unsigned int num_tx_queues;
1677 unsigned int real_num_tx_queues;
1678 struct Qdisc *qdisc;
1679 unsigned long tx_queue_len;
1680 spinlock_t tx_global_lock;
1683 struct xps_dev_maps __rcu *xps_maps;
1685 #ifdef CONFIG_RFS_ACCEL
1686 struct cpu_rmap *rx_cpu_rmap;
1689 /* These may be needed for future network-power-down code. */
1692 * trans_start here is expensive for high speed devices on SMP,
1693 * please use netdev_queue->trans_start instead.
1695 unsigned long trans_start;
1698 struct timer_list watchdog_timer;
1700 int __percpu *pcpu_refcnt;
1701 struct list_head todo_list;
1703 struct hlist_node index_hlist;
1704 struct list_head link_watch_list;
1706 enum { NETREG_UNINITIALIZED=0,
1707 NETREG_REGISTERED, /* completed register_netdevice */
1708 NETREG_UNREGISTERING, /* called unregister_netdevice */
1709 NETREG_UNREGISTERED, /* completed unregister todo */
1710 NETREG_RELEASED, /* called free_netdev */
1711 NETREG_DUMMY, /* dummy device for NAPI poll */
1717 RTNL_LINK_INITIALIZED,
1718 RTNL_LINK_INITIALIZING,
1719 } rtnl_link_state:16;
1721 void (*destructor)(struct net_device *dev);
1723 #ifdef CONFIG_NETPOLL
1724 struct netpoll_info __rcu *npinfo;
1727 possible_net_t nd_net;
1729 /* mid-layer private */
1732 struct pcpu_lstats __percpu *lstats;
1733 struct pcpu_sw_netstats __percpu *tstats;
1734 struct pcpu_dstats __percpu *dstats;
1735 struct pcpu_vstats __percpu *vstats;
1738 struct garp_port __rcu *garp_port;
1739 struct mrp_port __rcu *mrp_port;
1742 const struct attribute_group *sysfs_groups[4];
1743 const struct attribute_group *sysfs_rx_queue_group;
1745 const struct rtnl_link_ops *rtnl_link_ops;
1747 /* for setting kernel sock attribute on TCP connection setup */
1748 #define GSO_MAX_SIZE 65536
1749 unsigned int gso_max_size;
1750 #define GSO_MAX_SEGS 65535
1754 const struct dcbnl_rtnl_ops *dcbnl_ops;
1757 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1758 u8 prio_tc_map[TC_BITMASK + 1];
1760 #if IS_ENABLED(CONFIG_FCOE)
1761 unsigned int fcoe_ddp_xid;
1763 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1764 struct netprio_map __rcu *priomap;
1766 struct phy_device *phydev;
1767 struct lock_class_key *qdisc_tx_busylock;
1769 struct pm_qos_request pm_qos_req;
1771 #define to_net_dev(d) container_of(d, struct net_device, dev)
1773 #define NETDEV_ALIGN 32
1776 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1778 return dev->prio_tc_map[prio & TC_BITMASK];
1782 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1784 if (tc >= dev->num_tc)
1787 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1792 void netdev_reset_tc(struct net_device *dev)
1795 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1796 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1800 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1802 if (tc >= dev->num_tc)
1805 dev->tc_to_txq[tc].count = count;
1806 dev->tc_to_txq[tc].offset = offset;
1811 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1813 if (num_tc > TC_MAX_QUEUE)
1816 dev->num_tc = num_tc;
1821 int netdev_get_num_tc(struct net_device *dev)
1827 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1830 return &dev->_tx[index];
1833 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1834 const struct sk_buff *skb)
1836 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1839 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1840 void (*f)(struct net_device *,
1841 struct netdev_queue *,
1847 for (i = 0; i < dev->num_tx_queues; i++)
1848 f(dev, &dev->_tx[i], arg);
1851 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1852 struct sk_buff *skb,
1856 * Net namespace inlines
1859 struct net *dev_net(const struct net_device *dev)
1861 return read_pnet(&dev->nd_net);
1865 void dev_net_set(struct net_device *dev, struct net *net)
1867 write_pnet(&dev->nd_net, net);
1870 static inline bool netdev_uses_dsa(struct net_device *dev)
1872 #if IS_ENABLED(CONFIG_NET_DSA)
1873 if (dev->dsa_ptr != NULL)
1874 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1880 * netdev_priv - access network device private data
1881 * @dev: network device
1883 * Get network device private data
1885 static inline void *netdev_priv(const struct net_device *dev)
1887 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1890 /* Set the sysfs physical device reference for the network logical device
1891 * if set prior to registration will cause a symlink during initialization.
1893 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1895 /* Set the sysfs device type for the network logical device to allow
1896 * fine-grained identification of different network device types. For
1897 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1899 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1901 /* Default NAPI poll() weight
1902 * Device drivers are strongly advised to not use bigger value
1904 #define NAPI_POLL_WEIGHT 64
1907 * netif_napi_add - initialize a napi context
1908 * @dev: network device
1909 * @napi: napi context
1910 * @poll: polling function
1911 * @weight: default weight
1913 * netif_napi_add() must be used to initialize a napi context prior to calling
1914 * *any* of the other napi related functions.
1916 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1917 int (*poll)(struct napi_struct *, int), int weight);
1920 * netif_napi_del - remove a napi context
1921 * @napi: napi context
1923 * netif_napi_del() removes a napi context from the network device napi list
1925 void netif_napi_del(struct napi_struct *napi);
1927 struct napi_gro_cb {
1928 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1931 /* Length of frag0. */
1932 unsigned int frag0_len;
1934 /* This indicates where we are processing relative to skb->data. */
1937 /* This is non-zero if the packet cannot be merged with the new skb. */
1940 /* Save the IP ID here and check when we get to the transport layer */
1943 /* Number of segments aggregated. */
1946 /* Start offset for remote checksum offload */
1947 u16 gro_remcsum_start;
1949 /* jiffies when first packet was created/queued */
1952 /* Used in ipv6_gro_receive() and foo-over-udp */
1955 /* This is non-zero if the packet may be of the same flow. */
1958 /* Used in udp_gro_receive */
1961 /* GRO checksum is valid */
1964 /* Number of checksums via CHECKSUM_UNNECESSARY */
1969 #define NAPI_GRO_FREE 1
1970 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1972 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1977 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1980 /* used in skb_gro_receive() slow path */
1981 struct sk_buff *last;
1984 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1986 struct packet_type {
1987 __be16 type; /* This is really htons(ether_type). */
1988 struct net_device *dev; /* NULL is wildcarded here */
1989 int (*func) (struct sk_buff *,
1990 struct net_device *,
1991 struct packet_type *,
1992 struct net_device *);
1993 bool (*id_match)(struct packet_type *ptype,
1995 void *af_packet_priv;
1996 struct list_head list;
1999 struct offload_callbacks {
2000 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2001 netdev_features_t features);
2002 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2003 struct sk_buff *skb);
2004 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2007 struct packet_offload {
2008 __be16 type; /* This is really htons(ether_type). */
2009 struct offload_callbacks callbacks;
2010 struct list_head list;
2015 struct udp_offload_callbacks {
2016 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2017 struct sk_buff *skb,
2018 struct udp_offload *uoff);
2019 int (*gro_complete)(struct sk_buff *skb,
2021 struct udp_offload *uoff);
2024 struct udp_offload {
2027 struct udp_offload_callbacks callbacks;
2030 /* often modified stats are per cpu, other are shared (netdev->stats) */
2031 struct pcpu_sw_netstats {
2036 struct u64_stats_sync syncp;
2039 #define netdev_alloc_pcpu_stats(type) \
2041 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2044 for_each_possible_cpu(i) { \
2045 typeof(type) *stat; \
2046 stat = per_cpu_ptr(pcpu_stats, i); \
2047 u64_stats_init(&stat->syncp); \
2053 #include <linux/notifier.h>
2055 /* netdevice notifier chain. Please remember to update the rtnetlink
2056 * notification exclusion list in rtnetlink_event() when adding new
2059 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2060 #define NETDEV_DOWN 0x0002
2061 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2062 detected a hardware crash and restarted
2063 - we can use this eg to kick tcp sessions
2065 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2066 #define NETDEV_REGISTER 0x0005
2067 #define NETDEV_UNREGISTER 0x0006
2068 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2069 #define NETDEV_CHANGEADDR 0x0008
2070 #define NETDEV_GOING_DOWN 0x0009
2071 #define NETDEV_CHANGENAME 0x000A
2072 #define NETDEV_FEAT_CHANGE 0x000B
2073 #define NETDEV_BONDING_FAILOVER 0x000C
2074 #define NETDEV_PRE_UP 0x000D
2075 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2076 #define NETDEV_POST_TYPE_CHANGE 0x000F
2077 #define NETDEV_POST_INIT 0x0010
2078 #define NETDEV_UNREGISTER_FINAL 0x0011
2079 #define NETDEV_RELEASE 0x0012
2080 #define NETDEV_NOTIFY_PEERS 0x0013
2081 #define NETDEV_JOIN 0x0014
2082 #define NETDEV_CHANGEUPPER 0x0015
2083 #define NETDEV_RESEND_IGMP 0x0016
2084 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2085 #define NETDEV_CHANGEINFODATA 0x0018
2086 #define NETDEV_BONDING_INFO 0x0019
2088 int register_netdevice_notifier(struct notifier_block *nb);
2089 int unregister_netdevice_notifier(struct notifier_block *nb);
2091 struct netdev_notifier_info {
2092 struct net_device *dev;
2095 struct netdev_notifier_change_info {
2096 struct netdev_notifier_info info; /* must be first */
2097 unsigned int flags_changed;
2100 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2101 struct net_device *dev)
2106 static inline struct net_device *
2107 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2112 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2115 extern rwlock_t dev_base_lock; /* Device list lock */
2117 #define for_each_netdev(net, d) \
2118 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2119 #define for_each_netdev_reverse(net, d) \
2120 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2121 #define for_each_netdev_rcu(net, d) \
2122 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2123 #define for_each_netdev_safe(net, d, n) \
2124 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2125 #define for_each_netdev_continue(net, d) \
2126 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2127 #define for_each_netdev_continue_rcu(net, d) \
2128 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2129 #define for_each_netdev_in_bond_rcu(bond, slave) \
2130 for_each_netdev_rcu(&init_net, slave) \
2131 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2132 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2134 static inline struct net_device *next_net_device(struct net_device *dev)
2136 struct list_head *lh;
2140 lh = dev->dev_list.next;
2141 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2144 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2146 struct list_head *lh;
2150 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2151 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2154 static inline struct net_device *first_net_device(struct net *net)
2156 return list_empty(&net->dev_base_head) ? NULL :
2157 net_device_entry(net->dev_base_head.next);
2160 static inline struct net_device *first_net_device_rcu(struct net *net)
2162 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2164 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2167 int netdev_boot_setup_check(struct net_device *dev);
2168 unsigned long netdev_boot_base(const char *prefix, int unit);
2169 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2170 const char *hwaddr);
2171 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2172 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2173 void dev_add_pack(struct packet_type *pt);
2174 void dev_remove_pack(struct packet_type *pt);
2175 void __dev_remove_pack(struct packet_type *pt);
2176 void dev_add_offload(struct packet_offload *po);
2177 void dev_remove_offload(struct packet_offload *po);
2179 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2180 unsigned short mask);
2181 struct net_device *dev_get_by_name(struct net *net, const char *name);
2182 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2183 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2184 int dev_alloc_name(struct net_device *dev, const char *name);
2185 int dev_open(struct net_device *dev);
2186 int dev_close(struct net_device *dev);
2187 void dev_disable_lro(struct net_device *dev);
2188 int dev_loopback_xmit(struct sk_buff *newskb);
2189 int dev_queue_xmit(struct sk_buff *skb);
2190 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2191 int register_netdevice(struct net_device *dev);
2192 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2193 void unregister_netdevice_many(struct list_head *head);
2194 static inline void unregister_netdevice(struct net_device *dev)
2196 unregister_netdevice_queue(dev, NULL);
2199 int netdev_refcnt_read(const struct net_device *dev);
2200 void free_netdev(struct net_device *dev);
2201 void netdev_freemem(struct net_device *dev);
2202 void synchronize_net(void);
2203 int init_dummy_netdev(struct net_device *dev);
2205 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2206 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2207 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2208 int netdev_get_name(struct net *net, char *name, int ifindex);
2209 int dev_restart(struct net_device *dev);
2210 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2212 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2214 return NAPI_GRO_CB(skb)->data_offset;
2217 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2219 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2222 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2224 NAPI_GRO_CB(skb)->data_offset += len;
2227 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2228 unsigned int offset)
2230 return NAPI_GRO_CB(skb)->frag0 + offset;
2233 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2235 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2238 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2239 unsigned int offset)
2241 if (!pskb_may_pull(skb, hlen))
2244 NAPI_GRO_CB(skb)->frag0 = NULL;
2245 NAPI_GRO_CB(skb)->frag0_len = 0;
2246 return skb->data + offset;
2249 static inline void *skb_gro_network_header(struct sk_buff *skb)
2251 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2252 skb_network_offset(skb);
2255 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2256 const void *start, unsigned int len)
2258 if (NAPI_GRO_CB(skb)->csum_valid)
2259 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2260 csum_partial(start, len, 0));
2263 /* GRO checksum functions. These are logical equivalents of the normal
2264 * checksum functions (in skbuff.h) except that they operate on the GRO
2265 * offsets and fields in sk_buff.
2268 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2270 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2272 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2273 skb_gro_offset(skb));
2276 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2280 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2281 skb_checksum_start_offset(skb) <
2282 skb_gro_offset(skb)) &&
2283 !skb_at_gro_remcsum_start(skb) &&
2284 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2285 (!zero_okay || check));
2288 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2291 if (NAPI_GRO_CB(skb)->csum_valid &&
2292 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2295 NAPI_GRO_CB(skb)->csum = psum;
2297 return __skb_gro_checksum_complete(skb);
2300 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2302 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2303 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2304 NAPI_GRO_CB(skb)->csum_cnt--;
2306 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2307 * verified a new top level checksum or an encapsulated one
2308 * during GRO. This saves work if we fallback to normal path.
2310 __skb_incr_checksum_unnecessary(skb);
2314 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2317 __sum16 __ret = 0; \
2318 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2319 __ret = __skb_gro_checksum_validate_complete(skb, \
2320 compute_pseudo(skb, proto)); \
2322 __skb_mark_checksum_bad(skb); \
2324 skb_gro_incr_csum_unnecessary(skb); \
2328 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2329 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2331 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2333 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2335 #define skb_gro_checksum_simple_validate(skb) \
2336 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2338 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2340 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2341 !NAPI_GRO_CB(skb)->csum_valid);
2344 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2345 __sum16 check, __wsum pseudo)
2347 NAPI_GRO_CB(skb)->csum = ~pseudo;
2348 NAPI_GRO_CB(skb)->csum_valid = 1;
2351 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2353 if (__skb_gro_checksum_convert_check(skb)) \
2354 __skb_gro_checksum_convert(skb, check, \
2355 compute_pseudo(skb, proto)); \
2358 struct gro_remcsum {
2363 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2369 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2370 int start, int offset,
2371 struct gro_remcsum *grc,
2376 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2379 NAPI_GRO_CB(skb)->gro_remcsum_start =
2380 ((unsigned char *)ptr + start) - skb->head;
2384 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2386 /* Adjust skb->csum since we changed the packet */
2387 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2389 grc->offset = (ptr + offset) - (void *)skb->head;
2393 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2394 struct gro_remcsum *grc)
2399 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2402 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2403 unsigned short type,
2404 const void *daddr, const void *saddr,
2407 if (!dev->header_ops || !dev->header_ops->create)
2410 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2413 static inline int dev_parse_header(const struct sk_buff *skb,
2414 unsigned char *haddr)
2416 const struct net_device *dev = skb->dev;
2418 if (!dev->header_ops || !dev->header_ops->parse)
2420 return dev->header_ops->parse(skb, haddr);
2423 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2424 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2425 static inline int unregister_gifconf(unsigned int family)
2427 return register_gifconf(family, NULL);
2430 #ifdef CONFIG_NET_FLOW_LIMIT
2431 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2432 struct sd_flow_limit {
2434 unsigned int num_buckets;
2435 unsigned int history_head;
2436 u16 history[FLOW_LIMIT_HISTORY];
2440 extern int netdev_flow_limit_table_len;
2441 #endif /* CONFIG_NET_FLOW_LIMIT */
2444 * Incoming packets are placed on per-cpu queues
2446 struct softnet_data {
2447 struct list_head poll_list;
2448 struct sk_buff_head process_queue;
2451 unsigned int processed;
2452 unsigned int time_squeeze;
2453 unsigned int cpu_collision;
2454 unsigned int received_rps;
2456 struct softnet_data *rps_ipi_list;
2458 #ifdef CONFIG_NET_FLOW_LIMIT
2459 struct sd_flow_limit __rcu *flow_limit;
2461 struct Qdisc *output_queue;
2462 struct Qdisc **output_queue_tailp;
2463 struct sk_buff *completion_queue;
2466 /* Elements below can be accessed between CPUs for RPS */
2467 struct call_single_data csd ____cacheline_aligned_in_smp;
2468 struct softnet_data *rps_ipi_next;
2470 unsigned int input_queue_head;
2471 unsigned int input_queue_tail;
2473 unsigned int dropped;
2474 struct sk_buff_head input_pkt_queue;
2475 struct napi_struct backlog;
2479 static inline void input_queue_head_incr(struct softnet_data *sd)
2482 sd->input_queue_head++;
2486 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2487 unsigned int *qtail)
2490 *qtail = ++sd->input_queue_tail;
2494 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2496 void __netif_schedule(struct Qdisc *q);
2497 void netif_schedule_queue(struct netdev_queue *txq);
2499 static inline void netif_tx_schedule_all(struct net_device *dev)
2503 for (i = 0; i < dev->num_tx_queues; i++)
2504 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2507 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2509 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2513 * netif_start_queue - allow transmit
2514 * @dev: network device
2516 * Allow upper layers to call the device hard_start_xmit routine.
2518 static inline void netif_start_queue(struct net_device *dev)
2520 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2523 static inline void netif_tx_start_all_queues(struct net_device *dev)
2527 for (i = 0; i < dev->num_tx_queues; i++) {
2528 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2529 netif_tx_start_queue(txq);
2533 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2536 * netif_wake_queue - restart transmit
2537 * @dev: network device
2539 * Allow upper layers to call the device hard_start_xmit routine.
2540 * Used for flow control when transmit resources are available.
2542 static inline void netif_wake_queue(struct net_device *dev)
2544 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2547 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2551 for (i = 0; i < dev->num_tx_queues; i++) {
2552 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2553 netif_tx_wake_queue(txq);
2557 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2559 if (WARN_ON(!dev_queue)) {
2560 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2563 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2567 * netif_stop_queue - stop transmitted packets
2568 * @dev: network device
2570 * Stop upper layers calling the device hard_start_xmit routine.
2571 * Used for flow control when transmit resources are unavailable.
2573 static inline void netif_stop_queue(struct net_device *dev)
2575 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2578 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2582 for (i = 0; i < dev->num_tx_queues; i++) {
2583 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2584 netif_tx_stop_queue(txq);
2588 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2590 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2594 * netif_queue_stopped - test if transmit queue is flowblocked
2595 * @dev: network device
2597 * Test if transmit queue on device is currently unable to send.
2599 static inline bool netif_queue_stopped(const struct net_device *dev)
2601 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2604 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2606 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2610 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2612 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2616 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2618 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2622 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2623 * @dev_queue: pointer to transmit queue
2625 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2626 * to give appropriate hint to the cpu.
2628 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2631 prefetchw(&dev_queue->dql.num_queued);
2636 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2637 * @dev_queue: pointer to transmit queue
2639 * BQL enabled drivers might use this helper in their TX completion path,
2640 * to give appropriate hint to the cpu.
2642 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2645 prefetchw(&dev_queue->dql.limit);
2649 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2653 dql_queued(&dev_queue->dql, bytes);
2655 if (likely(dql_avail(&dev_queue->dql) >= 0))
2658 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2661 * The XOFF flag must be set before checking the dql_avail below,
2662 * because in netdev_tx_completed_queue we update the dql_completed
2663 * before checking the XOFF flag.
2667 /* check again in case another CPU has just made room avail */
2668 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2669 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2674 * netdev_sent_queue - report the number of bytes queued to hardware
2675 * @dev: network device
2676 * @bytes: number of bytes queued to the hardware device queue
2678 * Report the number of bytes queued for sending/completion to the network
2679 * device hardware queue. @bytes should be a good approximation and should
2680 * exactly match netdev_completed_queue() @bytes
2682 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2684 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2687 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2688 unsigned int pkts, unsigned int bytes)
2691 if (unlikely(!bytes))
2694 dql_completed(&dev_queue->dql, bytes);
2697 * Without the memory barrier there is a small possiblity that
2698 * netdev_tx_sent_queue will miss the update and cause the queue to
2699 * be stopped forever
2703 if (dql_avail(&dev_queue->dql) < 0)
2706 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2707 netif_schedule_queue(dev_queue);
2712 * netdev_completed_queue - report bytes and packets completed by device
2713 * @dev: network device
2714 * @pkts: actual number of packets sent over the medium
2715 * @bytes: actual number of bytes sent over the medium
2717 * Report the number of bytes and packets transmitted by the network device
2718 * hardware queue over the physical medium, @bytes must exactly match the
2719 * @bytes amount passed to netdev_sent_queue()
2721 static inline void netdev_completed_queue(struct net_device *dev,
2722 unsigned int pkts, unsigned int bytes)
2724 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2727 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2730 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2736 * netdev_reset_queue - reset the packets and bytes count of a network device
2737 * @dev_queue: network device
2739 * Reset the bytes and packet count of a network device and clear the
2740 * software flow control OFF bit for this network device
2742 static inline void netdev_reset_queue(struct net_device *dev_queue)
2744 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2748 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2749 * @dev: network device
2750 * @queue_index: given tx queue index
2752 * Returns 0 if given tx queue index >= number of device tx queues,
2753 * otherwise returns the originally passed tx queue index.
2755 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2757 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2758 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2759 dev->name, queue_index,
2760 dev->real_num_tx_queues);
2768 * netif_running - test if up
2769 * @dev: network device
2771 * Test if the device has been brought up.
2773 static inline bool netif_running(const struct net_device *dev)
2775 return test_bit(__LINK_STATE_START, &dev->state);
2779 * Routines to manage the subqueues on a device. We only need start
2780 * stop, and a check if it's stopped. All other device management is
2781 * done at the overall netdevice level.
2782 * Also test the device if we're multiqueue.
2786 * netif_start_subqueue - allow sending packets on subqueue
2787 * @dev: network device
2788 * @queue_index: sub queue index
2790 * Start individual transmit queue of a device with multiple transmit queues.
2792 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2794 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2796 netif_tx_start_queue(txq);
2800 * netif_stop_subqueue - stop sending packets on subqueue
2801 * @dev: network device
2802 * @queue_index: sub queue index
2804 * Stop individual transmit queue of a device with multiple transmit queues.
2806 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2808 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2809 netif_tx_stop_queue(txq);
2813 * netif_subqueue_stopped - test status of subqueue
2814 * @dev: network device
2815 * @queue_index: sub queue index
2817 * Check individual transmit queue of a device with multiple transmit queues.
2819 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2822 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2824 return netif_tx_queue_stopped(txq);
2827 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2828 struct sk_buff *skb)
2830 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2833 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2836 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2839 static inline int netif_set_xps_queue(struct net_device *dev,
2840 const struct cpumask *mask,
2848 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2849 * as a distribution range limit for the returned value.
2851 static inline u16 skb_tx_hash(const struct net_device *dev,
2852 struct sk_buff *skb)
2854 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2858 * netif_is_multiqueue - test if device has multiple transmit queues
2859 * @dev: network device
2861 * Check if device has multiple transmit queues
2863 static inline bool netif_is_multiqueue(const struct net_device *dev)
2865 return dev->num_tx_queues > 1;
2868 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2871 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2873 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2881 static inline unsigned int get_netdev_rx_queue_index(
2882 struct netdev_rx_queue *queue)
2884 struct net_device *dev = queue->dev;
2885 int index = queue - dev->_rx;
2887 BUG_ON(index >= dev->num_rx_queues);
2892 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2893 int netif_get_num_default_rss_queues(void);
2895 enum skb_free_reason {
2896 SKB_REASON_CONSUMED,
2900 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2901 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2904 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2905 * interrupt context or with hardware interrupts being disabled.
2906 * (in_irq() || irqs_disabled())
2908 * We provide four helpers that can be used in following contexts :
2910 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2911 * replacing kfree_skb(skb)
2913 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2914 * Typically used in place of consume_skb(skb) in TX completion path
2916 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2917 * replacing kfree_skb(skb)
2919 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2920 * and consumed a packet. Used in place of consume_skb(skb)
2922 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2924 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2927 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2929 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2932 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2934 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2937 static inline void dev_consume_skb_any(struct sk_buff *skb)
2939 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2942 int netif_rx(struct sk_buff *skb);
2943 int netif_rx_ni(struct sk_buff *skb);
2944 int netif_receive_skb(struct sk_buff *skb);
2945 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2946 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2947 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2948 gro_result_t napi_gro_frags(struct napi_struct *napi);
2949 struct packet_offload *gro_find_receive_by_type(__be16 type);
2950 struct packet_offload *gro_find_complete_by_type(__be16 type);
2952 static inline void napi_free_frags(struct napi_struct *napi)
2954 kfree_skb(napi->skb);
2958 int netdev_rx_handler_register(struct net_device *dev,
2959 rx_handler_func_t *rx_handler,
2960 void *rx_handler_data);
2961 void netdev_rx_handler_unregister(struct net_device *dev);
2963 bool dev_valid_name(const char *name);
2964 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2965 int dev_ethtool(struct net *net, struct ifreq *);
2966 unsigned int dev_get_flags(const struct net_device *);
2967 int __dev_change_flags(struct net_device *, unsigned int flags);
2968 int dev_change_flags(struct net_device *, unsigned int);
2969 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2970 unsigned int gchanges);
2971 int dev_change_name(struct net_device *, const char *);
2972 int dev_set_alias(struct net_device *, const char *, size_t);
2973 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2974 int dev_set_mtu(struct net_device *, int);
2975 void dev_set_group(struct net_device *, int);
2976 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2977 int dev_change_carrier(struct net_device *, bool new_carrier);
2978 int dev_get_phys_port_id(struct net_device *dev,
2979 struct netdev_phys_item_id *ppid);
2980 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2981 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2982 struct netdev_queue *txq, int *ret);
2983 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2984 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2985 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2987 extern int netdev_budget;
2989 /* Called by rtnetlink.c:rtnl_unlock() */
2990 void netdev_run_todo(void);
2993 * dev_put - release reference to device
2994 * @dev: network device
2996 * Release reference to device to allow it to be freed.
2998 static inline void dev_put(struct net_device *dev)
3000 this_cpu_dec(*dev->pcpu_refcnt);
3004 * dev_hold - get reference to device
3005 * @dev: network device
3007 * Hold reference to device to keep it from being freed.
3009 static inline void dev_hold(struct net_device *dev)
3011 this_cpu_inc(*dev->pcpu_refcnt);
3014 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3015 * and _off may be called from IRQ context, but it is caller
3016 * who is responsible for serialization of these calls.
3018 * The name carrier is inappropriate, these functions should really be
3019 * called netif_lowerlayer_*() because they represent the state of any
3020 * kind of lower layer not just hardware media.
3023 void linkwatch_init_dev(struct net_device *dev);
3024 void linkwatch_fire_event(struct net_device *dev);
3025 void linkwatch_forget_dev(struct net_device *dev);
3028 * netif_carrier_ok - test if carrier present
3029 * @dev: network device
3031 * Check if carrier is present on device
3033 static inline bool netif_carrier_ok(const struct net_device *dev)
3035 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3038 unsigned long dev_trans_start(struct net_device *dev);
3040 void __netdev_watchdog_up(struct net_device *dev);
3042 void netif_carrier_on(struct net_device *dev);
3044 void netif_carrier_off(struct net_device *dev);
3047 * netif_dormant_on - mark device as dormant.
3048 * @dev: network device
3050 * Mark device as dormant (as per RFC2863).
3052 * The dormant state indicates that the relevant interface is not
3053 * actually in a condition to pass packets (i.e., it is not 'up') but is
3054 * in a "pending" state, waiting for some external event. For "on-
3055 * demand" interfaces, this new state identifies the situation where the
3056 * interface is waiting for events to place it in the up state.
3059 static inline void netif_dormant_on(struct net_device *dev)
3061 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3062 linkwatch_fire_event(dev);
3066 * netif_dormant_off - set device as not dormant.
3067 * @dev: network device
3069 * Device is not in dormant state.
3071 static inline void netif_dormant_off(struct net_device *dev)
3073 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3074 linkwatch_fire_event(dev);
3078 * netif_dormant - test if carrier present
3079 * @dev: network device
3081 * Check if carrier is present on device
3083 static inline bool netif_dormant(const struct net_device *dev)
3085 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3090 * netif_oper_up - test if device is operational
3091 * @dev: network device
3093 * Check if carrier is operational
3095 static inline bool netif_oper_up(const struct net_device *dev)
3097 return (dev->operstate == IF_OPER_UP ||
3098 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3102 * netif_device_present - is device available or removed
3103 * @dev: network device
3105 * Check if device has not been removed from system.
3107 static inline bool netif_device_present(struct net_device *dev)
3109 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3112 void netif_device_detach(struct net_device *dev);
3114 void netif_device_attach(struct net_device *dev);
3117 * Network interface message level settings
3121 NETIF_MSG_DRV = 0x0001,
3122 NETIF_MSG_PROBE = 0x0002,
3123 NETIF_MSG_LINK = 0x0004,
3124 NETIF_MSG_TIMER = 0x0008,
3125 NETIF_MSG_IFDOWN = 0x0010,
3126 NETIF_MSG_IFUP = 0x0020,
3127 NETIF_MSG_RX_ERR = 0x0040,
3128 NETIF_MSG_TX_ERR = 0x0080,
3129 NETIF_MSG_TX_QUEUED = 0x0100,
3130 NETIF_MSG_INTR = 0x0200,
3131 NETIF_MSG_TX_DONE = 0x0400,
3132 NETIF_MSG_RX_STATUS = 0x0800,
3133 NETIF_MSG_PKTDATA = 0x1000,
3134 NETIF_MSG_HW = 0x2000,
3135 NETIF_MSG_WOL = 0x4000,
3138 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3139 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3140 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3141 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3142 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3143 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3144 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3145 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3146 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3147 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3148 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3149 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3150 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3151 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3152 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3154 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3157 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3158 return default_msg_enable_bits;
3159 if (debug_value == 0) /* no output */
3161 /* set low N bits */
3162 return (1 << debug_value) - 1;
3165 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3167 spin_lock(&txq->_xmit_lock);
3168 txq->xmit_lock_owner = cpu;
3171 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3173 spin_lock_bh(&txq->_xmit_lock);
3174 txq->xmit_lock_owner = smp_processor_id();
3177 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3179 bool ok = spin_trylock(&txq->_xmit_lock);
3181 txq->xmit_lock_owner = smp_processor_id();
3185 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3187 txq->xmit_lock_owner = -1;
3188 spin_unlock(&txq->_xmit_lock);
3191 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3193 txq->xmit_lock_owner = -1;
3194 spin_unlock_bh(&txq->_xmit_lock);
3197 static inline void txq_trans_update(struct netdev_queue *txq)
3199 if (txq->xmit_lock_owner != -1)
3200 txq->trans_start = jiffies;
3204 * netif_tx_lock - grab network device transmit lock
3205 * @dev: network device
3207 * Get network device transmit lock
3209 static inline void netif_tx_lock(struct net_device *dev)
3214 spin_lock(&dev->tx_global_lock);
3215 cpu = smp_processor_id();
3216 for (i = 0; i < dev->num_tx_queues; i++) {
3217 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3219 /* We are the only thread of execution doing a
3220 * freeze, but we have to grab the _xmit_lock in
3221 * order to synchronize with threads which are in
3222 * the ->hard_start_xmit() handler and already
3223 * checked the frozen bit.
3225 __netif_tx_lock(txq, cpu);
3226 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3227 __netif_tx_unlock(txq);
3231 static inline void netif_tx_lock_bh(struct net_device *dev)
3237 static inline void netif_tx_unlock(struct net_device *dev)
3241 for (i = 0; i < dev->num_tx_queues; i++) {
3242 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3244 /* No need to grab the _xmit_lock here. If the
3245 * queue is not stopped for another reason, we
3248 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3249 netif_schedule_queue(txq);
3251 spin_unlock(&dev->tx_global_lock);
3254 static inline void netif_tx_unlock_bh(struct net_device *dev)
3256 netif_tx_unlock(dev);
3260 #define HARD_TX_LOCK(dev, txq, cpu) { \
3261 if ((dev->features & NETIF_F_LLTX) == 0) { \
3262 __netif_tx_lock(txq, cpu); \
3266 #define HARD_TX_TRYLOCK(dev, txq) \
3267 (((dev->features & NETIF_F_LLTX) == 0) ? \
3268 __netif_tx_trylock(txq) : \
3271 #define HARD_TX_UNLOCK(dev, txq) { \
3272 if ((dev->features & NETIF_F_LLTX) == 0) { \
3273 __netif_tx_unlock(txq); \
3277 static inline void netif_tx_disable(struct net_device *dev)
3283 cpu = smp_processor_id();
3284 for (i = 0; i < dev->num_tx_queues; i++) {
3285 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3287 __netif_tx_lock(txq, cpu);
3288 netif_tx_stop_queue(txq);
3289 __netif_tx_unlock(txq);
3294 static inline void netif_addr_lock(struct net_device *dev)
3296 spin_lock(&dev->addr_list_lock);
3299 static inline void netif_addr_lock_nested(struct net_device *dev)
3301 int subclass = SINGLE_DEPTH_NESTING;
3303 if (dev->netdev_ops->ndo_get_lock_subclass)
3304 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3306 spin_lock_nested(&dev->addr_list_lock, subclass);
3309 static inline void netif_addr_lock_bh(struct net_device *dev)
3311 spin_lock_bh(&dev->addr_list_lock);
3314 static inline void netif_addr_unlock(struct net_device *dev)
3316 spin_unlock(&dev->addr_list_lock);
3319 static inline void netif_addr_unlock_bh(struct net_device *dev)
3321 spin_unlock_bh(&dev->addr_list_lock);
3325 * dev_addrs walker. Should be used only for read access. Call with
3326 * rcu_read_lock held.
3328 #define for_each_dev_addr(dev, ha) \
3329 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3331 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3333 void ether_setup(struct net_device *dev);
3335 /* Support for loadable net-drivers */
3336 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3337 unsigned char name_assign_type,
3338 void (*setup)(struct net_device *),
3339 unsigned int txqs, unsigned int rxqs);
3340 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3341 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3343 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3344 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3347 int register_netdev(struct net_device *dev);
3348 void unregister_netdev(struct net_device *dev);
3350 /* General hardware address lists handling functions */
3351 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3352 struct netdev_hw_addr_list *from_list, int addr_len);
3353 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3354 struct netdev_hw_addr_list *from_list, int addr_len);
3355 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3356 struct net_device *dev,
3357 int (*sync)(struct net_device *, const unsigned char *),
3358 int (*unsync)(struct net_device *,
3359 const unsigned char *));
3360 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3361 struct net_device *dev,
3362 int (*unsync)(struct net_device *,
3363 const unsigned char *));
3364 void __hw_addr_init(struct netdev_hw_addr_list *list);
3366 /* Functions used for device addresses handling */
3367 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3368 unsigned char addr_type);
3369 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3370 unsigned char addr_type);
3371 void dev_addr_flush(struct net_device *dev);
3372 int dev_addr_init(struct net_device *dev);
3374 /* Functions used for unicast addresses handling */
3375 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3376 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3377 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3378 int dev_uc_sync(struct net_device *to, struct net_device *from);
3379 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3380 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3381 void dev_uc_flush(struct net_device *dev);
3382 void dev_uc_init(struct net_device *dev);
3385 * __dev_uc_sync - Synchonize device's unicast list
3386 * @dev: device to sync
3387 * @sync: function to call if address should be added
3388 * @unsync: function to call if address should be removed
3390 * Add newly added addresses to the interface, and release
3391 * addresses that have been deleted.
3393 static inline int __dev_uc_sync(struct net_device *dev,
3394 int (*sync)(struct net_device *,
3395 const unsigned char *),
3396 int (*unsync)(struct net_device *,
3397 const unsigned char *))
3399 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3403 * __dev_uc_unsync - Remove synchronized addresses from device
3404 * @dev: device to sync
3405 * @unsync: function to call if address should be removed
3407 * Remove all addresses that were added to the device by dev_uc_sync().
3409 static inline void __dev_uc_unsync(struct net_device *dev,
3410 int (*unsync)(struct net_device *,
3411 const unsigned char *))
3413 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3416 /* Functions used for multicast addresses handling */
3417 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3418 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3419 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3420 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3421 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3422 int dev_mc_sync(struct net_device *to, struct net_device *from);
3423 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3424 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3425 void dev_mc_flush(struct net_device *dev);
3426 void dev_mc_init(struct net_device *dev);
3429 * __dev_mc_sync - Synchonize device's multicast list
3430 * @dev: device to sync
3431 * @sync: function to call if address should be added
3432 * @unsync: function to call if address should be removed
3434 * Add newly added addresses to the interface, and release
3435 * addresses that have been deleted.
3437 static inline int __dev_mc_sync(struct net_device *dev,
3438 int (*sync)(struct net_device *,
3439 const unsigned char *),
3440 int (*unsync)(struct net_device *,
3441 const unsigned char *))
3443 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3447 * __dev_mc_unsync - Remove synchronized addresses from device
3448 * @dev: device to sync
3449 * @unsync: function to call if address should be removed
3451 * Remove all addresses that were added to the device by dev_mc_sync().
3453 static inline void __dev_mc_unsync(struct net_device *dev,
3454 int (*unsync)(struct net_device *,
3455 const unsigned char *))
3457 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3460 /* Functions used for secondary unicast and multicast support */
3461 void dev_set_rx_mode(struct net_device *dev);
3462 void __dev_set_rx_mode(struct net_device *dev);
3463 int dev_set_promiscuity(struct net_device *dev, int inc);
3464 int dev_set_allmulti(struct net_device *dev, int inc);
3465 void netdev_state_change(struct net_device *dev);
3466 void netdev_notify_peers(struct net_device *dev);
3467 void netdev_features_change(struct net_device *dev);
3468 /* Load a device via the kmod */
3469 void dev_load(struct net *net, const char *name);
3470 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3471 struct rtnl_link_stats64 *storage);
3472 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3473 const struct net_device_stats *netdev_stats);
3475 extern int netdev_max_backlog;
3476 extern int netdev_tstamp_prequeue;
3477 extern int weight_p;
3478 extern int bpf_jit_enable;
3480 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3481 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3482 struct list_head **iter);
3483 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3484 struct list_head **iter);
3486 /* iterate through upper list, must be called under RCU read lock */
3487 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3488 for (iter = &(dev)->adj_list.upper, \
3489 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3491 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3493 /* iterate through upper list, must be called under RCU read lock */
3494 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3495 for (iter = &(dev)->all_adj_list.upper, \
3496 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3498 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3500 void *netdev_lower_get_next_private(struct net_device *dev,
3501 struct list_head **iter);
3502 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3503 struct list_head **iter);
3505 #define netdev_for_each_lower_private(dev, priv, iter) \
3506 for (iter = (dev)->adj_list.lower.next, \
3507 priv = netdev_lower_get_next_private(dev, &(iter)); \
3509 priv = netdev_lower_get_next_private(dev, &(iter)))
3511 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3512 for (iter = &(dev)->adj_list.lower, \
3513 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3515 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3517 void *netdev_lower_get_next(struct net_device *dev,
3518 struct list_head **iter);
3519 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3520 for (iter = &(dev)->adj_list.lower, \
3521 ldev = netdev_lower_get_next(dev, &(iter)); \
3523 ldev = netdev_lower_get_next(dev, &(iter)))
3525 void *netdev_adjacent_get_private(struct list_head *adj_list);
3526 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3527 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3528 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3529 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3530 int netdev_master_upper_dev_link(struct net_device *dev,
3531 struct net_device *upper_dev);
3532 int netdev_master_upper_dev_link_private(struct net_device *dev,
3533 struct net_device *upper_dev,
3535 void netdev_upper_dev_unlink(struct net_device *dev,
3536 struct net_device *upper_dev);
3537 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3538 void *netdev_lower_dev_get_private(struct net_device *dev,
3539 struct net_device *lower_dev);
3541 /* RSS keys are 40 or 52 bytes long */
3542 #define NETDEV_RSS_KEY_LEN 52
3543 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3544 void netdev_rss_key_fill(void *buffer, size_t len);
3546 int dev_get_nest_level(struct net_device *dev,
3547 bool (*type_check)(struct net_device *dev));
3548 int skb_checksum_help(struct sk_buff *skb);
3549 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3550 netdev_features_t features, bool tx_path);
3551 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3552 netdev_features_t features);
3554 struct netdev_bonding_info {
3559 struct netdev_notifier_bonding_info {
3560 struct netdev_notifier_info info; /* must be first */
3561 struct netdev_bonding_info bonding_info;
3564 void netdev_bonding_info_change(struct net_device *dev,
3565 struct netdev_bonding_info *bonding_info);
3568 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3570 return __skb_gso_segment(skb, features, true);
3572 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3574 static inline bool can_checksum_protocol(netdev_features_t features,
3577 return ((features & NETIF_F_GEN_CSUM) ||
3578 ((features & NETIF_F_V4_CSUM) &&
3579 protocol == htons(ETH_P_IP)) ||
3580 ((features & NETIF_F_V6_CSUM) &&
3581 protocol == htons(ETH_P_IPV6)) ||
3582 ((features & NETIF_F_FCOE_CRC) &&
3583 protocol == htons(ETH_P_FCOE)));
3587 void netdev_rx_csum_fault(struct net_device *dev);
3589 static inline void netdev_rx_csum_fault(struct net_device *dev)
3593 /* rx skb timestamps */
3594 void net_enable_timestamp(void);
3595 void net_disable_timestamp(void);
3597 #ifdef CONFIG_PROC_FS
3598 int __init dev_proc_init(void);
3600 #define dev_proc_init() 0
3603 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3604 struct sk_buff *skb, struct net_device *dev,
3607 skb->xmit_more = more ? 1 : 0;
3608 return ops->ndo_start_xmit(skb, dev);
3611 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3612 struct netdev_queue *txq, bool more)
3614 const struct net_device_ops *ops = dev->netdev_ops;
3617 rc = __netdev_start_xmit(ops, skb, dev, more);
3618 if (rc == NETDEV_TX_OK)
3619 txq_trans_update(txq);
3624 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3626 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3629 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3631 return netdev_class_create_file_ns(class_attr, NULL);
3634 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3636 netdev_class_remove_file_ns(class_attr, NULL);
3639 extern struct kobj_ns_type_operations net_ns_type_operations;
3641 const char *netdev_drivername(const struct net_device *dev);
3643 void linkwatch_run_queue(void);
3645 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3646 netdev_features_t f2)
3648 if (f1 & NETIF_F_GEN_CSUM)
3649 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3650 if (f2 & NETIF_F_GEN_CSUM)
3651 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3653 if (f1 & NETIF_F_GEN_CSUM)
3654 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3659 static inline netdev_features_t netdev_get_wanted_features(
3660 struct net_device *dev)
3662 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3664 netdev_features_t netdev_increment_features(netdev_features_t all,
3665 netdev_features_t one, netdev_features_t mask);
3667 /* Allow TSO being used on stacked device :
3668 * Performing the GSO segmentation before last device
3669 * is a performance improvement.
3671 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3672 netdev_features_t mask)
3674 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3677 int __netdev_update_features(struct net_device *dev);
3678 void netdev_update_features(struct net_device *dev);
3679 void netdev_change_features(struct net_device *dev);
3681 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3682 struct net_device *dev);
3684 netdev_features_t netif_skb_features(struct sk_buff *skb);
3686 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3688 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3690 /* check flags correspondence */
3691 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3692 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3693 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3694 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3695 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3696 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3697 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3698 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3699 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3700 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3701 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3702 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3703 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3705 return (features & feature) == feature;
3708 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3710 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3711 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3714 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3715 netdev_features_t features)
3717 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3718 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3719 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3722 static inline void netif_set_gso_max_size(struct net_device *dev,
3725 dev->gso_max_size = size;
3728 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3729 int pulled_hlen, u16 mac_offset,
3732 skb->protocol = protocol;
3733 skb->encapsulation = 1;
3734 skb_push(skb, pulled_hlen);
3735 skb_reset_transport_header(skb);
3736 skb->mac_header = mac_offset;
3737 skb->network_header = skb->mac_header + mac_len;
3738 skb->mac_len = mac_len;
3741 static inline bool netif_is_macvlan(struct net_device *dev)
3743 return dev->priv_flags & IFF_MACVLAN;
3746 static inline bool netif_is_macvlan_port(struct net_device *dev)
3748 return dev->priv_flags & IFF_MACVLAN_PORT;
3751 static inline bool netif_is_ipvlan(struct net_device *dev)
3753 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3756 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3758 return dev->priv_flags & IFF_IPVLAN_MASTER;
3761 static inline bool netif_is_bond_master(struct net_device *dev)
3763 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3766 static inline bool netif_is_bond_slave(struct net_device *dev)
3768 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3771 static inline bool netif_supports_nofcs(struct net_device *dev)
3773 return dev->priv_flags & IFF_SUPP_NOFCS;
3776 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3777 static inline void netif_keep_dst(struct net_device *dev)
3779 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3782 extern struct pernet_operations __net_initdata loopback_net_ops;
3784 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3786 /* netdev_printk helpers, similar to dev_printk */
3788 static inline const char *netdev_name(const struct net_device *dev)
3790 if (!dev->name[0] || strchr(dev->name, '%'))
3791 return "(unnamed net_device)";
3795 static inline const char *netdev_reg_state(const struct net_device *dev)
3797 switch (dev->reg_state) {
3798 case NETREG_UNINITIALIZED: return " (uninitialized)";
3799 case NETREG_REGISTERED: return "";
3800 case NETREG_UNREGISTERING: return " (unregistering)";
3801 case NETREG_UNREGISTERED: return " (unregistered)";
3802 case NETREG_RELEASED: return " (released)";
3803 case NETREG_DUMMY: return " (dummy)";
3806 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3807 return " (unknown)";
3811 void netdev_printk(const char *level, const struct net_device *dev,
3812 const char *format, ...);
3814 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3816 void netdev_alert(const struct net_device *dev, const char *format, ...);
3818 void netdev_crit(const struct net_device *dev, const char *format, ...);
3820 void netdev_err(const struct net_device *dev, const char *format, ...);
3822 void netdev_warn(const struct net_device *dev, const char *format, ...);
3824 void netdev_notice(const struct net_device *dev, const char *format, ...);
3826 void netdev_info(const struct net_device *dev, const char *format, ...);
3828 #define MODULE_ALIAS_NETDEV(device) \
3829 MODULE_ALIAS("netdev-" device)
3831 #if defined(CONFIG_DYNAMIC_DEBUG)
3832 #define netdev_dbg(__dev, format, args...) \
3834 dynamic_netdev_dbg(__dev, format, ##args); \
3836 #elif defined(DEBUG)
3837 #define netdev_dbg(__dev, format, args...) \
3838 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3840 #define netdev_dbg(__dev, format, args...) \
3843 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3847 #if defined(VERBOSE_DEBUG)
3848 #define netdev_vdbg netdev_dbg
3851 #define netdev_vdbg(dev, format, args...) \
3854 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3860 * netdev_WARN() acts like dev_printk(), but with the key difference
3861 * of using a WARN/WARN_ON to get the message out, including the
3862 * file/line information and a backtrace.
3864 #define netdev_WARN(dev, format, args...) \
3865 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3866 netdev_reg_state(dev), ##args)
3868 /* netif printk helpers, similar to netdev_printk */
3870 #define netif_printk(priv, type, level, dev, fmt, args...) \
3872 if (netif_msg_##type(priv)) \
3873 netdev_printk(level, (dev), fmt, ##args); \
3876 #define netif_level(level, priv, type, dev, fmt, args...) \
3878 if (netif_msg_##type(priv)) \
3879 netdev_##level(dev, fmt, ##args); \
3882 #define netif_emerg(priv, type, dev, fmt, args...) \
3883 netif_level(emerg, priv, type, dev, fmt, ##args)
3884 #define netif_alert(priv, type, dev, fmt, args...) \
3885 netif_level(alert, priv, type, dev, fmt, ##args)
3886 #define netif_crit(priv, type, dev, fmt, args...) \
3887 netif_level(crit, priv, type, dev, fmt, ##args)
3888 #define netif_err(priv, type, dev, fmt, args...) \
3889 netif_level(err, priv, type, dev, fmt, ##args)
3890 #define netif_warn(priv, type, dev, fmt, args...) \
3891 netif_level(warn, priv, type, dev, fmt, ##args)
3892 #define netif_notice(priv, type, dev, fmt, args...) \
3893 netif_level(notice, priv, type, dev, fmt, ##args)
3894 #define netif_info(priv, type, dev, fmt, args...) \
3895 netif_level(info, priv, type, dev, fmt, ##args)
3897 #if defined(CONFIG_DYNAMIC_DEBUG)
3898 #define netif_dbg(priv, type, netdev, format, args...) \
3900 if (netif_msg_##type(priv)) \
3901 dynamic_netdev_dbg(netdev, format, ##args); \
3903 #elif defined(DEBUG)
3904 #define netif_dbg(priv, type, dev, format, args...) \
3905 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3907 #define netif_dbg(priv, type, dev, format, args...) \
3910 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3915 #if defined(VERBOSE_DEBUG)
3916 #define netif_vdbg netif_dbg
3918 #define netif_vdbg(priv, type, dev, format, args...) \
3921 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3927 * The list of packet types we will receive (as opposed to discard)
3928 * and the routines to invoke.
3930 * Why 16. Because with 16 the only overlap we get on a hash of the
3931 * low nibble of the protocol value is RARP/SNAP/X.25.
3933 * NOTE: That is no longer true with the addition of VLAN tags. Not
3934 * sure which should go first, but I bet it won't make much
3935 * difference if we are running VLANs. The good news is that
3936 * this protocol won't be in the list unless compiled in, so
3937 * the average user (w/out VLANs) will not be adversely affected.
3953 #define PTYPE_HASH_SIZE (16)
3954 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3956 #endif /* _LINUX_NETDEVICE_H */