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/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
61 /* 802.15.4 specific */
64 /* UDP Tunnel offloads */
65 struct udp_tunnel_info;
68 void netdev_set_default_ethtool_ops(struct net_device *dev,
69 const struct ethtool_ops *ops);
71 /* Backlog congestion levels */
72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73 #define NET_RX_DROP 1 /* packet dropped */
76 * Transmit return codes: transmit return codes originate from three different
79 * - qdisc return codes
80 * - driver transmit return codes
83 * Drivers are allowed to return any one of those in their hard_start_xmit()
84 * function. Real network devices commonly used with qdiscs should only return
85 * the driver transmit return codes though - when qdiscs are used, the actual
86 * transmission happens asynchronously, so the value is not propagated to
87 * higher layers. Virtual network devices transmit synchronously; in this case
88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
89 * others are propagated to higher layers.
92 /* qdisc ->enqueue() return codes. */
93 #define NET_XMIT_SUCCESS 0x00
94 #define NET_XMIT_DROP 0x01 /* skb dropped */
95 #define NET_XMIT_CN 0x02 /* congestion notification */
96 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
99 * indicates that the device will soon be dropping packets, or already drops
100 * some packets of the same priority; prompting us to send less aggressively. */
101 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
102 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104 /* Driver transmit return codes */
105 #define NETDEV_TX_MASK 0xf0
108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
109 NETDEV_TX_OK = 0x00, /* driver took care of packet */
110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
112 typedef enum netdev_tx netdev_tx_t;
115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 static inline bool dev_xmit_complete(int rc)
121 * Positive cases with an skb consumed by a driver:
122 * - successful transmission (rc == NETDEV_TX_OK)
123 * - error while transmitting (rc < 0)
124 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 if (likely(rc < NET_XMIT_MASK))
133 * Compute the worst-case header length according to the protocols
137 #if defined(CONFIG_HYPERV_NET)
138 # define LL_MAX_HEADER 128
139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
140 # if defined(CONFIG_MAC80211_MESH)
141 # define LL_MAX_HEADER 128
143 # define LL_MAX_HEADER 96
146 # define LL_MAX_HEADER 32
149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
151 #define MAX_HEADER LL_MAX_HEADER
153 #define MAX_HEADER (LL_MAX_HEADER + 48)
157 * Old network device statistics. Fields are native words
158 * (unsigned long) so they can be read and written atomically.
161 struct net_device_stats {
162 unsigned long rx_packets;
163 unsigned long tx_packets;
164 unsigned long rx_bytes;
165 unsigned long tx_bytes;
166 unsigned long rx_errors;
167 unsigned long tx_errors;
168 unsigned long rx_dropped;
169 unsigned long tx_dropped;
170 unsigned long multicast;
171 unsigned long collisions;
172 unsigned long rx_length_errors;
173 unsigned long rx_over_errors;
174 unsigned long rx_crc_errors;
175 unsigned long rx_frame_errors;
176 unsigned long rx_fifo_errors;
177 unsigned long rx_missed_errors;
178 unsigned long tx_aborted_errors;
179 unsigned long tx_carrier_errors;
180 unsigned long tx_fifo_errors;
181 unsigned long tx_heartbeat_errors;
182 unsigned long tx_window_errors;
183 unsigned long rx_compressed;
184 unsigned long tx_compressed;
188 #include <linux/cache.h>
189 #include <linux/skbuff.h>
192 #include <linux/static_key.h>
193 extern struct static_key rps_needed;
200 struct netdev_hw_addr {
201 struct list_head list;
202 unsigned char addr[MAX_ADDR_LEN];
204 #define NETDEV_HW_ADDR_T_LAN 1
205 #define NETDEV_HW_ADDR_T_SAN 2
206 #define NETDEV_HW_ADDR_T_SLAVE 3
207 #define NETDEV_HW_ADDR_T_UNICAST 4
208 #define NETDEV_HW_ADDR_T_MULTICAST 5
213 struct rcu_head rcu_head;
216 struct netdev_hw_addr_list {
217 struct list_head list;
221 #define netdev_hw_addr_list_count(l) ((l)->count)
222 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
223 #define netdev_hw_addr_list_for_each(ha, l) \
224 list_for_each_entry(ha, &(l)->list, list)
226 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
227 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
228 #define netdev_for_each_uc_addr(ha, dev) \
229 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
231 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
232 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
233 #define netdev_for_each_mc_addr(ha, dev) \
234 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
241 /* cached hardware header; allow for machine alignment needs. */
242 #define HH_DATA_MOD 16
243 #define HH_DATA_OFF(__len) \
244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
245 #define HH_DATA_ALIGN(__len) \
246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
252 * dev->hard_header_len ? (dev->hard_header_len +
253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
255 * We could use other alignment values, but we must maintain the
256 * relationship HH alignment <= LL alignment.
258 #define LL_RESERVED_SPACE(dev) \
259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 int (*create) (struct sk_buff *skb, struct net_device *dev,
265 unsigned short type, const void *daddr,
266 const void *saddr, unsigned int len);
267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
269 void (*cache_update)(struct hh_cache *hh,
270 const struct net_device *dev,
271 const unsigned char *haddr);
272 bool (*validate)(const char *ll_header, unsigned int len);
275 /* These flag bits are private to the generic network queueing
276 * layer; they may not be explicitly referenced by any other
280 enum netdev_state_t {
282 __LINK_STATE_PRESENT,
283 __LINK_STATE_NOCARRIER,
284 __LINK_STATE_LINKWATCH_PENDING,
285 __LINK_STATE_DORMANT,
290 * This structure holds boot-time configured netdevice settings. They
291 * are then used in the device probing.
293 struct netdev_boot_setup {
297 #define NETDEV_BOOT_SETUP_MAX 8
299 int __init netdev_boot_setup(char *str);
302 * Structure for NAPI scheduling similar to tasklet but with weighting
305 /* The poll_list must only be managed by the entity which
306 * changes the state of the NAPI_STATE_SCHED bit. This means
307 * whoever atomically sets that bit can add this napi_struct
308 * to the per-CPU poll_list, and whoever clears that bit
309 * can remove from the list right before clearing the bit.
311 struct list_head poll_list;
315 unsigned int gro_count;
316 int (*poll)(struct napi_struct *, int);
317 #ifdef CONFIG_NETPOLL
318 spinlock_t poll_lock;
321 struct net_device *dev;
322 struct sk_buff *gro_list;
324 struct hrtimer timer;
325 struct list_head dev_list;
326 struct hlist_node napi_hash_node;
327 unsigned int napi_id;
331 NAPI_STATE_SCHED, /* Poll is scheduled */
332 NAPI_STATE_DISABLE, /* Disable pending */
333 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
334 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
335 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
345 typedef enum gro_result gro_result_t;
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
369 * If the rx_handler consumed the skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other way.
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
377 * If the rx_handler decides the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
379 * are registered on exact device (ptype->dev == skb->dev).
381 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
388 enum rx_handler_result {
394 typedef enum rx_handler_result rx_handler_result_t;
395 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
397 void __napi_schedule(struct napi_struct *n);
398 void __napi_schedule_irqoff(struct napi_struct *n);
400 static inline bool napi_disable_pending(struct napi_struct *n)
402 return test_bit(NAPI_STATE_DISABLE, &n->state);
406 * napi_schedule_prep - check if NAPI can be scheduled
409 * Test if NAPI routine is already running, and if not mark
410 * it as running. This is used as a condition variable to
411 * insure only one NAPI poll instance runs. We also make
412 * sure there is no pending NAPI disable.
414 static inline bool napi_schedule_prep(struct napi_struct *n)
416 return !napi_disable_pending(n) &&
417 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
421 * napi_schedule - schedule NAPI poll
424 * Schedule NAPI poll routine to be called if it is not already
427 static inline void napi_schedule(struct napi_struct *n)
429 if (napi_schedule_prep(n))
434 * napi_schedule_irqoff - schedule NAPI poll
437 * Variant of napi_schedule(), assuming hard irqs are masked.
439 static inline void napi_schedule_irqoff(struct napi_struct *n)
441 if (napi_schedule_prep(n))
442 __napi_schedule_irqoff(n);
445 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
446 static inline bool napi_reschedule(struct napi_struct *napi)
448 if (napi_schedule_prep(napi)) {
449 __napi_schedule(napi);
455 void __napi_complete(struct napi_struct *n);
456 void napi_complete_done(struct napi_struct *n, int work_done);
458 * napi_complete - NAPI processing complete
461 * Mark NAPI processing as complete.
462 * Consider using napi_complete_done() instead.
464 static inline void napi_complete(struct napi_struct *n)
466 return napi_complete_done(n, 0);
470 * napi_hash_add - add a NAPI to global hashtable
471 * @napi: NAPI context
473 * Generate a new napi_id and store a @napi under it in napi_hash.
474 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
475 * Note: This is normally automatically done from netif_napi_add(),
476 * so might disappear in a future Linux version.
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, if
486 * this function returns true.
487 * Note: core networking stack automatically calls it
488 * from netif_napi_del().
489 * Drivers might want to call this helper to combine all
490 * the needed RCU grace periods into a single one.
492 bool napi_hash_del(struct napi_struct *napi);
495 * napi_disable - prevent NAPI from scheduling
498 * Stop NAPI from being scheduled on this context.
499 * Waits till any outstanding processing completes.
501 void napi_disable(struct napi_struct *n);
504 * napi_enable - enable NAPI scheduling
507 * Resume NAPI from being scheduled on this context.
508 * Must be paired with napi_disable.
510 static inline void napi_enable(struct napi_struct *n)
512 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
513 smp_mb__before_atomic();
514 clear_bit(NAPI_STATE_SCHED, &n->state);
515 clear_bit(NAPI_STATE_NPSVC, &n->state);
519 * napi_synchronize - wait until NAPI is not running
522 * Wait until NAPI is done being scheduled on this context.
523 * Waits till any outstanding processing completes but
524 * does not disable future activations.
526 static inline void napi_synchronize(const struct napi_struct *n)
528 if (IS_ENABLED(CONFIG_SMP))
529 while (test_bit(NAPI_STATE_SCHED, &n->state))
535 enum netdev_queue_state_t {
536 __QUEUE_STATE_DRV_XOFF,
537 __QUEUE_STATE_STACK_XOFF,
538 __QUEUE_STATE_FROZEN,
541 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
542 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
543 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
546 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
552 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
553 * netif_tx_* functions below are used to manipulate this flag. The
554 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
555 * queue independently. The netif_xmit_*stopped functions below are called
556 * to check if the queue has been stopped by the driver or stack (either
557 * of the XOFF bits are set in the state). Drivers should not need to call
558 * netif_xmit*stopped functions, they should only be using netif_tx_*.
561 struct netdev_queue {
565 struct net_device *dev;
566 struct Qdisc __rcu *qdisc;
567 struct Qdisc *qdisc_sleeping;
571 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
574 unsigned long tx_maxrate;
576 * Number of TX timeouts for this queue
577 * (/sys/class/net/DEV/Q/trans_timeout)
579 unsigned long trans_timeout;
583 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
586 * Time (in jiffies) of last Tx
588 unsigned long trans_start;
595 } ____cacheline_aligned_in_smp;
597 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
623 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
630 struct rps_dev_flow {
633 unsigned int last_qtail;
635 #define RPS_NO_FILTER 0xffff
638 * The rps_dev_flow_table structure contains a table of flow mappings.
640 struct rps_dev_flow_table {
643 struct rps_dev_flow flows[0];
645 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
646 ((_num) * sizeof(struct rps_dev_flow)))
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
651 * Each entry is a 32bit value. Upper part is the high-order bits
652 * of flow hash, lower part is CPU number.
653 * rps_cpu_mask is used to partition the space, depending on number of
654 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
656 * meaning we use 32-6=26 bits for the hash.
658 struct rps_sock_flow_table {
661 u32 ents[0] ____cacheline_aligned_in_smp;
663 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
665 #define RPS_NO_CPU 0xffff
667 extern u32 rps_cpu_mask;
668 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
677 /* We only give a hint, preemption can change CPU under us */
678 val |= raw_smp_processor_id();
680 if (table->ents[index] != val)
681 table->ents[index] = val;
685 #ifdef CONFIG_RFS_ACCEL
686 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
689 #endif /* CONFIG_RPS */
691 /* This structure contains an instance of an RX queue. */
692 struct netdev_rx_queue {
694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
698 struct net_device *dev;
699 } ____cacheline_aligned_in_smp;
702 * RX queue sysfs structures and functions.
704 struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
719 unsigned int alloc_len;
723 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
724 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 struct xps_dev_maps {
732 struct xps_map __rcu *cpu_map[0];
734 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * sizeof(struct xps_map *)))
736 #endif /* CONFIG_XPS */
738 #define TC_MAX_QUEUE 16
739 #define TC_BITMASK 15
740 /* HW offloaded queuing disciplines txq count and offset maps */
741 struct netdev_tc_txq {
746 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
751 struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
759 char model_description[256];
763 #define MAX_PHYS_ITEM_ID_LEN 32
765 /* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
768 struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
770 unsigned char id_len;
773 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
780 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
783 /* These structures hold the attributes of qdisc and classifiers
784 * that are being passed to the netdevice through the setup_tc op.
792 struct tc_cls_u32_offload;
794 struct tc_to_netdev {
798 struct tc_cls_u32_offload *cls_u32;
799 struct tc_cls_flower_offload *cls_flower;
803 /* These structures hold the attributes of xdp state that are being passed
804 * to the netdevice through the xdp op.
806 enum xdp_netdev_command {
807 /* Set or clear a bpf program used in the earliest stages of packet
808 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
809 * is responsible for calling bpf_prog_put on any old progs that are
810 * stored. In case of error, the callee need not release the new prog
811 * reference, but on success it takes ownership and must bpf_prog_put
812 * when it is no longer used.
815 /* Check if a bpf program is set on the device. The callee should
816 * return true if a program is currently attached and running.
822 enum xdp_netdev_command command;
825 struct bpf_prog *prog;
832 * This structure defines the management hooks for network devices.
833 * The following hooks can be defined; unless noted otherwise, they are
834 * optional and can be filled with a null pointer.
836 * int (*ndo_init)(struct net_device *dev);
837 * This function is called once when a network device is registered.
838 * The network device can use this for any late stage initialization
839 * or semantic validation. It can fail with an error code which will
840 * be propagated back to register_netdev.
842 * void (*ndo_uninit)(struct net_device *dev);
843 * This function is called when device is unregistered or when registration
844 * fails. It is not called if init fails.
846 * int (*ndo_open)(struct net_device *dev);
847 * This function is called when a network device transitions to the up
850 * int (*ndo_stop)(struct net_device *dev);
851 * This function is called when a network device transitions to the down
854 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
855 * struct net_device *dev);
856 * Called when a packet needs to be transmitted.
857 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
858 * the queue before that can happen; it's for obsolete devices and weird
859 * corner cases, but the stack really does a non-trivial amount
860 * of useless work if you return NETDEV_TX_BUSY.
861 * Required; cannot be NULL.
863 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
864 * netdev_features_t features);
865 * Adjusts the requested feature flags according to device-specific
866 * constraints, and returns the resulting flags. Must not modify
869 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
870 * void *accel_priv, select_queue_fallback_t fallback);
871 * Called to decide which queue to use when device supports multiple
874 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
875 * This function is called to allow device receiver to make
876 * changes to configuration when multicast or promiscuous is enabled.
878 * void (*ndo_set_rx_mode)(struct net_device *dev);
879 * This function is called device changes address list filtering.
880 * If driver handles unicast address filtering, it should set
881 * IFF_UNICAST_FLT in its priv_flags.
883 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
884 * This function is called when the Media Access Control address
885 * needs to be changed. If this interface is not defined, the
886 * MAC address can not be changed.
888 * int (*ndo_validate_addr)(struct net_device *dev);
889 * Test if Media Access Control address is valid for the device.
891 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
892 * Called when a user requests an ioctl which can't be handled by
893 * the generic interface code. If not defined ioctls return
894 * not supported error code.
896 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
897 * Used to set network devices bus interface parameters. This interface
898 * is retained for legacy reasons; new devices should use the bus
899 * interface (PCI) for low level management.
901 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
902 * Called when a user wants to change the Maximum Transfer Unit
903 * of a device. If not defined, any request to change MTU will
904 * will return an error.
906 * void (*ndo_tx_timeout)(struct net_device *dev);
907 * Callback used when the transmitter has not made any progress
908 * for dev->watchdog ticks.
910 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
911 * struct rtnl_link_stats64 *storage);
912 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
913 * Called when a user wants to get the network device usage
914 * statistics. Drivers must do one of the following:
915 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
916 * rtnl_link_stats64 structure passed by the caller.
917 * 2. Define @ndo_get_stats to update a net_device_stats structure
918 * (which should normally be dev->stats) and return a pointer to
919 * it. The structure may be changed asynchronously only if each
920 * field is written atomically.
921 * 3. Update dev->stats asynchronously and atomically, and define
924 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
925 * If device supports VLAN filtering this function is called when a
926 * VLAN id is registered.
928 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
929 * If device supports VLAN filtering this function is called when a
930 * VLAN id is unregistered.
932 * void (*ndo_poll_controller)(struct net_device *dev);
934 * SR-IOV management functions.
935 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
936 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
937 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
939 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
940 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
941 * int (*ndo_get_vf_config)(struct net_device *dev,
942 * int vf, struct ifla_vf_info *ivf);
943 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
944 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
945 * struct nlattr *port[]);
947 * Enable or disable the VF ability to query its RSS Redirection Table and
948 * Hash Key. This is needed since on some devices VF share this information
949 * with PF and querying it may introduce a theoretical security risk.
950 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
951 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
952 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
953 * Called to setup 'tc' number of traffic classes in the net device. This
954 * is always called from the stack with the rtnl lock held and netif tx
955 * queues stopped. This allows the netdevice to perform queue management
958 * Fiber Channel over Ethernet (FCoE) offload functions.
959 * int (*ndo_fcoe_enable)(struct net_device *dev);
960 * Called when the FCoE protocol stack wants to start using LLD for FCoE
961 * so the underlying device can perform whatever needed configuration or
962 * initialization to support acceleration of FCoE traffic.
964 * int (*ndo_fcoe_disable)(struct net_device *dev);
965 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
966 * so the underlying device can perform whatever needed clean-ups to
967 * stop supporting acceleration of FCoE traffic.
969 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
970 * struct scatterlist *sgl, unsigned int sgc);
971 * Called when the FCoE Initiator wants to initialize an I/O that
972 * is a possible candidate for Direct Data Placement (DDP). The LLD can
973 * perform necessary setup and returns 1 to indicate the device is set up
974 * successfully to perform DDP on this I/O, otherwise this returns 0.
976 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
977 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
978 * indicated by the FC exchange id 'xid', so the underlying device can
979 * clean up and reuse resources for later DDP requests.
981 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
982 * struct scatterlist *sgl, unsigned int sgc);
983 * Called when the FCoE Target wants to initialize an I/O that
984 * is a possible candidate for Direct Data Placement (DDP). The LLD can
985 * perform necessary setup and returns 1 to indicate the device is set up
986 * successfully to perform DDP on this I/O, otherwise this returns 0.
988 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
989 * struct netdev_fcoe_hbainfo *hbainfo);
990 * Called when the FCoE Protocol stack wants information on the underlying
991 * device. This information is utilized by the FCoE protocol stack to
992 * register attributes with Fiber Channel management service as per the
993 * FC-GS Fabric Device Management Information(FDMI) specification.
995 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
996 * Called when the underlying device wants to override default World Wide
997 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
998 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
999 * protocol stack to use.
1002 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1003 * u16 rxq_index, u32 flow_id);
1004 * Set hardware filter for RFS. rxq_index is the target queue index;
1005 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1006 * Return the filter ID on success, or a negative error code.
1008 * Slave management functions (for bridge, bonding, etc).
1009 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1010 * Called to make another netdev an underling.
1012 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1013 * Called to release previously enslaved netdev.
1015 * Feature/offload setting functions.
1016 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1017 * Called to update device configuration to new features. Passed
1018 * feature set might be less than what was returned by ndo_fix_features()).
1019 * Must return >0 or -errno if it changed dev->features itself.
1021 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1022 * struct net_device *dev,
1023 * const unsigned char *addr, u16 vid, u16 flags)
1024 * Adds an FDB entry to dev for addr.
1025 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1026 * struct net_device *dev,
1027 * const unsigned char *addr, u16 vid)
1028 * Deletes the FDB entry from dev coresponding to addr.
1029 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1030 * struct net_device *dev, struct net_device *filter_dev,
1032 * Used to add FDB entries to dump requests. Implementers should add
1033 * entries to skb and update idx with the number of entries.
1035 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1037 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1038 * struct net_device *dev, u32 filter_mask,
1040 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1043 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1044 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1045 * which do not represent real hardware may define this to allow their
1046 * userspace components to manage their virtual carrier state. Devices
1047 * that determine carrier state from physical hardware properties (eg
1048 * network cables) or protocol-dependent mechanisms (eg
1049 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1051 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1052 * struct netdev_phys_item_id *ppid);
1053 * Called to get ID of physical port of this device. If driver does
1054 * not implement this, it is assumed that the hw is not able to have
1055 * multiple net devices on single physical port.
1057 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1058 * struct udp_tunnel_info *ti);
1059 * Called by UDP tunnel to notify a driver about the UDP port and socket
1060 * address family that a UDP tunnel is listnening to. It is called only
1061 * when a new port starts listening. The operation is protected by the
1064 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1065 * struct udp_tunnel_info *ti);
1066 * Called by UDP tunnel to notify the driver about a UDP port and socket
1067 * address family that the UDP tunnel is not listening to anymore. The
1068 * operation is protected by the RTNL.
1070 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1071 * struct net_device *dev)
1072 * Called by upper layer devices to accelerate switching or other
1073 * station functionality into hardware. 'pdev is the lowerdev
1074 * to use for the offload and 'dev' is the net device that will
1075 * back the offload. Returns a pointer to the private structure
1076 * the upper layer will maintain.
1077 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1078 * Called by upper layer device to delete the station created
1079 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1080 * the station and priv is the structure returned by the add
1082 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1083 * struct net_device *dev,
1085 * Callback to use for xmit over the accelerated station. This
1086 * is used in place of ndo_start_xmit on accelerated net
1088 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1089 * struct net_device *dev
1090 * netdev_features_t features);
1091 * Called by core transmit path to determine if device is capable of
1092 * performing offload operations on a given packet. This is to give
1093 * the device an opportunity to implement any restrictions that cannot
1094 * be otherwise expressed by feature flags. The check is called with
1095 * the set of features that the stack has calculated and it returns
1096 * those the driver believes to be appropriate.
1097 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1098 * int queue_index, u32 maxrate);
1099 * Called when a user wants to set a max-rate limitation of specific
1101 * int (*ndo_get_iflink)(const struct net_device *dev);
1102 * Called to get the iflink value of this device.
1103 * void (*ndo_change_proto_down)(struct net_device *dev,
1105 * This function is used to pass protocol port error state information
1106 * to the switch driver. The switch driver can react to the proto_down
1107 * by doing a phys down on the associated switch port.
1108 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1109 * This function is used to get egress tunnel information for given skb.
1110 * This is useful for retrieving outer tunnel header parameters while
1112 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1113 * This function is used to specify the headroom that the skb must
1114 * consider when allocation skb during packet reception. Setting
1115 * appropriate rx headroom value allows avoiding skb head copy on
1116 * forward. Setting a negative value resets the rx headroom to the
1118 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1119 * This function is used to set or query state related to XDP on the
1120 * netdevice. See definition of enum xdp_netdev_command for details.
1123 struct net_device_ops {
1124 int (*ndo_init)(struct net_device *dev);
1125 void (*ndo_uninit)(struct net_device *dev);
1126 int (*ndo_open)(struct net_device *dev);
1127 int (*ndo_stop)(struct net_device *dev);
1128 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1129 struct net_device *dev);
1130 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1131 struct net_device *dev,
1132 netdev_features_t features);
1133 u16 (*ndo_select_queue)(struct net_device *dev,
1134 struct sk_buff *skb,
1136 select_queue_fallback_t fallback);
1137 void (*ndo_change_rx_flags)(struct net_device *dev,
1139 void (*ndo_set_rx_mode)(struct net_device *dev);
1140 int (*ndo_set_mac_address)(struct net_device *dev,
1142 int (*ndo_validate_addr)(struct net_device *dev);
1143 int (*ndo_do_ioctl)(struct net_device *dev,
1144 struct ifreq *ifr, int cmd);
1145 int (*ndo_set_config)(struct net_device *dev,
1147 int (*ndo_change_mtu)(struct net_device *dev,
1149 int (*ndo_neigh_setup)(struct net_device *dev,
1150 struct neigh_parms *);
1151 void (*ndo_tx_timeout) (struct net_device *dev);
1153 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1154 struct rtnl_link_stats64 *storage);
1155 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1157 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1158 __be16 proto, u16 vid);
1159 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1160 __be16 proto, u16 vid);
1161 #ifdef CONFIG_NET_POLL_CONTROLLER
1162 void (*ndo_poll_controller)(struct net_device *dev);
1163 int (*ndo_netpoll_setup)(struct net_device *dev,
1164 struct netpoll_info *info);
1165 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1167 #ifdef CONFIG_NET_RX_BUSY_POLL
1168 int (*ndo_busy_poll)(struct napi_struct *dev);
1170 int (*ndo_set_vf_mac)(struct net_device *dev,
1171 int queue, u8 *mac);
1172 int (*ndo_set_vf_vlan)(struct net_device *dev,
1173 int queue, u16 vlan, u8 qos);
1174 int (*ndo_set_vf_rate)(struct net_device *dev,
1175 int vf, int min_tx_rate,
1177 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1178 int vf, bool setting);
1179 int (*ndo_set_vf_trust)(struct net_device *dev,
1180 int vf, bool setting);
1181 int (*ndo_get_vf_config)(struct net_device *dev,
1183 struct ifla_vf_info *ivf);
1184 int (*ndo_set_vf_link_state)(struct net_device *dev,
1185 int vf, int link_state);
1186 int (*ndo_get_vf_stats)(struct net_device *dev,
1188 struct ifla_vf_stats
1190 int (*ndo_set_vf_port)(struct net_device *dev,
1192 struct nlattr *port[]);
1193 int (*ndo_get_vf_port)(struct net_device *dev,
1194 int vf, struct sk_buff *skb);
1195 int (*ndo_set_vf_guid)(struct net_device *dev,
1198 int (*ndo_set_vf_rss_query_en)(
1199 struct net_device *dev,
1200 int vf, bool setting);
1201 int (*ndo_setup_tc)(struct net_device *dev,
1204 struct tc_to_netdev *tc);
1205 #if IS_ENABLED(CONFIG_FCOE)
1206 int (*ndo_fcoe_enable)(struct net_device *dev);
1207 int (*ndo_fcoe_disable)(struct net_device *dev);
1208 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1210 struct scatterlist *sgl,
1212 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1214 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1216 struct scatterlist *sgl,
1218 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1219 struct netdev_fcoe_hbainfo *hbainfo);
1222 #if IS_ENABLED(CONFIG_LIBFCOE)
1223 #define NETDEV_FCOE_WWNN 0
1224 #define NETDEV_FCOE_WWPN 1
1225 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1226 u64 *wwn, int type);
1229 #ifdef CONFIG_RFS_ACCEL
1230 int (*ndo_rx_flow_steer)(struct net_device *dev,
1231 const struct sk_buff *skb,
1235 int (*ndo_add_slave)(struct net_device *dev,
1236 struct net_device *slave_dev);
1237 int (*ndo_del_slave)(struct net_device *dev,
1238 struct net_device *slave_dev);
1239 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1240 netdev_features_t features);
1241 int (*ndo_set_features)(struct net_device *dev,
1242 netdev_features_t features);
1243 int (*ndo_neigh_construct)(struct net_device *dev,
1244 struct neighbour *n);
1245 void (*ndo_neigh_destroy)(struct net_device *dev,
1246 struct neighbour *n);
1248 int (*ndo_fdb_add)(struct ndmsg *ndm,
1249 struct nlattr *tb[],
1250 struct net_device *dev,
1251 const unsigned char *addr,
1254 int (*ndo_fdb_del)(struct ndmsg *ndm,
1255 struct nlattr *tb[],
1256 struct net_device *dev,
1257 const unsigned char *addr,
1259 int (*ndo_fdb_dump)(struct sk_buff *skb,
1260 struct netlink_callback *cb,
1261 struct net_device *dev,
1262 struct net_device *filter_dev,
1265 int (*ndo_bridge_setlink)(struct net_device *dev,
1266 struct nlmsghdr *nlh,
1268 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1270 struct net_device *dev,
1273 int (*ndo_bridge_dellink)(struct net_device *dev,
1274 struct nlmsghdr *nlh,
1276 int (*ndo_change_carrier)(struct net_device *dev,
1278 int (*ndo_get_phys_port_id)(struct net_device *dev,
1279 struct netdev_phys_item_id *ppid);
1280 int (*ndo_get_phys_port_name)(struct net_device *dev,
1281 char *name, size_t len);
1282 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1283 struct udp_tunnel_info *ti);
1284 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1285 struct udp_tunnel_info *ti);
1286 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1287 struct net_device *dev);
1288 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1291 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1292 struct net_device *dev,
1294 int (*ndo_get_lock_subclass)(struct net_device *dev);
1295 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1298 int (*ndo_get_iflink)(const struct net_device *dev);
1299 int (*ndo_change_proto_down)(struct net_device *dev,
1301 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1302 struct sk_buff *skb);
1303 void (*ndo_set_rx_headroom)(struct net_device *dev,
1304 int needed_headroom);
1305 int (*ndo_xdp)(struct net_device *dev,
1306 struct netdev_xdp *xdp);
1310 * enum net_device_priv_flags - &struct net_device priv_flags
1312 * These are the &struct net_device, they are only set internally
1313 * by drivers and used in the kernel. These flags are invisible to
1314 * userspace; this means that the order of these flags can change
1315 * during any kernel release.
1317 * You should have a pretty good reason to be extending these flags.
1319 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1320 * @IFF_EBRIDGE: Ethernet bridging device
1321 * @IFF_BONDING: bonding master or slave
1322 * @IFF_ISATAP: ISATAP interface (RFC4214)
1323 * @IFF_WAN_HDLC: WAN HDLC device
1324 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1326 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1327 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1328 * @IFF_MACVLAN_PORT: device used as macvlan port
1329 * @IFF_BRIDGE_PORT: device used as bridge port
1330 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1331 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1332 * @IFF_UNICAST_FLT: Supports unicast filtering
1333 * @IFF_TEAM_PORT: device used as team port
1334 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1335 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1336 * change when it's running
1337 * @IFF_MACVLAN: Macvlan device
1338 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1339 * underlying stacked devices
1340 * @IFF_IPVLAN_MASTER: IPvlan master device
1341 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1342 * @IFF_L3MDEV_MASTER: device is an L3 master device
1343 * @IFF_NO_QUEUE: device can run without qdisc attached
1344 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1345 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1346 * @IFF_TEAM: device is a team device
1347 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1348 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1349 * entity (i.e. the master device for bridged veth)
1350 * @IFF_MACSEC: device is a MACsec device
1352 enum netdev_priv_flags {
1353 IFF_802_1Q_VLAN = 1<<0,
1357 IFF_WAN_HDLC = 1<<4,
1358 IFF_XMIT_DST_RELEASE = 1<<5,
1359 IFF_DONT_BRIDGE = 1<<6,
1360 IFF_DISABLE_NETPOLL = 1<<7,
1361 IFF_MACVLAN_PORT = 1<<8,
1362 IFF_BRIDGE_PORT = 1<<9,
1363 IFF_OVS_DATAPATH = 1<<10,
1364 IFF_TX_SKB_SHARING = 1<<11,
1365 IFF_UNICAST_FLT = 1<<12,
1366 IFF_TEAM_PORT = 1<<13,
1367 IFF_SUPP_NOFCS = 1<<14,
1368 IFF_LIVE_ADDR_CHANGE = 1<<15,
1369 IFF_MACVLAN = 1<<16,
1370 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1371 IFF_IPVLAN_MASTER = 1<<18,
1372 IFF_IPVLAN_SLAVE = 1<<19,
1373 IFF_L3MDEV_MASTER = 1<<20,
1374 IFF_NO_QUEUE = 1<<21,
1375 IFF_OPENVSWITCH = 1<<22,
1376 IFF_L3MDEV_SLAVE = 1<<23,
1378 IFF_RXFH_CONFIGURED = 1<<25,
1379 IFF_PHONY_HEADROOM = 1<<26,
1383 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1384 #define IFF_EBRIDGE IFF_EBRIDGE
1385 #define IFF_BONDING IFF_BONDING
1386 #define IFF_ISATAP IFF_ISATAP
1387 #define IFF_WAN_HDLC IFF_WAN_HDLC
1388 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1389 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1390 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1391 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1392 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1393 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1394 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1395 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1396 #define IFF_TEAM_PORT IFF_TEAM_PORT
1397 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1398 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1399 #define IFF_MACVLAN IFF_MACVLAN
1400 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1401 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1402 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1403 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1404 #define IFF_NO_QUEUE IFF_NO_QUEUE
1405 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1406 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1407 #define IFF_TEAM IFF_TEAM
1408 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1409 #define IFF_MACSEC IFF_MACSEC
1412 * struct net_device - The DEVICE structure.
1413 * Actually, this whole structure is a big mistake. It mixes I/O
1414 * data with strictly "high-level" data, and it has to know about
1415 * almost every data structure used in the INET module.
1417 * @name: This is the first field of the "visible" part of this structure
1418 * (i.e. as seen by users in the "Space.c" file). It is the name
1421 * @name_hlist: Device name hash chain, please keep it close to name[]
1422 * @ifalias: SNMP alias
1423 * @mem_end: Shared memory end
1424 * @mem_start: Shared memory start
1425 * @base_addr: Device I/O address
1426 * @irq: Device IRQ number
1428 * @carrier_changes: Stats to monitor carrier on<->off transitions
1430 * @state: Generic network queuing layer state, see netdev_state_t
1431 * @dev_list: The global list of network devices
1432 * @napi_list: List entry used for polling NAPI devices
1433 * @unreg_list: List entry when we are unregistering the
1434 * device; see the function unregister_netdev
1435 * @close_list: List entry used when we are closing the device
1436 * @ptype_all: Device-specific packet handlers for all protocols
1437 * @ptype_specific: Device-specific, protocol-specific packet handlers
1439 * @adj_list: Directly linked devices, like slaves for bonding
1440 * @all_adj_list: All linked devices, *including* neighbours
1441 * @features: Currently active device features
1442 * @hw_features: User-changeable features
1444 * @wanted_features: User-requested features
1445 * @vlan_features: Mask of features inheritable by VLAN devices
1447 * @hw_enc_features: Mask of features inherited by encapsulating devices
1448 * This field indicates what encapsulation
1449 * offloads the hardware is capable of doing,
1450 * and drivers will need to set them appropriately.
1452 * @mpls_features: Mask of features inheritable by MPLS
1454 * @ifindex: interface index
1455 * @group: The group the device belongs to
1457 * @stats: Statistics struct, which was left as a legacy, use
1458 * rtnl_link_stats64 instead
1460 * @rx_dropped: Dropped packets by core network,
1461 * do not use this in drivers
1462 * @tx_dropped: Dropped packets by core network,
1463 * do not use this in drivers
1464 * @rx_nohandler: nohandler dropped packets by core network on
1465 * inactive devices, do not use this in drivers
1467 * @wireless_handlers: List of functions to handle Wireless Extensions,
1469 * see <net/iw_handler.h> for details.
1470 * @wireless_data: Instance data managed by the core of wireless extensions
1472 * @netdev_ops: Includes several pointers to callbacks,
1473 * if one wants to override the ndo_*() functions
1474 * @ethtool_ops: Management operations
1475 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1476 * discovery handling. Necessary for e.g. 6LoWPAN.
1477 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1478 * of Layer 2 headers.
1480 * @flags: Interface flags (a la BSD)
1481 * @priv_flags: Like 'flags' but invisible to userspace,
1482 * see if.h for the definitions
1483 * @gflags: Global flags ( kept as legacy )
1484 * @padded: How much padding added by alloc_netdev()
1485 * @operstate: RFC2863 operstate
1486 * @link_mode: Mapping policy to operstate
1487 * @if_port: Selectable AUI, TP, ...
1489 * @mtu: Interface MTU value
1490 * @type: Interface hardware type
1491 * @hard_header_len: Maximum hardware header length.
1493 * @needed_headroom: Extra headroom the hardware may need, but not in all
1494 * cases can this be guaranteed
1495 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1496 * cases can this be guaranteed. Some cases also use
1497 * LL_MAX_HEADER instead to allocate the skb
1499 * interface address info:
1501 * @perm_addr: Permanent hw address
1502 * @addr_assign_type: Hw address assignment type
1503 * @addr_len: Hardware address length
1504 * @neigh_priv_len: Used in neigh_alloc()
1505 * @dev_id: Used to differentiate devices that share
1506 * the same link layer address
1507 * @dev_port: Used to differentiate devices that share
1509 * @addr_list_lock: XXX: need comments on this one
1510 * @uc_promisc: Counter that indicates promiscuous mode
1511 * has been enabled due to the need to listen to
1512 * additional unicast addresses in a device that
1513 * does not implement ndo_set_rx_mode()
1514 * @uc: unicast mac addresses
1515 * @mc: multicast mac addresses
1516 * @dev_addrs: list of device hw addresses
1517 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1518 * @promiscuity: Number of times the NIC is told to work in
1519 * promiscuous mode; if it becomes 0 the NIC will
1520 * exit promiscuous mode
1521 * @allmulti: Counter, enables or disables allmulticast mode
1523 * @vlan_info: VLAN info
1524 * @dsa_ptr: dsa specific data
1525 * @tipc_ptr: TIPC specific data
1526 * @atalk_ptr: AppleTalk link
1527 * @ip_ptr: IPv4 specific data
1528 * @dn_ptr: DECnet specific data
1529 * @ip6_ptr: IPv6 specific data
1530 * @ax25_ptr: AX.25 specific data
1531 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1533 * @last_rx: Time of last Rx
1534 * @dev_addr: Hw address (before bcast,
1535 * because most packets are unicast)
1537 * @_rx: Array of RX queues
1538 * @num_rx_queues: Number of RX queues
1539 * allocated at register_netdev() time
1540 * @real_num_rx_queues: Number of RX queues currently active in device
1542 * @rx_handler: handler for received packets
1543 * @rx_handler_data: XXX: need comments on this one
1544 * @ingress_queue: XXX: need comments on this one
1545 * @broadcast: hw bcast address
1547 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1548 * indexed by RX queue number. Assigned by driver.
1549 * This must only be set if the ndo_rx_flow_steer
1550 * operation is defined
1551 * @index_hlist: Device index hash chain
1553 * @_tx: Array of TX queues
1554 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1555 * @real_num_tx_queues: Number of TX queues currently active in device
1556 * @qdisc: Root qdisc from userspace point of view
1557 * @tx_queue_len: Max frames per queue allowed
1558 * @tx_global_lock: XXX: need comments on this one
1560 * @xps_maps: XXX: need comments on this one
1562 * @offload_fwd_mark: Offload device fwding mark
1564 * @watchdog_timeo: Represents the timeout that is used by
1565 * the watchdog (see dev_watchdog())
1566 * @watchdog_timer: List of timers
1568 * @pcpu_refcnt: Number of references to this device
1569 * @todo_list: Delayed register/unregister
1570 * @link_watch_list: XXX: need comments on this one
1572 * @reg_state: Register/unregister state machine
1573 * @dismantle: Device is going to be freed
1574 * @rtnl_link_state: This enum represents the phases of creating
1577 * @destructor: Called from unregister,
1578 * can be used to call free_netdev
1579 * @npinfo: XXX: need comments on this one
1580 * @nd_net: Network namespace this network device is inside
1582 * @ml_priv: Mid-layer private
1583 * @lstats: Loopback statistics
1584 * @tstats: Tunnel statistics
1585 * @dstats: Dummy statistics
1586 * @vstats: Virtual ethernet statistics
1591 * @dev: Class/net/name entry
1592 * @sysfs_groups: Space for optional device, statistics and wireless
1595 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1596 * @rtnl_link_ops: Rtnl_link_ops
1598 * @gso_max_size: Maximum size of generic segmentation offload
1599 * @gso_max_segs: Maximum number of segments that can be passed to the
1602 * @dcbnl_ops: Data Center Bridging netlink ops
1603 * @num_tc: Number of traffic classes in the net device
1604 * @tc_to_txq: XXX: need comments on this one
1605 * @prio_tc_map XXX: need comments on this one
1607 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1609 * @priomap: XXX: need comments on this one
1610 * @phydev: Physical device may attach itself
1611 * for hardware timestamping
1613 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1614 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1616 * @proto_down: protocol port state information can be sent to the
1617 * switch driver and used to set the phys state of the
1620 * FIXME: cleanup struct net_device such that network protocol info
1625 char name[IFNAMSIZ];
1626 struct hlist_node name_hlist;
1629 * I/O specific fields
1630 * FIXME: Merge these and struct ifmap into one
1632 unsigned long mem_end;
1633 unsigned long mem_start;
1634 unsigned long base_addr;
1637 atomic_t carrier_changes;
1640 * Some hardware also needs these fields (state,dev_list,
1641 * napi_list,unreg_list,close_list) but they are not
1642 * part of the usual set specified in Space.c.
1645 unsigned long state;
1647 struct list_head dev_list;
1648 struct list_head napi_list;
1649 struct list_head unreg_list;
1650 struct list_head close_list;
1651 struct list_head ptype_all;
1652 struct list_head ptype_specific;
1655 struct list_head upper;
1656 struct list_head lower;
1660 struct list_head upper;
1661 struct list_head lower;
1664 netdev_features_t features;
1665 netdev_features_t hw_features;
1666 netdev_features_t wanted_features;
1667 netdev_features_t vlan_features;
1668 netdev_features_t hw_enc_features;
1669 netdev_features_t mpls_features;
1670 netdev_features_t gso_partial_features;
1675 struct net_device_stats stats;
1677 atomic_long_t rx_dropped;
1678 atomic_long_t tx_dropped;
1679 atomic_long_t rx_nohandler;
1681 #ifdef CONFIG_WIRELESS_EXT
1682 const struct iw_handler_def *wireless_handlers;
1683 struct iw_public_data *wireless_data;
1685 const struct net_device_ops *netdev_ops;
1686 const struct ethtool_ops *ethtool_ops;
1687 #ifdef CONFIG_NET_SWITCHDEV
1688 const struct switchdev_ops *switchdev_ops;
1690 #ifdef CONFIG_NET_L3_MASTER_DEV
1691 const struct l3mdev_ops *l3mdev_ops;
1693 #if IS_ENABLED(CONFIG_IPV6)
1694 const struct ndisc_ops *ndisc_ops;
1697 const struct header_ops *header_ops;
1700 unsigned int priv_flags;
1702 unsigned short gflags;
1703 unsigned short padded;
1705 unsigned char operstate;
1706 unsigned char link_mode;
1708 unsigned char if_port;
1712 unsigned short type;
1713 unsigned short hard_header_len;
1715 unsigned short needed_headroom;
1716 unsigned short needed_tailroom;
1718 /* Interface address info. */
1719 unsigned char perm_addr[MAX_ADDR_LEN];
1720 unsigned char addr_assign_type;
1721 unsigned char addr_len;
1722 unsigned short neigh_priv_len;
1723 unsigned short dev_id;
1724 unsigned short dev_port;
1725 spinlock_t addr_list_lock;
1726 unsigned char name_assign_type;
1728 struct netdev_hw_addr_list uc;
1729 struct netdev_hw_addr_list mc;
1730 struct netdev_hw_addr_list dev_addrs;
1733 struct kset *queues_kset;
1735 unsigned int promiscuity;
1736 unsigned int allmulti;
1739 /* Protocol-specific pointers */
1741 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1742 struct vlan_info __rcu *vlan_info;
1744 #if IS_ENABLED(CONFIG_NET_DSA)
1745 struct dsa_switch_tree *dsa_ptr;
1747 #if IS_ENABLED(CONFIG_TIPC)
1748 struct tipc_bearer __rcu *tipc_ptr;
1751 struct in_device __rcu *ip_ptr;
1752 struct dn_dev __rcu *dn_ptr;
1753 struct inet6_dev __rcu *ip6_ptr;
1755 struct wireless_dev *ieee80211_ptr;
1756 struct wpan_dev *ieee802154_ptr;
1757 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1758 struct mpls_dev __rcu *mpls_ptr;
1762 * Cache lines mostly used on receive path (including eth_type_trans())
1764 unsigned long last_rx;
1766 /* Interface address info used in eth_type_trans() */
1767 unsigned char *dev_addr;
1770 struct netdev_rx_queue *_rx;
1772 unsigned int num_rx_queues;
1773 unsigned int real_num_rx_queues;
1776 unsigned long gro_flush_timeout;
1777 rx_handler_func_t __rcu *rx_handler;
1778 void __rcu *rx_handler_data;
1780 #ifdef CONFIG_NET_CLS_ACT
1781 struct tcf_proto __rcu *ingress_cl_list;
1783 struct netdev_queue __rcu *ingress_queue;
1784 #ifdef CONFIG_NETFILTER_INGRESS
1785 struct list_head nf_hooks_ingress;
1788 unsigned char broadcast[MAX_ADDR_LEN];
1789 #ifdef CONFIG_RFS_ACCEL
1790 struct cpu_rmap *rx_cpu_rmap;
1792 struct hlist_node index_hlist;
1795 * Cache lines mostly used on transmit path
1797 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1798 unsigned int num_tx_queues;
1799 unsigned int real_num_tx_queues;
1800 struct Qdisc *qdisc;
1801 unsigned long tx_queue_len;
1802 spinlock_t tx_global_lock;
1806 struct xps_dev_maps __rcu *xps_maps;
1808 #ifdef CONFIG_NET_CLS_ACT
1809 struct tcf_proto __rcu *egress_cl_list;
1811 #ifdef CONFIG_NET_SWITCHDEV
1812 u32 offload_fwd_mark;
1815 /* These may be needed for future network-power-down code. */
1816 struct timer_list watchdog_timer;
1818 int __percpu *pcpu_refcnt;
1819 struct list_head todo_list;
1821 struct list_head link_watch_list;
1823 enum { NETREG_UNINITIALIZED=0,
1824 NETREG_REGISTERED, /* completed register_netdevice */
1825 NETREG_UNREGISTERING, /* called unregister_netdevice */
1826 NETREG_UNREGISTERED, /* completed unregister todo */
1827 NETREG_RELEASED, /* called free_netdev */
1828 NETREG_DUMMY, /* dummy device for NAPI poll */
1834 RTNL_LINK_INITIALIZED,
1835 RTNL_LINK_INITIALIZING,
1836 } rtnl_link_state:16;
1838 void (*destructor)(struct net_device *dev);
1840 #ifdef CONFIG_NETPOLL
1841 struct netpoll_info __rcu *npinfo;
1844 possible_net_t nd_net;
1846 /* mid-layer private */
1849 struct pcpu_lstats __percpu *lstats;
1850 struct pcpu_sw_netstats __percpu *tstats;
1851 struct pcpu_dstats __percpu *dstats;
1852 struct pcpu_vstats __percpu *vstats;
1855 struct garp_port __rcu *garp_port;
1856 struct mrp_port __rcu *mrp_port;
1859 const struct attribute_group *sysfs_groups[4];
1860 const struct attribute_group *sysfs_rx_queue_group;
1862 const struct rtnl_link_ops *rtnl_link_ops;
1864 /* for setting kernel sock attribute on TCP connection setup */
1865 #define GSO_MAX_SIZE 65536
1866 unsigned int gso_max_size;
1867 #define GSO_MAX_SEGS 65535
1871 const struct dcbnl_rtnl_ops *dcbnl_ops;
1874 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1875 u8 prio_tc_map[TC_BITMASK + 1];
1877 #if IS_ENABLED(CONFIG_FCOE)
1878 unsigned int fcoe_ddp_xid;
1880 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1881 struct netprio_map __rcu *priomap;
1883 struct phy_device *phydev;
1884 struct lock_class_key *qdisc_tx_busylock;
1885 struct lock_class_key *qdisc_running_key;
1888 #define to_net_dev(d) container_of(d, struct net_device, dev)
1890 #define NETDEV_ALIGN 32
1893 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1895 return dev->prio_tc_map[prio & TC_BITMASK];
1899 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1901 if (tc >= dev->num_tc)
1904 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1909 void netdev_reset_tc(struct net_device *dev)
1912 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1913 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1917 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1919 if (tc >= dev->num_tc)
1922 dev->tc_to_txq[tc].count = count;
1923 dev->tc_to_txq[tc].offset = offset;
1928 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1930 if (num_tc > TC_MAX_QUEUE)
1933 dev->num_tc = num_tc;
1938 int netdev_get_num_tc(struct net_device *dev)
1944 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1947 return &dev->_tx[index];
1950 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1951 const struct sk_buff *skb)
1953 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1956 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1957 void (*f)(struct net_device *,
1958 struct netdev_queue *,
1964 for (i = 0; i < dev->num_tx_queues; i++)
1965 f(dev, &dev->_tx[i], arg);
1968 #define netdev_lockdep_set_classes(dev) \
1970 static struct lock_class_key qdisc_tx_busylock_key; \
1971 static struct lock_class_key qdisc_running_key; \
1972 static struct lock_class_key qdisc_xmit_lock_key; \
1973 static struct lock_class_key dev_addr_list_lock_key; \
1976 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1977 (dev)->qdisc_running_key = &qdisc_running_key; \
1978 lockdep_set_class(&(dev)->addr_list_lock, \
1979 &dev_addr_list_lock_key); \
1980 for (i = 0; i < (dev)->num_tx_queues; i++) \
1981 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1982 &qdisc_xmit_lock_key); \
1985 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1986 struct sk_buff *skb,
1989 /* returns the headroom that the master device needs to take in account
1990 * when forwarding to this dev
1992 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1994 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1997 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1999 if (dev->netdev_ops->ndo_set_rx_headroom)
2000 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2003 /* set the device rx headroom to the dev's default */
2004 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2006 netdev_set_rx_headroom(dev, -1);
2010 * Net namespace inlines
2013 struct net *dev_net(const struct net_device *dev)
2015 return read_pnet(&dev->nd_net);
2019 void dev_net_set(struct net_device *dev, struct net *net)
2021 write_pnet(&dev->nd_net, net);
2024 static inline bool netdev_uses_dsa(struct net_device *dev)
2026 #if IS_ENABLED(CONFIG_NET_DSA)
2027 if (dev->dsa_ptr != NULL)
2028 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2034 * netdev_priv - access network device private data
2035 * @dev: network device
2037 * Get network device private data
2039 static inline void *netdev_priv(const struct net_device *dev)
2041 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2044 /* Set the sysfs physical device reference for the network logical device
2045 * if set prior to registration will cause a symlink during initialization.
2047 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2049 /* Set the sysfs device type for the network logical device to allow
2050 * fine-grained identification of different network device types. For
2051 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2053 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2055 /* Default NAPI poll() weight
2056 * Device drivers are strongly advised to not use bigger value
2058 #define NAPI_POLL_WEIGHT 64
2061 * netif_napi_add - initialize a NAPI context
2062 * @dev: network device
2063 * @napi: NAPI context
2064 * @poll: polling function
2065 * @weight: default weight
2067 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2068 * *any* of the other NAPI-related functions.
2070 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2071 int (*poll)(struct napi_struct *, int), int weight);
2074 * netif_tx_napi_add - initialize a NAPI context
2075 * @dev: network device
2076 * @napi: NAPI context
2077 * @poll: polling function
2078 * @weight: default weight
2080 * This variant of netif_napi_add() should be used from drivers using NAPI
2081 * to exclusively poll a TX queue.
2082 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2084 static inline void netif_tx_napi_add(struct net_device *dev,
2085 struct napi_struct *napi,
2086 int (*poll)(struct napi_struct *, int),
2089 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2090 netif_napi_add(dev, napi, poll, weight);
2094 * netif_napi_del - remove a NAPI context
2095 * @napi: NAPI context
2097 * netif_napi_del() removes a NAPI context from the network device NAPI list
2099 void netif_napi_del(struct napi_struct *napi);
2101 struct napi_gro_cb {
2102 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2105 /* Length of frag0. */
2106 unsigned int frag0_len;
2108 /* This indicates where we are processing relative to skb->data. */
2111 /* This is non-zero if the packet cannot be merged with the new skb. */
2114 /* Save the IP ID here and check when we get to the transport layer */
2117 /* Number of segments aggregated. */
2120 /* Start offset for remote checksum offload */
2121 u16 gro_remcsum_start;
2123 /* jiffies when first packet was created/queued */
2126 /* Used in ipv6_gro_receive() and foo-over-udp */
2129 /* This is non-zero if the packet may be of the same flow. */
2132 /* Used in tunnel GRO receive */
2135 /* GRO checksum is valid */
2138 /* Number of checksums via CHECKSUM_UNNECESSARY */
2143 #define NAPI_GRO_FREE 1
2144 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2146 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2149 /* Used in GRE, set in fou/gue_gro_receive */
2152 /* Used to determine if flush_id can be ignored */
2157 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2160 /* used in skb_gro_receive() slow path */
2161 struct sk_buff *last;
2164 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2166 struct packet_type {
2167 __be16 type; /* This is really htons(ether_type). */
2168 struct net_device *dev; /* NULL is wildcarded here */
2169 int (*func) (struct sk_buff *,
2170 struct net_device *,
2171 struct packet_type *,
2172 struct net_device *);
2173 bool (*id_match)(struct packet_type *ptype,
2175 void *af_packet_priv;
2176 struct list_head list;
2179 struct offload_callbacks {
2180 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2181 netdev_features_t features);
2182 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2183 struct sk_buff *skb);
2184 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2187 struct packet_offload {
2188 __be16 type; /* This is really htons(ether_type). */
2190 struct offload_callbacks callbacks;
2191 struct list_head list;
2194 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2195 struct pcpu_sw_netstats {
2200 struct u64_stats_sync syncp;
2203 #define __netdev_alloc_pcpu_stats(type, gfp) \
2205 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2208 for_each_possible_cpu(__cpu) { \
2209 typeof(type) *stat; \
2210 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2211 u64_stats_init(&stat->syncp); \
2217 #define netdev_alloc_pcpu_stats(type) \
2218 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2220 enum netdev_lag_tx_type {
2221 NETDEV_LAG_TX_TYPE_UNKNOWN,
2222 NETDEV_LAG_TX_TYPE_RANDOM,
2223 NETDEV_LAG_TX_TYPE_BROADCAST,
2224 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2225 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2226 NETDEV_LAG_TX_TYPE_HASH,
2229 struct netdev_lag_upper_info {
2230 enum netdev_lag_tx_type tx_type;
2233 struct netdev_lag_lower_state_info {
2238 #include <linux/notifier.h>
2240 /* netdevice notifier chain. Please remember to update the rtnetlink
2241 * notification exclusion list in rtnetlink_event() when adding new
2244 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2245 #define NETDEV_DOWN 0x0002
2246 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2247 detected a hardware crash and restarted
2248 - we can use this eg to kick tcp sessions
2250 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2251 #define NETDEV_REGISTER 0x0005
2252 #define NETDEV_UNREGISTER 0x0006
2253 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2254 #define NETDEV_CHANGEADDR 0x0008
2255 #define NETDEV_GOING_DOWN 0x0009
2256 #define NETDEV_CHANGENAME 0x000A
2257 #define NETDEV_FEAT_CHANGE 0x000B
2258 #define NETDEV_BONDING_FAILOVER 0x000C
2259 #define NETDEV_PRE_UP 0x000D
2260 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2261 #define NETDEV_POST_TYPE_CHANGE 0x000F
2262 #define NETDEV_POST_INIT 0x0010
2263 #define NETDEV_UNREGISTER_FINAL 0x0011
2264 #define NETDEV_RELEASE 0x0012
2265 #define NETDEV_NOTIFY_PEERS 0x0013
2266 #define NETDEV_JOIN 0x0014
2267 #define NETDEV_CHANGEUPPER 0x0015
2268 #define NETDEV_RESEND_IGMP 0x0016
2269 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2270 #define NETDEV_CHANGEINFODATA 0x0018
2271 #define NETDEV_BONDING_INFO 0x0019
2272 #define NETDEV_PRECHANGEUPPER 0x001A
2273 #define NETDEV_CHANGELOWERSTATE 0x001B
2274 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2275 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2277 int register_netdevice_notifier(struct notifier_block *nb);
2278 int unregister_netdevice_notifier(struct notifier_block *nb);
2280 struct netdev_notifier_info {
2281 struct net_device *dev;
2284 struct netdev_notifier_change_info {
2285 struct netdev_notifier_info info; /* must be first */
2286 unsigned int flags_changed;
2289 struct netdev_notifier_changeupper_info {
2290 struct netdev_notifier_info info; /* must be first */
2291 struct net_device *upper_dev; /* new upper dev */
2292 bool master; /* is upper dev master */
2293 bool linking; /* is the notification for link or unlink */
2294 void *upper_info; /* upper dev info */
2297 struct netdev_notifier_changelowerstate_info {
2298 struct netdev_notifier_info info; /* must be first */
2299 void *lower_state_info; /* is lower dev state */
2302 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2303 struct net_device *dev)
2308 static inline struct net_device *
2309 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2314 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2317 extern rwlock_t dev_base_lock; /* Device list lock */
2319 #define for_each_netdev(net, d) \
2320 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2321 #define for_each_netdev_reverse(net, d) \
2322 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2323 #define for_each_netdev_rcu(net, d) \
2324 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2325 #define for_each_netdev_safe(net, d, n) \
2326 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2327 #define for_each_netdev_continue(net, d) \
2328 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2329 #define for_each_netdev_continue_rcu(net, d) \
2330 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2331 #define for_each_netdev_in_bond_rcu(bond, slave) \
2332 for_each_netdev_rcu(&init_net, slave) \
2333 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2334 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2336 static inline struct net_device *next_net_device(struct net_device *dev)
2338 struct list_head *lh;
2342 lh = dev->dev_list.next;
2343 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2346 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2348 struct list_head *lh;
2352 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2353 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2356 static inline struct net_device *first_net_device(struct net *net)
2358 return list_empty(&net->dev_base_head) ? NULL :
2359 net_device_entry(net->dev_base_head.next);
2362 static inline struct net_device *first_net_device_rcu(struct net *net)
2364 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2366 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2369 int netdev_boot_setup_check(struct net_device *dev);
2370 unsigned long netdev_boot_base(const char *prefix, int unit);
2371 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2372 const char *hwaddr);
2373 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2374 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2375 void dev_add_pack(struct packet_type *pt);
2376 void dev_remove_pack(struct packet_type *pt);
2377 void __dev_remove_pack(struct packet_type *pt);
2378 void dev_add_offload(struct packet_offload *po);
2379 void dev_remove_offload(struct packet_offload *po);
2381 int dev_get_iflink(const struct net_device *dev);
2382 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2383 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2384 unsigned short mask);
2385 struct net_device *dev_get_by_name(struct net *net, const char *name);
2386 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2387 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2388 int dev_alloc_name(struct net_device *dev, const char *name);
2389 int dev_open(struct net_device *dev);
2390 int dev_close(struct net_device *dev);
2391 int dev_close_many(struct list_head *head, bool unlink);
2392 void dev_disable_lro(struct net_device *dev);
2393 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2394 int dev_queue_xmit(struct sk_buff *skb);
2395 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2396 int register_netdevice(struct net_device *dev);
2397 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2398 void unregister_netdevice_many(struct list_head *head);
2399 static inline void unregister_netdevice(struct net_device *dev)
2401 unregister_netdevice_queue(dev, NULL);
2404 int netdev_refcnt_read(const struct net_device *dev);
2405 void free_netdev(struct net_device *dev);
2406 void netdev_freemem(struct net_device *dev);
2407 void synchronize_net(void);
2408 int init_dummy_netdev(struct net_device *dev);
2410 DECLARE_PER_CPU(int, xmit_recursion);
2411 #define XMIT_RECURSION_LIMIT 10
2413 static inline int dev_recursion_level(void)
2415 return this_cpu_read(xmit_recursion);
2418 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2419 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2420 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2421 int netdev_get_name(struct net *net, char *name, int ifindex);
2422 int dev_restart(struct net_device *dev);
2423 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2425 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2427 return NAPI_GRO_CB(skb)->data_offset;
2430 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2432 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2435 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2437 NAPI_GRO_CB(skb)->data_offset += len;
2440 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2441 unsigned int offset)
2443 return NAPI_GRO_CB(skb)->frag0 + offset;
2446 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2448 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2451 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2452 unsigned int offset)
2454 if (!pskb_may_pull(skb, hlen))
2457 NAPI_GRO_CB(skb)->frag0 = NULL;
2458 NAPI_GRO_CB(skb)->frag0_len = 0;
2459 return skb->data + offset;
2462 static inline void *skb_gro_network_header(struct sk_buff *skb)
2464 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2465 skb_network_offset(skb);
2468 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2469 const void *start, unsigned int len)
2471 if (NAPI_GRO_CB(skb)->csum_valid)
2472 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2473 csum_partial(start, len, 0));
2476 /* GRO checksum functions. These are logical equivalents of the normal
2477 * checksum functions (in skbuff.h) except that they operate on the GRO
2478 * offsets and fields in sk_buff.
2481 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2483 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2485 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2488 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2492 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2493 skb_checksum_start_offset(skb) <
2494 skb_gro_offset(skb)) &&
2495 !skb_at_gro_remcsum_start(skb) &&
2496 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2497 (!zero_okay || check));
2500 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2503 if (NAPI_GRO_CB(skb)->csum_valid &&
2504 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2507 NAPI_GRO_CB(skb)->csum = psum;
2509 return __skb_gro_checksum_complete(skb);
2512 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2514 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2515 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2516 NAPI_GRO_CB(skb)->csum_cnt--;
2518 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2519 * verified a new top level checksum or an encapsulated one
2520 * during GRO. This saves work if we fallback to normal path.
2522 __skb_incr_checksum_unnecessary(skb);
2526 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2529 __sum16 __ret = 0; \
2530 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2531 __ret = __skb_gro_checksum_validate_complete(skb, \
2532 compute_pseudo(skb, proto)); \
2534 __skb_mark_checksum_bad(skb); \
2536 skb_gro_incr_csum_unnecessary(skb); \
2540 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2541 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2543 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2545 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2547 #define skb_gro_checksum_simple_validate(skb) \
2548 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2550 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2552 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2553 !NAPI_GRO_CB(skb)->csum_valid);
2556 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2557 __sum16 check, __wsum pseudo)
2559 NAPI_GRO_CB(skb)->csum = ~pseudo;
2560 NAPI_GRO_CB(skb)->csum_valid = 1;
2563 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2565 if (__skb_gro_checksum_convert_check(skb)) \
2566 __skb_gro_checksum_convert(skb, check, \
2567 compute_pseudo(skb, proto)); \
2570 struct gro_remcsum {
2575 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2581 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2582 unsigned int off, size_t hdrlen,
2583 int start, int offset,
2584 struct gro_remcsum *grc,
2588 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2590 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2593 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2597 ptr = skb_gro_header_fast(skb, off);
2598 if (skb_gro_header_hard(skb, off + plen)) {
2599 ptr = skb_gro_header_slow(skb, off + plen, off);
2604 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2607 /* Adjust skb->csum since we changed the packet */
2608 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2610 grc->offset = off + hdrlen + offset;
2616 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2617 struct gro_remcsum *grc)
2620 size_t plen = grc->offset + sizeof(u16);
2625 ptr = skb_gro_header_fast(skb, grc->offset);
2626 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2627 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2632 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2635 struct skb_csum_offl_spec {
2649 bool __skb_csum_offload_chk(struct sk_buff *skb,
2650 const struct skb_csum_offl_spec *spec,
2651 bool *csum_encapped,
2654 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2655 const struct skb_csum_offl_spec *spec,
2656 bool *csum_encapped,
2659 if (skb->ip_summed != CHECKSUM_PARTIAL)
2662 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2665 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2666 const struct skb_csum_offl_spec *spec)
2670 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2673 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2675 static const struct skb_csum_offl_spec csum_offl_spec = {
2677 .ip_options_okay = 1,
2684 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2687 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2689 static const struct skb_csum_offl_spec csum_offl_spec = {
2691 .ip_options_okay = 1,
2697 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2700 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2701 unsigned short type,
2702 const void *daddr, const void *saddr,
2705 if (!dev->header_ops || !dev->header_ops->create)
2708 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2711 static inline int dev_parse_header(const struct sk_buff *skb,
2712 unsigned char *haddr)
2714 const struct net_device *dev = skb->dev;
2716 if (!dev->header_ops || !dev->header_ops->parse)
2718 return dev->header_ops->parse(skb, haddr);
2721 /* ll_header must have at least hard_header_len allocated */
2722 static inline bool dev_validate_header(const struct net_device *dev,
2723 char *ll_header, int len)
2725 if (likely(len >= dev->hard_header_len))
2728 if (capable(CAP_SYS_RAWIO)) {
2729 memset(ll_header + len, 0, dev->hard_header_len - len);
2733 if (dev->header_ops && dev->header_ops->validate)
2734 return dev->header_ops->validate(ll_header, len);
2739 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2740 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2741 static inline int unregister_gifconf(unsigned int family)
2743 return register_gifconf(family, NULL);
2746 #ifdef CONFIG_NET_FLOW_LIMIT
2747 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2748 struct sd_flow_limit {
2750 unsigned int num_buckets;
2751 unsigned int history_head;
2752 u16 history[FLOW_LIMIT_HISTORY];
2756 extern int netdev_flow_limit_table_len;
2757 #endif /* CONFIG_NET_FLOW_LIMIT */
2760 * Incoming packets are placed on per-CPU queues
2762 struct softnet_data {
2763 struct list_head poll_list;
2764 struct sk_buff_head process_queue;
2767 unsigned int processed;
2768 unsigned int time_squeeze;
2769 unsigned int received_rps;
2771 struct softnet_data *rps_ipi_list;
2773 #ifdef CONFIG_NET_FLOW_LIMIT
2774 struct sd_flow_limit __rcu *flow_limit;
2776 struct Qdisc *output_queue;
2777 struct Qdisc **output_queue_tailp;
2778 struct sk_buff *completion_queue;
2781 /* input_queue_head should be written by cpu owning this struct,
2782 * and only read by other cpus. Worth using a cache line.
2784 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2786 /* Elements below can be accessed between CPUs for RPS/RFS */
2787 struct call_single_data csd ____cacheline_aligned_in_smp;
2788 struct softnet_data *rps_ipi_next;
2790 unsigned int input_queue_tail;
2792 unsigned int dropped;
2793 struct sk_buff_head input_pkt_queue;
2794 struct napi_struct backlog;
2798 static inline void input_queue_head_incr(struct softnet_data *sd)
2801 sd->input_queue_head++;
2805 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2806 unsigned int *qtail)
2809 *qtail = ++sd->input_queue_tail;
2813 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2815 void __netif_schedule(struct Qdisc *q);
2816 void netif_schedule_queue(struct netdev_queue *txq);
2818 static inline void netif_tx_schedule_all(struct net_device *dev)
2822 for (i = 0; i < dev->num_tx_queues; i++)
2823 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2826 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2828 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2832 * netif_start_queue - allow transmit
2833 * @dev: network device
2835 * Allow upper layers to call the device hard_start_xmit routine.
2837 static inline void netif_start_queue(struct net_device *dev)
2839 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2842 static inline void netif_tx_start_all_queues(struct net_device *dev)
2846 for (i = 0; i < dev->num_tx_queues; i++) {
2847 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2848 netif_tx_start_queue(txq);
2852 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2855 * netif_wake_queue - restart transmit
2856 * @dev: network device
2858 * Allow upper layers to call the device hard_start_xmit routine.
2859 * Used for flow control when transmit resources are available.
2861 static inline void netif_wake_queue(struct net_device *dev)
2863 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2866 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2870 for (i = 0; i < dev->num_tx_queues; i++) {
2871 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2872 netif_tx_wake_queue(txq);
2876 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2878 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2882 * netif_stop_queue - stop transmitted packets
2883 * @dev: network device
2885 * Stop upper layers calling the device hard_start_xmit routine.
2886 * Used for flow control when transmit resources are unavailable.
2888 static inline void netif_stop_queue(struct net_device *dev)
2890 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2893 void netif_tx_stop_all_queues(struct net_device *dev);
2895 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2897 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2901 * netif_queue_stopped - test if transmit queue is flowblocked
2902 * @dev: network device
2904 * Test if transmit queue on device is currently unable to send.
2906 static inline bool netif_queue_stopped(const struct net_device *dev)
2908 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2911 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2913 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2917 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2919 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2923 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2925 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2929 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2930 * @dev_queue: pointer to transmit queue
2932 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2933 * to give appropriate hint to the CPU.
2935 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2938 prefetchw(&dev_queue->dql.num_queued);
2943 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2944 * @dev_queue: pointer to transmit queue
2946 * BQL enabled drivers might use this helper in their TX completion path,
2947 * to give appropriate hint to the CPU.
2949 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2952 prefetchw(&dev_queue->dql.limit);
2956 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2960 dql_queued(&dev_queue->dql, bytes);
2962 if (likely(dql_avail(&dev_queue->dql) >= 0))
2965 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2968 * The XOFF flag must be set before checking the dql_avail below,
2969 * because in netdev_tx_completed_queue we update the dql_completed
2970 * before checking the XOFF flag.
2974 /* check again in case another CPU has just made room avail */
2975 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2976 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2981 * netdev_sent_queue - report the number of bytes queued to hardware
2982 * @dev: network device
2983 * @bytes: number of bytes queued to the hardware device queue
2985 * Report the number of bytes queued for sending/completion to the network
2986 * device hardware queue. @bytes should be a good approximation and should
2987 * exactly match netdev_completed_queue() @bytes
2989 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2991 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2994 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2995 unsigned int pkts, unsigned int bytes)
2998 if (unlikely(!bytes))
3001 dql_completed(&dev_queue->dql, bytes);
3004 * Without the memory barrier there is a small possiblity that
3005 * netdev_tx_sent_queue will miss the update and cause the queue to
3006 * be stopped forever
3010 if (dql_avail(&dev_queue->dql) < 0)
3013 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3014 netif_schedule_queue(dev_queue);
3019 * netdev_completed_queue - report bytes and packets completed by device
3020 * @dev: network device
3021 * @pkts: actual number of packets sent over the medium
3022 * @bytes: actual number of bytes sent over the medium
3024 * Report the number of bytes and packets transmitted by the network device
3025 * hardware queue over the physical medium, @bytes must exactly match the
3026 * @bytes amount passed to netdev_sent_queue()
3028 static inline void netdev_completed_queue(struct net_device *dev,
3029 unsigned int pkts, unsigned int bytes)
3031 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3034 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3037 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3043 * netdev_reset_queue - reset the packets and bytes count of a network device
3044 * @dev_queue: network device
3046 * Reset the bytes and packet count of a network device and clear the
3047 * software flow control OFF bit for this network device
3049 static inline void netdev_reset_queue(struct net_device *dev_queue)
3051 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3055 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3056 * @dev: network device
3057 * @queue_index: given tx queue index
3059 * Returns 0 if given tx queue index >= number of device tx queues,
3060 * otherwise returns the originally passed tx queue index.
3062 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3064 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3065 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3066 dev->name, queue_index,
3067 dev->real_num_tx_queues);
3075 * netif_running - test if up
3076 * @dev: network device
3078 * Test if the device has been brought up.
3080 static inline bool netif_running(const struct net_device *dev)
3082 return test_bit(__LINK_STATE_START, &dev->state);
3086 * Routines to manage the subqueues on a device. We only need start,
3087 * stop, and a check if it's stopped. All other device management is
3088 * done at the overall netdevice level.
3089 * Also test the device if we're multiqueue.
3093 * netif_start_subqueue - allow sending packets on subqueue
3094 * @dev: network device
3095 * @queue_index: sub queue index
3097 * Start individual transmit queue of a device with multiple transmit queues.
3099 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3101 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3103 netif_tx_start_queue(txq);
3107 * netif_stop_subqueue - stop sending packets on subqueue
3108 * @dev: network device
3109 * @queue_index: sub queue index
3111 * Stop individual transmit queue of a device with multiple transmit queues.
3113 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3115 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3116 netif_tx_stop_queue(txq);
3120 * netif_subqueue_stopped - test status of subqueue
3121 * @dev: network device
3122 * @queue_index: sub queue index
3124 * Check individual transmit queue of a device with multiple transmit queues.
3126 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3129 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3131 return netif_tx_queue_stopped(txq);
3134 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3135 struct sk_buff *skb)
3137 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3140 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3143 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3146 static inline int netif_set_xps_queue(struct net_device *dev,
3147 const struct cpumask *mask,
3154 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3155 unsigned int num_tx_queues);
3158 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3159 * as a distribution range limit for the returned value.
3161 static inline u16 skb_tx_hash(const struct net_device *dev,
3162 struct sk_buff *skb)
3164 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3168 * netif_is_multiqueue - test if device has multiple transmit queues
3169 * @dev: network device
3171 * Check if device has multiple transmit queues
3173 static inline bool netif_is_multiqueue(const struct net_device *dev)
3175 return dev->num_tx_queues > 1;
3178 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3181 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3183 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3191 static inline unsigned int get_netdev_rx_queue_index(
3192 struct netdev_rx_queue *queue)
3194 struct net_device *dev = queue->dev;
3195 int index = queue - dev->_rx;
3197 BUG_ON(index >= dev->num_rx_queues);
3202 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3203 int netif_get_num_default_rss_queues(void);
3205 enum skb_free_reason {
3206 SKB_REASON_CONSUMED,
3210 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3211 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3214 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3215 * interrupt context or with hardware interrupts being disabled.
3216 * (in_irq() || irqs_disabled())
3218 * We provide four helpers that can be used in following contexts :
3220 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3221 * replacing kfree_skb(skb)
3223 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3224 * Typically used in place of consume_skb(skb) in TX completion path
3226 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3227 * replacing kfree_skb(skb)
3229 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3230 * and consumed a packet. Used in place of consume_skb(skb)
3232 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3234 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3237 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3239 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3242 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3244 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3247 static inline void dev_consume_skb_any(struct sk_buff *skb)
3249 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3252 int netif_rx(struct sk_buff *skb);
3253 int netif_rx_ni(struct sk_buff *skb);
3254 int netif_receive_skb(struct sk_buff *skb);
3255 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3256 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3257 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3258 gro_result_t napi_gro_frags(struct napi_struct *napi);
3259 struct packet_offload *gro_find_receive_by_type(__be16 type);
3260 struct packet_offload *gro_find_complete_by_type(__be16 type);
3262 static inline void napi_free_frags(struct napi_struct *napi)
3264 kfree_skb(napi->skb);
3268 int netdev_rx_handler_register(struct net_device *dev,
3269 rx_handler_func_t *rx_handler,
3270 void *rx_handler_data);
3271 void netdev_rx_handler_unregister(struct net_device *dev);
3273 bool dev_valid_name(const char *name);
3274 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3275 int dev_ethtool(struct net *net, struct ifreq *);
3276 unsigned int dev_get_flags(const struct net_device *);
3277 int __dev_change_flags(struct net_device *, unsigned int flags);
3278 int dev_change_flags(struct net_device *, unsigned int);
3279 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3280 unsigned int gchanges);
3281 int dev_change_name(struct net_device *, const char *);
3282 int dev_set_alias(struct net_device *, const char *, size_t);
3283 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3284 int dev_set_mtu(struct net_device *, int);
3285 void dev_set_group(struct net_device *, int);
3286 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3287 int dev_change_carrier(struct net_device *, bool new_carrier);
3288 int dev_get_phys_port_id(struct net_device *dev,
3289 struct netdev_phys_item_id *ppid);
3290 int dev_get_phys_port_name(struct net_device *dev,
3291 char *name, size_t len);
3292 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3293 int dev_change_xdp_fd(struct net_device *dev, int fd);
3294 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3295 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3296 struct netdev_queue *txq, int *ret);
3297 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3298 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3299 bool is_skb_forwardable(const struct net_device *dev,
3300 const struct sk_buff *skb);
3302 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3304 extern int netdev_budget;
3306 /* Called by rtnetlink.c:rtnl_unlock() */
3307 void netdev_run_todo(void);
3310 * dev_put - release reference to device
3311 * @dev: network device
3313 * Release reference to device to allow it to be freed.
3315 static inline void dev_put(struct net_device *dev)
3317 this_cpu_dec(*dev->pcpu_refcnt);
3321 * dev_hold - get reference to device
3322 * @dev: network device
3324 * Hold reference to device to keep it from being freed.
3326 static inline void dev_hold(struct net_device *dev)
3328 this_cpu_inc(*dev->pcpu_refcnt);
3331 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3332 * and _off may be called from IRQ context, but it is caller
3333 * who is responsible for serialization of these calls.
3335 * The name carrier is inappropriate, these functions should really be
3336 * called netif_lowerlayer_*() because they represent the state of any
3337 * kind of lower layer not just hardware media.
3340 void linkwatch_init_dev(struct net_device *dev);
3341 void linkwatch_fire_event(struct net_device *dev);
3342 void linkwatch_forget_dev(struct net_device *dev);
3345 * netif_carrier_ok - test if carrier present
3346 * @dev: network device
3348 * Check if carrier is present on device
3350 static inline bool netif_carrier_ok(const struct net_device *dev)
3352 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3355 unsigned long dev_trans_start(struct net_device *dev);
3357 void __netdev_watchdog_up(struct net_device *dev);
3359 void netif_carrier_on(struct net_device *dev);
3361 void netif_carrier_off(struct net_device *dev);
3364 * netif_dormant_on - mark device as dormant.
3365 * @dev: network device
3367 * Mark device as dormant (as per RFC2863).
3369 * The dormant state indicates that the relevant interface is not
3370 * actually in a condition to pass packets (i.e., it is not 'up') but is
3371 * in a "pending" state, waiting for some external event. For "on-
3372 * demand" interfaces, this new state identifies the situation where the
3373 * interface is waiting for events to place it in the up state.
3375 static inline void netif_dormant_on(struct net_device *dev)
3377 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3378 linkwatch_fire_event(dev);
3382 * netif_dormant_off - set device as not dormant.
3383 * @dev: network device
3385 * Device is not in dormant state.
3387 static inline void netif_dormant_off(struct net_device *dev)
3389 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3390 linkwatch_fire_event(dev);
3394 * netif_dormant - test if carrier present
3395 * @dev: network device
3397 * Check if carrier is present on device
3399 static inline bool netif_dormant(const struct net_device *dev)
3401 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3406 * netif_oper_up - test if device is operational
3407 * @dev: network device
3409 * Check if carrier is operational
3411 static inline bool netif_oper_up(const struct net_device *dev)
3413 return (dev->operstate == IF_OPER_UP ||
3414 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3418 * netif_device_present - is device available or removed
3419 * @dev: network device
3421 * Check if device has not been removed from system.
3423 static inline bool netif_device_present(struct net_device *dev)
3425 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3428 void netif_device_detach(struct net_device *dev);
3430 void netif_device_attach(struct net_device *dev);
3433 * Network interface message level settings
3437 NETIF_MSG_DRV = 0x0001,
3438 NETIF_MSG_PROBE = 0x0002,
3439 NETIF_MSG_LINK = 0x0004,
3440 NETIF_MSG_TIMER = 0x0008,
3441 NETIF_MSG_IFDOWN = 0x0010,
3442 NETIF_MSG_IFUP = 0x0020,
3443 NETIF_MSG_RX_ERR = 0x0040,
3444 NETIF_MSG_TX_ERR = 0x0080,
3445 NETIF_MSG_TX_QUEUED = 0x0100,
3446 NETIF_MSG_INTR = 0x0200,
3447 NETIF_MSG_TX_DONE = 0x0400,
3448 NETIF_MSG_RX_STATUS = 0x0800,
3449 NETIF_MSG_PKTDATA = 0x1000,
3450 NETIF_MSG_HW = 0x2000,
3451 NETIF_MSG_WOL = 0x4000,
3454 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3455 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3456 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3457 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3458 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3459 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3460 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3461 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3462 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3463 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3464 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3465 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3466 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3467 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3468 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3470 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3473 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3474 return default_msg_enable_bits;
3475 if (debug_value == 0) /* no output */
3477 /* set low N bits */
3478 return (1 << debug_value) - 1;
3481 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3483 spin_lock(&txq->_xmit_lock);
3484 txq->xmit_lock_owner = cpu;
3487 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3489 spin_lock_bh(&txq->_xmit_lock);
3490 txq->xmit_lock_owner = smp_processor_id();
3493 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3495 bool ok = spin_trylock(&txq->_xmit_lock);
3497 txq->xmit_lock_owner = smp_processor_id();
3501 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3503 txq->xmit_lock_owner = -1;
3504 spin_unlock(&txq->_xmit_lock);
3507 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3509 txq->xmit_lock_owner = -1;
3510 spin_unlock_bh(&txq->_xmit_lock);
3513 static inline void txq_trans_update(struct netdev_queue *txq)
3515 if (txq->xmit_lock_owner != -1)
3516 txq->trans_start = jiffies;
3519 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3520 static inline void netif_trans_update(struct net_device *dev)
3522 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3524 if (txq->trans_start != jiffies)
3525 txq->trans_start = jiffies;
3529 * netif_tx_lock - grab network device transmit lock
3530 * @dev: network device
3532 * Get network device transmit lock
3534 static inline void netif_tx_lock(struct net_device *dev)
3539 spin_lock(&dev->tx_global_lock);
3540 cpu = smp_processor_id();
3541 for (i = 0; i < dev->num_tx_queues; i++) {
3542 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3544 /* We are the only thread of execution doing a
3545 * freeze, but we have to grab the _xmit_lock in
3546 * order to synchronize with threads which are in
3547 * the ->hard_start_xmit() handler and already
3548 * checked the frozen bit.
3550 __netif_tx_lock(txq, cpu);
3551 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3552 __netif_tx_unlock(txq);
3556 static inline void netif_tx_lock_bh(struct net_device *dev)
3562 static inline void netif_tx_unlock(struct net_device *dev)
3566 for (i = 0; i < dev->num_tx_queues; i++) {
3567 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3569 /* No need to grab the _xmit_lock here. If the
3570 * queue is not stopped for another reason, we
3573 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3574 netif_schedule_queue(txq);
3576 spin_unlock(&dev->tx_global_lock);
3579 static inline void netif_tx_unlock_bh(struct net_device *dev)
3581 netif_tx_unlock(dev);
3585 #define HARD_TX_LOCK(dev, txq, cpu) { \
3586 if ((dev->features & NETIF_F_LLTX) == 0) { \
3587 __netif_tx_lock(txq, cpu); \
3591 #define HARD_TX_TRYLOCK(dev, txq) \
3592 (((dev->features & NETIF_F_LLTX) == 0) ? \
3593 __netif_tx_trylock(txq) : \
3596 #define HARD_TX_UNLOCK(dev, txq) { \
3597 if ((dev->features & NETIF_F_LLTX) == 0) { \
3598 __netif_tx_unlock(txq); \
3602 static inline void netif_tx_disable(struct net_device *dev)
3608 cpu = smp_processor_id();
3609 for (i = 0; i < dev->num_tx_queues; i++) {
3610 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3612 __netif_tx_lock(txq, cpu);
3613 netif_tx_stop_queue(txq);
3614 __netif_tx_unlock(txq);
3619 static inline void netif_addr_lock(struct net_device *dev)
3621 spin_lock(&dev->addr_list_lock);
3624 static inline void netif_addr_lock_nested(struct net_device *dev)
3626 int subclass = SINGLE_DEPTH_NESTING;
3628 if (dev->netdev_ops->ndo_get_lock_subclass)
3629 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3631 spin_lock_nested(&dev->addr_list_lock, subclass);
3634 static inline void netif_addr_lock_bh(struct net_device *dev)
3636 spin_lock_bh(&dev->addr_list_lock);
3639 static inline void netif_addr_unlock(struct net_device *dev)
3641 spin_unlock(&dev->addr_list_lock);
3644 static inline void netif_addr_unlock_bh(struct net_device *dev)
3646 spin_unlock_bh(&dev->addr_list_lock);
3650 * dev_addrs walker. Should be used only for read access. Call with
3651 * rcu_read_lock held.
3653 #define for_each_dev_addr(dev, ha) \
3654 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3656 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3658 void ether_setup(struct net_device *dev);
3660 /* Support for loadable net-drivers */
3661 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3662 unsigned char name_assign_type,
3663 void (*setup)(struct net_device *),
3664 unsigned int txqs, unsigned int rxqs);
3665 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3666 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3668 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3669 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3672 int register_netdev(struct net_device *dev);
3673 void unregister_netdev(struct net_device *dev);
3675 /* General hardware address lists handling functions */
3676 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3677 struct netdev_hw_addr_list *from_list, int addr_len);
3678 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3679 struct netdev_hw_addr_list *from_list, int addr_len);
3680 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3681 struct net_device *dev,
3682 int (*sync)(struct net_device *, const unsigned char *),
3683 int (*unsync)(struct net_device *,
3684 const unsigned char *));
3685 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3686 struct net_device *dev,
3687 int (*unsync)(struct net_device *,
3688 const unsigned char *));
3689 void __hw_addr_init(struct netdev_hw_addr_list *list);
3691 /* Functions used for device addresses handling */
3692 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3693 unsigned char addr_type);
3694 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3695 unsigned char addr_type);
3696 void dev_addr_flush(struct net_device *dev);
3697 int dev_addr_init(struct net_device *dev);
3699 /* Functions used for unicast addresses handling */
3700 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3701 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3702 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3703 int dev_uc_sync(struct net_device *to, struct net_device *from);
3704 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3705 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3706 void dev_uc_flush(struct net_device *dev);
3707 void dev_uc_init(struct net_device *dev);
3710 * __dev_uc_sync - Synchonize device's unicast list
3711 * @dev: device to sync
3712 * @sync: function to call if address should be added
3713 * @unsync: function to call if address should be removed
3715 * Add newly added addresses to the interface, and release
3716 * addresses that have been deleted.
3718 static inline int __dev_uc_sync(struct net_device *dev,
3719 int (*sync)(struct net_device *,
3720 const unsigned char *),
3721 int (*unsync)(struct net_device *,
3722 const unsigned char *))
3724 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3728 * __dev_uc_unsync - Remove synchronized addresses from device
3729 * @dev: device to sync
3730 * @unsync: function to call if address should be removed
3732 * Remove all addresses that were added to the device by dev_uc_sync().
3734 static inline void __dev_uc_unsync(struct net_device *dev,
3735 int (*unsync)(struct net_device *,
3736 const unsigned char *))
3738 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3741 /* Functions used for multicast addresses handling */
3742 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3743 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3744 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3745 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3746 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3747 int dev_mc_sync(struct net_device *to, struct net_device *from);
3748 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3749 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3750 void dev_mc_flush(struct net_device *dev);
3751 void dev_mc_init(struct net_device *dev);
3754 * __dev_mc_sync - Synchonize device's multicast list
3755 * @dev: device to sync
3756 * @sync: function to call if address should be added
3757 * @unsync: function to call if address should be removed
3759 * Add newly added addresses to the interface, and release
3760 * addresses that have been deleted.
3762 static inline int __dev_mc_sync(struct net_device *dev,
3763 int (*sync)(struct net_device *,
3764 const unsigned char *),
3765 int (*unsync)(struct net_device *,
3766 const unsigned char *))
3768 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3772 * __dev_mc_unsync - Remove synchronized addresses from device
3773 * @dev: device to sync
3774 * @unsync: function to call if address should be removed
3776 * Remove all addresses that were added to the device by dev_mc_sync().
3778 static inline void __dev_mc_unsync(struct net_device *dev,
3779 int (*unsync)(struct net_device *,
3780 const unsigned char *))
3782 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3785 /* Functions used for secondary unicast and multicast support */
3786 void dev_set_rx_mode(struct net_device *dev);
3787 void __dev_set_rx_mode(struct net_device *dev);
3788 int dev_set_promiscuity(struct net_device *dev, int inc);
3789 int dev_set_allmulti(struct net_device *dev, int inc);
3790 void netdev_state_change(struct net_device *dev);
3791 void netdev_notify_peers(struct net_device *dev);
3792 void netdev_features_change(struct net_device *dev);
3793 /* Load a device via the kmod */
3794 void dev_load(struct net *net, const char *name);
3795 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3796 struct rtnl_link_stats64 *storage);
3797 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3798 const struct net_device_stats *netdev_stats);
3800 extern int netdev_max_backlog;
3801 extern int netdev_tstamp_prequeue;
3802 extern int weight_p;
3804 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3805 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3806 struct list_head **iter);
3807 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3808 struct list_head **iter);
3810 /* iterate through upper list, must be called under RCU read lock */
3811 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3812 for (iter = &(dev)->adj_list.upper, \
3813 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3815 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3817 /* iterate through upper list, must be called under RCU read lock */
3818 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3819 for (iter = &(dev)->all_adj_list.upper, \
3820 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3822 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3824 void *netdev_lower_get_next_private(struct net_device *dev,
3825 struct list_head **iter);
3826 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3827 struct list_head **iter);
3829 #define netdev_for_each_lower_private(dev, priv, iter) \
3830 for (iter = (dev)->adj_list.lower.next, \
3831 priv = netdev_lower_get_next_private(dev, &(iter)); \
3833 priv = netdev_lower_get_next_private(dev, &(iter)))
3835 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3836 for (iter = &(dev)->adj_list.lower, \
3837 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3839 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3841 void *netdev_lower_get_next(struct net_device *dev,
3842 struct list_head **iter);
3844 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3845 for (iter = (dev)->adj_list.lower.next, \
3846 ldev = netdev_lower_get_next(dev, &(iter)); \
3848 ldev = netdev_lower_get_next(dev, &(iter)))
3850 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3851 struct list_head **iter);
3852 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3853 struct list_head **iter);
3855 #define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3856 for (iter = (dev)->all_adj_list.lower.next, \
3857 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3859 ldev = netdev_all_lower_get_next(dev, &(iter)))
3861 #define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3862 for (iter = (dev)->all_adj_list.lower.next, \
3863 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3865 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3867 void *netdev_adjacent_get_private(struct list_head *adj_list);
3868 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3869 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3870 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3871 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3872 int netdev_master_upper_dev_link(struct net_device *dev,
3873 struct net_device *upper_dev,
3874 void *upper_priv, void *upper_info);
3875 void netdev_upper_dev_unlink(struct net_device *dev,
3876 struct net_device *upper_dev);
3877 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3878 void *netdev_lower_dev_get_private(struct net_device *dev,
3879 struct net_device *lower_dev);
3880 void netdev_lower_state_changed(struct net_device *lower_dev,
3881 void *lower_state_info);
3882 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3883 struct neighbour *n);
3884 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3885 struct neighbour *n);
3887 /* RSS keys are 40 or 52 bytes long */
3888 #define NETDEV_RSS_KEY_LEN 52
3889 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3890 void netdev_rss_key_fill(void *buffer, size_t len);
3892 int dev_get_nest_level(struct net_device *dev,
3893 bool (*type_check)(const struct net_device *dev));
3894 int skb_checksum_help(struct sk_buff *skb);
3895 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3896 netdev_features_t features, bool tx_path);
3897 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3898 netdev_features_t features);
3900 struct netdev_bonding_info {
3905 struct netdev_notifier_bonding_info {
3906 struct netdev_notifier_info info; /* must be first */
3907 struct netdev_bonding_info bonding_info;
3910 void netdev_bonding_info_change(struct net_device *dev,
3911 struct netdev_bonding_info *bonding_info);
3914 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3916 return __skb_gso_segment(skb, features, true);
3918 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3920 static inline bool can_checksum_protocol(netdev_features_t features,
3923 if (protocol == htons(ETH_P_FCOE))
3924 return !!(features & NETIF_F_FCOE_CRC);
3926 /* Assume this is an IP checksum (not SCTP CRC) */
3928 if (features & NETIF_F_HW_CSUM) {
3929 /* Can checksum everything */
3934 case htons(ETH_P_IP):
3935 return !!(features & NETIF_F_IP_CSUM);
3936 case htons(ETH_P_IPV6):
3937 return !!(features & NETIF_F_IPV6_CSUM);
3943 /* Map an ethertype into IP protocol if possible */
3944 static inline int eproto_to_ipproto(int eproto)
3947 case htons(ETH_P_IP):
3949 case htons(ETH_P_IPV6):
3950 return IPPROTO_IPV6;
3957 void netdev_rx_csum_fault(struct net_device *dev);
3959 static inline void netdev_rx_csum_fault(struct net_device *dev)
3963 /* rx skb timestamps */
3964 void net_enable_timestamp(void);
3965 void net_disable_timestamp(void);
3967 #ifdef CONFIG_PROC_FS
3968 int __init dev_proc_init(void);
3970 #define dev_proc_init() 0
3973 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3974 struct sk_buff *skb, struct net_device *dev,
3977 skb->xmit_more = more ? 1 : 0;
3978 return ops->ndo_start_xmit(skb, dev);
3981 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3982 struct netdev_queue *txq, bool more)
3984 const struct net_device_ops *ops = dev->netdev_ops;
3987 rc = __netdev_start_xmit(ops, skb, dev, more);
3988 if (rc == NETDEV_TX_OK)
3989 txq_trans_update(txq);
3994 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3996 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3999 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4001 return netdev_class_create_file_ns(class_attr, NULL);
4004 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4006 netdev_class_remove_file_ns(class_attr, NULL);
4009 extern struct kobj_ns_type_operations net_ns_type_operations;
4011 const char *netdev_drivername(const struct net_device *dev);
4013 void linkwatch_run_queue(void);
4015 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4016 netdev_features_t f2)
4018 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4019 if (f1 & NETIF_F_HW_CSUM)
4020 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4022 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4028 static inline netdev_features_t netdev_get_wanted_features(
4029 struct net_device *dev)
4031 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4033 netdev_features_t netdev_increment_features(netdev_features_t all,
4034 netdev_features_t one, netdev_features_t mask);
4036 /* Allow TSO being used on stacked device :
4037 * Performing the GSO segmentation before last device
4038 * is a performance improvement.
4040 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4041 netdev_features_t mask)
4043 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4046 int __netdev_update_features(struct net_device *dev);
4047 void netdev_update_features(struct net_device *dev);
4048 void netdev_change_features(struct net_device *dev);
4050 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4051 struct net_device *dev);
4053 netdev_features_t passthru_features_check(struct sk_buff *skb,
4054 struct net_device *dev,
4055 netdev_features_t features);
4056 netdev_features_t netif_skb_features(struct sk_buff *skb);
4058 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4060 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4062 /* check flags correspondence */
4063 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4064 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4065 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4066 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4067 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4068 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4069 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4070 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4071 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4072 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4073 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4074 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4075 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4076 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4077 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4078 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4080 return (features & feature) == feature;
4083 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4085 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4086 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4089 static inline bool netif_needs_gso(struct sk_buff *skb,
4090 netdev_features_t features)
4092 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4093 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4094 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4097 static inline void netif_set_gso_max_size(struct net_device *dev,
4100 dev->gso_max_size = size;
4103 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4104 int pulled_hlen, u16 mac_offset,
4107 skb->protocol = protocol;
4108 skb->encapsulation = 1;
4109 skb_push(skb, pulled_hlen);
4110 skb_reset_transport_header(skb);
4111 skb->mac_header = mac_offset;
4112 skb->network_header = skb->mac_header + mac_len;
4113 skb->mac_len = mac_len;
4116 static inline bool netif_is_macsec(const struct net_device *dev)
4118 return dev->priv_flags & IFF_MACSEC;
4121 static inline bool netif_is_macvlan(const struct net_device *dev)
4123 return dev->priv_flags & IFF_MACVLAN;
4126 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4128 return dev->priv_flags & IFF_MACVLAN_PORT;
4131 static inline bool netif_is_ipvlan(const struct net_device *dev)
4133 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4136 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4138 return dev->priv_flags & IFF_IPVLAN_MASTER;
4141 static inline bool netif_is_bond_master(const struct net_device *dev)
4143 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4146 static inline bool netif_is_bond_slave(const struct net_device *dev)
4148 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4151 static inline bool netif_supports_nofcs(struct net_device *dev)
4153 return dev->priv_flags & IFF_SUPP_NOFCS;
4156 static inline bool netif_is_l3_master(const struct net_device *dev)
4158 return dev->priv_flags & IFF_L3MDEV_MASTER;
4161 static inline bool netif_is_l3_slave(const struct net_device *dev)
4163 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4166 static inline bool netif_is_bridge_master(const struct net_device *dev)
4168 return dev->priv_flags & IFF_EBRIDGE;
4171 static inline bool netif_is_bridge_port(const struct net_device *dev)
4173 return dev->priv_flags & IFF_BRIDGE_PORT;
4176 static inline bool netif_is_ovs_master(const struct net_device *dev)
4178 return dev->priv_flags & IFF_OPENVSWITCH;
4181 static inline bool netif_is_team_master(const struct net_device *dev)
4183 return dev->priv_flags & IFF_TEAM;
4186 static inline bool netif_is_team_port(const struct net_device *dev)
4188 return dev->priv_flags & IFF_TEAM_PORT;
4191 static inline bool netif_is_lag_master(const struct net_device *dev)
4193 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4196 static inline bool netif_is_lag_port(const struct net_device *dev)
4198 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4201 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4203 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4206 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4207 static inline void netif_keep_dst(struct net_device *dev)
4209 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4212 extern struct pernet_operations __net_initdata loopback_net_ops;
4214 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4216 /* netdev_printk helpers, similar to dev_printk */
4218 static inline const char *netdev_name(const struct net_device *dev)
4220 if (!dev->name[0] || strchr(dev->name, '%'))
4221 return "(unnamed net_device)";
4225 static inline const char *netdev_reg_state(const struct net_device *dev)
4227 switch (dev->reg_state) {
4228 case NETREG_UNINITIALIZED: return " (uninitialized)";
4229 case NETREG_REGISTERED: return "";
4230 case NETREG_UNREGISTERING: return " (unregistering)";
4231 case NETREG_UNREGISTERED: return " (unregistered)";
4232 case NETREG_RELEASED: return " (released)";
4233 case NETREG_DUMMY: return " (dummy)";
4236 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4237 return " (unknown)";
4241 void netdev_printk(const char *level, const struct net_device *dev,
4242 const char *format, ...);
4244 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4246 void netdev_alert(const struct net_device *dev, const char *format, ...);
4248 void netdev_crit(const struct net_device *dev, const char *format, ...);
4250 void netdev_err(const struct net_device *dev, const char *format, ...);
4252 void netdev_warn(const struct net_device *dev, const char *format, ...);
4254 void netdev_notice(const struct net_device *dev, const char *format, ...);
4256 void netdev_info(const struct net_device *dev, const char *format, ...);
4258 #define MODULE_ALIAS_NETDEV(device) \
4259 MODULE_ALIAS("netdev-" device)
4261 #if defined(CONFIG_DYNAMIC_DEBUG)
4262 #define netdev_dbg(__dev, format, args...) \
4264 dynamic_netdev_dbg(__dev, format, ##args); \
4266 #elif defined(DEBUG)
4267 #define netdev_dbg(__dev, format, args...) \
4268 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4270 #define netdev_dbg(__dev, format, args...) \
4273 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4277 #if defined(VERBOSE_DEBUG)
4278 #define netdev_vdbg netdev_dbg
4281 #define netdev_vdbg(dev, format, args...) \
4284 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4290 * netdev_WARN() acts like dev_printk(), but with the key difference
4291 * of using a WARN/WARN_ON to get the message out, including the
4292 * file/line information and a backtrace.
4294 #define netdev_WARN(dev, format, args...) \
4295 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4296 netdev_reg_state(dev), ##args)
4298 /* netif printk helpers, similar to netdev_printk */
4300 #define netif_printk(priv, type, level, dev, fmt, args...) \
4302 if (netif_msg_##type(priv)) \
4303 netdev_printk(level, (dev), fmt, ##args); \
4306 #define netif_level(level, priv, type, dev, fmt, args...) \
4308 if (netif_msg_##type(priv)) \
4309 netdev_##level(dev, fmt, ##args); \
4312 #define netif_emerg(priv, type, dev, fmt, args...) \
4313 netif_level(emerg, priv, type, dev, fmt, ##args)
4314 #define netif_alert(priv, type, dev, fmt, args...) \
4315 netif_level(alert, priv, type, dev, fmt, ##args)
4316 #define netif_crit(priv, type, dev, fmt, args...) \
4317 netif_level(crit, priv, type, dev, fmt, ##args)
4318 #define netif_err(priv, type, dev, fmt, args...) \
4319 netif_level(err, priv, type, dev, fmt, ##args)
4320 #define netif_warn(priv, type, dev, fmt, args...) \
4321 netif_level(warn, priv, type, dev, fmt, ##args)
4322 #define netif_notice(priv, type, dev, fmt, args...) \
4323 netif_level(notice, priv, type, dev, fmt, ##args)
4324 #define netif_info(priv, type, dev, fmt, args...) \
4325 netif_level(info, priv, type, dev, fmt, ##args)
4327 #if defined(CONFIG_DYNAMIC_DEBUG)
4328 #define netif_dbg(priv, type, netdev, format, args...) \
4330 if (netif_msg_##type(priv)) \
4331 dynamic_netdev_dbg(netdev, format, ##args); \
4333 #elif defined(DEBUG)
4334 #define netif_dbg(priv, type, dev, format, args...) \
4335 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4337 #define netif_dbg(priv, type, dev, format, args...) \
4340 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4345 #if defined(VERBOSE_DEBUG)
4346 #define netif_vdbg netif_dbg
4348 #define netif_vdbg(priv, type, dev, format, args...) \
4351 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4357 * The list of packet types we will receive (as opposed to discard)
4358 * and the routines to invoke.
4360 * Why 16. Because with 16 the only overlap we get on a hash of the
4361 * low nibble of the protocol value is RARP/SNAP/X.25.
4363 * NOTE: That is no longer true with the addition of VLAN tags. Not
4364 * sure which should go first, but I bet it won't make much
4365 * difference if we are running VLANs. The good news is that
4366 * this protocol won't be in the list unless compiled in, so
4367 * the average user (w/out VLANs) will not be adversely affected.
4383 #define PTYPE_HASH_SIZE (16)
4384 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4386 #endif /* _LINUX_NETDEVICE_H */