blob: ec8f9b5f650058295c7518f2c055cba1aee4404f [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * 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>
Alan Cox113aa832008-10-13 19:01:08 -070014 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
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.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
Jean Pihete8db0be2011-08-25 15:35:03 +020028#include <linux/pm_qos.h>
Al Virod7fe0f22006-12-03 23:15:30 -050029#include <linux/timer.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050030#include <linux/bug.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070031#include <linux/delay.h>
Arun Sharma600634972011-07-26 16:09:06 -070032#include <linux/atomic.h>
Eric Dumazet53511452014-10-08 08:19:27 -070033#include <linux/prefetch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <asm/cache.h>
35#include <asm/byteorder.h>
36
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <linux/percpu.h>
David S. Miller4d5b78c2009-05-06 16:52:51 -070038#include <linux/rculist.h>
Chris Leechdb217332006-06-17 21:24:58 -070039#include <linux/dmaengine.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070040#include <linux/workqueue.h>
Tom Herbert114cf582011-11-28 16:33:09 +000041#include <linux/dynamic_queue_limits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
Patrick McHardyb1b67dd2009-04-20 04:49:28 +000043#include <linux/ethtool.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020044#include <net/net_namespace.h>
Lennert Buytenhekcf85d082008-10-07 13:45:02 +000045#include <net/dsa.h>
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -080046#ifdef CONFIG_DCB
Alexander Duyck2f90b862008-11-20 20:52:10 -080047#include <net/dcbnl.h>
48#endif
Neil Horman5bc14212011-11-22 05:10:51 +000049#include <net/netprio_cgroup.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020050
Michał Mirosława59e2ec2011-11-15 15:29:55 +000051#include <linux/netdev_features.h>
John Fastabend77162022012-04-15 06:43:56 +000052#include <linux/neighbour.h>
David Howells607ca462012-10-13 10:46:48 +010053#include <uapi/linux/netdevice.h>
Moni Shoua61bd3852015-02-03 16:48:29 +020054#include <uapi/linux/if_bonding.h>
Michał Mirosława59e2ec2011-11-15 15:29:55 +000055
Jeff Moyer115c1d62005-06-22 22:05:31 -070056struct netpoll_info;
Paul Gortmaker313162d2012-01-30 11:46:54 -050057struct device;
Richard Cochranc1f19b52010-07-17 08:49:36 +000058struct phy_device;
Johannes Berg704232c2007-04-23 12:20:05 -070059/* 802.11 specific */
60struct wireless_dev;
Alexander Aring98a18b62014-11-02 06:44:54 +010061/* 802.15.4 specific */
62struct wpan_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
Joe Perchesf629d202013-09-26 14:48:15 -070064void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000066
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000067/* Backlog congestion levels */
68#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69#define NET_RX_DROP 1 /* packet dropped */
70
Patrick McHardy572a9d72009-11-10 06:14:14 +000071/*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88/* qdisc ->enqueue() return codes. */
89#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000090#define NET_XMIT_DROP 0x01 /* skb dropped */
91#define NET_XMIT_CN 0x02 /* congestion notification */
92#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -020095/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +000098#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000101/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000102#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000103
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000104enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000109};
110typedef enum netdev_tx netdev_tx_t;
111
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000112/*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116static inline bool dev_xmit_complete(int rc)
117{
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128}
129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800134
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000135#if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700136# if defined(CONFIG_MAC80211_MESH)
137# define LL_MAX_HEADER 128
138# else
139# define LL_MAX_HEADER 96
140# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700142# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143#endif
144
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000145#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147#define MAX_HEADER LL_MAX_HEADER
148#else
149#define MAX_HEADER (LL_MAX_HEADER + 48)
150#endif
151
152/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800156
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800157struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181};
182
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183
184#include <linux/cache.h>
185#include <linux/skbuff.h>
186
Eric Dumazetadc93002011-11-17 03:13:26 +0000187#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100188#include <linux/static_key.h>
189extern struct static_key rps_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000190#endif
191
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192struct neighbour;
193struct neigh_parms;
194struct sk_buff;
195
Jiri Pirkof001fde2009-05-05 02:48:28 +0000196struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000200#define NETDEV_HW_ADDR_T_LAN 1
201#define NETDEV_HW_ADDR_T_SAN 2
202#define NETDEV_HW_ADDR_T_SLAVE 3
203#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000204#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000205 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000206 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700207 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400208 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000209 struct rcu_head rcu_head;
210};
211
Jiri Pirko31278e72009-06-17 01:12:19 +0000212struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215};
216
Jiri Pirko22bedad32010-04-01 21:22:57 +0000217#define netdev_hw_addr_list_count(l) ((l)->count)
218#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219#define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
221
222#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800224#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800226
Jiri Pirko22bedad32010-04-01 21:22:57 +0000227#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700229#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece32010-02-04 10:22:25 -0800231
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800232struct hh_cache {
David S. Millerf6b72b622011-07-14 07:53:20 -0700233 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700234 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800235 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
237 /* cached hardware header; allow for machine alignment needs. */
238#define HH_DATA_MOD 16
239#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241#define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244};
245
246/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700259struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000262 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268};
269
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270/* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800275enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281};
282
283
284/*
285 * This structure holds at boot time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800286 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 */
288struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291};
292#define NETDEV_BOOT_SETUP_MAX 8
293
Joe Perchesf629d202013-09-26 14:48:15 -0700294int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295
296/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000310 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700311 int (*poll)(struct napi_struct *, int);
312#ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700315#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800316 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800317 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800318 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800319 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000320 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700323};
324
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800325enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700326 NAPI_STATE_SCHED, /* Poll is scheduled */
David S. Millera0a46192008-01-07 20:35:07 -0800327 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300329 NAPI_STATE_HASHED, /* In NAPI hash */
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700330};
331
Ben Hutchings5b252f02009-10-29 07:17:09 +0000332enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000339typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700340
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000341/*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400373 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387};
388typedef enum rx_handler_result rx_handler_result_t;
389typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000390
Joe Perchesf629d202013-09-26 14:48:15 -0700391void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700392void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700393
David S. Miller4d295152012-03-07 21:02:35 -0500394static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800395{
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397}
398
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700399/**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
David S. Millera0a46192008-01-07 20:35:07 -0800405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700407 */
David S. Miller4d295152012-03-07 21:02:35 -0500408static inline bool napi_schedule_prep(struct napi_struct *n)
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700409{
David S. Millera0a46192008-01-07 20:35:07 -0800410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700412}
413
414/**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421static inline void napi_schedule(struct napi_struct *n)
422{
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425}
426
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700427/**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433static inline void napi_schedule_irqoff(struct napi_struct *n)
434{
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437}
438
Roland Dreierbfe13f52007-10-09 15:47:37 -0700439/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500440static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700441{
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500444 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700445 }
David S. Miller4d295152012-03-07 21:02:35 -0500446 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700447}
448
Eric Dumazet3b47d302014-11-06 21:09:44 -0800449void __napi_complete(struct napi_struct *n);
450void napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700451/**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800456 * Consider using napi_complete_done() instead.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700457 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800458static inline void napi_complete(struct napi_struct *n)
459{
460 return napi_complete_done(n, 0);
461}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700462
463/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
Joe Perchesf629d202013-09-26 14:48:15 -0700470struct napi_struct *napi_by_id(unsigned int napi_id);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300471
472/**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
Joe Perchesf629d202013-09-26 14:48:15 -0700478void napi_hash_add(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300479
480/**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
Joe Perchesf629d202013-09-26 14:48:15 -0700487void napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300488
489/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800496void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700497
498/**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505static inline void napi_enable(struct napi_struct *n)
506{
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100508 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700509 clear_bit(NAPI_STATE_SCHED, &n->state);
510}
511
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700512#ifdef CONFIG_SMP
513/**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: napi context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521static inline void napi_synchronize(const struct napi_struct *n)
522{
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
524 msleep(1);
525}
526#else
527# define napi_synchronize(n) barrier()
528#endif
529
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800530enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700533 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700534};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200535
536#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
539
540#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
542 QUEUE_STATE_FROZEN)
543#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
544 QUEUE_STATE_FROZEN)
545
Tom Herbert734664982011-11-28 16:32:44 +0000546/*
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 */
David S. Miller79d16382008-07-08 23:14:46 -0700555
David S. Millerbb949fb2008-07-08 16:55:56 -0700556struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700557/*
558 * read mostly part
559 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700560 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700561 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700562 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000563#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000564 struct kobject kobj;
565#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000566#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 int numa_node;
568#endif
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700569/*
570 * write mostly part
571 */
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700574 /*
575 * please use this field instead of dev->trans_start
576 */
577 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000578
579 /*
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
582 */
583 unsigned long trans_timeout;
Tom Herbert114cf582011-11-28 16:33:09 +0000584
585 unsigned long state;
586
587#ifdef CONFIG_BQL
588 struct dql dql;
589#endif
John Fastabend822b3b22015-03-18 14:57:33 +0200590 unsigned long tx_maxrate;
David S. Millere8a04642008-07-17 00:34:19 -0700591} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700592
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000593static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594{
595#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 return q->numa_node;
597#else
Changli Gaob236da62010-12-14 03:09:15 +0000598 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000599#endif
600}
601
602static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603{
604#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 q->numa_node = node;
606#endif
607}
608
Eric Dumazetdf334542010-03-24 19:13:54 +0000609#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000610/*
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
613 */
614struct rps_map {
615 unsigned int len;
616 struct rcu_head rcu;
617 u16 cpus[0];
618};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000619#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000620
Tom Herbertfec5e652010-04-16 16:01:27 -0700621/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700625 */
626struct rps_dev_flow {
627 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000628 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700629 unsigned int last_qtail;
630};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000631#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700632
633/*
634 * The rps_dev_flow_table structure contains a table of flow mappings.
635 */
636struct rps_dev_flow_table {
637 unsigned int mask;
638 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700639 struct rps_dev_flow flows[0];
640};
641#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000642 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700643
644/*
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
Eric Dumazet567e4b72015-02-06 12:59:01 -0800647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700653 */
654struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800655 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800656
657 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700658};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800659#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700660
661#define RPS_NO_CPU 0xffff
662
Eric Dumazet567e4b72015-02-06 12:59:01 -0800663extern u32 rps_cpu_mask;
664extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
Tom Herbertfec5e652010-04-16 16:01:27 -0700666static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 u32 hash)
668{
669 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800670 unsigned int index = hash & table->mask;
671 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700672
673 /* We only give a hint, preemption can change cpu under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800674 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700675
Eric Dumazet567e4b72015-02-06 12:59:01 -0800676 if (table->ents[index] != val)
677 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700678 }
679}
680
Ben Hutchingsc4454772011-01-19 11:03:53 +0000681#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700682bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
683 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000684#endif
Michael Daltona953be52014-01-16 22:23:28 -0800685#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000686
Tom Herbert0a9627f2010-03-16 08:03:29 +0000687/* This structure contains an instance of an RX queue. */
688struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800689#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000690 struct rps_map __rcu *rps_map;
691 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800692#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000693 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000694 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000695} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800696
697/*
698 * RX queue sysfs structures and functions.
699 */
700struct rx_queue_attribute {
701 struct attribute attr;
702 ssize_t (*show)(struct netdev_rx_queue *queue,
703 struct rx_queue_attribute *attr, char *buf);
704 ssize_t (*store)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, const char *buf, size_t len);
706};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800707
Tom Herbertbf264142010-11-26 08:36:09 +0000708#ifdef CONFIG_XPS
709/*
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
712 */
713struct xps_map {
714 unsigned int len;
715 unsigned int alloc_len;
716 struct rcu_head rcu;
717 u16 queues[0];
718};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000719#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Tom Herbertbf264142010-11-26 08:36:09 +0000720#define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
721 / sizeof(u16))
722
723/*
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 */
726struct xps_dev_maps {
727 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000728 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000729};
730#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732#endif /* CONFIG_XPS */
733
John Fastabend4f57c082011-01-17 08:06:04 +0000734#define TC_MAX_QUEUE 16
735#define TC_BITMASK 15
736/* HW offloaded queuing disciplines txq count and offset maps */
737struct netdev_tc_txq {
738 u16 count;
739 u16 offset;
740};
741
Neerav Parikh68bad942012-01-04 20:23:39 +0000742#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743/*
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
746 */
747struct netdev_fcoe_hbainfo {
748 char manufacturer[64];
749 char serial_number[64];
750 char hardware_version[64];
751 char driver_version[64];
752 char optionrom_version[64];
753 char firmware_version[64];
754 char model[256];
755 char model_description[256];
756};
757#endif
758
Jiri Pirko02637fc2014-11-28 14:34:16 +0100759#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200760
Jiri Pirko02637fc2014-11-28 14:34:16 +0100761/* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200763 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100764struct netdev_phys_item_id {
765 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200766 unsigned char id_len;
767};
768
Daniel Borkmann99932d42014-02-16 15:55:20 +0100769typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800772/*
773 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800774 * The following hooks can be defined; unless noted otherwise, they are
775 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800776 *
777 * int (*ndo_init)(struct net_device *dev);
778 * This function is called once when network device is registered.
779 * The network device can use this to any late stage initializaton
780 * or semantic validattion. It can fail with an error code which will
781 * be propogated back to register_netdev
782 *
783 * void (*ndo_uninit)(struct net_device *dev);
784 * This function is called when device is unregistered or when registration
785 * fails. It is not called if init fails.
786 *
787 * int (*ndo_open)(struct net_device *dev);
788 * This function is called when network device transistions to the up
789 * state.
790 *
791 * int (*ndo_stop)(struct net_device *dev);
792 * This function is called when network device transistions to the down
793 * state.
794 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
796 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800797 * Called when a packet needs to be transmitted.
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000798 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
799 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
Stephen Hemminger00829822008-11-20 20:14:53 -0800800 * Required can not be NULL.
801 *
Jason Wangf663dd92014-01-10 16:18:26 +0800802 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100803 * void *accel_priv, select_queue_fallback_t fallback);
Stephen Hemminger00829822008-11-20 20:14:53 -0800804 * Called to decide which queue to when device supports multiple
805 * transmit queues.
806 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800807 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
808 * This function is called to allow device receiver to make
809 * changes to configuration when multicast or promiscious is enabled.
810 *
811 * void (*ndo_set_rx_mode)(struct net_device *dev);
812 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000813 * If driver handles unicast address filtering, it should set
814 * IFF_UNICAST_FLT to its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800815 *
816 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
817 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700818 * needs to be changed. If this interface is not defined, the
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800819 * mac address can not be changed.
820 *
821 * int (*ndo_validate_addr)(struct net_device *dev);
822 * Test if Media Access Control address is valid for the device.
823 *
824 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
825 * Called when a user request an ioctl which can't be handled by
826 * the generic interface code. If not defined ioctl's return
827 * not supported error code.
828 *
829 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
830 * Used to set network devices bus interface parameters. This interface
831 * is retained for legacy reason, new devices should use the bus
832 * interface (PCI) for low level management.
833 *
834 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
835 * Called when a user wants to change the Maximum Transfer Unit
836 * of a device. If not defined, any request to change MTU will
837 * will return an error.
838 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800839 * void (*ndo_tx_timeout)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800840 * Callback uses when the transmitter has not made any progress
841 * for dev->watchdog ticks.
842 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000843 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700844 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700845 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800846 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000847 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000848 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
849 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700850 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000851 * (which should normally be dev->stats) and return a pointer to
852 * it. The structure may be changed asynchronously only if each
853 * field is written atomically.
854 * 3. Update dev->stats asynchronously and atomically, and define
855 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800856 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530857 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Patrick McHardy80d5c362013-04-19 02:04:28 +0000858 * If device support VLAN filtering this function is called when a
859 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800860 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530861 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Patrick McHardy80d5c362013-04-19 02:04:28 +0000862 * If device support VLAN filtering this function is called when a
863 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800864 *
865 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000866 *
867 * SR-IOV management functions.
868 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
869 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400870 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
871 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000872 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000873 * int (*ndo_get_vf_config)(struct net_device *dev,
874 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300875 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700876 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
877 * struct nlattr *port[]);
878 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000879 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
880 * Called to setup 'tc' number of traffic classes in the net device. This
881 * is always called from the stack with the rtnl lock held and netif tx
882 * queues stopped. This allows the netdevice to perform queue management
883 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000884 *
Yi Zoue9bce842011-03-09 08:48:03 +0000885 * Fiber Channel over Ethernet (FCoE) offload functions.
886 * int (*ndo_fcoe_enable)(struct net_device *dev);
887 * Called when the FCoE protocol stack wants to start using LLD for FCoE
888 * so the underlying device can perform whatever needed configuration or
889 * initialization to support acceleration of FCoE traffic.
890 *
891 * int (*ndo_fcoe_disable)(struct net_device *dev);
892 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
893 * so the underlying device can perform whatever needed clean-ups to
894 * stop supporting acceleration of FCoE traffic.
895 *
896 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
897 * struct scatterlist *sgl, unsigned int sgc);
898 * Called when the FCoE Initiator wants to initialize an I/O that
899 * is a possible candidate for Direct Data Placement (DDP). The LLD can
900 * perform necessary setup and returns 1 to indicate the device is set up
901 * successfully to perform DDP on this I/O, otherwise this returns 0.
902 *
903 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
904 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
905 * indicated by the FC exchange id 'xid', so the underlying device can
906 * clean up and reuse resources for later DDP requests.
907 *
908 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
909 * struct scatterlist *sgl, unsigned int sgc);
910 * Called when the FCoE Target wants to initialize an I/O that
911 * is a possible candidate for Direct Data Placement (DDP). The LLD can
912 * perform necessary setup and returns 1 to indicate the device is set up
913 * successfully to perform DDP on this I/O, otherwise this returns 0.
914 *
Neerav Parikh68bad942012-01-04 20:23:39 +0000915 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
916 * struct netdev_fcoe_hbainfo *hbainfo);
917 * Called when the FCoE Protocol stack wants information on the underlying
918 * device. This information is utilized by the FCoE protocol stack to
919 * register attributes with Fiber Channel management service as per the
920 * FC-GS Fabric Device Management Information(FDMI) specification.
921 *
Yi Zoue9bce842011-03-09 08:48:03 +0000922 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
923 * Called when the underlying device wants to override default World Wide
924 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
925 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
926 * protocol stack to use.
927 *
Ben Hutchingsc4454772011-01-19 11:03:53 +0000928 * RFS acceleration.
929 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
930 * u16 rxq_index, u32 flow_id);
931 * Set hardware filter for RFS. rxq_index is the target queue index;
932 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
933 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000934 *
Jiri Pirko8b98a702013-01-03 22:49:02 +0000935 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000936 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
937 * Called to make another netdev an underling.
938 *
939 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
940 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +0000941 *
942 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000943 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
944 * netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +0000945 * Adjusts the requested feature flags according to device-specific
946 * constraints, and returns the resulting flags. Must not modify
947 * the device state.
948 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000949 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +0000950 * Called to update device configuration to new features. Passed
951 * feature set might be less than what was returned by ndo_fix_features()).
952 * Must return >0 or -errno if it changed dev->features itself.
953 *
stephen hemmingeredc7d572012-10-01 12:32:33 +0000954 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
955 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100956 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +0000957 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +0000958 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
959 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100960 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +0000961 * Deletes the FDB entry from dev coresponding to addr.
962 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -0400963 * struct net_device *dev, struct net_device *filter_dev,
964 * int idx)
John Fastabend77162022012-04-15 06:43:56 +0000965 * Used to add FDB entries to dump requests. Implementers should add
966 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +0000967 *
968 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
969 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Dmitry Kravkov24f11a52013-03-27 06:54:00 +0000970 * struct net_device *dev, u32 filter_mask)
Jiri Pirko4bf84c32012-12-27 23:49:37 +0000971 *
972 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
973 * Called to change device carrier. Soft-devices (like dummy, team, etc)
974 * which do not represent real hardware may define this to allow their
975 * userspace components to manage their virtual carrier state. Devices
976 * that determine carrier state from physical hardware properties (eg
977 * network cables) or protocol-dependent mechanisms (eg
978 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200979 *
980 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +0100981 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +0200982 * Called to get ID of physical port of this device. If driver does
983 * not implement this, it is assumed that the hw is not able to have
984 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700985 *
986 * void (*ndo_add_vxlan_port)(struct net_device *dev,
Joseph Gasparakis35e42372013-09-13 07:34:13 -0700987 * sa_family_t sa_family, __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700988 * Called by vxlan to notiy a driver about the UDP port and socket
989 * address family that vxlan is listnening to. It is called only when
990 * a new port starts listening. The operation is protected by the
991 * vxlan_net->sock_lock.
992 *
993 * void (*ndo_del_vxlan_port)(struct net_device *dev,
Joseph Gasparakis35e42372013-09-13 07:34:13 -0700994 * sa_family_t sa_family, __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700995 * Called by vxlan to notify the driver about a UDP port and socket
996 * address family that vxlan is not listening to anymore. The operation
997 * is protected by the vxlan_net->sock_lock.
John Fastabenda6cc0cf2013-11-06 09:54:46 -0800998 *
999 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1000 * struct net_device *dev)
1001 * Called by upper layer devices to accelerate switching or other
1002 * station functionality into hardware. 'pdev is the lowerdev
1003 * to use for the offload and 'dev' is the net device that will
1004 * back the offload. Returns a pointer to the private structure
1005 * the upper layer will maintain.
1006 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1007 * Called by upper layer device to delete the station created
1008 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1009 * the station and priv is the structure returned by the add
1010 * operation.
1011 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1012 * struct net_device *dev,
1013 * void *priv);
1014 * Callback to use for xmit over the accelerated station. This
1015 * is used in place of ndo_start_xmit on accelerated net
1016 * devices.
Jesse Gross5f352272014-12-23 22:37:26 -08001017 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1018 * struct net_device *dev
1019 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001020 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001021 * performing offload operations on a given packet. This is to give
1022 * the device an opportunity to implement any restrictions that cannot
1023 * be otherwise expressed by feature flags. The check is called with
1024 * the set of features that the stack has calculated and it returns
1025 * those the driver believes to be appropriate.
John Fastabend822b3b22015-03-18 14:57:33 +02001026 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1027 * int queue_index, u32 maxrate);
1028 * Called when a user wants to set a max-rate limitation of specific
1029 * TX queue.
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001030 */
1031struct net_device_ops {
1032 int (*ndo_init)(struct net_device *dev);
1033 void (*ndo_uninit)(struct net_device *dev);
1034 int (*ndo_open)(struct net_device *dev);
1035 int (*ndo_stop)(struct net_device *dev);
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +00001036 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
Stephen Hemminger00829822008-11-20 20:14:53 -08001037 struct net_device *dev);
1038 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001039 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001040 void *accel_priv,
1041 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001042 void (*ndo_change_rx_flags)(struct net_device *dev,
1043 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001044 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001045 int (*ndo_set_mac_address)(struct net_device *dev,
1046 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001047 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001048 int (*ndo_do_ioctl)(struct net_device *dev,
1049 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001050 int (*ndo_set_config)(struct net_device *dev,
1051 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001052 int (*ndo_change_mtu)(struct net_device *dev,
1053 int new_mtu);
1054 int (*ndo_neigh_setup)(struct net_device *dev,
1055 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001056 void (*ndo_tx_timeout) (struct net_device *dev);
1057
Eric Dumazet28172732010-07-07 14:58:56 -07001058 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1059 struct rtnl_link_stats64 *storage);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001060 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1061
Jiri Pirko8e586132011-12-08 19:52:37 -05001062 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001063 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001064 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001065 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001066#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001067 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001068 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001069 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001070 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001071#endif
Cong Wange0d10952013-08-01 11:10:25 +08001072#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001073 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001074#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001075 int (*ndo_set_vf_mac)(struct net_device *dev,
1076 int queue, u8 *mac);
1077 int (*ndo_set_vf_vlan)(struct net_device *dev,
1078 int queue, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001079 int (*ndo_set_vf_rate)(struct net_device *dev,
1080 int vf, int min_tx_rate,
1081 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001082 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1083 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001084 int (*ndo_get_vf_config)(struct net_device *dev,
1085 int vf,
1086 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001087 int (*ndo_set_vf_link_state)(struct net_device *dev,
1088 int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -07001089 int (*ndo_set_vf_port)(struct net_device *dev,
1090 int vf,
1091 struct nlattr *port[]);
1092 int (*ndo_get_vf_port)(struct net_device *dev,
1093 int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +00001094 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001095#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001096 int (*ndo_fcoe_enable)(struct net_device *dev);
1097 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001098 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1099 u16 xid,
1100 struct scatterlist *sgl,
1101 unsigned int sgc);
1102 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1103 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001104 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1105 u16 xid,
1106 struct scatterlist *sgl,
1107 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001108 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1109 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001110#endif
1111
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001112#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001113#define NETDEV_FCOE_WWNN 0
1114#define NETDEV_FCOE_WWPN 1
1115 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1116 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001117#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001118
Ben Hutchingsc4454772011-01-19 11:03:53 +00001119#ifdef CONFIG_RFS_ACCEL
1120 int (*ndo_rx_flow_steer)(struct net_device *dev,
1121 const struct sk_buff *skb,
1122 u16 rxq_index,
1123 u32 flow_id);
1124#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001125 int (*ndo_add_slave)(struct net_device *dev,
1126 struct net_device *slave_dev);
1127 int (*ndo_del_slave)(struct net_device *dev,
1128 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001129 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1130 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001131 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001132 netdev_features_t features);
David Millerda6a8fa2011-07-25 00:01:38 +00001133 int (*ndo_neigh_construct)(struct neighbour *n);
David S. Miller447f2192011-12-19 15:04:41 -05001134 void (*ndo_neigh_destroy)(struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001135
1136 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001137 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001138 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001139 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001140 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001141 u16 flags);
1142 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001143 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001144 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001145 const unsigned char *addr,
1146 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001147 int (*ndo_fdb_dump)(struct sk_buff *skb,
1148 struct netlink_callback *cb,
1149 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001150 struct net_device *filter_dev,
John Fastabend77162022012-04-15 06:43:56 +00001151 int idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001152
1153 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001154 struct nlmsghdr *nlh,
1155 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001156 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1157 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001158 struct net_device *dev,
1159 u32 filter_mask);
Vlad Yasevich407af322013-02-13 12:00:12 +00001160 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001161 struct nlmsghdr *nlh,
1162 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001163 int (*ndo_change_carrier)(struct net_device *dev,
1164 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001165 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001166 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001167 int (*ndo_get_phys_port_name)(struct net_device *dev,
1168 char *name, size_t len);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001169 void (*ndo_add_vxlan_port)(struct net_device *dev,
1170 sa_family_t sa_family,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001171 __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001172 void (*ndo_del_vxlan_port)(struct net_device *dev,
1173 sa_family_t sa_family,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001174 __be16 port);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001175
1176 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1177 struct net_device *dev);
1178 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1179 void *priv);
1180
1181 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1182 struct net_device *dev,
1183 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001184 int (*ndo_get_lock_subclass)(struct net_device *dev);
Jesse Gross5f352272014-12-23 22:37:26 -08001185 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1186 struct net_device *dev,
1187 netdev_features_t features);
John Fastabend822b3b22015-03-18 14:57:33 +02001188 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1189 int queue_index,
1190 u32 maxrate);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001191};
1192
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001193/**
1194 * enum net_device_priv_flags - &struct net_device priv_flags
1195 *
1196 * These are the &struct net_device, they are only set internally
1197 * by drivers and used in the kernel. These flags are invisible to
1198 * userspace, this means that the order of these flags can change
1199 * during any kernel release.
1200 *
1201 * You should have a pretty good reason to be extending these flags.
1202 *
1203 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1204 * @IFF_EBRIDGE: Ethernet bridging device
1205 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1206 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1207 * @IFF_MASTER_ALB: bonding master, balance-alb
1208 * @IFF_BONDING: bonding master or slave
1209 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1210 * @IFF_ISATAP: ISATAP interface (RFC4214)
1211 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1212 * @IFF_WAN_HDLC: WAN HDLC device
1213 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1214 * release skb->dst
1215 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1216 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1217 * @IFF_MACVLAN_PORT: device used as macvlan port
1218 * @IFF_BRIDGE_PORT: device used as bridge port
1219 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1220 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1221 * @IFF_UNICAST_FLT: Supports unicast filtering
1222 * @IFF_TEAM_PORT: device used as team port
1223 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1224 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1225 * change when it's running
1226 * @IFF_MACVLAN: Macvlan device
1227 */
1228enum netdev_priv_flags {
1229 IFF_802_1Q_VLAN = 1<<0,
1230 IFF_EBRIDGE = 1<<1,
1231 IFF_SLAVE_INACTIVE = 1<<2,
1232 IFF_MASTER_8023AD = 1<<3,
1233 IFF_MASTER_ALB = 1<<4,
1234 IFF_BONDING = 1<<5,
1235 IFF_SLAVE_NEEDARP = 1<<6,
1236 IFF_ISATAP = 1<<7,
1237 IFF_MASTER_ARPMON = 1<<8,
1238 IFF_WAN_HDLC = 1<<9,
1239 IFF_XMIT_DST_RELEASE = 1<<10,
1240 IFF_DONT_BRIDGE = 1<<11,
1241 IFF_DISABLE_NETPOLL = 1<<12,
1242 IFF_MACVLAN_PORT = 1<<13,
1243 IFF_BRIDGE_PORT = 1<<14,
1244 IFF_OVS_DATAPATH = 1<<15,
1245 IFF_TX_SKB_SHARING = 1<<16,
1246 IFF_UNICAST_FLT = 1<<17,
1247 IFF_TEAM_PORT = 1<<18,
1248 IFF_SUPP_NOFCS = 1<<19,
1249 IFF_LIVE_ADDR_CHANGE = 1<<20,
1250 IFF_MACVLAN = 1<<21,
Eric Dumazet02875872014-10-05 18:38:35 -07001251 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001252 IFF_IPVLAN_MASTER = 1<<23,
1253 IFF_IPVLAN_SLAVE = 1<<24,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001254};
1255
1256#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1257#define IFF_EBRIDGE IFF_EBRIDGE
1258#define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1259#define IFF_MASTER_8023AD IFF_MASTER_8023AD
1260#define IFF_MASTER_ALB IFF_MASTER_ALB
1261#define IFF_BONDING IFF_BONDING
1262#define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1263#define IFF_ISATAP IFF_ISATAP
1264#define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1265#define IFF_WAN_HDLC IFF_WAN_HDLC
1266#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1267#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1268#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1269#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1270#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1271#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1272#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1273#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1274#define IFF_TEAM_PORT IFF_TEAM_PORT
1275#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1276#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1277#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001278#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001279#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1280#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001281
Karoly Kemeny536721b2014-07-30 20:27:36 +02001282/**
1283 * struct net_device - The DEVICE structure.
1284 * Actually, this whole structure is a big mistake. It mixes I/O
1285 * data with strictly "high-level" data, and it has to know about
1286 * almost every data structure used in the INET module.
1287 *
1288 * @name: This is the first field of the "visible" part of this structure
1289 * (i.e. as seen by users in the "Space.c" file). It is the name
1290 * of the interface.
1291 *
1292 * @name_hlist: Device name hash chain, please keep it close to name[]
1293 * @ifalias: SNMP alias
1294 * @mem_end: Shared memory end
1295 * @mem_start: Shared memory start
1296 * @base_addr: Device I/O address
1297 * @irq: Device IRQ number
1298 *
1299 * @state: Generic network queuing layer state, see netdev_state_t
1300 * @dev_list: The global list of network devices
1301 * @napi_list: List entry, that is used for polling napi devices
1302 * @unreg_list: List entry, that is used, when we are unregistering the
1303 * device, see the function unregister_netdev
1304 * @close_list: List entry, that is used, when we are closing the device
1305 *
1306 * @adj_list: Directly linked devices, like slaves for bonding
1307 * @all_adj_list: All linked devices, *including* neighbours
1308 * @features: Currently active device features
1309 * @hw_features: User-changeable features
1310 *
1311 * @wanted_features: User-requested features
1312 * @vlan_features: Mask of features inheritable by VLAN devices
1313 *
1314 * @hw_enc_features: Mask of features inherited by encapsulating devices
1315 * This field indicates what encapsulation
1316 * offloads the hardware is capable of doing,
1317 * and drivers will need to set them appropriately.
1318 *
1319 * @mpls_features: Mask of features inheritable by MPLS
1320 *
1321 * @ifindex: interface index
1322 * @iflink: unique device identifier
1323 *
1324 * @stats: Statistics struct, which was left as a legacy, use
1325 * rtnl_link_stats64 instead
1326 *
1327 * @rx_dropped: Dropped packets by core network,
1328 * do not use this in drivers
1329 * @tx_dropped: Dropped packets by core network,
1330 * do not use this in drivers
1331 *
1332 * @carrier_changes: Stats to monitor carrier on<->off transitions
1333 *
1334 * @wireless_handlers: List of functions to handle Wireless Extensions,
1335 * instead of ioctl,
1336 * see <net/iw_handler.h> for details.
1337 * @wireless_data: Instance data managed by the core of wireless extensions
1338 *
1339 * @netdev_ops: Includes several pointers to callbacks,
1340 * if one wants to override the ndo_*() functions
1341 * @ethtool_ops: Management operations
1342 * @fwd_ops: Management operations
Eric W. Biedermand4760592015-03-02 00:11:09 -06001343 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001344 * of Layer 2 headers.
1345 *
1346 * @flags: Interface flags (a la BSD)
1347 * @priv_flags: Like 'flags' but invisible to userspace,
1348 * see if.h for the definitions
1349 * @gflags: Global flags ( kept as legacy )
1350 * @padded: How much padding added by alloc_netdev()
1351 * @operstate: RFC2863 operstate
1352 * @link_mode: Mapping policy to operstate
1353 * @if_port: Selectable AUI, TP, ...
1354 * @dma: DMA channel
1355 * @mtu: Interface MTU value
1356 * @type: Interface hardware type
1357 * @hard_header_len: Hardware header length
1358 *
1359 * @needed_headroom: Extra headroom the hardware may need, but not in all
1360 * cases can this be guaranteed
1361 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1362 * cases can this be guaranteed. Some cases also use
1363 * LL_MAX_HEADER instead to allocate the skb
1364 *
1365 * interface address info:
1366 *
1367 * @perm_addr: Permanent hw address
1368 * @addr_assign_type: Hw address assignment type
1369 * @addr_len: Hardware address length
1370 * @neigh_priv_len; Used in neigh_alloc(),
1371 * initialized only in atm/clip.c
1372 * @dev_id: Used to differentiate devices that share
1373 * the same link layer address
1374 * @dev_port: Used to differentiate devices that share
1375 * the same function
1376 * @addr_list_lock: XXX: need comments on this one
1377 * @uc: unicast mac addresses
1378 * @mc: multicast mac addresses
1379 * @dev_addrs: list of device hw addresses
1380 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1381 * @uc_promisc: Counter, that indicates, that promiscuous mode
1382 * has been enabled due to the need to listen to
1383 * additional unicast addresses in a device that
1384 * does not implement ndo_set_rx_mode()
1385 * @promiscuity: Number of times, the NIC is told to work in
1386 * Promiscuous mode, if it becomes 0 the NIC will
1387 * exit from working in Promiscuous mode
1388 * @allmulti: Counter, enables or disables allmulticast mode
1389 *
1390 * @vlan_info: VLAN info
1391 * @dsa_ptr: dsa specific data
1392 * @tipc_ptr: TIPC specific data
1393 * @atalk_ptr: AppleTalk link
1394 * @ip_ptr: IPv4 specific data
1395 * @dn_ptr: DECnet specific data
1396 * @ip6_ptr: IPv6 specific data
1397 * @ax25_ptr: AX.25 specific data
1398 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1399 *
1400 * @last_rx: Time of last Rx
1401 * @dev_addr: Hw address (before bcast,
1402 * because most packets are unicast)
1403 *
1404 * @_rx: Array of RX queues
1405 * @num_rx_queues: Number of RX queues
1406 * allocated at register_netdev() time
1407 * @real_num_rx_queues: Number of RX queues currently active in device
1408 *
1409 * @rx_handler: handler for received packets
1410 * @rx_handler_data: XXX: need comments on this one
1411 * @ingress_queue: XXX: need comments on this one
1412 * @broadcast: hw bcast address
1413 *
1414 * @_tx: Array of TX queues
1415 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1416 * @real_num_tx_queues: Number of TX queues currently active in device
1417 * @qdisc: Root qdisc from userspace point of view
1418 * @tx_queue_len: Max frames per queue allowed
1419 * @tx_global_lock: XXX: need comments on this one
1420 *
1421 * @xps_maps: XXX: need comments on this one
1422 *
1423 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1424 * indexed by RX queue number. Assigned by driver.
1425 * This must only be set if the ndo_rx_flow_steer
1426 * operation is defined
1427 *
1428 * @trans_start: Time (in jiffies) of last Tx
1429 * @watchdog_timeo: Represents the timeout that is used by
1430 * the watchdog ( see dev_watchdog() )
1431 * @watchdog_timer: List of timers
1432 *
1433 * @pcpu_refcnt: Number of references to this device
1434 * @todo_list: Delayed register/unregister
1435 * @index_hlist: Device index hash chain
1436 * @link_watch_list: XXX: need comments on this one
1437 *
1438 * @reg_state: Register/unregister state machine
1439 * @dismantle: Device is going to be freed
1440 * @rtnl_link_state: This enum represents the phases of creating
1441 * a new link
1442 *
1443 * @destructor: Called from unregister,
1444 * can be used to call free_netdev
1445 * @npinfo: XXX: need comments on this one
1446 * @nd_net: Network namespace this network device is inside
1447 *
1448 * @ml_priv: Mid-layer private
1449 * @lstats: Loopback statistics
1450 * @tstats: Tunnel statistics
1451 * @dstats: Dummy statistics
1452 * @vstats: Virtual ethernet statistics
1453 *
1454 * @garp_port: GARP
1455 * @mrp_port: MRP
1456 *
1457 * @dev: Class/net/name entry
1458 * @sysfs_groups: Space for optional device, statistics and wireless
1459 * sysfs groups
1460 *
1461 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1462 * @rtnl_link_ops: Rtnl_link_ops
1463 *
1464 * @gso_max_size: Maximum size of generic segmentation offload
1465 * @gso_max_segs: Maximum number of segments that can be passed to the
1466 * NIC for GSO
Eric Dumazetfcbeb972014-10-05 10:11:27 -07001467 * @gso_min_segs: Minimum number of segments that can be passed to the
1468 * NIC for GSO
Karoly Kemeny536721b2014-07-30 20:27:36 +02001469 *
1470 * @dcbnl_ops: Data Center Bridging netlink ops
1471 * @num_tc: Number of traffic classes in the net device
1472 * @tc_to_txq: XXX: need comments on this one
1473 * @prio_tc_map XXX: need comments on this one
1474 *
1475 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1476 *
1477 * @priomap: XXX: need comments on this one
1478 * @phydev: Physical device may attach itself
1479 * for hardware timestamping
1480 *
1481 * @qdisc_tx_busylock: XXX: need comments on this one
1482 *
1483 * @group: The group, that the device belongs to
1484 * @pm_qos_req: Power Management QoS object
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 *
1486 * FIXME: cleanup struct net_device such that network protocol info
1487 * moves out.
1488 */
1489
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001490struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001492 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001493 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 /*
1495 * I/O specific fields
1496 * FIXME: Merge these and struct ifmap into one
1497 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001498 unsigned long mem_end;
1499 unsigned long mem_start;
1500 unsigned long base_addr;
1501 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502
1503 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001504 * Some hardware also needs these fields (state,dev_list,
1505 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 * part of the usual set specified in Space.c.
1507 */
1508
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 unsigned long state;
1510
Pavel Emelianov7562f872007-05-03 15:13:45 -07001511 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001512 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001513 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001514 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001515 struct list_head ptype_all;
1516 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001517
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001518 struct {
1519 struct list_head upper;
1520 struct list_head lower;
1521 } adj_list;
1522
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001523 struct {
1524 struct list_head upper;
1525 struct list_head lower;
1526 } all_adj_list;
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001527
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001528 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001529 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001530 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001531 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001532 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001533 netdev_features_t mpls_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001534
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 int ifindex;
1536 int iflink;
1537
Rusty Russellc45d2862007-03-28 14:29:08 -07001538 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001539
Eric Dumazet015f0682014-03-27 08:45:56 -07001540 atomic_long_t rx_dropped;
1541 atomic_long_t tx_dropped;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542
david decotigny2d3b4792014-03-29 09:48:35 -07001543 atomic_t carrier_changes;
1544
Johannes Bergb86e0282007-04-26 20:48:23 -07001545#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 const struct iw_handler_def * wireless_handlers;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 struct iw_public_data * wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001548#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001549 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001550 const struct ethtool_ops *ethtool_ops;
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001551 const struct forwarding_accel_ops *fwd_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001552#ifdef CONFIG_NET_SWITCHDEV
1553 const struct swdev_ops *swdev_ops;
1554#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001556 const struct header_ops *header_ops;
1557
Karoly Kemeny536721b2014-07-30 20:27:36 +02001558 unsigned int flags;
1559 unsigned int priv_flags;
1560
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001562 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563
Karoly Kemeny536721b2014-07-30 20:27:36 +02001564 unsigned char operstate;
1565 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001566
Karoly Kemeny536721b2014-07-30 20:27:36 +02001567 unsigned char if_port;
1568 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001569
Karoly Kemeny536721b2014-07-30 20:27:36 +02001570 unsigned int mtu;
1571 unsigned short type;
1572 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573
Johannes Bergf5184d22008-05-12 20:48:31 -07001574 unsigned short needed_headroom;
1575 unsigned short needed_tailroom;
1576
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001578 unsigned char perm_addr[MAX_ADDR_LEN];
1579 unsigned char addr_assign_type;
1580 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001581 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001582 unsigned short dev_id;
1583 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001584 spinlock_t addr_list_lock;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001585 struct netdev_hw_addr_list uc;
1586 struct netdev_hw_addr_list mc;
1587 struct netdev_hw_addr_list dev_addrs;
1588
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001589#ifdef CONFIG_SYSFS
1590 struct kset *queues_kset;
1591#endif
1592
Tom Gundersen685343f2014-07-14 16:37:22 +02001593 unsigned char name_assign_type;
1594
Joe Perches2d348d12011-07-25 16:17:35 -07001595 bool uc_promisc;
Wang Chen9d45abe2008-06-17 21:12:48 -07001596 unsigned int promiscuity;
1597 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599
1600 /* Protocol specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001601
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001602#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001603 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001604#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001605#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001606 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001607#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001608#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001609 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001610#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001611 void *atalk_ptr;
1612 struct in_device __rcu *ip_ptr;
1613 struct dn_dev __rcu *dn_ptr;
1614 struct inet6_dev __rcu *ip6_ptr;
1615 void *ax25_ptr;
1616 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001617 struct wpan_dev *ieee802154_ptr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001619/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001620 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001621 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001622 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001623
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001624 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001625 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001626
Tom Herbert0a9627f2010-03-16 08:03:29 +00001627
Michael Daltona953be52014-01-16 22:23:28 -08001628#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001629 struct netdev_rx_queue *_rx;
1630
Tom Herbert0a9627f2010-03-16 08:03:29 +00001631 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001632 unsigned int real_num_rx_queues;
Ben Hutchingsc4454772011-01-19 11:03:53 +00001633
Eric Dumazetdf334542010-03-24 19:13:54 +00001634#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001635
Eric Dumazet3b47d302014-11-06 21:09:44 -08001636 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001637 rx_handler_func_t __rcu *rx_handler;
1638 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001639
Eric Dumazet24824a02010-10-02 06:11:55 +00001640 struct netdev_queue __rcu *ingress_queue;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001641 unsigned char broadcast[MAX_ADDR_LEN];
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001642
Eric Dumazetcd135392010-09-16 02:58:13 +00001643
1644/*
1645 * Cache lines mostly used on transmit path
1646 */
David S. Millere8a04642008-07-17 00:34:19 -07001647 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1648 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001649 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001650 struct Qdisc *qdisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001651 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001652 spinlock_t tx_global_lock;
Eric Dumazetcd135392010-09-16 02:58:13 +00001653
Tom Herbertbf264142010-11-26 08:36:09 +00001654#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001655 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001656#endif
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001657#ifdef CONFIG_RFS_ACCEL
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001658 struct cpu_rmap *rx_cpu_rmap;
1659#endif
Tom Herbert1d24eb42010-11-21 13:17:27 +00001660
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001661 /* These may be needed for future network-power-down code. */
Eric Dumazet9d214932009-05-17 20:55:16 -07001662
1663 /*
1664 * trans_start here is expensive for high speed devices on SMP,
1665 * please use netdev_queue->trans_start instead.
1666 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001667 unsigned long trans_start;
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001668
Karoly Kemeny536721b2014-07-30 20:27:36 +02001669 int watchdog_timeo;
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001670 struct timer_list watchdog_timer;
1671
Eric Dumazet29b44332010-10-11 10:22:12 +00001672 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674
Karoly Kemeny536721b2014-07-30 20:27:36 +02001675 struct hlist_node index_hlist;
Eric Dumazete014deb2009-11-17 05:59:21 +00001676 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001677
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001679 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 NETREG_UNREGISTERING, /* called unregister_netdevice */
1681 NETREG_UNREGISTERED, /* completed unregister todo */
1682 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001683 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001684 } reg_state:8;
1685
Karoly Kemeny536721b2014-07-30 20:27:36 +02001686 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001687
1688 enum {
1689 RTNL_LINK_INITIALIZED,
1690 RTNL_LINK_INITIALIZING,
1691 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001693 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001696 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697#endif
David S. Millereae792b2008-07-15 03:03:33 -07001698
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001699 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001700
David S. Miller49517042008-05-12 03:29:11 -07001701 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001702 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001703 void *ml_priv;
1704 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001705 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001706 struct pcpu_dstats __percpu *dstats;
1707 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001708 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001709
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001710 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001711 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712
Karoly Kemeny536721b2014-07-30 20:27:36 +02001713 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001714 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001715 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001716
Patrick McHardy38f7b872007-06-13 12:03:51 -07001717 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001718
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001719 /* for setting kernel sock attribute on TCP connection setup */
1720#define GSO_MAX_SIZE 65536
1721 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001722#define GSO_MAX_SEGS 65535
1723 u16 gso_max_segs;
Eric Dumazetfcbeb972014-10-05 10:11:27 -07001724 u16 gso_min_segs;
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001725#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001726 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001727#endif
John Fastabend4f57c082011-01-17 08:06:04 +00001728 u8 num_tc;
1729 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1730 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001731
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001732#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001733 unsigned int fcoe_ddp_xid;
1734#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001735#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001736 struct netprio_map __rcu *priomap;
1737#endif
Richard Cochranc1f19b52010-07-17 08:49:36 +00001738 struct phy_device *phydev;
Eric Dumazet23d3b8b2012-09-05 01:02:56 +00001739 struct lock_class_key *qdisc_tx_busylock;
Vlad Dogarucbda10f2011-01-13 23:38:30 +00001740 int group;
Eric Dumazet91364612012-06-11 06:36:13 +00001741 struct pm_qos_request pm_qos_req;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001743#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744
1745#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746
David S. Millere8a04642008-07-17 00:34:19 -07001747static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001748int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1749{
1750 return dev->prio_tc_map[prio & TC_BITMASK];
1751}
1752
1753static inline
1754int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1755{
1756 if (tc >= dev->num_tc)
1757 return -EINVAL;
1758
1759 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1760 return 0;
1761}
1762
1763static inline
1764void netdev_reset_tc(struct net_device *dev)
1765{
1766 dev->num_tc = 0;
1767 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1768 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1769}
1770
1771static inline
1772int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1773{
1774 if (tc >= dev->num_tc)
1775 return -EINVAL;
1776
1777 dev->tc_to_txq[tc].count = count;
1778 dev->tc_to_txq[tc].offset = offset;
1779 return 0;
1780}
1781
1782static inline
1783int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1784{
1785 if (num_tc > TC_MAX_QUEUE)
1786 return -EINVAL;
1787
1788 dev->num_tc = num_tc;
1789 return 0;
1790}
1791
1792static inline
1793int netdev_get_num_tc(struct net_device *dev)
1794{
1795 return dev->num_tc;
1796}
1797
1798static inline
David S. Millere8a04642008-07-17 00:34:19 -07001799struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1800 unsigned int index)
1801{
1802 return &dev->_tx[index];
1803}
1804
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001805static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1806 const struct sk_buff *skb)
1807{
1808 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1809}
1810
David S. Millere8a04642008-07-17 00:34:19 -07001811static inline void netdev_for_each_tx_queue(struct net_device *dev,
1812 void (*f)(struct net_device *,
1813 struct netdev_queue *,
1814 void *),
1815 void *arg)
1816{
1817 unsigned int i;
1818
1819 for (i = 0; i < dev->num_tx_queues; i++)
1820 f(dev, &dev->_tx[i], arg);
1821}
1822
Joe Perchesf629d202013-09-26 14:48:15 -07001823struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001824 struct sk_buff *skb,
1825 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001826
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001827/*
1828 * Net namespace inlines
1829 */
1830static inline
1831struct net *dev_net(const struct net_device *dev)
1832{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001833 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001834}
1835
1836static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001837void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001838{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001839 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001840}
1841
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07001842static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001843{
Alexander Duyck3fc88672014-09-12 23:12:46 -07001844#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001845 if (dev->dsa_ptr != NULL)
1846 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001847#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001848 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00001849}
1850
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001851/**
1852 * netdev_priv - access network device private data
1853 * @dev: network device
1854 *
1855 * Get network device private data
1856 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07001857static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00001859 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860}
1861
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862/* Set the sysfs physical device reference for the network logical device
1863 * if set prior to registration will cause a symlink during initialization.
1864 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001865#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866
Marcel Holtmann384912e2009-08-31 21:08:19 +00001867/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02001868 * fine-grained identification of different network device types. For
Marcel Holtmann384912e2009-08-31 21:08:19 +00001869 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1870 */
1871#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1872
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00001873/* Default NAPI poll() weight
1874 * Device drivers are strongly advised to not use bigger value
1875 */
1876#define NAPI_POLL_WEIGHT 64
1877
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07001878/**
1879 * netif_napi_add - initialize a napi context
1880 * @dev: network device
1881 * @napi: napi context
1882 * @poll: polling function
1883 * @weight: default weight
1884 *
1885 * netif_napi_add() must be used to initialize a napi context prior to calling
1886 * *any* of the other napi related functions.
1887 */
Herbert Xud565b0a2008-12-15 23:38:52 -08001888void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1889 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001890
Alexander Duyckd8156532008-07-08 15:13:05 -07001891/**
1892 * netif_napi_del - remove a napi context
1893 * @napi: napi context
1894 *
1895 * netif_napi_del() removes a napi context from the network device napi list
1896 */
Herbert Xud565b0a2008-12-15 23:38:52 -08001897void netif_napi_del(struct napi_struct *napi);
1898
1899struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00001900 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1901 void *frag0;
1902
Herbert Xu74895942009-05-26 18:50:27 +00001903 /* Length of frag0. */
1904 unsigned int frag0_len;
1905
Herbert Xu86911732009-01-29 14:19:50 +00001906 /* This indicates where we are processing relative to skb->data. */
1907 int data_offset;
1908
Herbert Xud565b0a2008-12-15 23:38:52 -08001909 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08001910 u16 flush;
1911
1912 /* Save the IP ID here and check when we get to the transport layer */
1913 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08001914
1915 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00001916 u16 count;
1917
Tom Herbert15e23962015-02-10 16:30:31 -08001918 /* Start offset for remote checksum offload */
1919 u16 gro_remcsum_start;
1920
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00001921 /* jiffies when first packet was created/queued */
1922 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02001923
Tom Herbertafe93322014-09-17 12:25:57 -07001924 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02001925 u16 proto;
1926
Tom Herbertbaa32ff2015-02-10 16:30:30 -08001927 /* This is non-zero if the packet may be of the same flow. */
1928 u8 same_flow:1;
1929
Or Gerlitzb582ef02014-01-20 13:59:19 +02001930 /* Used in udp_gro_receive */
Tom Herbert573e8fc2014-08-22 13:33:47 -07001931 u8 udp_mark:1;
1932
1933 /* GRO checksum is valid */
1934 u8 csum_valid:1;
1935
Tom Herbert662880f2014-08-27 21:26:56 -07001936 /* Number of checksums via CHECKSUM_UNNECESSARY */
1937 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00001938
Tom Herbertbaa32ff2015-02-10 16:30:30 -08001939 /* Free the skb? */
1940 u8 free:2;
1941#define NAPI_GRO_FREE 1
1942#define NAPI_GRO_FREE_STOLEN_HEAD 2
1943
Tom Herbertefc98d02014-10-03 15:48:08 -07001944 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1945 u8 is_ipv6:1;
1946
Tom Herbertbaa32ff2015-02-10 16:30:30 -08001947 /* 7 bit hole */
1948
Jerry Chubf5a7552014-01-07 10:23:19 -08001949 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1950 __wsum csum;
1951
Eric Dumazetc3c7c252012-12-06 13:54:59 +00001952 /* used in skb_gro_receive() slow path */
1953 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08001954};
1955
1956#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07001957
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07001959 __be16 type; /* This is really htons(ether_type). */
1960 struct net_device *dev; /* NULL is wildcarded here */
1961 int (*func) (struct sk_buff *,
1962 struct net_device *,
1963 struct packet_type *,
1964 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00001965 bool (*id_match)(struct packet_type *ptype,
1966 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967 void *af_packet_priv;
1968 struct list_head list;
1969};
1970
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001971struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1973 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08001975 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08001976 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001977};
1978
1979struct packet_offload {
1980 __be16 type; /* This is really htons(ether_type). */
1981 struct offload_callbacks callbacks;
1982 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983};
1984
Tom Herberta2b12f32015-01-12 17:00:37 -08001985struct udp_offload;
1986
1987struct udp_offload_callbacks {
1988 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1989 struct sk_buff *skb,
1990 struct udp_offload *uoff);
1991 int (*gro_complete)(struct sk_buff *skb,
1992 int nhoff,
1993 struct udp_offload *uoff);
1994};
1995
Or Gerlitzb582ef02014-01-20 13:59:19 +02001996struct udp_offload {
1997 __be16 port;
Tom Herbertafe93322014-09-17 12:25:57 -07001998 u8 ipproto;
Tom Herberta2b12f32015-01-12 17:00:37 -08001999 struct udp_offload_callbacks callbacks;
Or Gerlitzb582ef02014-01-20 13:59:19 +02002000};
2001
Li RongQing8f849852014-01-04 13:57:59 +08002002/* often modified stats are per cpu, other are shared (netdev->stats) */
2003struct pcpu_sw_netstats {
2004 u64 rx_packets;
2005 u64 rx_bytes;
2006 u64 tx_packets;
2007 u64 tx_bytes;
2008 struct u64_stats_sync syncp;
2009};
2010
WANG Cong1c213bd2014-02-13 11:46:28 -08002011#define netdev_alloc_pcpu_stats(type) \
2012({ \
stephen hemminger693350c2014-03-10 09:41:46 -07002013 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
WANG Cong1c213bd2014-02-13 11:46:28 -08002014 if (pcpu_stats) { \
2015 int i; \
2016 for_each_possible_cpu(i) { \
2017 typeof(type) *stat; \
2018 stat = per_cpu_ptr(pcpu_stats, i); \
2019 u64_stats_init(&stat->syncp); \
2020 } \
2021 } \
2022 pcpu_stats; \
2023})
2024
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025#include <linux/notifier.h>
2026
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002027/* netdevice notifier chain. Please remember to update the rtnetlink
2028 * notification exclusion list in rtnetlink_event() when adding new
2029 * types.
2030 */
2031#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2032#define NETDEV_DOWN 0x0002
2033#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2034 detected a hardware crash and restarted
2035 - we can use this eg to kick tcp sessions
2036 once done */
2037#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2038#define NETDEV_REGISTER 0x0005
2039#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002040#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002041#define NETDEV_CHANGEADDR 0x0008
2042#define NETDEV_GOING_DOWN 0x0009
2043#define NETDEV_CHANGENAME 0x000A
2044#define NETDEV_FEAT_CHANGE 0x000B
2045#define NETDEV_BONDING_FAILOVER 0x000C
2046#define NETDEV_PRE_UP 0x000D
2047#define NETDEV_PRE_TYPE_CHANGE 0x000E
2048#define NETDEV_POST_TYPE_CHANGE 0x000F
2049#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002050#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002051#define NETDEV_RELEASE 0x0012
2052#define NETDEV_NOTIFY_PEERS 0x0013
2053#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002054#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002055#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002056#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002057#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002058#define NETDEV_BONDING_INFO 0x0019
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002059
Joe Perchesf629d202013-09-26 14:48:15 -07002060int register_netdevice_notifier(struct notifier_block *nb);
2061int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002062
2063struct netdev_notifier_info {
2064 struct net_device *dev;
2065};
2066
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002067struct netdev_notifier_change_info {
2068 struct netdev_notifier_info info; /* must be first */
2069 unsigned int flags_changed;
2070};
2071
Cong Wang75538c22013-05-29 11:30:50 +08002072static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2073 struct net_device *dev)
2074{
2075 info->dev = dev;
2076}
2077
Jiri Pirko351638e2013-05-28 01:30:21 +00002078static inline struct net_device *
2079netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2080{
2081 return info->dev;
2082}
2083
Joe Perchesf629d202013-09-26 14:48:15 -07002084int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002085
2086
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087extern rwlock_t dev_base_lock; /* Device list lock */
2088
Eric W. Biederman881d9662007-09-17 11:56:21 -07002089#define for_each_netdev(net, d) \
2090 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002091#define for_each_netdev_reverse(net, d) \
2092 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002093#define for_each_netdev_rcu(net, d) \
2094 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002095#define for_each_netdev_safe(net, d, n) \
2096 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2097#define for_each_netdev_continue(net, d) \
2098 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002099#define for_each_netdev_continue_rcu(net, d) \
2100 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002101#define for_each_netdev_in_bond_rcu(bond, slave) \
2102 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002103 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002104#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2105
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002106static inline struct net_device *next_net_device(struct net_device *dev)
2107{
2108 struct list_head *lh;
2109 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002110
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002111 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002112 lh = dev->dev_list.next;
2113 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2114}
2115
Eric Dumazetce81b762009-11-11 17:34:30 +00002116static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2117{
2118 struct list_head *lh;
2119 struct net *net;
2120
2121 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002122 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002123 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2124}
2125
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002126static inline struct net_device *first_net_device(struct net *net)
2127{
2128 return list_empty(&net->dev_base_head) ? NULL :
2129 net_device_entry(net->dev_base_head.next);
2130}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002131
Eric Dumazetccf43432011-01-26 18:08:02 +00002132static inline struct net_device *first_net_device_rcu(struct net *net)
2133{
2134 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2135
2136 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2137}
2138
Joe Perchesf629d202013-09-26 14:48:15 -07002139int netdev_boot_setup_check(struct net_device *dev);
2140unsigned long netdev_boot_base(const char *prefix, int unit);
2141struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2142 const char *hwaddr);
2143struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2144struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2145void dev_add_pack(struct packet_type *pt);
2146void dev_remove_pack(struct packet_type *pt);
2147void __dev_remove_pack(struct packet_type *pt);
2148void dev_add_offload(struct packet_offload *po);
2149void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150
WANG Cong6c555492014-09-11 15:35:09 -07002151struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2152 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002153struct net_device *dev_get_by_name(struct net *net, const char *name);
2154struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2155struct net_device *__dev_get_by_name(struct net *net, const char *name);
2156int dev_alloc_name(struct net_device *dev, const char *name);
2157int dev_open(struct net_device *dev);
2158int dev_close(struct net_device *dev);
2159void dev_disable_lro(struct net_device *dev);
2160int dev_loopback_xmit(struct sk_buff *newskb);
2161int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002162int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002163int register_netdevice(struct net_device *dev);
2164void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2165void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002166static inline void unregister_netdevice(struct net_device *dev)
2167{
2168 unregister_netdevice_queue(dev, NULL);
2169}
2170
Joe Perchesf629d202013-09-26 14:48:15 -07002171int netdev_refcnt_read(const struct net_device *dev);
2172void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002173void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002174void synchronize_net(void);
2175int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002176
Joe Perchesf629d202013-09-26 14:48:15 -07002177struct net_device *dev_get_by_index(struct net *net, int ifindex);
2178struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2179struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2180int netdev_get_name(struct net *net, char *name, int ifindex);
2181int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002182int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002183
2184static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2185{
2186 return NAPI_GRO_CB(skb)->data_offset;
2187}
2188
2189static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2190{
2191 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2192}
2193
2194static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2195{
2196 NAPI_GRO_CB(skb)->data_offset += len;
2197}
2198
Herbert Xua5b1cf22009-05-26 18:50:28 +00002199static inline void *skb_gro_header_fast(struct sk_buff *skb,
2200 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002201{
Herbert Xu78a478d2009-05-26 18:50:21 +00002202 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002203}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204
Herbert Xua5b1cf22009-05-26 18:50:28 +00002205static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2206{
2207 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2208}
2209
2210static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2211 unsigned int offset)
2212{
Herbert Xu17dd7592011-07-27 06:16:28 -07002213 if (!pskb_may_pull(skb, hlen))
2214 return NULL;
2215
Herbert Xua5b1cf22009-05-26 18:50:28 +00002216 NAPI_GRO_CB(skb)->frag0 = NULL;
2217 NAPI_GRO_CB(skb)->frag0_len = 0;
Herbert Xu17dd7592011-07-27 06:16:28 -07002218 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002219}
2220
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002221static inline void *skb_gro_network_header(struct sk_buff *skb)
2222{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002223 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2224 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002225}
2226
Jerry Chubf5a7552014-01-07 10:23:19 -08002227static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2228 const void *start, unsigned int len)
2229{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002230 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002231 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2232 csum_partial(start, len, 0));
2233}
2234
Tom Herbert573e8fc2014-08-22 13:33:47 -07002235/* GRO checksum functions. These are logical equivalents of the normal
2236 * checksum functions (in skbuff.h) except that they operate on the GRO
2237 * offsets and fields in sk_buff.
2238 */
2239
2240__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2241
Tom Herbert15e23962015-02-10 16:30:31 -08002242static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2243{
2244 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2245 skb_gro_offset(skb));
2246}
2247
Tom Herbert573e8fc2014-08-22 13:33:47 -07002248static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2249 bool zero_okay,
2250 __sum16 check)
2251{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002252 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2253 skb_checksum_start_offset(skb) <
2254 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002255 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002256 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002257 (!zero_okay || check));
2258}
2259
2260static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2261 __wsum psum)
2262{
2263 if (NAPI_GRO_CB(skb)->csum_valid &&
2264 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2265 return 0;
2266
2267 NAPI_GRO_CB(skb)->csum = psum;
2268
2269 return __skb_gro_checksum_complete(skb);
2270}
2271
Tom Herbert573e8fc2014-08-22 13:33:47 -07002272static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2273{
Tom Herbert662880f2014-08-27 21:26:56 -07002274 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2275 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2276 NAPI_GRO_CB(skb)->csum_cnt--;
2277 } else {
2278 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2279 * verified a new top level checksum or an encapsulated one
2280 * during GRO. This saves work if we fallback to normal path.
2281 */
2282 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002283 }
2284}
2285
2286#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2287 compute_pseudo) \
2288({ \
2289 __sum16 __ret = 0; \
2290 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2291 __ret = __skb_gro_checksum_validate_complete(skb, \
2292 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002293 if (__ret) \
2294 __skb_mark_checksum_bad(skb); \
2295 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002296 skb_gro_incr_csum_unnecessary(skb); \
2297 __ret; \
2298})
2299
2300#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2301 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2302
2303#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2304 compute_pseudo) \
2305 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2306
2307#define skb_gro_checksum_simple_validate(skb) \
2308 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2309
Tom Herbertd96535a2014-08-31 15:12:42 -07002310static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2311{
2312 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2313 !NAPI_GRO_CB(skb)->csum_valid);
2314}
2315
2316static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2317 __sum16 check, __wsum pseudo)
2318{
2319 NAPI_GRO_CB(skb)->csum = ~pseudo;
2320 NAPI_GRO_CB(skb)->csum_valid = 1;
2321}
2322
2323#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2324do { \
2325 if (__skb_gro_checksum_convert_check(skb)) \
2326 __skb_gro_checksum_convert(skb, check, \
2327 compute_pseudo(skb, proto)); \
2328} while (0)
2329
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002330struct gro_remcsum {
2331 int offset;
2332 __wsum delta;
2333};
2334
2335static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2336{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002337 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002338 grc->delta = 0;
2339}
2340
Tom Herbertdcdc8992015-02-02 16:07:34 -08002341static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002342 int start, int offset,
Tom Herbert15e23962015-02-10 16:30:31 -08002343 struct gro_remcsum *grc,
2344 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002345{
2346 __wsum delta;
2347
2348 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2349
Tom Herbert15e23962015-02-10 16:30:31 -08002350 if (!nopartial) {
2351 NAPI_GRO_CB(skb)->gro_remcsum_start =
2352 ((unsigned char *)ptr + start) - skb->head;
2353 return;
2354 }
2355
Tom Herbertdcdc8992015-02-02 16:07:34 -08002356 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2357
2358 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002359 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002360
2361 grc->offset = (ptr + offset) - (void *)skb->head;
2362 grc->delta = delta;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002363}
2364
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002365static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2366 struct gro_remcsum *grc)
2367{
2368 if (!grc->delta)
2369 return;
2370
2371 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2372}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002373
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002374static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2375 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002376 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002377 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002378{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002379 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002380 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002381
2382 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002383}
2384
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002385static inline int dev_parse_header(const struct sk_buff *skb,
2386 unsigned char *haddr)
2387{
2388 const struct net_device *dev = skb->dev;
2389
Patrick McHardy1b833362007-10-18 05:09:28 -07002390 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002391 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002392 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002393}
2394
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002396int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397static inline int unregister_gifconf(unsigned int family)
2398{
2399 return register_gifconf(family, NULL);
2400}
2401
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002402#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002403#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002404struct sd_flow_limit {
2405 u64 count;
2406 unsigned int num_buckets;
2407 unsigned int history_head;
2408 u16 history[FLOW_LIMIT_HISTORY];
2409 u8 buckets[];
2410};
2411
2412extern int netdev_flow_limit_table_len;
2413#endif /* CONFIG_NET_FLOW_LIMIT */
2414
Linus Torvalds1da177e2005-04-16 15:20:36 -07002415/*
Eric Dumazet88751272010-04-19 05:07:33 +00002416 * Incoming packets are placed on per-cpu queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002417 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002418struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002419 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002420 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421
Changli Gaodee42872010-05-02 05:42:16 +00002422 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002423 unsigned int processed;
2424 unsigned int time_squeeze;
2425 unsigned int cpu_collision;
2426 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002427#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002428 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002429#endif
2430#ifdef CONFIG_NET_FLOW_LIMIT
2431 struct sd_flow_limit __rcu *flow_limit;
2432#endif
2433 struct Qdisc *output_queue;
2434 struct Qdisc **output_queue_tailp;
2435 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002436
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002437#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002438 /* Elements below can be accessed between CPUs for RPS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002439 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002440 struct softnet_data *rps_ipi_next;
2441 unsigned int cpu;
Tom Herbertfec5e652010-04-16 16:01:27 -07002442 unsigned int input_queue_head;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002443 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002444#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002445 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002446 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002447 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002448
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449};
2450
Tom Herbert76cc8b12010-05-20 18:37:59 +00002451static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002452{
2453#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002454 sd->input_queue_head++;
2455#endif
2456}
2457
2458static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2459 unsigned int *qtail)
2460{
2461#ifdef CONFIG_RPS
2462 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002463#endif
2464}
2465
Tom Herbert0a9627f2010-03-16 08:03:29 +00002466DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467
Joe Perchesf629d202013-09-26 14:48:15 -07002468void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002469void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002471static inline void netif_tx_schedule_all(struct net_device *dev)
2472{
2473 unsigned int i;
2474
2475 for (i = 0; i < dev->num_tx_queues; i++)
2476 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2477}
2478
Dave Jonesd29f7492008-07-22 14:09:06 -07002479static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2480{
Tom Herbert734664982011-11-28 16:32:44 +00002481 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002482}
2483
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002484/**
2485 * netif_start_queue - allow transmit
2486 * @dev: network device
2487 *
2488 * Allow upper layers to call the device hard_start_xmit routine.
2489 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002490static inline void netif_start_queue(struct net_device *dev)
2491{
David S. Millere8a04642008-07-17 00:34:19 -07002492 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002493}
2494
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002495static inline void netif_tx_start_all_queues(struct net_device *dev)
2496{
2497 unsigned int i;
2498
2499 for (i = 0; i < dev->num_tx_queues; i++) {
2500 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2501 netif_tx_start_queue(txq);
2502 }
2503}
2504
John Fastabend46e5da40a2014-09-12 20:04:52 -07002505void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002506
Dave Jonesd29f7492008-07-22 14:09:06 -07002507/**
2508 * netif_wake_queue - restart transmit
2509 * @dev: network device
2510 *
2511 * Allow upper layers to call the device hard_start_xmit routine.
2512 * Used for flow control when transmit resources are available.
2513 */
David S. Miller79d16382008-07-08 23:14:46 -07002514static inline void netif_wake_queue(struct net_device *dev)
2515{
David S. Millere8a04642008-07-17 00:34:19 -07002516 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002517}
2518
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002519static inline void netif_tx_wake_all_queues(struct net_device *dev)
2520{
2521 unsigned int i;
2522
2523 for (i = 0; i < dev->num_tx_queues; i++) {
2524 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2525 netif_tx_wake_queue(txq);
2526 }
2527}
2528
Dave Jonesd29f7492008-07-22 14:09:06 -07002529static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2530{
Guillaume Chazarain18543a62010-11-06 06:39:32 +00002531 if (WARN_ON(!dev_queue)) {
Joe Perches256ee432011-03-01 07:06:12 +00002532 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
Guillaume Chazarain18543a62010-11-06 06:39:32 +00002533 return;
2534 }
Tom Herbert734664982011-11-28 16:32:44 +00002535 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002536}
2537
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002538/**
2539 * netif_stop_queue - stop transmitted packets
2540 * @dev: network device
2541 *
2542 * Stop upper layers calling the device hard_start_xmit routine.
2543 * Used for flow control when transmit resources are unavailable.
2544 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002545static inline void netif_stop_queue(struct net_device *dev)
2546{
David S. Millere8a04642008-07-17 00:34:19 -07002547 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548}
2549
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002550static inline void netif_tx_stop_all_queues(struct net_device *dev)
2551{
2552 unsigned int i;
2553
2554 for (i = 0; i < dev->num_tx_queues; i++) {
2555 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2556 netif_tx_stop_queue(txq);
2557 }
2558}
2559
David S. Miller4d295152012-03-07 21:02:35 -05002560static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002561{
Tom Herbert734664982011-11-28 16:32:44 +00002562 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002563}
2564
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002565/**
2566 * netif_queue_stopped - test if transmit queue is flowblocked
2567 * @dev: network device
2568 *
2569 * Test if transmit queue on device is currently unable to send.
2570 */
David S. Miller4d295152012-03-07 21:02:35 -05002571static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572{
David S. Millere8a04642008-07-17 00:34:19 -07002573 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002574}
2575
David S. Miller4d295152012-03-07 21:02:35 -05002576static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002577{
Tom Herbert734664982011-11-28 16:32:44 +00002578 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2579}
2580
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002581static inline bool
2582netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002583{
2584 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2585}
2586
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002587static inline bool
2588netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2589{
2590 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2591}
2592
Eric Dumazet53511452014-10-08 08:19:27 -07002593/**
2594 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2595 * @dev_queue: pointer to transmit queue
2596 *
2597 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2598 * to give appropriate hint to the cpu.
2599 */
2600static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2601{
2602#ifdef CONFIG_BQL
2603 prefetchw(&dev_queue->dql.num_queued);
2604#endif
2605}
2606
2607/**
2608 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2609 * @dev_queue: pointer to transmit queue
2610 *
2611 * BQL enabled drivers might use this helper in their TX completion path,
2612 * to give appropriate hint to the cpu.
2613 */
2614static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2615{
2616#ifdef CONFIG_BQL
2617 prefetchw(&dev_queue->dql.limit);
2618#endif
2619}
2620
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002621static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2622 unsigned int bytes)
2623{
Tom Herbert114cf582011-11-28 16:33:09 +00002624#ifdef CONFIG_BQL
2625 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002626
2627 if (likely(dql_avail(&dev_queue->dql) >= 0))
2628 return;
2629
2630 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2631
2632 /*
2633 * The XOFF flag must be set before checking the dql_avail below,
2634 * because in netdev_tx_completed_queue we update the dql_completed
2635 * before checking the XOFF flag.
2636 */
2637 smp_mb();
2638
2639 /* check again in case another CPU has just made room avail */
2640 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2641 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002642#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002643}
2644
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002645/**
2646 * netdev_sent_queue - report the number of bytes queued to hardware
2647 * @dev: network device
2648 * @bytes: number of bytes queued to the hardware device queue
2649 *
2650 * Report the number of bytes queued for sending/completion to the network
2651 * device hardware queue. @bytes should be a good approximation and should
2652 * exactly match netdev_completed_queue() @bytes
2653 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002654static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2655{
2656 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2657}
2658
2659static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002660 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002661{
Tom Herbert114cf582011-11-28 16:33:09 +00002662#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002663 if (unlikely(!bytes))
2664 return;
2665
2666 dql_completed(&dev_queue->dql, bytes);
2667
2668 /*
2669 * Without the memory barrier there is a small possiblity that
2670 * netdev_tx_sent_queue will miss the update and cause the queue to
2671 * be stopped forever
2672 */
2673 smp_mb();
2674
2675 if (dql_avail(&dev_queue->dql) < 0)
2676 return;
2677
2678 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2679 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002680#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002681}
2682
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002683/**
2684 * netdev_completed_queue - report bytes and packets completed by device
2685 * @dev: network device
2686 * @pkts: actual number of packets sent over the medium
2687 * @bytes: actual number of bytes sent over the medium
2688 *
2689 * Report the number of bytes and packets transmitted by the network device
2690 * hardware queue over the physical medium, @bytes must exactly match the
2691 * @bytes amount passed to netdev_sent_queue()
2692 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002693static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00002694 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002695{
2696 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2697}
2698
2699static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2700{
Tom Herbert114cf582011-11-28 16:33:09 +00002701#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00002702 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002703 dql_reset(&q->dql);
2704#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002705}
2706
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002707/**
2708 * netdev_reset_queue - reset the packets and bytes count of a network device
2709 * @dev_queue: network device
2710 *
2711 * Reset the bytes and packet count of a network device and clear the
2712 * software flow control OFF bit for this network device
2713 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002714static inline void netdev_reset_queue(struct net_device *dev_queue)
2715{
2716 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07002717}
2718
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002719/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01002720 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2721 * @dev: network device
2722 * @queue_index: given tx queue index
2723 *
2724 * Returns 0 if given tx queue index >= number of device tx queues,
2725 * otherwise returns the originally passed tx queue index.
2726 */
2727static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2728{
2729 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2730 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2731 dev->name, queue_index,
2732 dev->real_num_tx_queues);
2733 return 0;
2734 }
2735
2736 return queue_index;
2737}
2738
2739/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002740 * netif_running - test if up
2741 * @dev: network device
2742 *
2743 * Test if the device has been brought up.
2744 */
David S. Miller4d295152012-03-07 21:02:35 -05002745static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002746{
2747 return test_bit(__LINK_STATE_START, &dev->state);
2748}
2749
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002750/*
2751 * Routines to manage the subqueues on a device. We only need start
2752 * stop, and a check if it's stopped. All other device management is
2753 * done at the overall netdevice level.
2754 * Also test the device if we're multiqueue.
2755 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002756
2757/**
2758 * netif_start_subqueue - allow sending packets on subqueue
2759 * @dev: network device
2760 * @queue_index: sub queue index
2761 *
2762 * Start individual transmit queue of a device with multiple transmit queues.
2763 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002764static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2765{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002766 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00002767
2768 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002769}
2770
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002771/**
2772 * netif_stop_subqueue - stop sending packets on subqueue
2773 * @dev: network device
2774 * @queue_index: sub queue index
2775 *
2776 * Stop individual transmit queue of a device with multiple transmit queues.
2777 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002778static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2779{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002780 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00002781 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002782}
2783
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002784/**
2785 * netif_subqueue_stopped - test status of subqueue
2786 * @dev: network device
2787 * @queue_index: sub queue index
2788 *
2789 * Check individual transmit queue of a device with multiple transmit queues.
2790 */
David S. Miller4d295152012-03-07 21:02:35 -05002791static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2792 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002793{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002794 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00002795
2796 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002797}
2798
David S. Miller4d295152012-03-07 21:02:35 -05002799static inline bool netif_subqueue_stopped(const struct net_device *dev,
2800 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07002801{
2802 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2803}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002804
John Fastabend46e5da40a2014-09-12 20:04:52 -07002805void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002806
Alexander Duyck537c00d2013-01-10 08:57:02 +00002807#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04002808int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07002809 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00002810#else
2811static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03002812 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00002813 u16 index)
2814{
2815 return 0;
2816}
2817#endif
2818
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002819/*
2820 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2821 * as a distribution range limit for the returned value.
2822 */
2823static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07002824 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002825{
2826 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2827}
2828
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002829/**
2830 * netif_is_multiqueue - test if device has multiple transmit queues
2831 * @dev: network device
2832 *
2833 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002834 */
David S. Miller4d295152012-03-07 21:02:35 -05002835static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002836{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002837 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002838}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002839
Joe Perchesf629d202013-09-26 14:48:15 -07002840int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00002841
Michael Daltona953be52014-01-16 22:23:28 -08002842#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07002843int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00002844#else
2845static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2846 unsigned int rxq)
2847{
2848 return 0;
2849}
2850#endif
2851
Michael Daltona953be52014-01-16 22:23:28 -08002852#ifdef CONFIG_SYSFS
2853static inline unsigned int get_netdev_rx_queue_index(
2854 struct netdev_rx_queue *queue)
2855{
2856 struct net_device *dev = queue->dev;
2857 int index = queue - dev->_rx;
2858
2859 BUG_ON(index >= dev->num_rx_queues);
2860 return index;
2861}
2862#endif
2863
Yuval Mintz16917b82012-07-01 03:18:50 +00002864#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07002865int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00002866
Eric Dumazete6247022013-12-05 04:45:08 -08002867enum skb_free_reason {
2868 SKB_REASON_CONSUMED,
2869 SKB_REASON_DROPPED,
2870};
Linus Torvalds1da177e2005-04-16 15:20:36 -07002871
Eric Dumazete6247022013-12-05 04:45:08 -08002872void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2873void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2874
2875/*
2876 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2877 * interrupt context or with hardware interrupts being disabled.
2878 * (in_irq() || irqs_disabled())
2879 *
2880 * We provide four helpers that can be used in following contexts :
2881 *
2882 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2883 * replacing kfree_skb(skb)
2884 *
2885 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2886 * Typically used in place of consume_skb(skb) in TX completion path
2887 *
2888 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2889 * replacing kfree_skb(skb)
2890 *
2891 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2892 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002893 */
Eric Dumazete6247022013-12-05 04:45:08 -08002894static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2895{
2896 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2897}
2898
2899static inline void dev_consume_skb_irq(struct sk_buff *skb)
2900{
2901 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2902}
2903
2904static inline void dev_kfree_skb_any(struct sk_buff *skb)
2905{
2906 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2907}
2908
2909static inline void dev_consume_skb_any(struct sk_buff *skb)
2910{
2911 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2912}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002913
Joe Perchesf629d202013-09-26 14:48:15 -07002914int netif_rx(struct sk_buff *skb);
2915int netif_rx_ni(struct sk_buff *skb);
2916int netif_receive_skb(struct sk_buff *skb);
2917gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2918void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2919struct sk_buff *napi_get_frags(struct napi_struct *napi);
2920gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08002921struct packet_offload *gro_find_receive_by_type(__be16 type);
2922struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07002923
2924static inline void napi_free_frags(struct napi_struct *napi)
2925{
2926 kfree_skb(napi->skb);
2927 napi->skb = NULL;
2928}
2929
Joe Perchesf629d202013-09-26 14:48:15 -07002930int netdev_rx_handler_register(struct net_device *dev,
2931 rx_handler_func_t *rx_handler,
2932 void *rx_handler_data);
2933void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00002934
Joe Perchesf629d202013-09-26 14:48:15 -07002935bool dev_valid_name(const char *name);
2936int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2937int dev_ethtool(struct net *net, struct ifreq *);
2938unsigned int dev_get_flags(const struct net_device *);
2939int __dev_change_flags(struct net_device *, unsigned int flags);
2940int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04002941void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2942 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07002943int dev_change_name(struct net_device *, const char *);
2944int dev_set_alias(struct net_device *, const char *, size_t);
2945int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2946int dev_set_mtu(struct net_device *, int);
2947void dev_set_group(struct net_device *, int);
2948int dev_set_mac_address(struct net_device *, struct sockaddr *);
2949int dev_change_carrier(struct net_device *, bool new_carrier);
2950int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01002951 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06002952int dev_get_phys_port_name(struct net_device *dev,
2953 char *name, size_t len);
Eric Dumazet55a93b32014-10-03 15:31:07 -07002954struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07002955struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2956 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08002957int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07002958int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Vlad Yasevich1ee481f2014-03-27 17:32:29 -04002959bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002960
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002961extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002962
2963/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07002964void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002965
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002966/**
2967 * dev_put - release reference to device
2968 * @dev: network device
2969 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07002970 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002971 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002972static inline void dev_put(struct net_device *dev)
2973{
Christoph Lameter933393f2011-12-22 11:58:51 -06002974 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002975}
2976
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002977/**
2978 * dev_hold - get reference to device
2979 * @dev: network device
2980 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07002981 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002982 */
Stephen Hemminger15333062006-03-20 22:32:28 -08002983static inline void dev_hold(struct net_device *dev)
2984{
Christoph Lameter933393f2011-12-22 11:58:51 -06002985 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08002986}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002987
2988/* Carrier loss detection, dial on demand. The functions netif_carrier_on
2989 * and _off may be called from IRQ context, but it is caller
2990 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08002991 *
2992 * The name carrier is inappropriate, these functions should really be
2993 * called netif_lowerlayer_*() because they represent the state of any
2994 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002995 */
2996
Joe Perchesf629d202013-09-26 14:48:15 -07002997void linkwatch_init_dev(struct net_device *dev);
2998void linkwatch_fire_event(struct net_device *dev);
2999void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003000
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003001/**
3002 * netif_carrier_ok - test if carrier present
3003 * @dev: network device
3004 *
3005 * Check if carrier is present on device
3006 */
David S. Miller4d295152012-03-07 21:02:35 -05003007static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003008{
3009 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3010}
3011
Joe Perchesf629d202013-09-26 14:48:15 -07003012unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003013
Joe Perchesf629d202013-09-26 14:48:15 -07003014void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003015
Joe Perchesf629d202013-09-26 14:48:15 -07003016void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003017
Joe Perchesf629d202013-09-26 14:48:15 -07003018void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003019
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003020/**
3021 * netif_dormant_on - mark device as dormant.
3022 * @dev: network device
3023 *
3024 * Mark device as dormant (as per RFC2863).
3025 *
3026 * The dormant state indicates that the relevant interface is not
3027 * actually in a condition to pass packets (i.e., it is not 'up') but is
3028 * in a "pending" state, waiting for some external event. For "on-
3029 * demand" interfaces, this new state identifies the situation where the
3030 * interface is waiting for events to place it in the up state.
3031 *
3032 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003033static inline void netif_dormant_on(struct net_device *dev)
3034{
3035 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3036 linkwatch_fire_event(dev);
3037}
3038
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003039/**
3040 * netif_dormant_off - set device as not dormant.
3041 * @dev: network device
3042 *
3043 * Device is not in dormant state.
3044 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003045static inline void netif_dormant_off(struct net_device *dev)
3046{
3047 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3048 linkwatch_fire_event(dev);
3049}
3050
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003051/**
3052 * netif_dormant - test if carrier present
3053 * @dev: network device
3054 *
3055 * Check if carrier is present on device
3056 */
David S. Miller4d295152012-03-07 21:02:35 -05003057static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003058{
3059 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3060}
3061
3062
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003063/**
3064 * netif_oper_up - test if device is operational
3065 * @dev: network device
3066 *
3067 * Check if carrier is operational
3068 */
David S. Miller4d295152012-03-07 21:02:35 -05003069static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003070{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003071 return (dev->operstate == IF_OPER_UP ||
3072 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3073}
3074
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003075/**
3076 * netif_device_present - is device available or removed
3077 * @dev: network device
3078 *
3079 * Check if device has not been removed from system.
3080 */
David S. Miller4d295152012-03-07 21:02:35 -05003081static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003082{
3083 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3084}
3085
Joe Perchesf629d202013-09-26 14:48:15 -07003086void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003087
Joe Perchesf629d202013-09-26 14:48:15 -07003088void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003089
3090/*
3091 * Network interface message level settings
3092 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003093
3094enum {
3095 NETIF_MSG_DRV = 0x0001,
3096 NETIF_MSG_PROBE = 0x0002,
3097 NETIF_MSG_LINK = 0x0004,
3098 NETIF_MSG_TIMER = 0x0008,
3099 NETIF_MSG_IFDOWN = 0x0010,
3100 NETIF_MSG_IFUP = 0x0020,
3101 NETIF_MSG_RX_ERR = 0x0040,
3102 NETIF_MSG_TX_ERR = 0x0080,
3103 NETIF_MSG_TX_QUEUED = 0x0100,
3104 NETIF_MSG_INTR = 0x0200,
3105 NETIF_MSG_TX_DONE = 0x0400,
3106 NETIF_MSG_RX_STATUS = 0x0800,
3107 NETIF_MSG_PKTDATA = 0x1000,
3108 NETIF_MSG_HW = 0x2000,
3109 NETIF_MSG_WOL = 0x4000,
3110};
3111
3112#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3113#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3114#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3115#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3116#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3117#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3118#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3119#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3120#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3121#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3122#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3123#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3124#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3125#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3126#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3127
3128static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3129{
3130 /* use default */
3131 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3132 return default_msg_enable_bits;
3133 if (debug_value == 0) /* no output */
3134 return 0;
3135 /* set low N bits */
3136 return (1 << debug_value) - 1;
3137}
3138
David S. Millerc773e842008-07-08 23:13:53 -07003139static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003140{
David S. Millerc773e842008-07-08 23:13:53 -07003141 spin_lock(&txq->_xmit_lock);
3142 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003143}
3144
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003145static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3146{
3147 spin_lock_bh(&txq->_xmit_lock);
3148 txq->xmit_lock_owner = smp_processor_id();
3149}
3150
David S. Miller4d295152012-03-07 21:02:35 -05003151static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003152{
David S. Miller4d295152012-03-07 21:02:35 -05003153 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003154 if (likely(ok))
3155 txq->xmit_lock_owner = smp_processor_id();
3156 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003157}
3158
David S. Millerc773e842008-07-08 23:13:53 -07003159static inline void __netif_tx_unlock(struct netdev_queue *txq)
3160{
3161 txq->xmit_lock_owner = -1;
3162 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003163}
3164
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003165static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3166{
3167 txq->xmit_lock_owner = -1;
3168 spin_unlock_bh(&txq->_xmit_lock);
3169}
3170
Eric Dumazet08baf562009-05-25 22:58:01 -07003171static inline void txq_trans_update(struct netdev_queue *txq)
3172{
3173 if (txq->xmit_lock_owner != -1)
3174 txq->trans_start = jiffies;
3175}
3176
David S. Millerc3f26a22008-07-31 16:58:50 -07003177/**
3178 * netif_tx_lock - grab network device transmit lock
3179 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003180 *
3181 * Get network device transmit lock
3182 */
3183static inline void netif_tx_lock(struct net_device *dev)
3184{
3185 unsigned int i;
3186 int cpu;
3187
3188 spin_lock(&dev->tx_global_lock);
3189 cpu = smp_processor_id();
3190 for (i = 0; i < dev->num_tx_queues; i++) {
3191 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3192
3193 /* We are the only thread of execution doing a
3194 * freeze, but we have to grab the _xmit_lock in
3195 * order to synchronize with threads which are in
3196 * the ->hard_start_xmit() handler and already
3197 * checked the frozen bit.
3198 */
3199 __netif_tx_lock(txq, cpu);
3200 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3201 __netif_tx_unlock(txq);
3202 }
3203}
3204
3205static inline void netif_tx_lock_bh(struct net_device *dev)
3206{
3207 local_bh_disable();
3208 netif_tx_lock(dev);
3209}
3210
Herbert Xu932ff272006-06-09 12:20:56 -07003211static inline void netif_tx_unlock(struct net_device *dev)
3212{
David S. Millere8a04642008-07-17 00:34:19 -07003213 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003214
David S. Millere8a04642008-07-17 00:34:19 -07003215 for (i = 0; i < dev->num_tx_queues; i++) {
3216 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003217
David S. Millerc3f26a22008-07-31 16:58:50 -07003218 /* No need to grab the _xmit_lock here. If the
3219 * queue is not stopped for another reason, we
3220 * force a schedule.
3221 */
3222 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003223 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003224 }
3225 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003226}
3227
3228static inline void netif_tx_unlock_bh(struct net_device *dev)
3229{
David S. Millere8a04642008-07-17 00:34:19 -07003230 netif_tx_unlock(dev);
3231 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003232}
3233
David S. Millerc773e842008-07-08 23:13:53 -07003234#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003235 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003236 __netif_tx_lock(txq, cpu); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003237 } \
3238}
3239
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003240#define HARD_TX_TRYLOCK(dev, txq) \
3241 (((dev->features & NETIF_F_LLTX) == 0) ? \
3242 __netif_tx_trylock(txq) : \
3243 true )
3244
David S. Millerc773e842008-07-08 23:13:53 -07003245#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003246 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003247 __netif_tx_unlock(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003248 } \
3249}
3250
Linus Torvalds1da177e2005-04-16 15:20:36 -07003251static inline void netif_tx_disable(struct net_device *dev)
3252{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003253 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003254 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003255
David S. Millerc3f26a22008-07-31 16:58:50 -07003256 local_bh_disable();
3257 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003258 for (i = 0; i < dev->num_tx_queues; i++) {
3259 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003260
3261 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003262 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003263 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003264 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003265 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003266}
3267
David S. Millere308a5d2008-07-15 00:13:44 -07003268static inline void netif_addr_lock(struct net_device *dev)
3269{
3270 spin_lock(&dev->addr_list_lock);
3271}
3272
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003273static inline void netif_addr_lock_nested(struct net_device *dev)
3274{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003275 int subclass = SINGLE_DEPTH_NESTING;
3276
3277 if (dev->netdev_ops->ndo_get_lock_subclass)
3278 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3279
3280 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003281}
3282
David S. Millere308a5d2008-07-15 00:13:44 -07003283static inline void netif_addr_lock_bh(struct net_device *dev)
3284{
3285 spin_lock_bh(&dev->addr_list_lock);
3286}
3287
3288static inline void netif_addr_unlock(struct net_device *dev)
3289{
3290 spin_unlock(&dev->addr_list_lock);
3291}
3292
3293static inline void netif_addr_unlock_bh(struct net_device *dev)
3294{
3295 spin_unlock_bh(&dev->addr_list_lock);
3296}
3297
Jiri Pirkof001fde2009-05-05 02:48:28 +00003298/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003299 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003300 * rcu_read_lock held.
3301 */
3302#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003303 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003304
Linus Torvalds1da177e2005-04-16 15:20:36 -07003305/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3306
Joe Perchesf629d202013-09-26 14:48:15 -07003307void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003308
3309/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003310struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003311 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003312 void (*setup)(struct net_device *),
3313 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003314#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3315 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003316
Tom Gundersenc835a672014-07-14 16:37:24 +02003317#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3318 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3319 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003320
Joe Perchesf629d202013-09-26 14:48:15 -07003321int register_netdev(struct net_device *dev);
3322void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003323
Jiri Pirko22bedad32010-04-01 21:22:57 +00003324/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003325int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3326 struct netdev_hw_addr_list *from_list, int addr_len);
3327void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3328 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003329int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3330 struct net_device *dev,
3331 int (*sync)(struct net_device *, const unsigned char *),
3332 int (*unsync)(struct net_device *,
3333 const unsigned char *));
3334void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3335 struct net_device *dev,
3336 int (*unsync)(struct net_device *,
3337 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003338void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003339
Jiri Pirkof001fde2009-05-05 02:48:28 +00003340/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003341int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3342 unsigned char addr_type);
3343int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3344 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003345void dev_addr_flush(struct net_device *dev);
3346int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003347
3348/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003349int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3350int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3351int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3352int dev_uc_sync(struct net_device *to, struct net_device *from);
3353int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3354void dev_uc_unsync(struct net_device *to, struct net_device *from);
3355void dev_uc_flush(struct net_device *dev);
3356void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003357
Alexander Duyck670e5b82014-05-28 18:44:46 -07003358/**
3359 * __dev_uc_sync - Synchonize device's unicast list
3360 * @dev: device to sync
3361 * @sync: function to call if address should be added
3362 * @unsync: function to call if address should be removed
3363 *
3364 * Add newly added addresses to the interface, and release
3365 * addresses that have been deleted.
3366 **/
3367static inline int __dev_uc_sync(struct net_device *dev,
3368 int (*sync)(struct net_device *,
3369 const unsigned char *),
3370 int (*unsync)(struct net_device *,
3371 const unsigned char *))
3372{
3373 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3374}
3375
3376/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003377 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003378 * @dev: device to sync
3379 * @unsync: function to call if address should be removed
3380 *
3381 * Remove all addresses that were added to the device by dev_uc_sync().
3382 **/
3383static inline void __dev_uc_unsync(struct net_device *dev,
3384 int (*unsync)(struct net_device *,
3385 const unsigned char *))
3386{
3387 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3388}
3389
Jiri Pirko22bedad32010-04-01 21:22:57 +00003390/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003391int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3392int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3393int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3394int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3395int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3396int dev_mc_sync(struct net_device *to, struct net_device *from);
3397int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3398void dev_mc_unsync(struct net_device *to, struct net_device *from);
3399void dev_mc_flush(struct net_device *dev);
3400void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003401
Alexander Duyck670e5b82014-05-28 18:44:46 -07003402/**
3403 * __dev_mc_sync - Synchonize device's multicast list
3404 * @dev: device to sync
3405 * @sync: function to call if address should be added
3406 * @unsync: function to call if address should be removed
3407 *
3408 * Add newly added addresses to the interface, and release
3409 * addresses that have been deleted.
3410 **/
3411static inline int __dev_mc_sync(struct net_device *dev,
3412 int (*sync)(struct net_device *,
3413 const unsigned char *),
3414 int (*unsync)(struct net_device *,
3415 const unsigned char *))
3416{
3417 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3418}
3419
3420/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003421 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003422 * @dev: device to sync
3423 * @unsync: function to call if address should be removed
3424 *
3425 * Remove all addresses that were added to the device by dev_mc_sync().
3426 **/
3427static inline void __dev_mc_unsync(struct net_device *dev,
3428 int (*unsync)(struct net_device *,
3429 const unsigned char *))
3430{
3431 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3432}
3433
Linus Torvalds1da177e2005-04-16 15:20:36 -07003434/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003435void dev_set_rx_mode(struct net_device *dev);
3436void __dev_set_rx_mode(struct net_device *dev);
3437int dev_set_promiscuity(struct net_device *dev, int inc);
3438int dev_set_allmulti(struct net_device *dev, int inc);
3439void netdev_state_change(struct net_device *dev);
3440void netdev_notify_peers(struct net_device *dev);
3441void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003442/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003443void dev_load(struct net *net, const char *name);
3444struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3445 struct rtnl_link_stats64 *storage);
3446void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3447 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003448
3449extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003450extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003451extern int weight_p;
Eric Dumazet0a148422011-04-20 09:27:32 +00003452extern int bpf_jit_enable;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003453
Joe Perchesf629d202013-09-26 14:48:15 -07003454bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003455struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3456 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003457struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3458 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003459
3460/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003461#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3462 for (iter = &(dev)->adj_list.upper, \
3463 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3464 updev; \
3465 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3466
3467/* iterate through upper list, must be called under RCU read lock */
Veaceslav Falico2f268f12013-09-25 09:20:07 +02003468#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3469 for (iter = &(dev)->all_adj_list.upper, \
3470 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3471 updev; \
3472 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003473
Joe Perchesf629d202013-09-26 14:48:15 -07003474void *netdev_lower_get_next_private(struct net_device *dev,
3475 struct list_head **iter);
3476void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3477 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003478
3479#define netdev_for_each_lower_private(dev, priv, iter) \
3480 for (iter = (dev)->adj_list.lower.next, \
3481 priv = netdev_lower_get_next_private(dev, &(iter)); \
3482 priv; \
3483 priv = netdev_lower_get_next_private(dev, &(iter)))
3484
3485#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3486 for (iter = &(dev)->adj_list.lower, \
3487 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3488 priv; \
3489 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3490
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003491void *netdev_lower_get_next(struct net_device *dev,
3492 struct list_head **iter);
3493#define netdev_for_each_lower_dev(dev, ldev, iter) \
3494 for (iter = &(dev)->adj_list.lower, \
3495 ldev = netdev_lower_get_next(dev, &(iter)); \
3496 ldev; \
3497 ldev = netdev_lower_get_next(dev, &(iter)))
3498
Joe Perchesf629d202013-09-26 14:48:15 -07003499void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003500void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003501struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3502struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3503int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3504int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003505 struct net_device *upper_dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003506int netdev_master_upper_dev_link_private(struct net_device *dev,
3507 struct net_device *upper_dev,
3508 void *private);
3509void netdev_upper_dev_unlink(struct net_device *dev,
3510 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003511void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003512void *netdev_lower_dev_get_private(struct net_device *dev,
3513 struct net_device *lower_dev);
Eric Dumazet960fb622014-11-16 06:23:05 -08003514
3515/* RSS keys are 40 or 52 bytes long */
3516#define NETDEV_RSS_KEY_LEN 52
3517extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3518void netdev_rss_key_fill(void *buffer, size_t len);
3519
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003520int dev_get_nest_level(struct net_device *dev,
3521 bool (*type_check)(struct net_device *dev));
Joe Perchesf629d202013-09-26 14:48:15 -07003522int skb_checksum_help(struct sk_buff *skb);
3523struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3524 netdev_features_t features, bool tx_path);
3525struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3526 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003527
Moni Shoua61bd3852015-02-03 16:48:29 +02003528struct netdev_bonding_info {
3529 ifslave slave;
3530 ifbond master;
3531};
3532
3533struct netdev_notifier_bonding_info {
3534 struct netdev_notifier_info info; /* must be first */
3535 struct netdev_bonding_info bonding_info;
3536};
3537
3538void netdev_bonding_info_change(struct net_device *dev,
3539 struct netdev_bonding_info *bonding_info);
3540
Cong Wang12b00042013-02-05 16:36:38 +00003541static inline
3542struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3543{
3544 return __skb_gso_segment(skb, features, true);
3545}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003546__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003547
3548static inline bool can_checksum_protocol(netdev_features_t features,
3549 __be16 protocol)
3550{
3551 return ((features & NETIF_F_GEN_CSUM) ||
3552 ((features & NETIF_F_V4_CSUM) &&
3553 protocol == htons(ETH_P_IP)) ||
3554 ((features & NETIF_F_V6_CSUM) &&
3555 protocol == htons(ETH_P_IPV6)) ||
3556 ((features & NETIF_F_FCOE_CRC) &&
3557 protocol == htons(ETH_P_FCOE)));
3558}
Cong Wang12b00042013-02-05 16:36:38 +00003559
Herbert Xufb286bb2005-11-10 13:01:24 -08003560#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003561void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003562#else
3563static inline void netdev_rx_csum_fault(struct net_device *dev)
3564{
3565}
3566#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003567/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003568void net_enable_timestamp(void);
3569void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003570
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003571#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003572int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003573#else
3574#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003575#endif
3576
David S. Miller47982482014-08-22 16:21:53 -07003577static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003578 struct sk_buff *skb, struct net_device *dev,
3579 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003580{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003581 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003582 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003583}
3584
David S. Miller10b3ad82014-08-29 21:07:24 -07003585static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003586 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003587{
3588 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003589 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003590
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003591 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003592 if (rc == NETDEV_TX_OK)
3593 txq_trans_update(txq);
3594
3595 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003596}
3597
Linus Torvalds42a2d922013-11-13 17:40:34 +09003598int netdev_class_create_file_ns(struct class_attribute *class_attr,
3599 const void *ns);
3600void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3601 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003602
3603static inline int netdev_class_create_file(struct class_attribute *class_attr)
3604{
3605 return netdev_class_create_file_ns(class_attr, NULL);
3606}
3607
3608static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3609{
3610 netdev_class_remove_file_ns(class_attr, NULL);
3611}
Jay Vosburghb8a97872008-06-13 18:12:04 -07003612
Johannes Berg04600792010-08-05 17:45:15 +02003613extern struct kobj_ns_type_operations net_ns_type_operations;
3614
Joe Perchesf629d202013-09-26 14:48:15 -07003615const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07003616
Joe Perchesf629d202013-09-26 14:48:15 -07003617void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003618
Michal Kubečekda081432014-05-20 08:29:25 +02003619static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3620 netdev_features_t f2)
3621{
3622 if (f1 & NETIF_F_GEN_CSUM)
3623 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3624 if (f2 & NETIF_F_GEN_CSUM)
3625 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3626 f1 &= f2;
3627 if (f1 & NETIF_F_GEN_CSUM)
3628 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3629
3630 return f1;
3631}
3632
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003633static inline netdev_features_t netdev_get_wanted_features(
3634 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00003635{
3636 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3637}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003638netdev_features_t netdev_increment_features(netdev_features_t all,
3639 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00003640
3641/* Allow TSO being used on stacked device :
3642 * Performing the GSO segmentation before last device
3643 * is a performance improvement.
3644 */
3645static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3646 netdev_features_t mask)
3647{
3648 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3649}
3650
Michał Mirosław6cb6a272011-04-02 22:48:47 -07003651int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00003652void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00003653void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07003654
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08003655void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3656 struct net_device *dev);
3657
Florian Westphalc1e756b2014-05-05 15:00:44 +02003658netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00003659
David S. Miller4d295152012-03-07 21:02:35 -05003660static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07003661{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003662 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00003663
3664 /* check flags correspondence */
3665 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3666 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3667 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3668 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3669 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3670 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07003671 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3672 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3673 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3674 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3675 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3676 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08003677 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00003678
Herbert Xubcd76112006-06-30 13:36:35 -07003679 return (features & feature) == feature;
3680}
3681
David S. Miller4d295152012-03-07 21:02:35 -05003682static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07003683{
Herbert Xu278b2512009-06-03 21:20:51 -07003684 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07003685 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07003686}
3687
Tom Herbert04ffcb22014-10-14 15:19:06 -07003688static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05003689 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07003690{
Jesse Grossfc741212011-01-09 06:23:32 +00003691 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00003692 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3693 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07003694}
3695
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07003696static inline void netif_set_gso_max_size(struct net_device *dev,
3697 unsigned int size)
3698{
3699 dev->gso_max_size = size;
3700}
3701
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08003702static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3703 int pulled_hlen, u16 mac_offset,
3704 int mac_len)
3705{
3706 skb->protocol = protocol;
3707 skb->encapsulation = 1;
3708 skb_push(skb, pulled_hlen);
3709 skb_reset_transport_header(skb);
3710 skb->mac_header = mac_offset;
3711 skb->network_header = skb->mac_header + mac_len;
3712 skb->mac_len = mac_len;
3713}
3714
John Fastabenda6cc0cf2013-11-06 09:54:46 -08003715static inline bool netif_is_macvlan(struct net_device *dev)
3716{
3717 return dev->priv_flags & IFF_MACVLAN;
3718}
3719
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08003720static inline bool netif_is_macvlan_port(struct net_device *dev)
3721{
3722 return dev->priv_flags & IFF_MACVLAN_PORT;
3723}
3724
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08003725static inline bool netif_is_ipvlan(struct net_device *dev)
3726{
3727 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3728}
3729
3730static inline bool netif_is_ipvlan_port(struct net_device *dev)
3731{
3732 return dev->priv_flags & IFF_IPVLAN_MASTER;
3733}
3734
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00003735static inline bool netif_is_bond_master(struct net_device *dev)
3736{
3737 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3738}
3739
David S. Miller4d295152012-03-07 21:02:35 -05003740static inline bool netif_is_bond_slave(struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00003741{
3742 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3743}
3744
Ben Greear3bdc0eb2012-02-11 15:39:30 +00003745static inline bool netif_supports_nofcs(struct net_device *dev)
3746{
3747 return dev->priv_flags & IFF_SUPP_NOFCS;
3748}
3749
Eric Dumazet02875872014-10-05 18:38:35 -07003750/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3751static inline void netif_keep_dst(struct net_device *dev)
3752{
3753 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3754}
3755
Eric W. Biederman505d4f72008-11-07 22:54:20 -08003756extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00003757
Joe Perches571ba422010-02-09 11:49:47 +00003758/* Logging, debugging and troubleshooting/diagnostic helpers. */
3759
3760/* netdev_printk helpers, similar to dev_printk */
3761
3762static inline const char *netdev_name(const struct net_device *dev)
3763{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02003764 if (!dev->name[0] || strchr(dev->name, '%'))
3765 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00003766 return dev->name;
3767}
3768
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02003769static inline const char *netdev_reg_state(const struct net_device *dev)
3770{
3771 switch (dev->reg_state) {
3772 case NETREG_UNINITIALIZED: return " (uninitialized)";
3773 case NETREG_REGISTERED: return "";
3774 case NETREG_UNREGISTERING: return " (unregistering)";
3775 case NETREG_UNREGISTERED: return " (unregistered)";
3776 case NETREG_RELEASED: return " (released)";
3777 case NETREG_DUMMY: return " (dummy)";
3778 }
3779
3780 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3781 return " (unknown)";
3782}
3783
Joe Perchesf629d202013-09-26 14:48:15 -07003784__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07003785void netdev_printk(const char *level, const struct net_device *dev,
3786 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003787__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003788void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003789__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003790void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003791__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003792void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003793__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003794void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003795__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003796void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003797__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003798void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07003799__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07003800void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00003801
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03003802#define MODULE_ALIAS_NETDEV(device) \
3803 MODULE_ALIAS("netdev-" device)
3804
Jim Cromieb558c962011-12-19 17:11:18 -05003805#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00003806#define netdev_dbg(__dev, format, args...) \
3807do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04003808 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00003809} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05003810#elif defined(DEBUG)
3811#define netdev_dbg(__dev, format, args...) \
3812 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00003813#else
3814#define netdev_dbg(__dev, format, args...) \
3815({ \
3816 if (0) \
3817 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00003818})
3819#endif
3820
3821#if defined(VERBOSE_DEBUG)
3822#define netdev_vdbg netdev_dbg
3823#else
3824
3825#define netdev_vdbg(dev, format, args...) \
3826({ \
3827 if (0) \
3828 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3829 0; \
3830})
3831#endif
3832
3833/*
3834 * netdev_WARN() acts like dev_printk(), but with the key difference
3835 * of using a WARN/WARN_ON to get the message out, including the
3836 * file/line information and a backtrace.
3837 */
3838#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02003839 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3840 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00003841
Joe Perchesb3d95c52010-02-09 11:49:49 +00003842/* netif printk helpers, similar to netdev_printk */
3843
3844#define netif_printk(priv, type, level, dev, fmt, args...) \
3845do { \
3846 if (netif_msg_##type(priv)) \
3847 netdev_printk(level, (dev), fmt, ##args); \
3848} while (0)
3849
Joe Perchesf45f4322010-06-27 01:02:36 +00003850#define netif_level(level, priv, type, dev, fmt, args...) \
3851do { \
3852 if (netif_msg_##type(priv)) \
3853 netdev_##level(dev, fmt, ##args); \
3854} while (0)
3855
Joe Perchesb3d95c52010-02-09 11:49:49 +00003856#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003857 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003858#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003859 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003860#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003861 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003862#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003863 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003864#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003865 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003866#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003867 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003868#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00003869 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003870
Joe Perches0053ea92012-05-30 07:43:34 +00003871#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003872#define netif_dbg(priv, type, netdev, format, args...) \
3873do { \
3874 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04003875 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00003876} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00003877#elif defined(DEBUG)
3878#define netif_dbg(priv, type, dev, format, args...) \
3879 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00003880#else
3881#define netif_dbg(priv, type, dev, format, args...) \
3882({ \
3883 if (0) \
3884 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3885 0; \
3886})
3887#endif
3888
3889#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00003890#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00003891#else
3892#define netif_vdbg(priv, type, dev, format, args...) \
3893({ \
3894 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00003895 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00003896 0; \
3897})
3898#endif
Joe Perches571ba422010-02-09 11:49:47 +00003899
Cong Wang900ff8c2013-02-18 19:20:33 +00003900/*
3901 * The list of packet types we will receive (as opposed to discard)
3902 * and the routines to invoke.
3903 *
3904 * Why 16. Because with 16 the only overlap we get on a hash of the
3905 * low nibble of the protocol value is RARP/SNAP/X.25.
3906 *
3907 * NOTE: That is no longer true with the addition of VLAN tags. Not
3908 * sure which should go first, but I bet it won't make much
3909 * difference if we are running VLANs. The good news is that
3910 * this protocol won't be in the list unless compiled in, so
3911 * the average user (w/out VLANs) will not be adversely affected.
3912 * --BLG
3913 *
3914 * 0800 IP
3915 * 8100 802.1Q VLAN
3916 * 0001 802.3
3917 * 0002 AX.25
3918 * 0004 802.2
3919 * 8035 RARP
3920 * 0005 SNAP
3921 * 0805 X.25
3922 * 0806 ARP
3923 * 8137 IPX
3924 * 0009 Localtalk
3925 * 86DD IPv6
3926 */
3927#define PTYPE_HASH_SIZE (16)
3928#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3929
Jiri Pirko385a1542009-05-27 15:48:07 -07003930#endif /* _LINUX_NETDEVICE_H */