blob: 41df0b450757e2821e7957da7face4ce32f99ac5 [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
Al Virod7fe0f22006-12-03 23:15:30 -050028#include <linux/timer.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050029#include <linux/bug.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070030#include <linux/delay.h>
Arun Sharma600634972011-07-26 16:09:06 -070031#include <linux/atomic.h>
Eric Dumazet53511452014-10-08 08:19:27 -070032#include <linux/prefetch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <asm/cache.h>
34#include <asm/byteorder.h>
35
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/percpu.h>
David S. Miller4d5b78c2009-05-06 16:52:51 -070037#include <linux/rculist.h>
Chris Leechdb217332006-06-17 21:24:58 -070038#include <linux/dmaengine.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070039#include <linux/workqueue.h>
Tom Herbert114cf582011-11-28 16:33:09 +000040#include <linux/dynamic_queue_limits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Patrick McHardyb1b67dd2009-04-20 04:49:28 +000042#include <linux/ethtool.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020043#include <net/net_namespace.h>
Lennert Buytenhekcf85d082008-10-07 13:45:02 +000044#include <net/dsa.h>
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -080045#ifdef CONFIG_DCB
Alexander Duyck2f90b862008-11-20 20:52:10 -080046#include <net/dcbnl.h>
47#endif
Neil Horman5bc14212011-11-22 05:10:51 +000048#include <net/netprio_cgroup.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020049
Michał Mirosława59e2ec2011-11-15 15:29:55 +000050#include <linux/netdev_features.h>
John Fastabend77162022012-04-15 06:43:56 +000051#include <linux/neighbour.h>
David Howells607ca462012-10-13 10:46:48 +010052#include <uapi/linux/netdevice.h>
Moni Shoua61bd3852015-02-03 16:48:29 +020053#include <uapi/linux/if_bonding.h>
John Fastabende4c67342016-02-16 21:16:15 -080054#include <uapi/linux/pkt_cls.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;
Robert Shearman03c57742015-04-22 11:14:37 +010063struct mpls_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
Joe Perchesf629d202013-09-26 14:48:15 -070065void netdev_set_default_ethtool_ops(struct net_device *dev,
66 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000067
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000068/* Backlog congestion levels */
69#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
70#define NET_RX_DROP 1 /* packet dropped */
71
Patrick McHardy572a9d72009-11-10 06:14:14 +000072/*
73 * Transmit return codes: transmit return codes originate from three different
74 * namespaces:
75 *
76 * - qdisc return codes
77 * - driver transmit return codes
78 * - errno values
79 *
80 * Drivers are allowed to return any one of those in their hard_start_xmit()
81 * function. Real network devices commonly used with qdiscs should only return
82 * the driver transmit return codes though - when qdiscs are used, the actual
83 * transmission happens asynchronously, so the value is not propagated to
84 * higher layers. Virtual network devices transmit synchronously, in this case
85 * the driver transmit return codes are consumed by dev_queue_xmit(), all
86 * others are propagated to higher layers.
87 */
88
89/* qdisc ->enqueue() return codes. */
90#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000091#define NET_XMIT_DROP 0x01 /* skb dropped */
92#define NET_XMIT_CN 0x02 /* congestion notification */
93#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
94#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -020096/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97 * indicates that the device will soon be dropping packets, or already drops
98 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +000099#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000102/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000103#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000104
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000105enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000106 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000107 NETDEV_TX_OK = 0x00, /* driver took care of packet */
108 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
109 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000110};
111typedef enum netdev_tx netdev_tx_t;
112
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000113/*
114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 */
117static inline bool dev_xmit_complete(int rc)
118{
119 /*
120 * Positive cases with an skb consumed by a driver:
121 * - successful transmission (rc == NETDEV_TX_OK)
122 * - error while transmitting (rc < 0)
123 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 */
125 if (likely(rc < NET_XMIT_MASK))
126 return true;
127
128 return false;
129}
130
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131/*
132 * Compute the worst case header length according to the protocols
133 * used.
134 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800135
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800136#if defined(CONFIG_HYPERV_NET)
137# define LL_MAX_HEADER 128
138#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700139# if defined(CONFIG_MAC80211_MESH)
140# define LL_MAX_HEADER 128
141# else
142# define LL_MAX_HEADER 96
143# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700145# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146#endif
147
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000148#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
149 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150#define MAX_HEADER LL_MAX_HEADER
151#else
152#define MAX_HEADER (LL_MAX_HEADER + 48)
153#endif
154
155/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000156 * Old network device statistics. Fields are native words
157 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800159
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800160struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000161 unsigned long rx_packets;
162 unsigned long tx_packets;
163 unsigned long rx_bytes;
164 unsigned long tx_bytes;
165 unsigned long rx_errors;
166 unsigned long tx_errors;
167 unsigned long rx_dropped;
168 unsigned long tx_dropped;
169 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000172 unsigned long rx_over_errors;
173 unsigned long rx_crc_errors;
174 unsigned long rx_frame_errors;
175 unsigned long rx_fifo_errors;
176 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 unsigned long tx_aborted_errors;
178 unsigned long tx_carrier_errors;
179 unsigned long tx_fifo_errors;
180 unsigned long tx_heartbeat_errors;
181 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 unsigned long rx_compressed;
183 unsigned long tx_compressed;
184};
185
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
187#include <linux/cache.h>
188#include <linux/skbuff.h>
189
Eric Dumazetadc93002011-11-17 03:13:26 +0000190#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100191#include <linux/static_key.h>
192extern struct static_key rps_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000193#endif
194
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195struct neighbour;
196struct neigh_parms;
197struct sk_buff;
198
Jiri Pirkof001fde2009-05-05 02:48:28 +0000199struct netdev_hw_addr {
200 struct list_head list;
201 unsigned char addr[MAX_ADDR_LEN];
202 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000203#define NETDEV_HW_ADDR_T_LAN 1
204#define NETDEV_HW_ADDR_T_SAN 2
205#define NETDEV_HW_ADDR_T_SLAVE 3
206#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000207#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000208 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000209 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700210 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400211 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000212 struct rcu_head rcu_head;
213};
214
Jiri Pirko31278e72009-06-17 01:12:19 +0000215struct netdev_hw_addr_list {
216 struct list_head list;
217 int count;
218};
219
Jiri Pirko22bedad32010-04-01 21:22:57 +0000220#define netdev_hw_addr_list_count(l) ((l)->count)
221#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
222#define netdev_hw_addr_list_for_each(ha, l) \
223 list_for_each_entry(ha, &(l)->list, list)
224
225#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
226#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800227#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000228 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800229
Jiri Pirko22bedad32010-04-01 21:22:57 +0000230#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
231#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700232#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000233 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece32010-02-04 10:22:25 -0800234
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800235struct hh_cache {
David S. Millerf6b72b622011-07-14 07:53:20 -0700236 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700237 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800238 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240 /* cached hardware header; allow for machine alignment needs. */
241#define HH_DATA_MOD 16
242#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700243 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244#define HH_DATA_ALIGN(__len) \
245 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
246 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
247};
248
249/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
250 * Alternative is:
251 * dev->hard_header_len ? (dev->hard_header_len +
252 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
253 *
254 * We could use other alignment values, but we must maintain the
255 * relationship HH alignment <= LL alignment.
256 */
257#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700258 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700260 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700262struct header_ops {
263 int (*create) (struct sk_buff *skb, struct net_device *dev,
264 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000265 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700266 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700267 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700268 void (*cache_update)(struct hh_cache *hh,
269 const struct net_device *dev,
270 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500271 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700272};
273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274/* These flag bits are private to the generic network queueing
275 * layer, they may not be explicitly referenced by any other
276 * code.
277 */
278
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800279enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 __LINK_STATE_START,
281 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800283 __LINK_STATE_LINKWATCH_PENDING,
284 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285};
286
287
288/*
289 * This structure holds at boot time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800290 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 */
292struct netdev_boot_setup {
293 char name[IFNAMSIZ];
294 struct ifmap map;
295};
296#define NETDEV_BOOT_SETUP_MAX 8
297
Joe Perchesf629d202013-09-26 14:48:15 -0700298int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
300/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700301 * Structure for NAPI scheduling similar to tasklet but with weighting
302 */
303struct napi_struct {
304 /* The poll_list must only be managed by the entity which
305 * changes the state of the NAPI_STATE_SCHED bit. This means
306 * whoever atomically sets that bit can add this napi_struct
307 * to the per-cpu poll_list, and whoever clears that bit
308 * can remove from the list right before clearing the bit.
309 */
310 struct list_head poll_list;
311
312 unsigned long state;
313 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000314 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700315 int (*poll)(struct napi_struct *, int);
316#ifdef CONFIG_NETPOLL
317 spinlock_t poll_lock;
318 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700319#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800320 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800321 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800322 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800323 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000324 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300325 struct hlist_node napi_hash_node;
326 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700327};
328
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800329enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700330 NAPI_STATE_SCHED, /* Poll is scheduled */
David S. Millera0a46192008-01-07 20:35:07 -0800331 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800332 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800333 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
334 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700335};
336
Ben Hutchings5b252f02009-10-29 07:17:09 +0000337enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700338 GRO_MERGED,
339 GRO_MERGED_FREE,
340 GRO_HELD,
341 GRO_NORMAL,
342 GRO_DROP,
343};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000344typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700345
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000346/*
347 * enum rx_handler_result - Possible return values for rx_handlers.
348 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
349 * further.
350 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
351 * case skb->dev was changed by rx_handler.
352 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
353 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
354 *
355 * rx_handlers are functions called from inside __netif_receive_skb(), to do
356 * special processing of the skb, prior to delivery to protocol handlers.
357 *
358 * Currently, a net_device can only have a single rx_handler registered. Trying
359 * to register a second rx_handler will return -EBUSY.
360 *
361 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
362 * To unregister a rx_handler on a net_device, use
363 * netdev_rx_handler_unregister().
364 *
365 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
366 * do with the skb.
367 *
368 * If the rx_handler consumed to skb in some way, it should return
369 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
370 * the skb to be delivered in some other ways.
371 *
372 * If the rx_handler changed skb->dev, to divert the skb to another
373 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
374 * new device will be called if it exists.
375 *
376 * If the rx_handler consider the skb should be ignored, it should return
377 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400378 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000379 *
380 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
381 * delivered, it should return RX_HANDLER_PASS.
382 *
383 * A device without a registered rx_handler will behave as if rx_handler
384 * returned RX_HANDLER_PASS.
385 */
386
387enum rx_handler_result {
388 RX_HANDLER_CONSUMED,
389 RX_HANDLER_ANOTHER,
390 RX_HANDLER_EXACT,
391 RX_HANDLER_PASS,
392};
393typedef enum rx_handler_result rx_handler_result_t;
394typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000395
Joe Perchesf629d202013-09-26 14:48:15 -0700396void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700397void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700398
David S. Miller4d295152012-03-07 21:02:35 -0500399static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800400{
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
402}
403
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700404/**
405 * napi_schedule_prep - check if napi can be scheduled
406 * @n: napi context
407 *
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
David S. Millera0a46192008-01-07 20:35:07 -0800410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700412 */
David S. Miller4d295152012-03-07 21:02:35 -0500413static inline bool napi_schedule_prep(struct napi_struct *n)
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700414{
David S. Millera0a46192008-01-07 20:35:07 -0800415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700417}
418
419/**
420 * napi_schedule - schedule NAPI poll
421 * @n: napi context
422 *
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
425 */
426static inline void napi_schedule(struct napi_struct *n)
427{
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
430}
431
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700432/**
433 * napi_schedule_irqoff - schedule NAPI poll
434 * @n: napi context
435 *
436 * Variant of napi_schedule(), assuming hard irqs are masked.
437 */
438static inline void napi_schedule_irqoff(struct napi_struct *n)
439{
440 if (napi_schedule_prep(n))
441 __napi_schedule_irqoff(n);
442}
443
Roland Dreierbfe13f52007-10-09 15:47:37 -0700444/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500445static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700446{
447 if (napi_schedule_prep(napi)) {
448 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500449 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700450 }
David S. Miller4d295152012-03-07 21:02:35 -0500451 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700452}
453
Eric Dumazet3b47d302014-11-06 21:09:44 -0800454void __napi_complete(struct napi_struct *n);
455void napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700456/**
457 * napi_complete - NAPI processing complete
458 * @n: napi context
459 *
460 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800461 * Consider using napi_complete_done() instead.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700462 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800463static inline void napi_complete(struct napi_struct *n)
464{
465 return napi_complete_done(n, 0);
466}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700467
468/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300469 * napi_hash_add - add a NAPI to global hashtable
470 * @napi: napi context
471 *
472 * generate a new napi_id and store a @napi under it in napi_hash
Eric Dumazet93d05d42015-11-18 06:31:03 -0800473 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL)
474 * Note: This is normally automatically done from netif_napi_add(),
475 * so might disappear in a future linux version.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300476 */
Joe Perchesf629d202013-09-26 14:48:15 -0700477void napi_hash_add(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300478
479/**
480 * napi_hash_del - remove a NAPI from global table
481 * @napi: napi context
482 *
483 * Warning: caller must observe rcu grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800484 * before freeing memory containing @napi, if
485 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800486 * Note: core networking stack automatically calls it
487 * from netif_napi_del()
488 * Drivers might want to call this helper to combine all
489 * the needed rcu grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300490 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800491bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300492
493/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700494 * napi_disable - prevent NAPI from scheduling
495 * @n: napi context
496 *
497 * Stop NAPI from being scheduled on this context.
498 * Waits till any outstanding processing completes.
499 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800500void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700501
502/**
503 * napi_enable - enable NAPI scheduling
504 * @n: napi context
505 *
506 * Resume NAPI from being scheduled on this context.
507 * Must be paired with napi_disable.
508 */
509static inline void napi_enable(struct napi_struct *n)
510{
511 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100512 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700513 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400514 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700515}
516
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700517/**
518 * napi_synchronize - wait until NAPI is not running
519 * @n: napi context
520 *
521 * Wait until NAPI is done being scheduled on this context.
522 * Waits till any outstanding processing completes but
523 * does not disable future activations.
524 */
525static inline void napi_synchronize(const struct napi_struct *n)
526{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100527 if (IS_ENABLED(CONFIG_SMP))
528 while (test_bit(NAPI_STATE_SCHED, &n->state))
529 msleep(1);
530 else
531 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700532}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700533
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800534enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000535 __QUEUE_STATE_DRV_XOFF,
536 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700537 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700538};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200539
540#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
541#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
542#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
543
544#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
545#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
546 QUEUE_STATE_FROZEN)
547#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 QUEUE_STATE_FROZEN)
549
Tom Herbert734664982011-11-28 16:32:44 +0000550/*
551 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
552 * netif_tx_* functions below are used to manipulate this flag. The
553 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
554 * queue independently. The netif_xmit_*stopped functions below are called
555 * to check if the queue has been stopped by the driver or stack (either
556 * of the XOFF bits are set in the state). Drivers should not need to call
557 * netif_xmit*stopped functions, they should only be using netif_tx_*.
558 */
David S. Miller79d16382008-07-08 23:14:46 -0700559
David S. Millerbb949fb2008-07-08 16:55:56 -0700560struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700561/*
562 * read mostly part
563 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700564 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700565 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700566 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000567#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000568 struct kobject kobj;
569#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000570#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
571 int numa_node;
572#endif
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700573/*
574 * write mostly part
575 */
576 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
577 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700578 /*
579 * please use this field instead of dev->trans_start
580 */
581 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000582
583 /*
584 * Number of TX timeouts for this queue
585 * (/sys/class/net/DEV/Q/trans_timeout)
586 */
587 unsigned long trans_timeout;
Tom Herbert114cf582011-11-28 16:33:09 +0000588
589 unsigned long state;
590
591#ifdef CONFIG_BQL
592 struct dql dql;
593#endif
John Fastabend822b3b22015-03-18 14:57:33 +0200594 unsigned long tx_maxrate;
David S. Millere8a04642008-07-17 00:34:19 -0700595} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700596
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000597static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 return q->numa_node;
601#else
Changli Gaob236da62010-12-14 03:09:15 +0000602 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000603#endif
604}
605
606static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
607{
608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 q->numa_node = node;
610#endif
611}
612
Eric Dumazetdf334542010-03-24 19:13:54 +0000613#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000614/*
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
617 */
618struct rps_map {
619 unsigned int len;
620 struct rcu_head rcu;
621 u16 cpus[0];
622};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000623#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000624
Tom Herbertfec5e652010-04-16 16:01:27 -0700625/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700629 */
630struct rps_dev_flow {
631 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000632 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700633 unsigned int last_qtail;
634};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000635#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700636
637/*
638 * The rps_dev_flow_table structure contains a table of flow mappings.
639 */
640struct rps_dev_flow_table {
641 unsigned int mask;
642 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700643 struct rps_dev_flow flows[0];
644};
645#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000646 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700647
648/*
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
Eric Dumazet567e4b72015-02-06 12:59:01 -0800651 * Each entry is a 32bit value. Upper part is the high order bits
652 * of flow hash, lower part is cpu number.
653 * rps_cpu_mask is used to partition the space, depending on number of
654 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
656 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700657 */
658struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800659 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800660
661 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700662};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800663#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700664
665#define RPS_NO_CPU 0xffff
666
Eric Dumazet567e4b72015-02-06 12:59:01 -0800667extern u32 rps_cpu_mask;
668extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
669
Tom Herbertfec5e652010-04-16 16:01:27 -0700670static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 u32 hash)
672{
673 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700676
677 /* We only give a hint, preemption can change cpu under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800678 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700679
Eric Dumazet567e4b72015-02-06 12:59:01 -0800680 if (table->ents[index] != val)
681 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700682 }
683}
684
Ben Hutchingsc4454772011-01-19 11:03:53 +0000685#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700686bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
687 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000688#endif
Michael Daltona953be52014-01-16 22:23:28 -0800689#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000690
Tom Herbert0a9627f2010-03-16 08:03:29 +0000691/* This structure contains an instance of an RX queue. */
692struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800693#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800696#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000697 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000698 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000699} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800700
701/*
702 * RX queue sysfs structures and functions.
703 */
704struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
710};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800711
Tom Herbertbf264142010-11-26 08:36:09 +0000712#ifdef CONFIG_XPS
713/*
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
716 */
717struct xps_map {
718 unsigned int len;
719 unsigned int alloc_len;
720 struct rcu_head rcu;
721 u16 queues[0];
722};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000723#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100724#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000726
727/*
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
729 */
730struct xps_dev_maps {
731 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000732 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000733};
734#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * sizeof(struct xps_map *)))
736#endif /* CONFIG_XPS */
737
John Fastabend4f57c082011-01-17 08:06:04 +0000738#define TC_MAX_QUEUE 16
739#define TC_BITMASK 15
740/* HW offloaded queuing disciplines txq count and offset maps */
741struct netdev_tc_txq {
742 u16 count;
743 u16 offset;
744};
745
Neerav Parikh68bad942012-01-04 20:23:39 +0000746#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
747/*
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
750 */
751struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
758 char model[256];
759 char model_description[256];
760};
761#endif
762
Jiri Pirko02637fc2014-11-28 14:34:16 +0100763#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200764
Jiri Pirko02637fc2014-11-28 14:34:16 +0100765/* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200767 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100768struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200770 unsigned char id_len;
771};
772
Scott Feldmand754f982015-07-18 18:24:49 -0700773static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
775{
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
778}
779
Daniel Borkmann99932d42014-02-16 15:55:20 +0100780typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
782
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800783/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800784 * that are being passed to the netdevice through the setup_tc op.
785 */
786enum {
787 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800788 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200789 TC_SETUP_CLSFLOWER,
John Fastabend16e5cc62016-02-16 21:16:43 -0800790};
791
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800792struct tc_cls_u32_offload;
793
John Fastabend16e5cc62016-02-16 21:16:43 -0800794struct tc_to_netdev {
795 unsigned int type;
796 union {
797 u8 tc;
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800798 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200799 struct tc_cls_flower_offload *cls_flower;
John Fastabend16e5cc62016-02-16 21:16:43 -0800800 };
801};
802
803
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800804/*
805 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800806 * The following hooks can be defined; unless noted otherwise, they are
807 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800808 *
809 * int (*ndo_init)(struct net_device *dev);
810 * This function is called once when network device is registered.
811 * The network device can use this to any late stage initializaton
812 * or semantic validattion. It can fail with an error code which will
813 * be propogated back to register_netdev
814 *
815 * void (*ndo_uninit)(struct net_device *dev);
816 * This function is called when device is unregistered or when registration
817 * fails. It is not called if init fails.
818 *
819 * int (*ndo_open)(struct net_device *dev);
820 * This function is called when network device transistions to the up
821 * state.
822 *
823 * int (*ndo_stop)(struct net_device *dev);
824 * This function is called when network device transistions to the down
825 * state.
826 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000827 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
828 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800829 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030830 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
831 * the queue before that can happen; it's for obsolete devices and weird
832 * corner cases, but the stack really does a non-trivial amount
833 * of useless work if you return NETDEV_TX_BUSY.
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000834 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
Stephen Hemminger00829822008-11-20 20:14:53 -0800835 * Required can not be NULL.
836 *
Eric Dumazetcdba7562016-01-06 06:53:50 -0800837 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
838 * netdev_features_t features);
839 * Adjusts the requested feature flags according to device-specific
840 * constraints, and returns the resulting flags. Must not modify
841 * the device state.
842 *
Jason Wangf663dd92014-01-10 16:18:26 +0800843 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100844 * void *accel_priv, select_queue_fallback_t fallback);
Stephen Hemminger00829822008-11-20 20:14:53 -0800845 * Called to decide which queue to when device supports multiple
846 * transmit queues.
847 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800848 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
849 * This function is called to allow device receiver to make
850 * changes to configuration when multicast or promiscious is enabled.
851 *
852 * void (*ndo_set_rx_mode)(struct net_device *dev);
853 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000854 * If driver handles unicast address filtering, it should set
855 * IFF_UNICAST_FLT to its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800856 *
857 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
858 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700859 * needs to be changed. If this interface is not defined, the
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800860 * mac address can not be changed.
861 *
862 * int (*ndo_validate_addr)(struct net_device *dev);
863 * Test if Media Access Control address is valid for the device.
864 *
865 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
866 * Called when a user request an ioctl which can't be handled by
867 * the generic interface code. If not defined ioctl's return
868 * not supported error code.
869 *
870 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
871 * Used to set network devices bus interface parameters. This interface
872 * is retained for legacy reason, new devices should use the bus
873 * interface (PCI) for low level management.
874 *
875 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
876 * Called when a user wants to change the Maximum Transfer Unit
877 * of a device. If not defined, any request to change MTU will
878 * will return an error.
879 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800880 * void (*ndo_tx_timeout)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800881 * Callback uses when the transmitter has not made any progress
882 * for dev->watchdog ticks.
883 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000884 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700885 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700886 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800887 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000888 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000889 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
890 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700891 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000892 * (which should normally be dev->stats) and return a pointer to
893 * it. The structure may be changed asynchronously only if each
894 * field is written atomically.
895 * 3. Update dev->stats asynchronously and atomically, and define
896 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800897 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530898 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Patrick McHardy80d5c362013-04-19 02:04:28 +0000899 * If device support VLAN filtering this function is called when a
900 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800901 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530902 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Patrick McHardy80d5c362013-04-19 02:04:28 +0000903 * If device support VLAN filtering this function is called when a
904 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800905 *
906 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000907 *
908 * SR-IOV management functions.
909 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
910 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400911 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
912 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000913 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000914 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000915 * int (*ndo_get_vf_config)(struct net_device *dev,
916 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300917 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700918 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
919 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300920 *
921 * Enable or disable the VF ability to query its RSS Redirection Table and
922 * Hash Key. This is needed since on some devices VF share this information
923 * with PF and querying it may adduce a theoretical security risk.
924 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700925 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000926 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
927 * Called to setup 'tc' number of traffic classes in the net device. This
928 * is always called from the stack with the rtnl lock held and netif tx
929 * queues stopped. This allows the netdevice to perform queue management
930 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000931 *
Yi Zoue9bce842011-03-09 08:48:03 +0000932 * Fiber Channel over Ethernet (FCoE) offload functions.
933 * int (*ndo_fcoe_enable)(struct net_device *dev);
934 * Called when the FCoE protocol stack wants to start using LLD for FCoE
935 * so the underlying device can perform whatever needed configuration or
936 * initialization to support acceleration of FCoE traffic.
937 *
938 * int (*ndo_fcoe_disable)(struct net_device *dev);
939 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
940 * so the underlying device can perform whatever needed clean-ups to
941 * stop supporting acceleration of FCoE traffic.
942 *
943 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
944 * struct scatterlist *sgl, unsigned int sgc);
945 * Called when the FCoE Initiator wants to initialize an I/O that
946 * is a possible candidate for Direct Data Placement (DDP). The LLD can
947 * perform necessary setup and returns 1 to indicate the device is set up
948 * successfully to perform DDP on this I/O, otherwise this returns 0.
949 *
950 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
951 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
952 * indicated by the FC exchange id 'xid', so the underlying device can
953 * clean up and reuse resources for later DDP requests.
954 *
955 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
956 * struct scatterlist *sgl, unsigned int sgc);
957 * Called when the FCoE Target wants to initialize an I/O that
958 * is a possible candidate for Direct Data Placement (DDP). The LLD can
959 * perform necessary setup and returns 1 to indicate the device is set up
960 * successfully to perform DDP on this I/O, otherwise this returns 0.
961 *
Neerav Parikh68bad942012-01-04 20:23:39 +0000962 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
963 * struct netdev_fcoe_hbainfo *hbainfo);
964 * Called when the FCoE Protocol stack wants information on the underlying
965 * device. This information is utilized by the FCoE protocol stack to
966 * register attributes with Fiber Channel management service as per the
967 * FC-GS Fabric Device Management Information(FDMI) specification.
968 *
Yi Zoue9bce842011-03-09 08:48:03 +0000969 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
970 * Called when the underlying device wants to override default World Wide
971 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
972 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
973 * protocol stack to use.
974 *
Ben Hutchingsc4454772011-01-19 11:03:53 +0000975 * RFS acceleration.
976 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
977 * u16 rxq_index, u32 flow_id);
978 * Set hardware filter for RFS. rxq_index is the target queue index;
979 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
980 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000981 *
Jiri Pirko8b98a702013-01-03 22:49:02 +0000982 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000983 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
984 * Called to make another netdev an underling.
985 *
986 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
987 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +0000988 *
989 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000990 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +0000991 * Called to update device configuration to new features. Passed
992 * feature set might be less than what was returned by ndo_fix_features()).
993 * Must return >0 or -errno if it changed dev->features itself.
994 *
stephen hemmingeredc7d572012-10-01 12:32:33 +0000995 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
996 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100997 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +0000998 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +0000999 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1000 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001001 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001002 * Deletes the FDB entry from dev coresponding to addr.
1003 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001004 * struct net_device *dev, struct net_device *filter_dev,
1005 * int idx)
John Fastabend77162022012-04-15 06:43:56 +00001006 * Used to add FDB entries to dump requests. Implementers should add
1007 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001008 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001009 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1010 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001011 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001012 * struct net_device *dev, u32 filter_mask,
1013 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001014 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1015 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001016 *
1017 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1018 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1019 * which do not represent real hardware may define this to allow their
1020 * userspace components to manage their virtual carrier state. Devices
1021 * that determine carrier state from physical hardware properties (eg
1022 * network cables) or protocol-dependent mechanisms (eg
1023 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001024 *
1025 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001026 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001027 * Called to get ID of physical port of this device. If driver does
1028 * not implement this, it is assumed that the hw is not able to have
1029 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001030 *
1031 * void (*ndo_add_vxlan_port)(struct net_device *dev,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001032 * sa_family_t sa_family, __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001033 * Called by vxlan to notiy a driver about the UDP port and socket
1034 * address family that vxlan is listnening to. It is called only when
1035 * a new port starts listening. The operation is protected by the
1036 * vxlan_net->sock_lock.
1037 *
Singhai, Anjalia8170d22015-12-14 12:21:17 -08001038 * void (*ndo_add_geneve_port)(struct net_device *dev,
1039 * sa_family_t sa_family, __be16 port);
1040 * Called by geneve to notify a driver about the UDP port and socket
1041 * address family that geneve is listnening to. It is called only when
1042 * a new port starts listening. The operation is protected by the
1043 * geneve_net->sock_lock.
1044 *
1045 * void (*ndo_del_geneve_port)(struct net_device *dev,
1046 * sa_family_t sa_family, __be16 port);
1047 * Called by geneve to notify the driver about a UDP port and socket
1048 * address family that geneve is not listening to anymore. The operation
1049 * is protected by the geneve_net->sock_lock.
1050 *
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001051 * void (*ndo_del_vxlan_port)(struct net_device *dev,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001052 * sa_family_t sa_family, __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001053 * Called by vxlan to notify the driver about a UDP port and socket
1054 * address family that vxlan is not listening to anymore. The operation
1055 * is protected by the vxlan_net->sock_lock.
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001056 *
1057 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1058 * struct net_device *dev)
1059 * Called by upper layer devices to accelerate switching or other
1060 * station functionality into hardware. 'pdev is the lowerdev
1061 * to use for the offload and 'dev' is the net device that will
1062 * back the offload. Returns a pointer to the private structure
1063 * the upper layer will maintain.
1064 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1065 * Called by upper layer device to delete the station created
1066 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1067 * the station and priv is the structure returned by the add
1068 * operation.
1069 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1070 * struct net_device *dev,
1071 * void *priv);
1072 * Callback to use for xmit over the accelerated station. This
1073 * is used in place of ndo_start_xmit on accelerated net
1074 * devices.
Jesse Gross5f352272014-12-23 22:37:26 -08001075 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1076 * struct net_device *dev
1077 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001078 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001079 * performing offload operations on a given packet. This is to give
1080 * the device an opportunity to implement any restrictions that cannot
1081 * be otherwise expressed by feature flags. The check is called with
1082 * the set of features that the stack has calculated and it returns
1083 * those the driver believes to be appropriate.
John Fastabend822b3b22015-03-18 14:57:33 +02001084 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1085 * int queue_index, u32 maxrate);
1086 * Called when a user wants to set a max-rate limitation of specific
1087 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001088 * int (*ndo_get_iflink)(const struct net_device *dev);
1089 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001090 * void (*ndo_change_proto_down)(struct net_device *dev,
1091 * bool proto_down);
1092 * This function is used to pass protocol port error state information
1093 * to the switch driver. The switch driver can react to the proto_down
1094 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001095 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1096 * This function is used to get egress tunnel information for given skb.
1097 * This is useful for retrieving outer tunnel header parameters while
1098 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001099 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1100 * This function is used to specify the headroom that the skb must
1101 * consider when allocation skb during packet reception. Setting
1102 * appropriate rx headroom value allows avoiding skb head copy on
1103 * forward. Setting a negative value reset the rx headroom to the
1104 * default value.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001105 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001106 */
1107struct net_device_ops {
1108 int (*ndo_init)(struct net_device *dev);
1109 void (*ndo_uninit)(struct net_device *dev);
1110 int (*ndo_open)(struct net_device *dev);
1111 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001112 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1113 struct net_device *dev);
1114 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1115 struct net_device *dev,
1116 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001117 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001118 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001119 void *accel_priv,
1120 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001121 void (*ndo_change_rx_flags)(struct net_device *dev,
1122 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001123 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001124 int (*ndo_set_mac_address)(struct net_device *dev,
1125 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001126 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001127 int (*ndo_do_ioctl)(struct net_device *dev,
1128 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001129 int (*ndo_set_config)(struct net_device *dev,
1130 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001131 int (*ndo_change_mtu)(struct net_device *dev,
1132 int new_mtu);
1133 int (*ndo_neigh_setup)(struct net_device *dev,
1134 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001135 void (*ndo_tx_timeout) (struct net_device *dev);
1136
Eric Dumazet28172732010-07-07 14:58:56 -07001137 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1138 struct rtnl_link_stats64 *storage);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001139 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1140
Jiri Pirko8e586132011-12-08 19:52:37 -05001141 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001142 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001143 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001144 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001145#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001146 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001147 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001148 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001149 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001150#endif
Cong Wange0d10952013-08-01 11:10:25 +08001151#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001152 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001153#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001154 int (*ndo_set_vf_mac)(struct net_device *dev,
1155 int queue, u8 *mac);
1156 int (*ndo_set_vf_vlan)(struct net_device *dev,
1157 int queue, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001158 int (*ndo_set_vf_rate)(struct net_device *dev,
1159 int vf, int min_tx_rate,
1160 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001161 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1162 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001163 int (*ndo_set_vf_trust)(struct net_device *dev,
1164 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001165 int (*ndo_get_vf_config)(struct net_device *dev,
1166 int vf,
1167 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001168 int (*ndo_set_vf_link_state)(struct net_device *dev,
1169 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001170 int (*ndo_get_vf_stats)(struct net_device *dev,
1171 int vf,
1172 struct ifla_vf_stats
1173 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001174 int (*ndo_set_vf_port)(struct net_device *dev,
1175 int vf,
1176 struct nlattr *port[]);
1177 int (*ndo_get_vf_port)(struct net_device *dev,
1178 int vf, struct sk_buff *skb);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001179 int (*ndo_set_vf_rss_query_en)(
1180 struct net_device *dev,
1181 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001182 int (*ndo_setup_tc)(struct net_device *dev,
1183 u32 handle,
1184 __be16 protocol,
1185 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001186#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001187 int (*ndo_fcoe_enable)(struct net_device *dev);
1188 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001189 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1190 u16 xid,
1191 struct scatterlist *sgl,
1192 unsigned int sgc);
1193 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1194 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001195 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1196 u16 xid,
1197 struct scatterlist *sgl,
1198 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001199 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1200 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001201#endif
1202
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001203#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001204#define NETDEV_FCOE_WWNN 0
1205#define NETDEV_FCOE_WWPN 1
1206 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1207 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001208#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001209
Ben Hutchingsc4454772011-01-19 11:03:53 +00001210#ifdef CONFIG_RFS_ACCEL
1211 int (*ndo_rx_flow_steer)(struct net_device *dev,
1212 const struct sk_buff *skb,
1213 u16 rxq_index,
1214 u32 flow_id);
1215#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001216 int (*ndo_add_slave)(struct net_device *dev,
1217 struct net_device *slave_dev);
1218 int (*ndo_del_slave)(struct net_device *dev,
1219 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001220 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1221 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001222 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001223 netdev_features_t features);
David Millerda6a8fa2011-07-25 00:01:38 +00001224 int (*ndo_neigh_construct)(struct neighbour *n);
David S. Miller447f2192011-12-19 15:04:41 -05001225 void (*ndo_neigh_destroy)(struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001226
1227 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001228 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001229 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001230 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001231 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001232 u16 flags);
1233 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001234 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001235 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001236 const unsigned char *addr,
1237 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001238 int (*ndo_fdb_dump)(struct sk_buff *skb,
1239 struct netlink_callback *cb,
1240 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001241 struct net_device *filter_dev,
John Fastabend77162022012-04-15 06:43:56 +00001242 int idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001243
1244 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001245 struct nlmsghdr *nlh,
1246 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001247 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1248 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001249 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001250 u32 filter_mask,
1251 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001252 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001253 struct nlmsghdr *nlh,
1254 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001255 int (*ndo_change_carrier)(struct net_device *dev,
1256 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001257 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001258 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001259 int (*ndo_get_phys_port_name)(struct net_device *dev,
1260 char *name, size_t len);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001261 void (*ndo_add_vxlan_port)(struct net_device *dev,
1262 sa_family_t sa_family,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001263 __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001264 void (*ndo_del_vxlan_port)(struct net_device *dev,
1265 sa_family_t sa_family,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001266 __be16 port);
Singhai, Anjalia8170d22015-12-14 12:21:17 -08001267 void (*ndo_add_geneve_port)(struct net_device *dev,
1268 sa_family_t sa_family,
1269 __be16 port);
1270 void (*ndo_del_geneve_port)(struct net_device *dev,
1271 sa_family_t sa_family,
1272 __be16 port);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001273 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1274 struct net_device *dev);
1275 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1276 void *priv);
1277
1278 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1279 struct net_device *dev,
1280 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001281 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001282 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1283 int queue_index,
1284 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001285 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001286 int (*ndo_change_proto_down)(struct net_device *dev,
1287 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001288 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1289 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001290 void (*ndo_set_rx_headroom)(struct net_device *dev,
1291 int needed_headroom);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001292};
1293
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001294/**
1295 * enum net_device_priv_flags - &struct net_device priv_flags
1296 *
1297 * These are the &struct net_device, they are only set internally
1298 * by drivers and used in the kernel. These flags are invisible to
1299 * userspace, this means that the order of these flags can change
1300 * during any kernel release.
1301 *
1302 * You should have a pretty good reason to be extending these flags.
1303 *
1304 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1305 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001306 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001307 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001308 * @IFF_WAN_HDLC: WAN HDLC device
1309 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1310 * release skb->dst
1311 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1312 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1313 * @IFF_MACVLAN_PORT: device used as macvlan port
1314 * @IFF_BRIDGE_PORT: device used as bridge port
1315 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1316 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1317 * @IFF_UNICAST_FLT: Supports unicast filtering
1318 * @IFF_TEAM_PORT: device used as team port
1319 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1320 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1321 * change when it's running
1322 * @IFF_MACVLAN: Macvlan device
David Ahern007979e2015-09-29 20:07:10 -07001323 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001324 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001325 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001326 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001327 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001328 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001329 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1330 * entity (i.e. the master device for bridged veth)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001331 */
1332enum netdev_priv_flags {
1333 IFF_802_1Q_VLAN = 1<<0,
1334 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001335 IFF_BONDING = 1<<2,
1336 IFF_ISATAP = 1<<3,
1337 IFF_WAN_HDLC = 1<<4,
1338 IFF_XMIT_DST_RELEASE = 1<<5,
1339 IFF_DONT_BRIDGE = 1<<6,
1340 IFF_DISABLE_NETPOLL = 1<<7,
1341 IFF_MACVLAN_PORT = 1<<8,
1342 IFF_BRIDGE_PORT = 1<<9,
1343 IFF_OVS_DATAPATH = 1<<10,
1344 IFF_TX_SKB_SHARING = 1<<11,
1345 IFF_UNICAST_FLT = 1<<12,
1346 IFF_TEAM_PORT = 1<<13,
1347 IFF_SUPP_NOFCS = 1<<14,
1348 IFF_LIVE_ADDR_CHANGE = 1<<15,
1349 IFF_MACVLAN = 1<<16,
1350 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1351 IFF_IPVLAN_MASTER = 1<<18,
1352 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001353 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001354 IFF_NO_QUEUE = 1<<21,
1355 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001356 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001357 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001358 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001359 IFF_PHONY_HEADROOM = 1<<26,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001360};
1361
1362#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1363#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001364#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001365#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001366#define IFF_WAN_HDLC IFF_WAN_HDLC
1367#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1368#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1369#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1370#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1371#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1372#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1373#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1374#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1375#define IFF_TEAM_PORT IFF_TEAM_PORT
1376#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1377#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1378#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001379#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001380#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1381#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001382#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001383#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001384#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001385#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001386#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001387#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001388
Karoly Kemeny536721b2014-07-30 20:27:36 +02001389/**
1390 * struct net_device - The DEVICE structure.
1391 * Actually, this whole structure is a big mistake. It mixes I/O
1392 * data with strictly "high-level" data, and it has to know about
1393 * almost every data structure used in the INET module.
1394 *
1395 * @name: This is the first field of the "visible" part of this structure
1396 * (i.e. as seen by users in the "Space.c" file). It is the name
1397 * of the interface.
1398 *
1399 * @name_hlist: Device name hash chain, please keep it close to name[]
1400 * @ifalias: SNMP alias
1401 * @mem_end: Shared memory end
1402 * @mem_start: Shared memory start
1403 * @base_addr: Device I/O address
1404 * @irq: Device IRQ number
1405 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001406 * @carrier_changes: Stats to monitor carrier on<->off transitions
1407 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001408 * @state: Generic network queuing layer state, see netdev_state_t
1409 * @dev_list: The global list of network devices
1410 * @napi_list: List entry, that is used for polling napi devices
1411 * @unreg_list: List entry, that is used, when we are unregistering the
1412 * device, see the function unregister_netdev
1413 * @close_list: List entry, that is used, when we are closing the device
1414 *
1415 * @adj_list: Directly linked devices, like slaves for bonding
1416 * @all_adj_list: All linked devices, *including* neighbours
1417 * @features: Currently active device features
1418 * @hw_features: User-changeable features
1419 *
1420 * @wanted_features: User-requested features
1421 * @vlan_features: Mask of features inheritable by VLAN devices
1422 *
1423 * @hw_enc_features: Mask of features inherited by encapsulating devices
1424 * This field indicates what encapsulation
1425 * offloads the hardware is capable of doing,
1426 * and drivers will need to set them appropriately.
1427 *
1428 * @mpls_features: Mask of features inheritable by MPLS
1429 *
1430 * @ifindex: interface index
Nicolas Dichtel388069d2015-04-07 12:10:16 +02001431 * @group: The group, that the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001432 *
1433 * @stats: Statistics struct, which was left as a legacy, use
1434 * rtnl_link_stats64 instead
1435 *
1436 * @rx_dropped: Dropped packets by core network,
1437 * do not use this in drivers
1438 * @tx_dropped: Dropped packets by core network,
1439 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001440 * @rx_nohandler: nohandler dropped packets by core network on
1441 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001442 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001443 * @wireless_handlers: List of functions to handle Wireless Extensions,
1444 * instead of ioctl,
1445 * see <net/iw_handler.h> for details.
1446 * @wireless_data: Instance data managed by the core of wireless extensions
1447 *
1448 * @netdev_ops: Includes several pointers to callbacks,
1449 * if one wants to override the ndo_*() functions
1450 * @ethtool_ops: Management operations
Eric W. Biedermand4760592015-03-02 00:11:09 -06001451 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001452 * of Layer 2 headers.
1453 *
1454 * @flags: Interface flags (a la BSD)
1455 * @priv_flags: Like 'flags' but invisible to userspace,
1456 * see if.h for the definitions
1457 * @gflags: Global flags ( kept as legacy )
1458 * @padded: How much padding added by alloc_netdev()
1459 * @operstate: RFC2863 operstate
1460 * @link_mode: Mapping policy to operstate
1461 * @if_port: Selectable AUI, TP, ...
1462 * @dma: DMA channel
1463 * @mtu: Interface MTU value
1464 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001465 * @hard_header_len: Maximum hardware header length.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001466 *
1467 * @needed_headroom: Extra headroom the hardware may need, but not in all
1468 * cases can this be guaranteed
1469 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1470 * cases can this be guaranteed. Some cases also use
1471 * LL_MAX_HEADER instead to allocate the skb
1472 *
1473 * interface address info:
1474 *
1475 * @perm_addr: Permanent hw address
1476 * @addr_assign_type: Hw address assignment type
1477 * @addr_len: Hardware address length
1478 * @neigh_priv_len; Used in neigh_alloc(),
1479 * initialized only in atm/clip.c
1480 * @dev_id: Used to differentiate devices that share
1481 * the same link layer address
1482 * @dev_port: Used to differentiate devices that share
1483 * the same function
1484 * @addr_list_lock: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001485 * @uc_promisc: Counter, that indicates, that promiscuous mode
1486 * has been enabled due to the need to listen to
1487 * additional unicast addresses in a device that
1488 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001489 * @uc: unicast mac addresses
1490 * @mc: multicast mac addresses
1491 * @dev_addrs: list of device hw addresses
1492 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Karoly Kemeny536721b2014-07-30 20:27:36 +02001493 * @promiscuity: Number of times, the NIC is told to work in
1494 * Promiscuous mode, if it becomes 0 the NIC will
1495 * exit from working in Promiscuous mode
1496 * @allmulti: Counter, enables or disables allmulticast mode
1497 *
1498 * @vlan_info: VLAN info
1499 * @dsa_ptr: dsa specific data
1500 * @tipc_ptr: TIPC specific data
1501 * @atalk_ptr: AppleTalk link
1502 * @ip_ptr: IPv4 specific data
1503 * @dn_ptr: DECnet specific data
1504 * @ip6_ptr: IPv6 specific data
1505 * @ax25_ptr: AX.25 specific data
1506 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1507 *
1508 * @last_rx: Time of last Rx
1509 * @dev_addr: Hw address (before bcast,
1510 * because most packets are unicast)
1511 *
1512 * @_rx: Array of RX queues
1513 * @num_rx_queues: Number of RX queues
1514 * allocated at register_netdev() time
1515 * @real_num_rx_queues: Number of RX queues currently active in device
1516 *
1517 * @rx_handler: handler for received packets
1518 * @rx_handler_data: XXX: need comments on this one
1519 * @ingress_queue: XXX: need comments on this one
1520 * @broadcast: hw bcast address
1521 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001522 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1523 * indexed by RX queue number. Assigned by driver.
1524 * This must only be set if the ndo_rx_flow_steer
1525 * operation is defined
1526 * @index_hlist: Device index hash chain
1527 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001528 * @_tx: Array of TX queues
1529 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1530 * @real_num_tx_queues: Number of TX queues currently active in device
1531 * @qdisc: Root qdisc from userspace point of view
1532 * @tx_queue_len: Max frames per queue allowed
1533 * @tx_global_lock: XXX: need comments on this one
1534 *
1535 * @xps_maps: XXX: need comments on this one
1536 *
Scott Feldman0c4f6912015-07-18 18:24:48 -07001537 * @offload_fwd_mark: Offload device fwding mark
1538 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001539 * @trans_start: Time (in jiffies) of last Tx
1540 * @watchdog_timeo: Represents the timeout that is used by
1541 * the watchdog ( see dev_watchdog() )
1542 * @watchdog_timer: List of timers
1543 *
1544 * @pcpu_refcnt: Number of references to this device
1545 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001546 * @link_watch_list: XXX: need comments on this one
1547 *
1548 * @reg_state: Register/unregister state machine
1549 * @dismantle: Device is going to be freed
1550 * @rtnl_link_state: This enum represents the phases of creating
1551 * a new link
1552 *
1553 * @destructor: Called from unregister,
1554 * can be used to call free_netdev
1555 * @npinfo: XXX: need comments on this one
1556 * @nd_net: Network namespace this network device is inside
1557 *
1558 * @ml_priv: Mid-layer private
1559 * @lstats: Loopback statistics
1560 * @tstats: Tunnel statistics
1561 * @dstats: Dummy statistics
1562 * @vstats: Virtual ethernet statistics
1563 *
1564 * @garp_port: GARP
1565 * @mrp_port: MRP
1566 *
1567 * @dev: Class/net/name entry
1568 * @sysfs_groups: Space for optional device, statistics and wireless
1569 * sysfs groups
1570 *
1571 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1572 * @rtnl_link_ops: Rtnl_link_ops
1573 *
1574 * @gso_max_size: Maximum size of generic segmentation offload
1575 * @gso_max_segs: Maximum number of segments that can be passed to the
1576 * NIC for GSO
Eric Dumazetfcbeb972014-10-05 10:11:27 -07001577 * @gso_min_segs: Minimum number of segments that can be passed to the
1578 * NIC for GSO
Karoly Kemeny536721b2014-07-30 20:27:36 +02001579 *
1580 * @dcbnl_ops: Data Center Bridging netlink ops
1581 * @num_tc: Number of traffic classes in the net device
1582 * @tc_to_txq: XXX: need comments on this one
1583 * @prio_tc_map XXX: need comments on this one
1584 *
1585 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1586 *
1587 * @priomap: XXX: need comments on this one
1588 * @phydev: Physical device may attach itself
1589 * for hardware timestamping
1590 *
1591 * @qdisc_tx_busylock: XXX: need comments on this one
1592 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001593 * @proto_down: protocol port state information can be sent to the
1594 * switch driver and used to set the phys state of the
1595 * switch port.
1596 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 * FIXME: cleanup struct net_device such that network protocol info
1598 * moves out.
1599 */
1600
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001601struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001603 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001604 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 /*
1606 * I/O specific fields
1607 * FIXME: Merge these and struct ifmap into one
1608 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001609 unsigned long mem_end;
1610 unsigned long mem_start;
1611 unsigned long base_addr;
1612 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001614 atomic_t carrier_changes;
1615
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001617 * Some hardware also needs these fields (state,dev_list,
1618 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 * part of the usual set specified in Space.c.
1620 */
1621
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 unsigned long state;
1623
Pavel Emelianov7562f872007-05-03 15:13:45 -07001624 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001625 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001626 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001627 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001628 struct list_head ptype_all;
1629 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001630
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001631 struct {
1632 struct list_head upper;
1633 struct list_head lower;
1634 } adj_list;
1635
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001636 struct {
1637 struct list_head upper;
1638 struct list_head lower;
1639 } all_adj_list;
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001640
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001641 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001642 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001643 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001644 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001645 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001646 netdev_features_t mpls_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001647
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001649 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650
Rusty Russellc45d2862007-03-28 14:29:08 -07001651 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001652
Eric Dumazet015f0682014-03-27 08:45:56 -07001653 atomic_long_t rx_dropped;
1654 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001655 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656
Johannes Bergb86e0282007-04-26 20:48:23 -07001657#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 const struct iw_handler_def * wireless_handlers;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 struct iw_public_data * wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001660#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001661 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001662 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001663#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001664 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001665#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001666#ifdef CONFIG_NET_L3_MASTER_DEV
1667 const struct l3mdev_ops *l3mdev_ops;
1668#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001670 const struct header_ops *header_ops;
1671
Karoly Kemeny536721b2014-07-30 20:27:36 +02001672 unsigned int flags;
1673 unsigned int priv_flags;
1674
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001676 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
Karoly Kemeny536721b2014-07-30 20:27:36 +02001678 unsigned char operstate;
1679 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001680
Karoly Kemeny536721b2014-07-30 20:27:36 +02001681 unsigned char if_port;
1682 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001683
Karoly Kemeny536721b2014-07-30 20:27:36 +02001684 unsigned int mtu;
1685 unsigned short type;
1686 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
Johannes Bergf5184d22008-05-12 20:48:31 -07001688 unsigned short needed_headroom;
1689 unsigned short needed_tailroom;
1690
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001692 unsigned char perm_addr[MAX_ADDR_LEN];
1693 unsigned char addr_assign_type;
1694 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001695 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001696 unsigned short dev_id;
1697 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001698 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001699 unsigned char name_assign_type;
1700 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001701 struct netdev_hw_addr_list uc;
1702 struct netdev_hw_addr_list mc;
1703 struct netdev_hw_addr_list dev_addrs;
1704
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001705#ifdef CONFIG_SYSFS
1706 struct kset *queues_kset;
1707#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001708 unsigned int promiscuity;
1709 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711
1712 /* Protocol specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001713
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001714#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001715 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001716#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001717#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001718 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001719#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001720#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001721 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001722#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001723 void *atalk_ptr;
1724 struct in_device __rcu *ip_ptr;
1725 struct dn_dev __rcu *dn_ptr;
1726 struct inet6_dev __rcu *ip6_ptr;
1727 void *ax25_ptr;
1728 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001729 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001730#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1731 struct mpls_dev __rcu *mpls_ptr;
1732#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001734/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001735 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001736 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001737 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001738
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001739 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001740 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001741
Tom Herbert0a9627f2010-03-16 08:03:29 +00001742
Michael Daltona953be52014-01-16 22:23:28 -08001743#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001744 struct netdev_rx_queue *_rx;
1745
Tom Herbert0a9627f2010-03-16 08:03:29 +00001746 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001747 unsigned int real_num_rx_queues;
Ben Hutchingsc4454772011-01-19 11:03:53 +00001748
Eric Dumazetdf334542010-03-24 19:13:54 +00001749#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001750
Eric Dumazet3b47d302014-11-06 21:09:44 -08001751 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001752 rx_handler_func_t __rcu *rx_handler;
1753 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001754
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001755#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001756 struct tcf_proto __rcu *ingress_cl_list;
1757#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001758 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001759#ifdef CONFIG_NETFILTER_INGRESS
1760 struct list_head nf_hooks_ingress;
1761#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001762
Karoly Kemeny536721b2014-07-30 20:27:36 +02001763 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001764#ifdef CONFIG_RFS_ACCEL
1765 struct cpu_rmap *rx_cpu_rmap;
1766#endif
1767 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001768
1769/*
1770 * Cache lines mostly used on transmit path
1771 */
David S. Millere8a04642008-07-17 00:34:19 -07001772 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1773 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001774 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001775 struct Qdisc *qdisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001776 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001777 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001778 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001779
Tom Herbertbf264142010-11-26 08:36:09 +00001780#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001781 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001782#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001783#ifdef CONFIG_NET_CLS_ACT
1784 struct tcf_proto __rcu *egress_cl_list;
1785#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001786#ifdef CONFIG_NET_SWITCHDEV
1787 u32 offload_fwd_mark;
1788#endif
1789
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001790 /* These may be needed for future network-power-down code. */
Eric Dumazet9d214932009-05-17 20:55:16 -07001791
1792 /*
1793 * trans_start here is expensive for high speed devices on SMP,
1794 * please use netdev_queue->trans_start instead.
1795 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001796 unsigned long trans_start;
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001797
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001798 struct timer_list watchdog_timer;
1799
Eric Dumazet29b44332010-10-11 10:22:12 +00001800 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802
Eric Dumazete014deb2009-11-17 05:59:21 +00001803 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001804
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001806 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 NETREG_UNREGISTERING, /* called unregister_netdevice */
1808 NETREG_UNREGISTERED, /* completed unregister todo */
1809 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001810 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001811 } reg_state:8;
1812
Karoly Kemeny536721b2014-07-30 20:27:36 +02001813 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001814
1815 enum {
1816 RTNL_LINK_INITIALIZED,
1817 RTNL_LINK_INITIALIZING,
1818 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001820 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001823 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824#endif
David S. Millereae792b2008-07-15 03:03:33 -07001825
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001826 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001827
David S. Miller49517042008-05-12 03:29:11 -07001828 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001829 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001830 void *ml_priv;
1831 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001832 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001833 struct pcpu_dstats __percpu *dstats;
1834 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001835 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001836
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001837 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001838 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839
Karoly Kemeny536721b2014-07-30 20:27:36 +02001840 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001841 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001842 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001843
Patrick McHardy38f7b872007-06-13 12:03:51 -07001844 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001845
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001846 /* for setting kernel sock attribute on TCP connection setup */
1847#define GSO_MAX_SIZE 65536
1848 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001849#define GSO_MAX_SEGS 65535
1850 u16 gso_max_segs;
Eric Dumazetfcbeb972014-10-05 10:11:27 -07001851 u16 gso_min_segs;
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001852#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001853 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001854#endif
John Fastabend4f57c082011-01-17 08:06:04 +00001855 u8 num_tc;
1856 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1857 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001858
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001859#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001860 unsigned int fcoe_ddp_xid;
1861#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001862#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001863 struct netprio_map __rcu *priomap;
1864#endif
Richard Cochranc1f19b52010-07-17 08:49:36 +00001865 struct phy_device *phydev;
Eric Dumazet23d3b8b2012-09-05 01:02:56 +00001866 struct lock_class_key *qdisc_tx_busylock;
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001867 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001869#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870
1871#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872
David S. Millere8a04642008-07-17 00:34:19 -07001873static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001874int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1875{
1876 return dev->prio_tc_map[prio & TC_BITMASK];
1877}
1878
1879static inline
1880int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1881{
1882 if (tc >= dev->num_tc)
1883 return -EINVAL;
1884
1885 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1886 return 0;
1887}
1888
1889static inline
1890void netdev_reset_tc(struct net_device *dev)
1891{
1892 dev->num_tc = 0;
1893 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1894 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1895}
1896
1897static inline
1898int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1899{
1900 if (tc >= dev->num_tc)
1901 return -EINVAL;
1902
1903 dev->tc_to_txq[tc].count = count;
1904 dev->tc_to_txq[tc].offset = offset;
1905 return 0;
1906}
1907
1908static inline
1909int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1910{
1911 if (num_tc > TC_MAX_QUEUE)
1912 return -EINVAL;
1913
1914 dev->num_tc = num_tc;
1915 return 0;
1916}
1917
1918static inline
1919int netdev_get_num_tc(struct net_device *dev)
1920{
1921 return dev->num_tc;
1922}
1923
1924static inline
David S. Millere8a04642008-07-17 00:34:19 -07001925struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1926 unsigned int index)
1927{
1928 return &dev->_tx[index];
1929}
1930
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001931static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1932 const struct sk_buff *skb)
1933{
1934 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1935}
1936
David S. Millere8a04642008-07-17 00:34:19 -07001937static inline void netdev_for_each_tx_queue(struct net_device *dev,
1938 void (*f)(struct net_device *,
1939 struct netdev_queue *,
1940 void *),
1941 void *arg)
1942{
1943 unsigned int i;
1944
1945 for (i = 0; i < dev->num_tx_queues; i++)
1946 f(dev, &dev->_tx[i], arg);
1947}
1948
Joe Perchesf629d202013-09-26 14:48:15 -07001949struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001950 struct sk_buff *skb,
1951 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001952
Paolo Abeni871b6422016-02-26 10:45:37 +01001953/* returns the headroom that the master device needs to take in account
1954 * when forwarding to this dev
1955 */
1956static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1957{
1958 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1959}
1960
1961static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1962{
1963 if (dev->netdev_ops->ndo_set_rx_headroom)
1964 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1965}
1966
1967/* set the device rx headroom to the dev's default */
1968static inline void netdev_reset_rx_headroom(struct net_device *dev)
1969{
1970 netdev_set_rx_headroom(dev, -1);
1971}
1972
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001973/*
1974 * Net namespace inlines
1975 */
1976static inline
1977struct net *dev_net(const struct net_device *dev)
1978{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001979 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001980}
1981
1982static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001983void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001984{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001985 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001986}
1987
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07001988static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001989{
Alexander Duyck3fc88672014-09-12 23:12:46 -07001990#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001991 if (dev->dsa_ptr != NULL)
1992 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001993#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001994 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00001995}
1996
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001997/**
1998 * netdev_priv - access network device private data
1999 * @dev: network device
2000 *
2001 * Get network device private data
2002 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002003static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002005 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006}
2007
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008/* Set the sysfs physical device reference for the network logical device
2009 * if set prior to registration will cause a symlink during initialization.
2010 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002011#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012
Marcel Holtmann384912e2009-08-31 21:08:19 +00002013/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002014 * fine-grained identification of different network device types. For
Marcel Holtmann384912e2009-08-31 21:08:19 +00002015 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
2016 */
2017#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2018
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002019/* Default NAPI poll() weight
2020 * Device drivers are strongly advised to not use bigger value
2021 */
2022#define NAPI_POLL_WEIGHT 64
2023
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002024/**
2025 * netif_napi_add - initialize a napi context
2026 * @dev: network device
2027 * @napi: napi context
2028 * @poll: polling function
2029 * @weight: default weight
2030 *
2031 * netif_napi_add() must be used to initialize a napi context prior to calling
2032 * *any* of the other napi related functions.
2033 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002034void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2035 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002036
Alexander Duyckd8156532008-07-08 15:13:05 -07002037/**
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002038 * netif_tx_napi_add - initialize a napi context
2039 * @dev: network device
2040 * @napi: napi context
2041 * @poll: polling function
2042 * @weight: default weight
2043 *
2044 * This variant of netif_napi_add() should be used from drivers using NAPI
2045 * to exclusively poll a TX queue.
2046 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2047 */
2048static inline void netif_tx_napi_add(struct net_device *dev,
2049 struct napi_struct *napi,
2050 int (*poll)(struct napi_struct *, int),
2051 int weight)
2052{
2053 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2054 netif_napi_add(dev, napi, poll, weight);
2055}
2056
2057/**
Alexander Duyckd8156532008-07-08 15:13:05 -07002058 * netif_napi_del - remove a napi context
2059 * @napi: napi context
2060 *
2061 * netif_napi_del() removes a napi context from the network device napi list
2062 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002063void netif_napi_del(struct napi_struct *napi);
2064
2065struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002066 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2067 void *frag0;
2068
Herbert Xu74895942009-05-26 18:50:27 +00002069 /* Length of frag0. */
2070 unsigned int frag0_len;
2071
Herbert Xu86911732009-01-29 14:19:50 +00002072 /* This indicates where we are processing relative to skb->data. */
2073 int data_offset;
2074
Herbert Xud565b0a2008-12-15 23:38:52 -08002075 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002076 u16 flush;
2077
2078 /* Save the IP ID here and check when we get to the transport layer */
2079 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002080
2081 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002082 u16 count;
2083
Tom Herbert15e23962015-02-10 16:30:31 -08002084 /* Start offset for remote checksum offload */
2085 u16 gro_remcsum_start;
2086
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002087 /* jiffies when first packet was created/queued */
2088 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002089
Tom Herbertafe93322014-09-17 12:25:57 -07002090 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002091 u16 proto;
2092
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002093 /* This is non-zero if the packet may be of the same flow. */
2094 u8 same_flow:1;
2095
Or Gerlitzb582ef02014-01-20 13:59:19 +02002096 /* Used in udp_gro_receive */
Tom Herbert573e8fc2014-08-22 13:33:47 -07002097 u8 udp_mark:1;
2098
2099 /* GRO checksum is valid */
2100 u8 csum_valid:1;
2101
Tom Herbert662880f2014-08-27 21:26:56 -07002102 /* Number of checksums via CHECKSUM_UNNECESSARY */
2103 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002104
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002105 /* Free the skb? */
2106 u8 free:2;
2107#define NAPI_GRO_FREE 1
2108#define NAPI_GRO_FREE_STOLEN_HEAD 2
2109
Tom Herbertefc98d02014-10-03 15:48:08 -07002110 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2111 u8 is_ipv6:1;
2112
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002113 /* 7 bit hole */
2114
Jerry Chubf5a7552014-01-07 10:23:19 -08002115 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2116 __wsum csum;
2117
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002118 /* used in skb_gro_receive() slow path */
2119 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002120};
2121
2122#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002123
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002125 __be16 type; /* This is really htons(ether_type). */
2126 struct net_device *dev; /* NULL is wildcarded here */
2127 int (*func) (struct sk_buff *,
2128 struct net_device *,
2129 struct packet_type *,
2130 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002131 bool (*id_match)(struct packet_type *ptype,
2132 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133 void *af_packet_priv;
2134 struct list_head list;
2135};
2136
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002137struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2139 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002141 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002142 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002143};
2144
2145struct packet_offload {
2146 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002147 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002148 struct offload_callbacks callbacks;
2149 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150};
2151
Tom Herberta2b12f32015-01-12 17:00:37 -08002152struct udp_offload;
2153
2154struct udp_offload_callbacks {
2155 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2156 struct sk_buff *skb,
2157 struct udp_offload *uoff);
2158 int (*gro_complete)(struct sk_buff *skb,
2159 int nhoff,
2160 struct udp_offload *uoff);
2161};
2162
Or Gerlitzb582ef02014-01-20 13:59:19 +02002163struct udp_offload {
2164 __be16 port;
Tom Herbertafe93322014-09-17 12:25:57 -07002165 u8 ipproto;
Tom Herberta2b12f32015-01-12 17:00:37 -08002166 struct udp_offload_callbacks callbacks;
Or Gerlitzb582ef02014-01-20 13:59:19 +02002167};
2168
Li RongQing8f849852014-01-04 13:57:59 +08002169/* often modified stats are per cpu, other are shared (netdev->stats) */
2170struct pcpu_sw_netstats {
2171 u64 rx_packets;
2172 u64 rx_bytes;
2173 u64 tx_packets;
2174 u64 tx_bytes;
2175 struct u64_stats_sync syncp;
2176};
2177
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002178#define __netdev_alloc_pcpu_stats(type, gfp) \
2179({ \
2180 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2181 if (pcpu_stats) { \
2182 int __cpu; \
2183 for_each_possible_cpu(__cpu) { \
2184 typeof(type) *stat; \
2185 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2186 u64_stats_init(&stat->syncp); \
2187 } \
2188 } \
2189 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002190})
2191
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002192#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002193 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002194
Jiri Pirko764f5e52015-12-03 12:12:12 +01002195enum netdev_lag_tx_type {
2196 NETDEV_LAG_TX_TYPE_UNKNOWN,
2197 NETDEV_LAG_TX_TYPE_RANDOM,
2198 NETDEV_LAG_TX_TYPE_BROADCAST,
2199 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2200 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2201 NETDEV_LAG_TX_TYPE_HASH,
2202};
2203
2204struct netdev_lag_upper_info {
2205 enum netdev_lag_tx_type tx_type;
2206};
2207
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002208struct netdev_lag_lower_state_info {
2209 u8 link_up : 1,
2210 tx_enabled : 1;
2211};
2212
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213#include <linux/notifier.h>
2214
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002215/* netdevice notifier chain. Please remember to update the rtnetlink
2216 * notification exclusion list in rtnetlink_event() when adding new
2217 * types.
2218 */
2219#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2220#define NETDEV_DOWN 0x0002
2221#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2222 detected a hardware crash and restarted
2223 - we can use this eg to kick tcp sessions
2224 once done */
2225#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2226#define NETDEV_REGISTER 0x0005
2227#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002228#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002229#define NETDEV_CHANGEADDR 0x0008
2230#define NETDEV_GOING_DOWN 0x0009
2231#define NETDEV_CHANGENAME 0x000A
2232#define NETDEV_FEAT_CHANGE 0x000B
2233#define NETDEV_BONDING_FAILOVER 0x000C
2234#define NETDEV_PRE_UP 0x000D
2235#define NETDEV_PRE_TYPE_CHANGE 0x000E
2236#define NETDEV_POST_TYPE_CHANGE 0x000F
2237#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002238#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002239#define NETDEV_RELEASE 0x0012
2240#define NETDEV_NOTIFY_PEERS 0x0013
2241#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002242#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002243#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002244#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002245#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002246#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002247#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002248#define NETDEV_CHANGELOWERSTATE 0x001B
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002249
Joe Perchesf629d202013-09-26 14:48:15 -07002250int register_netdevice_notifier(struct notifier_block *nb);
2251int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002252
2253struct netdev_notifier_info {
2254 struct net_device *dev;
2255};
2256
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002257struct netdev_notifier_change_info {
2258 struct netdev_notifier_info info; /* must be first */
2259 unsigned int flags_changed;
2260};
2261
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002262struct netdev_notifier_changeupper_info {
2263 struct netdev_notifier_info info; /* must be first */
2264 struct net_device *upper_dev; /* new upper dev */
2265 bool master; /* is upper dev master */
2266 bool linking; /* is the nofication for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002267 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002268};
2269
Jiri Pirko04d48262015-12-03 12:12:15 +01002270struct netdev_notifier_changelowerstate_info {
2271 struct netdev_notifier_info info; /* must be first */
2272 void *lower_state_info; /* is lower dev state */
2273};
2274
Cong Wang75538c22013-05-29 11:30:50 +08002275static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2276 struct net_device *dev)
2277{
2278 info->dev = dev;
2279}
2280
Jiri Pirko351638e2013-05-28 01:30:21 +00002281static inline struct net_device *
2282netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2283{
2284 return info->dev;
2285}
2286
Joe Perchesf629d202013-09-26 14:48:15 -07002287int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002288
2289
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290extern rwlock_t dev_base_lock; /* Device list lock */
2291
Eric W. Biederman881d9662007-09-17 11:56:21 -07002292#define for_each_netdev(net, d) \
2293 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002294#define for_each_netdev_reverse(net, d) \
2295 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002296#define for_each_netdev_rcu(net, d) \
2297 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002298#define for_each_netdev_safe(net, d, n) \
2299 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2300#define for_each_netdev_continue(net, d) \
2301 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002302#define for_each_netdev_continue_rcu(net, d) \
2303 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002304#define for_each_netdev_in_bond_rcu(bond, slave) \
2305 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002306 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002307#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2308
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002309static inline struct net_device *next_net_device(struct net_device *dev)
2310{
2311 struct list_head *lh;
2312 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002313
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002314 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002315 lh = dev->dev_list.next;
2316 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2317}
2318
Eric Dumazetce81b762009-11-11 17:34:30 +00002319static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2320{
2321 struct list_head *lh;
2322 struct net *net;
2323
2324 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002325 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002326 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2327}
2328
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002329static inline struct net_device *first_net_device(struct net *net)
2330{
2331 return list_empty(&net->dev_base_head) ? NULL :
2332 net_device_entry(net->dev_base_head.next);
2333}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002334
Eric Dumazetccf43432011-01-26 18:08:02 +00002335static inline struct net_device *first_net_device_rcu(struct net *net)
2336{
2337 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2338
2339 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2340}
2341
Joe Perchesf629d202013-09-26 14:48:15 -07002342int netdev_boot_setup_check(struct net_device *dev);
2343unsigned long netdev_boot_base(const char *prefix, int unit);
2344struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2345 const char *hwaddr);
2346struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2347struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2348void dev_add_pack(struct packet_type *pt);
2349void dev_remove_pack(struct packet_type *pt);
2350void __dev_remove_pack(struct packet_type *pt);
2351void dev_add_offload(struct packet_offload *po);
2352void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002354int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002355int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002356struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2357 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002358struct net_device *dev_get_by_name(struct net *net, const char *name);
2359struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2360struct net_device *__dev_get_by_name(struct net *net, const char *name);
2361int dev_alloc_name(struct net_device *dev, const char *name);
2362int dev_open(struct net_device *dev);
2363int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002364int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002365void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002366int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002367int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002368int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002369int register_netdevice(struct net_device *dev);
2370void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2371void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002372static inline void unregister_netdevice(struct net_device *dev)
2373{
2374 unregister_netdevice_queue(dev, NULL);
2375}
2376
Joe Perchesf629d202013-09-26 14:48:15 -07002377int netdev_refcnt_read(const struct net_device *dev);
2378void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002379void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002380void synchronize_net(void);
2381int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002382
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002383DECLARE_PER_CPU(int, xmit_recursion);
2384static inline int dev_recursion_level(void)
2385{
2386 return this_cpu_read(xmit_recursion);
2387}
2388
Joe Perchesf629d202013-09-26 14:48:15 -07002389struct net_device *dev_get_by_index(struct net *net, int ifindex);
2390struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2391struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2392int netdev_get_name(struct net *net, char *name, int ifindex);
2393int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002394int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002395
2396static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2397{
2398 return NAPI_GRO_CB(skb)->data_offset;
2399}
2400
2401static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2402{
2403 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2404}
2405
2406static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2407{
2408 NAPI_GRO_CB(skb)->data_offset += len;
2409}
2410
Herbert Xua5b1cf22009-05-26 18:50:28 +00002411static inline void *skb_gro_header_fast(struct sk_buff *skb,
2412 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002413{
Herbert Xu78a478d2009-05-26 18:50:21 +00002414 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002415}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002416
Herbert Xua5b1cf22009-05-26 18:50:28 +00002417static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2418{
2419 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2420}
2421
2422static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2423 unsigned int offset)
2424{
Herbert Xu17dd7592011-07-27 06:16:28 -07002425 if (!pskb_may_pull(skb, hlen))
2426 return NULL;
2427
Herbert Xua5b1cf22009-05-26 18:50:28 +00002428 NAPI_GRO_CB(skb)->frag0 = NULL;
2429 NAPI_GRO_CB(skb)->frag0_len = 0;
Herbert Xu17dd7592011-07-27 06:16:28 -07002430 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002431}
2432
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002433static inline void *skb_gro_network_header(struct sk_buff *skb)
2434{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002435 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2436 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002437}
2438
Jerry Chubf5a7552014-01-07 10:23:19 -08002439static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2440 const void *start, unsigned int len)
2441{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002442 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002443 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2444 csum_partial(start, len, 0));
2445}
2446
Tom Herbert573e8fc2014-08-22 13:33:47 -07002447/* GRO checksum functions. These are logical equivalents of the normal
2448 * checksum functions (in skbuff.h) except that they operate on the GRO
2449 * offsets and fields in sk_buff.
2450 */
2451
2452__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2453
Tom Herbert15e23962015-02-10 16:30:31 -08002454static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2455{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002456 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002457}
2458
Tom Herbert573e8fc2014-08-22 13:33:47 -07002459static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2460 bool zero_okay,
2461 __sum16 check)
2462{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002463 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2464 skb_checksum_start_offset(skb) <
2465 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002466 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002467 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002468 (!zero_okay || check));
2469}
2470
2471static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2472 __wsum psum)
2473{
2474 if (NAPI_GRO_CB(skb)->csum_valid &&
2475 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2476 return 0;
2477
2478 NAPI_GRO_CB(skb)->csum = psum;
2479
2480 return __skb_gro_checksum_complete(skb);
2481}
2482
Tom Herbert573e8fc2014-08-22 13:33:47 -07002483static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2484{
Tom Herbert662880f2014-08-27 21:26:56 -07002485 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2486 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2487 NAPI_GRO_CB(skb)->csum_cnt--;
2488 } else {
2489 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2490 * verified a new top level checksum or an encapsulated one
2491 * during GRO. This saves work if we fallback to normal path.
2492 */
2493 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002494 }
2495}
2496
2497#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2498 compute_pseudo) \
2499({ \
2500 __sum16 __ret = 0; \
2501 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2502 __ret = __skb_gro_checksum_validate_complete(skb, \
2503 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002504 if (__ret) \
2505 __skb_mark_checksum_bad(skb); \
2506 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002507 skb_gro_incr_csum_unnecessary(skb); \
2508 __ret; \
2509})
2510
2511#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2512 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2513
2514#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2515 compute_pseudo) \
2516 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2517
2518#define skb_gro_checksum_simple_validate(skb) \
2519 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2520
Tom Herbertd96535a2014-08-31 15:12:42 -07002521static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2522{
2523 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2524 !NAPI_GRO_CB(skb)->csum_valid);
2525}
2526
2527static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2528 __sum16 check, __wsum pseudo)
2529{
2530 NAPI_GRO_CB(skb)->csum = ~pseudo;
2531 NAPI_GRO_CB(skb)->csum_valid = 1;
2532}
2533
2534#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2535do { \
2536 if (__skb_gro_checksum_convert_check(skb)) \
2537 __skb_gro_checksum_convert(skb, check, \
2538 compute_pseudo(skb, proto)); \
2539} while (0)
2540
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002541struct gro_remcsum {
2542 int offset;
2543 __wsum delta;
2544};
2545
2546static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2547{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002548 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002549 grc->delta = 0;
2550}
2551
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002552static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2553 unsigned int off, size_t hdrlen,
2554 int start, int offset,
2555 struct gro_remcsum *grc,
2556 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002557{
2558 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002559 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002560
2561 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2562
Tom Herbert15e23962015-02-10 16:30:31 -08002563 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002564 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2565 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002566 }
2567
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002568 ptr = skb_gro_header_fast(skb, off);
2569 if (skb_gro_header_hard(skb, off + plen)) {
2570 ptr = skb_gro_header_slow(skb, off + plen, off);
2571 if (!ptr)
2572 return NULL;
2573 }
2574
2575 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2576 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002577
2578 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002579 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002580
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002581 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002582 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002583
2584 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002585}
2586
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002587static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2588 struct gro_remcsum *grc)
2589{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002590 void *ptr;
2591 size_t plen = grc->offset + sizeof(u16);
2592
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002593 if (!grc->delta)
2594 return;
2595
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002596 ptr = skb_gro_header_fast(skb, grc->offset);
2597 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2598 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2599 if (!ptr)
2600 return;
2601 }
2602
2603 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002604}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002605
Tom Herbert6ae23ad2015-12-14 11:19:46 -08002606struct skb_csum_offl_spec {
2607 __u16 ipv4_okay:1,
2608 ipv6_okay:1,
2609 encap_okay:1,
2610 ip_options_okay:1,
2611 ext_hdrs_okay:1,
2612 tcp_okay:1,
2613 udp_okay:1,
2614 sctp_okay:1,
2615 vlan_okay:1,
2616 no_encapped_ipv6:1,
2617 no_not_encapped:1;
2618};
2619
2620bool __skb_csum_offload_chk(struct sk_buff *skb,
2621 const struct skb_csum_offl_spec *spec,
2622 bool *csum_encapped,
2623 bool csum_help);
2624
2625static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2626 const struct skb_csum_offl_spec *spec,
2627 bool *csum_encapped,
2628 bool csum_help)
2629{
2630 if (skb->ip_summed != CHECKSUM_PARTIAL)
2631 return false;
2632
2633 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2634}
2635
2636static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2637 const struct skb_csum_offl_spec *spec)
2638{
2639 bool csum_encapped;
2640
2641 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2642}
2643
2644static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2645{
2646 static const struct skb_csum_offl_spec csum_offl_spec = {
2647 .ipv4_okay = 1,
2648 .ip_options_okay = 1,
2649 .ipv6_okay = 1,
2650 .vlan_okay = 1,
2651 .tcp_okay = 1,
2652 .udp_okay = 1,
2653 };
2654
2655 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2656}
2657
2658static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2659{
2660 static const struct skb_csum_offl_spec csum_offl_spec = {
2661 .ipv4_okay = 1,
2662 .ip_options_okay = 1,
2663 .tcp_okay = 1,
2664 .udp_okay = 1,
2665 .vlan_okay = 1,
2666 };
2667
2668 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2669}
2670
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002671static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2672 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002673 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002674 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002675{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002676 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002677 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002678
2679 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002680}
2681
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002682static inline int dev_parse_header(const struct sk_buff *skb,
2683 unsigned char *haddr)
2684{
2685 const struct net_device *dev = skb->dev;
2686
Patrick McHardy1b833362007-10-18 05:09:28 -07002687 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002688 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002689 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002690}
2691
Willem de Bruijn2793a232016-03-09 21:58:32 -05002692/* ll_header must have at least hard_header_len allocated */
2693static inline bool dev_validate_header(const struct net_device *dev,
2694 char *ll_header, int len)
2695{
2696 if (likely(len >= dev->hard_header_len))
2697 return true;
2698
2699 if (capable(CAP_SYS_RAWIO)) {
2700 memset(ll_header + len, 0, dev->hard_header_len - len);
2701 return true;
2702 }
2703
2704 if (dev->header_ops && dev->header_ops->validate)
2705 return dev->header_ops->validate(ll_header, len);
2706
2707 return false;
2708}
2709
Linus Torvalds1da177e2005-04-16 15:20:36 -07002710typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002711int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002712static inline int unregister_gifconf(unsigned int family)
2713{
2714 return register_gifconf(family, NULL);
2715}
2716
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002717#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002718#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002719struct sd_flow_limit {
2720 u64 count;
2721 unsigned int num_buckets;
2722 unsigned int history_head;
2723 u16 history[FLOW_LIMIT_HISTORY];
2724 u8 buckets[];
2725};
2726
2727extern int netdev_flow_limit_table_len;
2728#endif /* CONFIG_NET_FLOW_LIMIT */
2729
Linus Torvalds1da177e2005-04-16 15:20:36 -07002730/*
Eric Dumazet88751272010-04-19 05:07:33 +00002731 * Incoming packets are placed on per-cpu queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002733struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002734 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002735 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002736
Changli Gaodee42872010-05-02 05:42:16 +00002737 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002738 unsigned int processed;
2739 unsigned int time_squeeze;
2740 unsigned int cpu_collision;
2741 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002742#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002743 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002744#endif
2745#ifdef CONFIG_NET_FLOW_LIMIT
2746 struct sd_flow_limit __rcu *flow_limit;
2747#endif
2748 struct Qdisc *output_queue;
2749 struct Qdisc **output_queue_tailp;
2750 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002751
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002752#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002753 /* Elements below can be accessed between CPUs for RPS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002754 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002755 struct softnet_data *rps_ipi_next;
2756 unsigned int cpu;
Tom Herbertfec5e652010-04-16 16:01:27 -07002757 unsigned int input_queue_head;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002758 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002759#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002760 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002761 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002762 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002763
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764};
2765
Tom Herbert76cc8b12010-05-20 18:37:59 +00002766static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002767{
2768#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002769 sd->input_queue_head++;
2770#endif
2771}
2772
2773static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2774 unsigned int *qtail)
2775{
2776#ifdef CONFIG_RPS
2777 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002778#endif
2779}
2780
Tom Herbert0a9627f2010-03-16 08:03:29 +00002781DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002782
Joe Perchesf629d202013-09-26 14:48:15 -07002783void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002784void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002785
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002786static inline void netif_tx_schedule_all(struct net_device *dev)
2787{
2788 unsigned int i;
2789
2790 for (i = 0; i < dev->num_tx_queues; i++)
2791 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2792}
2793
Dave Jonesd29f7492008-07-22 14:09:06 -07002794static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2795{
Tom Herbert734664982011-11-28 16:32:44 +00002796 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002797}
2798
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002799/**
2800 * netif_start_queue - allow transmit
2801 * @dev: network device
2802 *
2803 * Allow upper layers to call the device hard_start_xmit routine.
2804 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002805static inline void netif_start_queue(struct net_device *dev)
2806{
David S. Millere8a04642008-07-17 00:34:19 -07002807 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002808}
2809
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002810static inline void netif_tx_start_all_queues(struct net_device *dev)
2811{
2812 unsigned int i;
2813
2814 for (i = 0; i < dev->num_tx_queues; i++) {
2815 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2816 netif_tx_start_queue(txq);
2817 }
2818}
2819
John Fastabend46e5da40a2014-09-12 20:04:52 -07002820void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002821
Dave Jonesd29f7492008-07-22 14:09:06 -07002822/**
2823 * netif_wake_queue - restart transmit
2824 * @dev: network device
2825 *
2826 * Allow upper layers to call the device hard_start_xmit routine.
2827 * Used for flow control when transmit resources are available.
2828 */
David S. Miller79d16382008-07-08 23:14:46 -07002829static inline void netif_wake_queue(struct net_device *dev)
2830{
David S. Millere8a04642008-07-17 00:34:19 -07002831 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002832}
2833
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002834static inline void netif_tx_wake_all_queues(struct net_device *dev)
2835{
2836 unsigned int i;
2837
2838 for (i = 0; i < dev->num_tx_queues; i++) {
2839 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2840 netif_tx_wake_queue(txq);
2841 }
2842}
2843
Dave Jonesd29f7492008-07-22 14:09:06 -07002844static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2845{
Tom Herbert734664982011-11-28 16:32:44 +00002846 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002847}
2848
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002849/**
2850 * netif_stop_queue - stop transmitted packets
2851 * @dev: network device
2852 *
2853 * Stop upper layers calling the device hard_start_xmit routine.
2854 * Used for flow control when transmit resources are unavailable.
2855 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002856static inline void netif_stop_queue(struct net_device *dev)
2857{
David S. Millere8a04642008-07-17 00:34:19 -07002858 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859}
2860
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002861void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002862
David S. Miller4d295152012-03-07 21:02:35 -05002863static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002864{
Tom Herbert734664982011-11-28 16:32:44 +00002865 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002866}
2867
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002868/**
2869 * netif_queue_stopped - test if transmit queue is flowblocked
2870 * @dev: network device
2871 *
2872 * Test if transmit queue on device is currently unable to send.
2873 */
David S. Miller4d295152012-03-07 21:02:35 -05002874static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002875{
David S. Millere8a04642008-07-17 00:34:19 -07002876 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002877}
2878
David S. Miller4d295152012-03-07 21:02:35 -05002879static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002880{
Tom Herbert734664982011-11-28 16:32:44 +00002881 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2882}
2883
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002884static inline bool
2885netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002886{
2887 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2888}
2889
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002890static inline bool
2891netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2892{
2893 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2894}
2895
Eric Dumazet53511452014-10-08 08:19:27 -07002896/**
2897 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2898 * @dev_queue: pointer to transmit queue
2899 *
2900 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2901 * to give appropriate hint to the cpu.
2902 */
2903static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2904{
2905#ifdef CONFIG_BQL
2906 prefetchw(&dev_queue->dql.num_queued);
2907#endif
2908}
2909
2910/**
2911 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2912 * @dev_queue: pointer to transmit queue
2913 *
2914 * BQL enabled drivers might use this helper in their TX completion path,
2915 * to give appropriate hint to the cpu.
2916 */
2917static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2918{
2919#ifdef CONFIG_BQL
2920 prefetchw(&dev_queue->dql.limit);
2921#endif
2922}
2923
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002924static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2925 unsigned int bytes)
2926{
Tom Herbert114cf582011-11-28 16:33:09 +00002927#ifdef CONFIG_BQL
2928 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002929
2930 if (likely(dql_avail(&dev_queue->dql) >= 0))
2931 return;
2932
2933 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2934
2935 /*
2936 * The XOFF flag must be set before checking the dql_avail below,
2937 * because in netdev_tx_completed_queue we update the dql_completed
2938 * before checking the XOFF flag.
2939 */
2940 smp_mb();
2941
2942 /* check again in case another CPU has just made room avail */
2943 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2944 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002945#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002946}
2947
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002948/**
2949 * netdev_sent_queue - report the number of bytes queued to hardware
2950 * @dev: network device
2951 * @bytes: number of bytes queued to the hardware device queue
2952 *
2953 * Report the number of bytes queued for sending/completion to the network
2954 * device hardware queue. @bytes should be a good approximation and should
2955 * exactly match netdev_completed_queue() @bytes
2956 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002957static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2958{
2959 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2960}
2961
2962static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002963 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002964{
Tom Herbert114cf582011-11-28 16:33:09 +00002965#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002966 if (unlikely(!bytes))
2967 return;
2968
2969 dql_completed(&dev_queue->dql, bytes);
2970
2971 /*
2972 * Without the memory barrier there is a small possiblity that
2973 * netdev_tx_sent_queue will miss the update and cause the queue to
2974 * be stopped forever
2975 */
2976 smp_mb();
2977
2978 if (dql_avail(&dev_queue->dql) < 0)
2979 return;
2980
2981 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2982 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002983#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002984}
2985
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002986/**
2987 * netdev_completed_queue - report bytes and packets completed by device
2988 * @dev: network device
2989 * @pkts: actual number of packets sent over the medium
2990 * @bytes: actual number of bytes sent over the medium
2991 *
2992 * Report the number of bytes and packets transmitted by the network device
2993 * hardware queue over the physical medium, @bytes must exactly match the
2994 * @bytes amount passed to netdev_sent_queue()
2995 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002996static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00002997 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002998{
2999 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3000}
3001
3002static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3003{
Tom Herbert114cf582011-11-28 16:33:09 +00003004#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003005 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003006 dql_reset(&q->dql);
3007#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003008}
3009
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003010/**
3011 * netdev_reset_queue - reset the packets and bytes count of a network device
3012 * @dev_queue: network device
3013 *
3014 * Reset the bytes and packet count of a network device and clear the
3015 * software flow control OFF bit for this network device
3016 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003017static inline void netdev_reset_queue(struct net_device *dev_queue)
3018{
3019 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003020}
3021
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003022/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003023 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3024 * @dev: network device
3025 * @queue_index: given tx queue index
3026 *
3027 * Returns 0 if given tx queue index >= number of device tx queues,
3028 * otherwise returns the originally passed tx queue index.
3029 */
3030static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3031{
3032 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3033 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3034 dev->name, queue_index,
3035 dev->real_num_tx_queues);
3036 return 0;
3037 }
3038
3039 return queue_index;
3040}
3041
3042/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003043 * netif_running - test if up
3044 * @dev: network device
3045 *
3046 * Test if the device has been brought up.
3047 */
David S. Miller4d295152012-03-07 21:02:35 -05003048static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003049{
3050 return test_bit(__LINK_STATE_START, &dev->state);
3051}
3052
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003053/*
3054 * Routines to manage the subqueues on a device. We only need start
3055 * stop, and a check if it's stopped. All other device management is
3056 * done at the overall netdevice level.
3057 * Also test the device if we're multiqueue.
3058 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003059
3060/**
3061 * netif_start_subqueue - allow sending packets on subqueue
3062 * @dev: network device
3063 * @queue_index: sub queue index
3064 *
3065 * Start individual transmit queue of a device with multiple transmit queues.
3066 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003067static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3068{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003069 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003070
3071 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003072}
3073
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003074/**
3075 * netif_stop_subqueue - stop sending packets on subqueue
3076 * @dev: network device
3077 * @queue_index: sub queue index
3078 *
3079 * Stop individual transmit queue of a device with multiple transmit queues.
3080 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003081static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3082{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003083 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003084 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003085}
3086
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003087/**
3088 * netif_subqueue_stopped - test status of subqueue
3089 * @dev: network device
3090 * @queue_index: sub queue index
3091 *
3092 * Check individual transmit queue of a device with multiple transmit queues.
3093 */
David S. Miller4d295152012-03-07 21:02:35 -05003094static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3095 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003096{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003097 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003098
3099 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003100}
3101
David S. Miller4d295152012-03-07 21:02:35 -05003102static inline bool netif_subqueue_stopped(const struct net_device *dev,
3103 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003104{
3105 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3106}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003107
John Fastabend46e5da40a2014-09-12 20:04:52 -07003108void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003109
Alexander Duyck537c00d2013-01-10 08:57:02 +00003110#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003111int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003112 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003113#else
3114static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003115 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003116 u16 index)
3117{
3118 return 0;
3119}
3120#endif
3121
Jiri Pirko5605c762015-05-12 14:56:12 +02003122u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3123 unsigned int num_tx_queues);
3124
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003125/*
3126 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3127 * as a distribution range limit for the returned value.
3128 */
3129static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003130 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003131{
3132 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3133}
3134
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003135/**
3136 * netif_is_multiqueue - test if device has multiple transmit queues
3137 * @dev: network device
3138 *
3139 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003140 */
David S. Miller4d295152012-03-07 21:02:35 -05003141static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003142{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003143 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003144}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003145
Joe Perchesf629d202013-09-26 14:48:15 -07003146int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003147
Michael Daltona953be52014-01-16 22:23:28 -08003148#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003149int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003150#else
3151static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3152 unsigned int rxq)
3153{
3154 return 0;
3155}
3156#endif
3157
Michael Daltona953be52014-01-16 22:23:28 -08003158#ifdef CONFIG_SYSFS
3159static inline unsigned int get_netdev_rx_queue_index(
3160 struct netdev_rx_queue *queue)
3161{
3162 struct net_device *dev = queue->dev;
3163 int index = queue - dev->_rx;
3164
3165 BUG_ON(index >= dev->num_rx_queues);
3166 return index;
3167}
3168#endif
3169
Yuval Mintz16917b82012-07-01 03:18:50 +00003170#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003171int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003172
Eric Dumazete6247022013-12-05 04:45:08 -08003173enum skb_free_reason {
3174 SKB_REASON_CONSUMED,
3175 SKB_REASON_DROPPED,
3176};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003177
Eric Dumazete6247022013-12-05 04:45:08 -08003178void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3179void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3180
3181/*
3182 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3183 * interrupt context or with hardware interrupts being disabled.
3184 * (in_irq() || irqs_disabled())
3185 *
3186 * We provide four helpers that can be used in following contexts :
3187 *
3188 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3189 * replacing kfree_skb(skb)
3190 *
3191 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3192 * Typically used in place of consume_skb(skb) in TX completion path
3193 *
3194 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3195 * replacing kfree_skb(skb)
3196 *
3197 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3198 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003199 */
Eric Dumazete6247022013-12-05 04:45:08 -08003200static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3201{
3202 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3203}
3204
3205static inline void dev_consume_skb_irq(struct sk_buff *skb)
3206{
3207 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3208}
3209
3210static inline void dev_kfree_skb_any(struct sk_buff *skb)
3211{
3212 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3213}
3214
3215static inline void dev_consume_skb_any(struct sk_buff *skb)
3216{
3217 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3218}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003219
Joe Perchesf629d202013-09-26 14:48:15 -07003220int netif_rx(struct sk_buff *skb);
3221int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003222int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003223gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3224void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3225struct sk_buff *napi_get_frags(struct napi_struct *napi);
3226gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003227struct packet_offload *gro_find_receive_by_type(__be16 type);
3228struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003229
3230static inline void napi_free_frags(struct napi_struct *napi)
3231{
3232 kfree_skb(napi->skb);
3233 napi->skb = NULL;
3234}
3235
Joe Perchesf629d202013-09-26 14:48:15 -07003236int netdev_rx_handler_register(struct net_device *dev,
3237 rx_handler_func_t *rx_handler,
3238 void *rx_handler_data);
3239void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003240
Joe Perchesf629d202013-09-26 14:48:15 -07003241bool dev_valid_name(const char *name);
3242int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3243int dev_ethtool(struct net *net, struct ifreq *);
3244unsigned int dev_get_flags(const struct net_device *);
3245int __dev_change_flags(struct net_device *, unsigned int flags);
3246int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003247void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3248 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003249int dev_change_name(struct net_device *, const char *);
3250int dev_set_alias(struct net_device *, const char *, size_t);
3251int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3252int dev_set_mtu(struct net_device *, int);
3253void dev_set_group(struct net_device *, int);
3254int dev_set_mac_address(struct net_device *, struct sockaddr *);
3255int dev_change_carrier(struct net_device *, bool new_carrier);
3256int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003257 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003258int dev_get_phys_port_name(struct net_device *dev,
3259 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003260int dev_change_proto_down(struct net_device *dev, bool proto_down);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003261struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003262struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3263 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003264int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003265int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Vlad Yasevich1ee481f2014-03-27 17:32:29 -04003266bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003267
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003268extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003269
3270/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003271void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003272
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003273/**
3274 * dev_put - release reference to device
3275 * @dev: network device
3276 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003277 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003278 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003279static inline void dev_put(struct net_device *dev)
3280{
Christoph Lameter933393f2011-12-22 11:58:51 -06003281 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003282}
3283
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003284/**
3285 * dev_hold - get reference to device
3286 * @dev: network device
3287 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003288 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003289 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003290static inline void dev_hold(struct net_device *dev)
3291{
Christoph Lameter933393f2011-12-22 11:58:51 -06003292 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003293}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003294
3295/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3296 * and _off may be called from IRQ context, but it is caller
3297 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003298 *
3299 * The name carrier is inappropriate, these functions should really be
3300 * called netif_lowerlayer_*() because they represent the state of any
3301 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003302 */
3303
Joe Perchesf629d202013-09-26 14:48:15 -07003304void linkwatch_init_dev(struct net_device *dev);
3305void linkwatch_fire_event(struct net_device *dev);
3306void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003307
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003308/**
3309 * netif_carrier_ok - test if carrier present
3310 * @dev: network device
3311 *
3312 * Check if carrier is present on device
3313 */
David S. Miller4d295152012-03-07 21:02:35 -05003314static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003315{
3316 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3317}
3318
Joe Perchesf629d202013-09-26 14:48:15 -07003319unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003320
Joe Perchesf629d202013-09-26 14:48:15 -07003321void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003322
Joe Perchesf629d202013-09-26 14:48:15 -07003323void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003324
Joe Perchesf629d202013-09-26 14:48:15 -07003325void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003326
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003327/**
3328 * netif_dormant_on - mark device as dormant.
3329 * @dev: network device
3330 *
3331 * Mark device as dormant (as per RFC2863).
3332 *
3333 * The dormant state indicates that the relevant interface is not
3334 * actually in a condition to pass packets (i.e., it is not 'up') but is
3335 * in a "pending" state, waiting for some external event. For "on-
3336 * demand" interfaces, this new state identifies the situation where the
3337 * interface is waiting for events to place it in the up state.
3338 *
3339 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003340static inline void netif_dormant_on(struct net_device *dev)
3341{
3342 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3343 linkwatch_fire_event(dev);
3344}
3345
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003346/**
3347 * netif_dormant_off - set device as not dormant.
3348 * @dev: network device
3349 *
3350 * Device is not in dormant state.
3351 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003352static inline void netif_dormant_off(struct net_device *dev)
3353{
3354 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3355 linkwatch_fire_event(dev);
3356}
3357
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003358/**
3359 * netif_dormant - test if carrier present
3360 * @dev: network device
3361 *
3362 * Check if carrier is present on device
3363 */
David S. Miller4d295152012-03-07 21:02:35 -05003364static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003365{
3366 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3367}
3368
3369
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003370/**
3371 * netif_oper_up - test if device is operational
3372 * @dev: network device
3373 *
3374 * Check if carrier is operational
3375 */
David S. Miller4d295152012-03-07 21:02:35 -05003376static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003377{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003378 return (dev->operstate == IF_OPER_UP ||
3379 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3380}
3381
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003382/**
3383 * netif_device_present - is device available or removed
3384 * @dev: network device
3385 *
3386 * Check if device has not been removed from system.
3387 */
David S. Miller4d295152012-03-07 21:02:35 -05003388static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003389{
3390 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3391}
3392
Joe Perchesf629d202013-09-26 14:48:15 -07003393void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003394
Joe Perchesf629d202013-09-26 14:48:15 -07003395void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003396
3397/*
3398 * Network interface message level settings
3399 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003400
3401enum {
3402 NETIF_MSG_DRV = 0x0001,
3403 NETIF_MSG_PROBE = 0x0002,
3404 NETIF_MSG_LINK = 0x0004,
3405 NETIF_MSG_TIMER = 0x0008,
3406 NETIF_MSG_IFDOWN = 0x0010,
3407 NETIF_MSG_IFUP = 0x0020,
3408 NETIF_MSG_RX_ERR = 0x0040,
3409 NETIF_MSG_TX_ERR = 0x0080,
3410 NETIF_MSG_TX_QUEUED = 0x0100,
3411 NETIF_MSG_INTR = 0x0200,
3412 NETIF_MSG_TX_DONE = 0x0400,
3413 NETIF_MSG_RX_STATUS = 0x0800,
3414 NETIF_MSG_PKTDATA = 0x1000,
3415 NETIF_MSG_HW = 0x2000,
3416 NETIF_MSG_WOL = 0x4000,
3417};
3418
3419#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3420#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3421#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3422#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3423#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3424#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3425#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3426#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3427#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3428#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3429#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3430#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3431#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3432#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3433#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3434
3435static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3436{
3437 /* use default */
3438 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3439 return default_msg_enable_bits;
3440 if (debug_value == 0) /* no output */
3441 return 0;
3442 /* set low N bits */
3443 return (1 << debug_value) - 1;
3444}
3445
David S. Millerc773e842008-07-08 23:13:53 -07003446static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003447{
David S. Millerc773e842008-07-08 23:13:53 -07003448 spin_lock(&txq->_xmit_lock);
3449 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003450}
3451
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003452static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3453{
3454 spin_lock_bh(&txq->_xmit_lock);
3455 txq->xmit_lock_owner = smp_processor_id();
3456}
3457
David S. Miller4d295152012-03-07 21:02:35 -05003458static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003459{
David S. Miller4d295152012-03-07 21:02:35 -05003460 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003461 if (likely(ok))
3462 txq->xmit_lock_owner = smp_processor_id();
3463 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003464}
3465
David S. Millerc773e842008-07-08 23:13:53 -07003466static inline void __netif_tx_unlock(struct netdev_queue *txq)
3467{
3468 txq->xmit_lock_owner = -1;
3469 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003470}
3471
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003472static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3473{
3474 txq->xmit_lock_owner = -1;
3475 spin_unlock_bh(&txq->_xmit_lock);
3476}
3477
Eric Dumazet08baf562009-05-25 22:58:01 -07003478static inline void txq_trans_update(struct netdev_queue *txq)
3479{
3480 if (txq->xmit_lock_owner != -1)
3481 txq->trans_start = jiffies;
3482}
3483
David S. Millerc3f26a22008-07-31 16:58:50 -07003484/**
3485 * netif_tx_lock - grab network device transmit lock
3486 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003487 *
3488 * Get network device transmit lock
3489 */
3490static inline void netif_tx_lock(struct net_device *dev)
3491{
3492 unsigned int i;
3493 int cpu;
3494
3495 spin_lock(&dev->tx_global_lock);
3496 cpu = smp_processor_id();
3497 for (i = 0; i < dev->num_tx_queues; i++) {
3498 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3499
3500 /* We are the only thread of execution doing a
3501 * freeze, but we have to grab the _xmit_lock in
3502 * order to synchronize with threads which are in
3503 * the ->hard_start_xmit() handler and already
3504 * checked the frozen bit.
3505 */
3506 __netif_tx_lock(txq, cpu);
3507 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3508 __netif_tx_unlock(txq);
3509 }
3510}
3511
3512static inline void netif_tx_lock_bh(struct net_device *dev)
3513{
3514 local_bh_disable();
3515 netif_tx_lock(dev);
3516}
3517
Herbert Xu932ff272006-06-09 12:20:56 -07003518static inline void netif_tx_unlock(struct net_device *dev)
3519{
David S. Millere8a04642008-07-17 00:34:19 -07003520 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003521
David S. Millere8a04642008-07-17 00:34:19 -07003522 for (i = 0; i < dev->num_tx_queues; i++) {
3523 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003524
David S. Millerc3f26a22008-07-31 16:58:50 -07003525 /* No need to grab the _xmit_lock here. If the
3526 * queue is not stopped for another reason, we
3527 * force a schedule.
3528 */
3529 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003530 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003531 }
3532 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003533}
3534
3535static inline void netif_tx_unlock_bh(struct net_device *dev)
3536{
David S. Millere8a04642008-07-17 00:34:19 -07003537 netif_tx_unlock(dev);
3538 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003539}
3540
David S. Millerc773e842008-07-08 23:13:53 -07003541#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003542 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003543 __netif_tx_lock(txq, cpu); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003544 } \
3545}
3546
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003547#define HARD_TX_TRYLOCK(dev, txq) \
3548 (((dev->features & NETIF_F_LLTX) == 0) ? \
3549 __netif_tx_trylock(txq) : \
3550 true )
3551
David S. Millerc773e842008-07-08 23:13:53 -07003552#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003553 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003554 __netif_tx_unlock(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003555 } \
3556}
3557
Linus Torvalds1da177e2005-04-16 15:20:36 -07003558static inline void netif_tx_disable(struct net_device *dev)
3559{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003560 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003561 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003562
David S. Millerc3f26a22008-07-31 16:58:50 -07003563 local_bh_disable();
3564 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003565 for (i = 0; i < dev->num_tx_queues; i++) {
3566 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003567
3568 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003569 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003570 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003571 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003572 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003573}
3574
David S. Millere308a5d2008-07-15 00:13:44 -07003575static inline void netif_addr_lock(struct net_device *dev)
3576{
3577 spin_lock(&dev->addr_list_lock);
3578}
3579
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003580static inline void netif_addr_lock_nested(struct net_device *dev)
3581{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003582 int subclass = SINGLE_DEPTH_NESTING;
3583
3584 if (dev->netdev_ops->ndo_get_lock_subclass)
3585 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3586
3587 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003588}
3589
David S. Millere308a5d2008-07-15 00:13:44 -07003590static inline void netif_addr_lock_bh(struct net_device *dev)
3591{
3592 spin_lock_bh(&dev->addr_list_lock);
3593}
3594
3595static inline void netif_addr_unlock(struct net_device *dev)
3596{
3597 spin_unlock(&dev->addr_list_lock);
3598}
3599
3600static inline void netif_addr_unlock_bh(struct net_device *dev)
3601{
3602 spin_unlock_bh(&dev->addr_list_lock);
3603}
3604
Jiri Pirkof001fde2009-05-05 02:48:28 +00003605/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003606 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003607 * rcu_read_lock held.
3608 */
3609#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003610 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003611
Linus Torvalds1da177e2005-04-16 15:20:36 -07003612/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3613
Joe Perchesf629d202013-09-26 14:48:15 -07003614void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003615
3616/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003617struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003618 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003619 void (*setup)(struct net_device *),
3620 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003621#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3622 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003623
Tom Gundersenc835a672014-07-14 16:37:24 +02003624#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3625 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3626 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003627
Joe Perchesf629d202013-09-26 14:48:15 -07003628int register_netdev(struct net_device *dev);
3629void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003630
Jiri Pirko22bedad32010-04-01 21:22:57 +00003631/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003632int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3633 struct netdev_hw_addr_list *from_list, int addr_len);
3634void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3635 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003636int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3637 struct net_device *dev,
3638 int (*sync)(struct net_device *, const unsigned char *),
3639 int (*unsync)(struct net_device *,
3640 const unsigned char *));
3641void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3642 struct net_device *dev,
3643 int (*unsync)(struct net_device *,
3644 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003645void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003646
Jiri Pirkof001fde2009-05-05 02:48:28 +00003647/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003648int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3649 unsigned char addr_type);
3650int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3651 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003652void dev_addr_flush(struct net_device *dev);
3653int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003654
3655/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003656int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3657int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3658int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3659int dev_uc_sync(struct net_device *to, struct net_device *from);
3660int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3661void dev_uc_unsync(struct net_device *to, struct net_device *from);
3662void dev_uc_flush(struct net_device *dev);
3663void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003664
Alexander Duyck670e5b82014-05-28 18:44:46 -07003665/**
3666 * __dev_uc_sync - Synchonize device's unicast list
3667 * @dev: device to sync
3668 * @sync: function to call if address should be added
3669 * @unsync: function to call if address should be removed
3670 *
3671 * Add newly added addresses to the interface, and release
3672 * addresses that have been deleted.
3673 **/
3674static inline int __dev_uc_sync(struct net_device *dev,
3675 int (*sync)(struct net_device *,
3676 const unsigned char *),
3677 int (*unsync)(struct net_device *,
3678 const unsigned char *))
3679{
3680 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3681}
3682
3683/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003684 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003685 * @dev: device to sync
3686 * @unsync: function to call if address should be removed
3687 *
3688 * Remove all addresses that were added to the device by dev_uc_sync().
3689 **/
3690static inline void __dev_uc_unsync(struct net_device *dev,
3691 int (*unsync)(struct net_device *,
3692 const unsigned char *))
3693{
3694 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3695}
3696
Jiri Pirko22bedad32010-04-01 21:22:57 +00003697/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003698int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3699int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3700int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3701int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3702int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3703int dev_mc_sync(struct net_device *to, struct net_device *from);
3704int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3705void dev_mc_unsync(struct net_device *to, struct net_device *from);
3706void dev_mc_flush(struct net_device *dev);
3707void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003708
Alexander Duyck670e5b82014-05-28 18:44:46 -07003709/**
3710 * __dev_mc_sync - Synchonize device's multicast list
3711 * @dev: device to sync
3712 * @sync: function to call if address should be added
3713 * @unsync: function to call if address should be removed
3714 *
3715 * Add newly added addresses to the interface, and release
3716 * addresses that have been deleted.
3717 **/
3718static inline int __dev_mc_sync(struct net_device *dev,
3719 int (*sync)(struct net_device *,
3720 const unsigned char *),
3721 int (*unsync)(struct net_device *,
3722 const unsigned char *))
3723{
3724 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3725}
3726
3727/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003728 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003729 * @dev: device to sync
3730 * @unsync: function to call if address should be removed
3731 *
3732 * Remove all addresses that were added to the device by dev_mc_sync().
3733 **/
3734static inline void __dev_mc_unsync(struct net_device *dev,
3735 int (*unsync)(struct net_device *,
3736 const unsigned char *))
3737{
3738 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3739}
3740
Linus Torvalds1da177e2005-04-16 15:20:36 -07003741/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003742void dev_set_rx_mode(struct net_device *dev);
3743void __dev_set_rx_mode(struct net_device *dev);
3744int dev_set_promiscuity(struct net_device *dev, int inc);
3745int dev_set_allmulti(struct net_device *dev, int inc);
3746void netdev_state_change(struct net_device *dev);
3747void netdev_notify_peers(struct net_device *dev);
3748void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003749/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003750void dev_load(struct net *net, const char *name);
3751struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3752 struct rtnl_link_stats64 *storage);
3753void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3754 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003755
3756extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003757extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003758extern int weight_p;
Eric Dumazet0a148422011-04-20 09:27:32 +00003759extern int bpf_jit_enable;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003760
Joe Perchesf629d202013-09-26 14:48:15 -07003761bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003762struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3763 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003764struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3765 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003766
3767/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003768#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3769 for (iter = &(dev)->adj_list.upper, \
3770 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3771 updev; \
3772 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3773
3774/* iterate through upper list, must be called under RCU read lock */
Veaceslav Falico2f268f12013-09-25 09:20:07 +02003775#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3776 for (iter = &(dev)->all_adj_list.upper, \
3777 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3778 updev; \
3779 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003780
Joe Perchesf629d202013-09-26 14:48:15 -07003781void *netdev_lower_get_next_private(struct net_device *dev,
3782 struct list_head **iter);
3783void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3784 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003785
3786#define netdev_for_each_lower_private(dev, priv, iter) \
3787 for (iter = (dev)->adj_list.lower.next, \
3788 priv = netdev_lower_get_next_private(dev, &(iter)); \
3789 priv; \
3790 priv = netdev_lower_get_next_private(dev, &(iter)))
3791
3792#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3793 for (iter = &(dev)->adj_list.lower, \
3794 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3795 priv; \
3796 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3797
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003798void *netdev_lower_get_next(struct net_device *dev,
3799 struct list_head **iter);
3800#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003801 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003802 ldev = netdev_lower_get_next(dev, &(iter)); \
3803 ldev; \
3804 ldev = netdev_lower_get_next(dev, &(iter)))
3805
Joe Perchesf629d202013-09-26 14:48:15 -07003806void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003807void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003808struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3809struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3810int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3811int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003812 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003813 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003814void netdev_upper_dev_unlink(struct net_device *dev,
3815 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003816void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003817void *netdev_lower_dev_get_private(struct net_device *dev,
3818 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003819void netdev_lower_state_changed(struct net_device *lower_dev,
3820 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003821
3822/* RSS keys are 40 or 52 bytes long */
3823#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003824extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003825void netdev_rss_key_fill(void *buffer, size_t len);
3826
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003827int dev_get_nest_level(struct net_device *dev,
Jiri Pirkob618aaa2015-12-04 15:01:31 +01003828 bool (*type_check)(const struct net_device *dev));
Joe Perchesf629d202013-09-26 14:48:15 -07003829int skb_checksum_help(struct sk_buff *skb);
3830struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3831 netdev_features_t features, bool tx_path);
3832struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3833 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003834
Moni Shoua61bd3852015-02-03 16:48:29 +02003835struct netdev_bonding_info {
3836 ifslave slave;
3837 ifbond master;
3838};
3839
3840struct netdev_notifier_bonding_info {
3841 struct netdev_notifier_info info; /* must be first */
3842 struct netdev_bonding_info bonding_info;
3843};
3844
3845void netdev_bonding_info_change(struct net_device *dev,
3846 struct netdev_bonding_info *bonding_info);
3847
Cong Wang12b00042013-02-05 16:36:38 +00003848static inline
3849struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3850{
3851 return __skb_gso_segment(skb, features, true);
3852}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003853__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003854
3855static inline bool can_checksum_protocol(netdev_features_t features,
3856 __be16 protocol)
3857{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003858 if (protocol == htons(ETH_P_FCOE))
3859 return !!(features & NETIF_F_FCOE_CRC);
3860
3861 /* Assume this is an IP checksum (not SCTP CRC) */
3862
3863 if (features & NETIF_F_HW_CSUM) {
3864 /* Can checksum everything */
3865 return true;
3866 }
3867
3868 switch (protocol) {
3869 case htons(ETH_P_IP):
3870 return !!(features & NETIF_F_IP_CSUM);
3871 case htons(ETH_P_IPV6):
3872 return !!(features & NETIF_F_IPV6_CSUM);
3873 default:
3874 return false;
3875 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003876}
Cong Wang12b00042013-02-05 16:36:38 +00003877
Tom Herbert6ae23ad2015-12-14 11:19:46 -08003878/* Map an ethertype into IP protocol if possible */
3879static inline int eproto_to_ipproto(int eproto)
3880{
3881 switch (eproto) {
3882 case htons(ETH_P_IP):
3883 return IPPROTO_IP;
3884 case htons(ETH_P_IPV6):
3885 return IPPROTO_IPV6;
3886 default:
3887 return -1;
3888 }
3889}
3890
Herbert Xufb286bb2005-11-10 13:01:24 -08003891#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003892void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003893#else
3894static inline void netdev_rx_csum_fault(struct net_device *dev)
3895{
3896}
3897#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003898/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003899void net_enable_timestamp(void);
3900void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003901
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003902#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003903int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003904#else
3905#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003906#endif
3907
David S. Miller47982482014-08-22 16:21:53 -07003908static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003909 struct sk_buff *skb, struct net_device *dev,
3910 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003911{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003912 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003913 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003914}
3915
David S. Miller10b3ad82014-08-29 21:07:24 -07003916static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003917 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003918{
3919 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003920 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003921
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003922 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003923 if (rc == NETDEV_TX_OK)
3924 txq_trans_update(txq);
3925
3926 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003927}
3928
Linus Torvalds42a2d922013-11-13 17:40:34 +09003929int netdev_class_create_file_ns(struct class_attribute *class_attr,
3930 const void *ns);
3931void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3932 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003933
3934static inline int netdev_class_create_file(struct class_attribute *class_attr)
3935{
3936 return netdev_class_create_file_ns(class_attr, NULL);
3937}
3938
3939static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3940{
3941 netdev_class_remove_file_ns(class_attr, NULL);
3942}
Jay Vosburghb8a97872008-06-13 18:12:04 -07003943
Johannes Berg04600792010-08-05 17:45:15 +02003944extern struct kobj_ns_type_operations net_ns_type_operations;
3945
Joe Perchesf629d202013-09-26 14:48:15 -07003946const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07003947
Joe Perchesf629d202013-09-26 14:48:15 -07003948void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003949
Michal Kubečekda081432014-05-20 08:29:25 +02003950static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3951 netdev_features_t f2)
3952{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003953 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3954 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08003955 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08003956 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08003957 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08003958 }
Michal Kubečekda081432014-05-20 08:29:25 +02003959
Tom Herbertc8cd0982015-12-14 11:19:44 -08003960 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02003961}
3962
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003963static inline netdev_features_t netdev_get_wanted_features(
3964 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00003965{
3966 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3967}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003968netdev_features_t netdev_increment_features(netdev_features_t all,
3969 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00003970
3971/* Allow TSO being used on stacked device :
3972 * Performing the GSO segmentation before last device
3973 * is a performance improvement.
3974 */
3975static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3976 netdev_features_t mask)
3977{
3978 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3979}
3980
Michał Mirosław6cb6a272011-04-02 22:48:47 -07003981int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00003982void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00003983void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07003984
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08003985void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3986 struct net_device *dev);
3987
Toshiaki Makitae38f3022015-03-27 14:31:13 +09003988netdev_features_t passthru_features_check(struct sk_buff *skb,
3989 struct net_device *dev,
3990 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02003991netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00003992
David S. Miller4d295152012-03-07 21:02:35 -05003993static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07003994{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003995 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00003996
3997 /* check flags correspondence */
3998 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3999 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4000 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4001 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4002 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4003 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004004 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4005 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4006 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
4007 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
4008 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4009 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 -08004010 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004011
Herbert Xubcd76112006-06-30 13:36:35 -07004012 return (features & feature) == feature;
4013}
4014
David S. Miller4d295152012-03-07 21:02:35 -05004015static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004016{
Herbert Xu278b2512009-06-03 21:20:51 -07004017 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004018 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004019}
4020
Johannes Berg8b86a612015-04-17 15:45:04 +02004021static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004022 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004023{
Jesse Grossfc741212011-01-09 06:23:32 +00004024 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004025 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4026 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004027}
4028
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004029static inline void netif_set_gso_max_size(struct net_device *dev,
4030 unsigned int size)
4031{
4032 dev->gso_max_size = size;
4033}
4034
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004035static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4036 int pulled_hlen, u16 mac_offset,
4037 int mac_len)
4038{
4039 skb->protocol = protocol;
4040 skb->encapsulation = 1;
4041 skb_push(skb, pulled_hlen);
4042 skb_reset_transport_header(skb);
4043 skb->mac_header = mac_offset;
4044 skb->network_header = skb->mac_header + mac_len;
4045 skb->mac_len = mac_len;
4046}
4047
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004048static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004049{
4050 return dev->priv_flags & IFF_MACVLAN;
4051}
4052
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004053static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004054{
4055 return dev->priv_flags & IFF_MACVLAN_PORT;
4056}
4057
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004058static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004059{
4060 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4061}
4062
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004063static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004064{
4065 return dev->priv_flags & IFF_IPVLAN_MASTER;
4066}
4067
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004068static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004069{
4070 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4071}
4072
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004073static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004074{
4075 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4076}
4077
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004078static inline bool netif_supports_nofcs(struct net_device *dev)
4079{
4080 return dev->priv_flags & IFF_SUPP_NOFCS;
4081}
4082
David Ahern007979e2015-09-29 20:07:10 -07004083static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004084{
David Ahern007979e2015-09-29 20:07:10 -07004085 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004086}
4087
David Ahernfee6d4c2015-10-05 08:51:24 -07004088static inline bool netif_is_l3_slave(const struct net_device *dev)
4089{
4090 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4091}
4092
Jiri Pirko0894ae32015-08-27 09:31:19 +02004093static inline bool netif_is_bridge_master(const struct net_device *dev)
4094{
4095 return dev->priv_flags & IFF_EBRIDGE;
4096}
4097
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004098static inline bool netif_is_bridge_port(const struct net_device *dev)
4099{
4100 return dev->priv_flags & IFF_BRIDGE_PORT;
4101}
4102
Jiri Pirko35d4e172015-08-27 09:31:20 +02004103static inline bool netif_is_ovs_master(const struct net_device *dev)
4104{
4105 return dev->priv_flags & IFF_OPENVSWITCH;
4106}
4107
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004108static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004109{
4110 return dev->priv_flags & IFF_TEAM;
4111}
4112
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004113static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004114{
4115 return dev->priv_flags & IFF_TEAM_PORT;
4116}
4117
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004118static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004119{
4120 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4121}
4122
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004123static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004124{
4125 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4126}
4127
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004128static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4129{
4130 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4131}
4132
Eric Dumazet02875872014-10-05 18:38:35 -07004133/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4134static inline void netif_keep_dst(struct net_device *dev)
4135{
4136 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4137}
4138
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004139extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004140
Joe Perches571ba422010-02-09 11:49:47 +00004141/* Logging, debugging and troubleshooting/diagnostic helpers. */
4142
4143/* netdev_printk helpers, similar to dev_printk */
4144
4145static inline const char *netdev_name(const struct net_device *dev)
4146{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004147 if (!dev->name[0] || strchr(dev->name, '%'))
4148 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004149 return dev->name;
4150}
4151
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004152static inline const char *netdev_reg_state(const struct net_device *dev)
4153{
4154 switch (dev->reg_state) {
4155 case NETREG_UNINITIALIZED: return " (uninitialized)";
4156 case NETREG_REGISTERED: return "";
4157 case NETREG_UNREGISTERING: return " (unregistering)";
4158 case NETREG_UNREGISTERED: return " (unregistered)";
4159 case NETREG_RELEASED: return " (released)";
4160 case NETREG_DUMMY: return " (dummy)";
4161 }
4162
4163 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4164 return " (unknown)";
4165}
4166
Joe Perchesf629d202013-09-26 14:48:15 -07004167__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004168void netdev_printk(const char *level, const struct net_device *dev,
4169 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004170__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004171void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004172__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004173void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004174__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004175void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004176__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004177void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004178__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004179void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004180__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004181void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004182__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004183void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004184
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004185#define MODULE_ALIAS_NETDEV(device) \
4186 MODULE_ALIAS("netdev-" device)
4187
Jim Cromieb558c962011-12-19 17:11:18 -05004188#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004189#define netdev_dbg(__dev, format, args...) \
4190do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004191 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004192} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004193#elif defined(DEBUG)
4194#define netdev_dbg(__dev, format, args...) \
4195 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004196#else
4197#define netdev_dbg(__dev, format, args...) \
4198({ \
4199 if (0) \
4200 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004201})
4202#endif
4203
4204#if defined(VERBOSE_DEBUG)
4205#define netdev_vdbg netdev_dbg
4206#else
4207
4208#define netdev_vdbg(dev, format, args...) \
4209({ \
4210 if (0) \
4211 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4212 0; \
4213})
4214#endif
4215
4216/*
4217 * netdev_WARN() acts like dev_printk(), but with the key difference
4218 * of using a WARN/WARN_ON to get the message out, including the
4219 * file/line information and a backtrace.
4220 */
4221#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004222 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4223 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004224
Joe Perchesb3d95c52010-02-09 11:49:49 +00004225/* netif printk helpers, similar to netdev_printk */
4226
4227#define netif_printk(priv, type, level, dev, fmt, args...) \
4228do { \
4229 if (netif_msg_##type(priv)) \
4230 netdev_printk(level, (dev), fmt, ##args); \
4231} while (0)
4232
Joe Perchesf45f4322010-06-27 01:02:36 +00004233#define netif_level(level, priv, type, dev, fmt, args...) \
4234do { \
4235 if (netif_msg_##type(priv)) \
4236 netdev_##level(dev, fmt, ##args); \
4237} while (0)
4238
Joe Perchesb3d95c52010-02-09 11:49:49 +00004239#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004240 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004241#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004242 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004243#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004244 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004245#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004246 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004247#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004248 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004249#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004250 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004251#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004252 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004253
Joe Perches0053ea92012-05-30 07:43:34 +00004254#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004255#define netif_dbg(priv, type, netdev, format, args...) \
4256do { \
4257 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004258 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004259} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004260#elif defined(DEBUG)
4261#define netif_dbg(priv, type, dev, format, args...) \
4262 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004263#else
4264#define netif_dbg(priv, type, dev, format, args...) \
4265({ \
4266 if (0) \
4267 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4268 0; \
4269})
4270#endif
4271
4272#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004273#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004274#else
4275#define netif_vdbg(priv, type, dev, format, args...) \
4276({ \
4277 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004278 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004279 0; \
4280})
4281#endif
Joe Perches571ba422010-02-09 11:49:47 +00004282
Cong Wang900ff8c2013-02-18 19:20:33 +00004283/*
4284 * The list of packet types we will receive (as opposed to discard)
4285 * and the routines to invoke.
4286 *
4287 * Why 16. Because with 16 the only overlap we get on a hash of the
4288 * low nibble of the protocol value is RARP/SNAP/X.25.
4289 *
4290 * NOTE: That is no longer true with the addition of VLAN tags. Not
4291 * sure which should go first, but I bet it won't make much
4292 * difference if we are running VLANs. The good news is that
4293 * this protocol won't be in the list unless compiled in, so
4294 * the average user (w/out VLANs) will not be adversely affected.
4295 * --BLG
4296 *
4297 * 0800 IP
4298 * 8100 802.1Q VLAN
4299 * 0001 802.3
4300 * 0002 AX.25
4301 * 0004 802.2
4302 * 8035 RARP
4303 * 0005 SNAP
4304 * 0805 X.25
4305 * 0806 ARP
4306 * 8137 IPX
4307 * 0009 Localtalk
4308 * 86DD IPv6
4309 */
4310#define PTYPE_HASH_SIZE (16)
4311#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4312
Jiri Pirko385a1542009-05-27 15:48:07 -07004313#endif /* _LINUX_NETDEVICE_H */