blob: 91ee3643ccc8df3f8d244b6dd5f47526ea688274 [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>
Jiri Kosina59cc1f62016-08-10 11:05:15 +020055#include <linux/hashtable.h>
Michał Mirosława59e2ec2011-11-15 15:29:55 +000056
Jeff Moyer115c1d62005-06-22 22:05:31 -070057struct netpoll_info;
Paul Gortmaker313162d2012-01-30 11:46:54 -050058struct device;
Richard Cochranc1f19b52010-07-17 08:49:36 +000059struct phy_device;
Johannes Berg704232c2007-04-23 12:20:05 -070060/* 802.11 specific */
61struct wireless_dev;
Alexander Aring98a18b62014-11-02 06:44:54 +010062/* 802.15.4 specific */
63struct wpan_dev;
Robert Shearman03c57742015-04-22 11:14:37 +010064struct mpls_dev;
Alexander Duyck7c46a642016-06-16 12:21:00 -070065/* UDP Tunnel offloads */
66struct udp_tunnel_info;
Brenden Blancoa7862b42016-07-19 12:16:48 -070067struct bpf_prog;
Linus Torvalds1da177e2005-04-16 15:20:36 -070068
Joe Perchesf629d202013-09-26 14:48:15 -070069void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000071
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000072/* Backlog congestion levels */
73#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74#define NET_RX_DROP 1 /* packet dropped */
75
Patrick McHardy572a9d72009-11-10 06:14:14 +000076/*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -050088 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
Patrick McHardy572a9d72009-11-10 06:14:14 +000090 * others are propagated to higher layers.
91 */
92
93/* qdisc ->enqueue() return codes. */
94#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000095#define NET_XMIT_DROP 0x01 /* skb dropped */
96#define NET_XMIT_CN 0x02 /* congestion notification */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000097#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -020099/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +0000102#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000105/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000106#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000107
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000108enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000112};
113typedef enum netdev_tx netdev_tx_t;
114
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000115/*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
119static inline bool dev_xmit_complete(int rc)
120{
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131}
132
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500134 * Compute the worst-case header length according to the protocols
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 * used.
136 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800137
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800138#if defined(CONFIG_HYPERV_NET)
139# define LL_MAX_HEADER 128
140#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700141# if defined(CONFIG_MAC80211_MESH)
142# define LL_MAX_HEADER 128
143# else
144# define LL_MAX_HEADER 96
145# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700147# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148#endif
149
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000150#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152#define MAX_HEADER LL_MAX_HEADER
153#else
154#define MAX_HEADER (LL_MAX_HEADER + 48)
155#endif
156
157/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800161
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800162struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186};
187
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
189#include <linux/cache.h>
190#include <linux/skbuff.h>
191
Eric Dumazetadc93002011-11-17 03:13:26 +0000192#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100193#include <linux/static_key.h>
194extern struct static_key rps_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000195#endif
196
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197struct neighbour;
198struct neigh_parms;
199struct sk_buff;
200
Jiri Pirkof001fde2009-05-05 02:48:28 +0000201struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000205#define NETDEV_HW_ADDR_T_LAN 1
206#define NETDEV_HW_ADDR_T_SAN 2
207#define NETDEV_HW_ADDR_T_SLAVE 3
208#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000209#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000210 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000211 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700212 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400213 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000214 struct rcu_head rcu_head;
215};
216
Jiri Pirko31278e72009-06-17 01:12:19 +0000217struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220};
221
Jiri Pirko22bedad32010-04-01 21:22:57 +0000222#define netdev_hw_addr_list_count(l) ((l)->count)
223#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224#define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
226
227#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800229#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800231
Jiri Pirko22bedad32010-04-01 21:22:57 +0000232#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700234#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece2010-02-04 10:22:25 -0800236
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800237struct hh_cache {
David S. Millerf6b72b622011-07-14 07:53:20 -0700238 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700239 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800240 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
242 /* cached hardware header; allow for machine alignment needs. */
243#define HH_DATA_MOD 16
244#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246#define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
249};
250
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500251/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 * Alternative is:
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
255 *
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
258 */
259#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700264struct header_ops {
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000267 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500273 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700274};
275
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276/* These flag bits are private to the generic network queueing
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500277 * layer; they may not be explicitly referenced by any other
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 * code.
279 */
280
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800281enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287};
288
289
290/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500291 * This structure holds boot-time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800292 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 */
294struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297};
298#define NETDEV_BOOT_SETUP_MAX 8
299
Joe Perchesf629d202013-09-26 14:48:15 -0700300int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
302/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700303 * Structure for NAPI scheduling similar to tasklet but with weighting
304 */
305struct napi_struct {
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500309 * to the per-CPU poll_list, and whoever clears that bit
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700310 * can remove from the list right before clearing the bit.
311 */
312 struct list_head poll_list;
313
314 unsigned long state;
315 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000316 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700317 int (*poll)(struct napi_struct *, int);
318#ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
320 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700321#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800322 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800323 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800324 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800325 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000326 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700329};
330
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800331enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700332 NAPI_STATE_SCHED, /* Poll is scheduled */
David S. Millera0a46192008-01-07 20:35:07 -0800333 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700337};
338
Ben Hutchings5b252f02009-10-29 07:17:09 +0000339enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700340 GRO_MERGED,
341 GRO_MERGED_FREE,
342 GRO_HELD,
343 GRO_NORMAL,
344 GRO_DROP,
345};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000346typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700347
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000348/*
349 * enum rx_handler_result - Possible return values for rx_handlers.
350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * further.
352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
353 * case skb->dev was changed by rx_handler.
354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000356 *
357 * rx_handlers are functions called from inside __netif_receive_skb(), to do
358 * special processing of the skb, prior to delivery to protocol handlers.
359 *
360 * Currently, a net_device can only have a single rx_handler registered. Trying
361 * to register a second rx_handler will return -EBUSY.
362 *
363 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
364 * To unregister a rx_handler on a net_device, use
365 * netdev_rx_handler_unregister().
366 *
367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
368 * do with the skb.
369 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500370 * If the rx_handler consumed the skb in some way, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500372 * the skb to be delivered in some other way.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000373 *
374 * If the rx_handler changed skb->dev, to divert the skb to another
375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
376 * new device will be called if it exists.
377 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500378 * If the rx_handler decides the skb should be ignored, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400380 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000381 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000383 * delivered, it should return RX_HANDLER_PASS.
384 *
385 * A device without a registered rx_handler will behave as if rx_handler
386 * returned RX_HANDLER_PASS.
387 */
388
389enum rx_handler_result {
390 RX_HANDLER_CONSUMED,
391 RX_HANDLER_ANOTHER,
392 RX_HANDLER_EXACT,
393 RX_HANDLER_PASS,
394};
395typedef enum rx_handler_result rx_handler_result_t;
396typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000397
Joe Perchesf629d202013-09-26 14:48:15 -0700398void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700399void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700400
David S. Miller4d295152012-03-07 21:02:35 -0500401static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800402{
403 return test_bit(NAPI_STATE_DISABLE, &n->state);
404}
405
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700406/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500407 * napi_schedule_prep - check if NAPI can be scheduled
408 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700409 *
410 * Test if NAPI routine is already running, and if not mark
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500411 * it as running. This is used as a condition variable to
David S. Millera0a46192008-01-07 20:35:07 -0800412 * insure only one NAPI poll instance runs. We also make
413 * sure there is no pending NAPI disable.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700414 */
David S. Miller4d295152012-03-07 21:02:35 -0500415static inline bool napi_schedule_prep(struct napi_struct *n)
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700416{
David S. Millera0a46192008-01-07 20:35:07 -0800417 return !napi_disable_pending(n) &&
418 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700419}
420
421/**
422 * napi_schedule - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500423 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428static inline void napi_schedule(struct napi_struct *n)
429{
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432}
433
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700434/**
435 * napi_schedule_irqoff - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500436 * @n: NAPI context
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440static inline void napi_schedule_irqoff(struct napi_struct *n)
441{
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444}
445
Roland Dreierbfe13f52007-10-09 15:47:37 -0700446/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500447static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700448{
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500451 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700452 }
David S. Miller4d295152012-03-07 21:02:35 -0500453 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700454}
455
Eric Dumazet3b47d302014-11-06 21:09:44 -0800456void __napi_complete(struct napi_struct *n);
457void napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700458/**
459 * napi_complete - NAPI processing complete
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500460 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700461 *
462 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800463 * Consider using napi_complete_done() instead.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700464 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800465static inline void napi_complete(struct napi_struct *n)
466{
467 return napi_complete_done(n, 0);
468}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700469
470/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300471 * napi_hash_add - add a NAPI to global hashtable
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500472 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300473 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500474 * Generate a new napi_id and store a @napi under it in napi_hash.
475 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
Eric Dumazet93d05d42015-11-18 06:31:03 -0800476 * Note: This is normally automatically done from netif_napi_add(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500477 * so might disappear in a future Linux version.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300478 */
Joe Perchesf629d202013-09-26 14:48:15 -0700479void napi_hash_add(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300480
481/**
482 * napi_hash_del - remove a NAPI from global table
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500483 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300484 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500485 * Warning: caller must observe RCU grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800486 * before freeing memory containing @napi, if
487 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800488 * Note: core networking stack automatically calls it
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500489 * from netif_napi_del().
Eric Dumazet93d05d42015-11-18 06:31:03 -0800490 * Drivers might want to call this helper to combine all
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500491 * the needed RCU grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300492 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800493bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300494
495/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700496 * napi_disable - prevent NAPI from scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500497 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700498 *
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
501 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800502void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700503
504/**
505 * napi_enable - enable NAPI scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500506 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700507 *
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
510 */
511static inline void napi_enable(struct napi_struct *n)
512{
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100514 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700515 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400516 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700517}
518
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700519/**
520 * napi_synchronize - wait until NAPI is not running
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500521 * @n: NAPI context
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700522 *
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
526 */
527static inline void napi_synchronize(const struct napi_struct *n)
528{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 msleep(1);
532 else
533 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700534}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700535
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800536enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700539 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700540};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200541
542#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545
546#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 QUEUE_STATE_FROZEN)
549#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 QUEUE_STATE_FROZEN)
551
Tom Herbert734664982011-11-28 16:32:44 +0000552/*
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
560 */
David S. Miller79d16382008-07-08 23:14:46 -0700561
David S. Millerbb949fb2008-07-08 16:55:56 -0700562struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700563/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500564 * read-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700565 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700566 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700567 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700568 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000569#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000570 struct kobject kobj;
571#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000572#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 int numa_node;
574#endif
Florian Westphalc0ef0792016-05-03 03:29:09 +0200575 unsigned long tx_maxrate;
576 /*
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
579 */
580 unsigned long trans_timeout;
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700581/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500582 * write-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700583 */
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
585 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700586 /*
Florian Westphal9b366272016-05-03 16:33:14 +0200587 * Time (in jiffies) of last Tx
Eric Dumazet9d214932009-05-17 20:55:16 -0700588 */
589 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000590
Tom Herbert114cf582011-11-28 16:33:09 +0000591 unsigned long state;
592
593#ifdef CONFIG_BQL
594 struct dql dql;
595#endif
David S. Millere8a04642008-07-17 00:34:19 -0700596} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700597
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000598static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599{
600#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602#else
Changli Gaob236da62010-12-14 03:09:15 +0000603 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000604#endif
605}
606
607static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608{
609#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611#endif
612}
613
Eric Dumazetdf334542010-03-24 19:13:54 +0000614#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000615/*
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
618 */
619struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
623};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000624#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000625
Tom Herbertfec5e652010-04-16 16:01:27 -0700626/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700630 */
631struct rps_dev_flow {
632 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000633 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700634 unsigned int last_qtail;
635};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000636#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700637
638/*
639 * The rps_dev_flow_table structure contains a table of flow mappings.
640 */
641struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700644 struct rps_dev_flow flows[0];
645};
646#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000647 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700648
649/*
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
Eric Dumazet567e4b72015-02-06 12:59:01 -0800654 * rps_cpu_mask is used to partition the space, depending on number of
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
Eric Dumazet567e4b72015-02-06 12:59:01 -0800657 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700658 */
659struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800660 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800661
662 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700663};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800664#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700665
666#define RPS_NO_CPU 0xffff
667
Eric Dumazet567e4b72015-02-06 12:59:01 -0800668extern u32 rps_cpu_mask;
669extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670
Tom Herbertfec5e652010-04-16 16:01:27 -0700671static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673{
674 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700677
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500678 /* We only give a hint, preemption can change CPU under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800679 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700680
Eric Dumazet567e4b72015-02-06 12:59:01 -0800681 if (table->ents[index] != val)
682 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700683 }
684}
685
Ben Hutchingsc4454772011-01-19 11:03:53 +0000686#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700687bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000689#endif
Michael Daltona953be52014-01-16 22:23:28 -0800690#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000691
Tom Herbert0a9627f2010-03-16 08:03:29 +0000692/* This structure contains an instance of an RX queue. */
693struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800694#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800697#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000698 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000699 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000700} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800701
702/*
703 * RX queue sysfs structures and functions.
704 */
705struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
711};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800712
Tom Herbertbf264142010-11-26 08:36:09 +0000713#ifdef CONFIG_XPS
714/*
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
717 */
718struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
723};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000724#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100725#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000727
728/*
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 */
731struct xps_dev_maps {
732 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000733 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000734};
735#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * sizeof(struct xps_map *)))
737#endif /* CONFIG_XPS */
738
John Fastabend4f57c082011-01-17 08:06:04 +0000739#define TC_MAX_QUEUE 16
740#define TC_BITMASK 15
741/* HW offloaded queuing disciplines txq count and offset maps */
742struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
745};
746
Neerav Parikh68bad942012-01-04 20:23:39 +0000747#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748/*
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
751 */
752struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
759 char model[256];
760 char model_description[256];
761};
762#endif
763
Jiri Pirko02637fc2014-11-28 14:34:16 +0100764#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200765
Jiri Pirko02637fc2014-11-28 14:34:16 +0100766/* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200768 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100769struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200771 unsigned char id_len;
772};
773
Scott Feldmand754f982015-07-18 18:24:49 -0700774static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
776{
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
779}
780
Daniel Borkmann99932d42014-02-16 15:55:20 +0100781typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
783
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800784/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800785 * that are being passed to the netdevice through the setup_tc op.
786 */
787enum {
788 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800789 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200790 TC_SETUP_CLSFLOWER,
Yotam Gigib87f7932016-07-21 12:03:12 +0200791 TC_SETUP_MATCHALL,
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100792 TC_SETUP_CLSBPF,
John Fastabend16e5cc62016-02-16 21:16:43 -0800793};
794
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800795struct tc_cls_u32_offload;
796
John Fastabend16e5cc62016-02-16 21:16:43 -0800797struct tc_to_netdev {
798 unsigned int type;
799 union {
800 u8 tc;
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800801 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200802 struct tc_cls_flower_offload *cls_flower;
Yotam Gigib87f7932016-07-21 12:03:12 +0200803 struct tc_cls_matchall_offload *cls_mall;
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100804 struct tc_cls_bpf_offload *cls_bpf;
John Fastabend16e5cc62016-02-16 21:16:43 -0800805 };
806};
807
Brenden Blancoa7862b42016-07-19 12:16:48 -0700808/* These structures hold the attributes of xdp state that are being passed
809 * to the netdevice through the xdp op.
810 */
811enum xdp_netdev_command {
812 /* Set or clear a bpf program used in the earliest stages of packet
813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
814 * is responsible for calling bpf_prog_put on any old progs that are
815 * stored. In case of error, the callee need not release the new prog
816 * reference, but on success it takes ownership and must bpf_prog_put
817 * when it is no longer used.
818 */
819 XDP_SETUP_PROG,
820 /* Check if a bpf program is set on the device. The callee should
821 * return true if a program is currently attached and running.
822 */
823 XDP_QUERY_PROG,
824};
825
826struct netdev_xdp {
827 enum xdp_netdev_command command;
828 union {
829 /* XDP_SETUP_PROG */
830 struct bpf_prog *prog;
831 /* XDP_QUERY_PROG */
832 bool prog_attached;
833 };
834};
John Fastabend16e5cc62016-02-16 21:16:43 -0800835
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800836/*
837 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800840 *
841 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500852 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500856 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800857 * state.
858 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800861 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500866 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800867 *
Eric Dumazetcdba7562016-01-06 06:53:50 -0800868 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
869 * netdev_features_t features);
870 * Adjusts the requested feature flags according to device-specific
871 * constraints, and returns the resulting flags. Must not modify
872 * the device state.
873 *
Jason Wangf663dd92014-01-10 16:18:26 +0800874 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100875 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500876 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800877 * transmit queues.
878 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800879 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
880 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500881 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800882 *
883 * void (*ndo_set_rx_mode)(struct net_device *dev);
884 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000885 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500886 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800887 *
888 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
889 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700890 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500891 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800892 *
893 * int (*ndo_validate_addr)(struct net_device *dev);
894 * Test if Media Access Control address is valid for the device.
895 *
896 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500897 * Called when a user requests an ioctl which can't be handled by
898 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800899 * not supported error code.
900 *
901 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
902 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500903 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800904 * interface (PCI) for low level management.
905 *
906 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
907 * Called when a user wants to change the Maximum Transfer Unit
908 * of a device. If not defined, any request to change MTU will
909 * will return an error.
910 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800911 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500912 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800913 * for dev->watchdog ticks.
914 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000915 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700916 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700917 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800918 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000919 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000920 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
921 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700922 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000923 * (which should normally be dev->stats) and return a pointer to
924 * it. The structure may be changed asynchronously only if each
925 * field is written atomically.
926 * 3. Update dev->stats asynchronously and atomically, and define
927 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800928 *
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200929 * bool (*ndo_has_offload_stats)(int attr_id)
930 * Return true if this device supports offload stats of this attr_id.
931 *
932 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
933 * void *attr_data)
934 * Get statistics for offload operations by attr_id. Write it into the
935 * attr_data pointer.
936 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530937 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500938 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000939 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800940 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530941 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500942 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000943 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800944 *
945 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000946 *
947 * SR-IOV management functions.
948 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300949 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
950 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400951 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
952 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000953 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000954 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000955 * int (*ndo_get_vf_config)(struct net_device *dev,
956 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300957 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700958 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
959 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300960 *
961 * Enable or disable the VF ability to query its RSS Redirection Table and
962 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500963 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300964 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700965 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000966 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
967 * Called to setup 'tc' number of traffic classes in the net device. This
968 * is always called from the stack with the rtnl lock held and netif tx
969 * queues stopped. This allows the netdevice to perform queue management
970 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000971 *
Yi Zoue9bce842011-03-09 08:48:03 +0000972 * Fiber Channel over Ethernet (FCoE) offload functions.
973 * int (*ndo_fcoe_enable)(struct net_device *dev);
974 * Called when the FCoE protocol stack wants to start using LLD for FCoE
975 * so the underlying device can perform whatever needed configuration or
976 * initialization to support acceleration of FCoE traffic.
977 *
978 * int (*ndo_fcoe_disable)(struct net_device *dev);
979 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
980 * so the underlying device can perform whatever needed clean-ups to
981 * stop supporting acceleration of FCoE traffic.
982 *
983 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
984 * struct scatterlist *sgl, unsigned int sgc);
985 * Called when the FCoE Initiator wants to initialize an I/O that
986 * is a possible candidate for Direct Data Placement (DDP). The LLD can
987 * perform necessary setup and returns 1 to indicate the device is set up
988 * successfully to perform DDP on this I/O, otherwise this returns 0.
989 *
990 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
991 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
992 * indicated by the FC exchange id 'xid', so the underlying device can
993 * clean up and reuse resources for later DDP requests.
994 *
995 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
996 * struct scatterlist *sgl, unsigned int sgc);
997 * Called when the FCoE Target wants to initialize an I/O that
998 * is a possible candidate for Direct Data Placement (DDP). The LLD can
999 * perform necessary setup and returns 1 to indicate the device is set up
1000 * successfully to perform DDP on this I/O, otherwise this returns 0.
1001 *
Neerav Parikh68bad942012-01-04 20:23:39 +00001002 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1003 * struct netdev_fcoe_hbainfo *hbainfo);
1004 * Called when the FCoE Protocol stack wants information on the underlying
1005 * device. This information is utilized by the FCoE protocol stack to
1006 * register attributes with Fiber Channel management service as per the
1007 * FC-GS Fabric Device Management Information(FDMI) specification.
1008 *
Yi Zoue9bce842011-03-09 08:48:03 +00001009 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1010 * Called when the underlying device wants to override default World Wide
1011 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1012 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1013 * protocol stack to use.
1014 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001015 * RFS acceleration.
1016 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1017 * u16 rxq_index, u32 flow_id);
1018 * Set hardware filter for RFS. rxq_index is the target queue index;
1019 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1020 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001021 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001022 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001023 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1024 * Called to make another netdev an underling.
1025 *
1026 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1027 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001028 *
1029 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001030 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001031 * Called to update device configuration to new features. Passed
1032 * feature set might be less than what was returned by ndo_fix_features()).
1033 * Must return >0 or -errno if it changed dev->features itself.
1034 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001035 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1036 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001037 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001038 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001039 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1040 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001041 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001042 * Deletes the FDB entry from dev coresponding to addr.
1043 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001044 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001045 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001046 * Used to add FDB entries to dump requests. Implementers should add
1047 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001048 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001049 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1050 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001051 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001052 * struct net_device *dev, u32 filter_mask,
1053 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001054 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1055 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001056 *
1057 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1058 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1059 * which do not represent real hardware may define this to allow their
1060 * userspace components to manage their virtual carrier state. Devices
1061 * that determine carrier state from physical hardware properties (eg
1062 * network cables) or protocol-dependent mechanisms (eg
1063 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001064 *
1065 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001066 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001067 * Called to get ID of physical port of this device. If driver does
1068 * not implement this, it is assumed that the hw is not able to have
1069 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001070 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001071 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1072 * struct udp_tunnel_info *ti);
1073 * Called by UDP tunnel to notify a driver about the UDP port and socket
1074 * address family that a UDP tunnel is listnening to. It is called only
1075 * when a new port starts listening. The operation is protected by the
1076 * RTNL.
1077 *
1078 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1079 * struct udp_tunnel_info *ti);
1080 * Called by UDP tunnel to notify the driver about a UDP port and socket
1081 * address family that the UDP tunnel is not listening to anymore. The
1082 * operation is protected by the RTNL.
1083 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001084 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1085 * struct net_device *dev)
1086 * Called by upper layer devices to accelerate switching or other
1087 * station functionality into hardware. 'pdev is the lowerdev
1088 * to use for the offload and 'dev' is the net device that will
1089 * back the offload. Returns a pointer to the private structure
1090 * the upper layer will maintain.
1091 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1092 * Called by upper layer device to delete the station created
1093 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1094 * the station and priv is the structure returned by the add
1095 * operation.
1096 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1097 * struct net_device *dev,
1098 * void *priv);
1099 * Callback to use for xmit over the accelerated station. This
1100 * is used in place of ndo_start_xmit on accelerated net
1101 * devices.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001102 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1103 * struct net_device *dev
1104 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001105 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001106 * performing offload operations on a given packet. This is to give
1107 * the device an opportunity to implement any restrictions that cannot
1108 * be otherwise expressed by feature flags. The check is called with
1109 * the set of features that the stack has calculated and it returns
1110 * those the driver believes to be appropriate.
John Fastabend822b3b22015-03-18 14:57:33 +02001111 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1112 * int queue_index, u32 maxrate);
1113 * Called when a user wants to set a max-rate limitation of specific
1114 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001115 * int (*ndo_get_iflink)(const struct net_device *dev);
1116 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001117 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001118 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001119 * This function is used to pass protocol port error state information
1120 * to the switch driver. The switch driver can react to the proto_down
1121 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001122 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1123 * This function is used to get egress tunnel information for given skb.
1124 * This is useful for retrieving outer tunnel header parameters while
1125 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001126 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1127 * This function is used to specify the headroom that the skb must
1128 * consider when allocation skb during packet reception. Setting
1129 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001130 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001131 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001132 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1133 * This function is used to set or query state related to XDP on the
1134 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001135 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001136 */
1137struct net_device_ops {
1138 int (*ndo_init)(struct net_device *dev);
1139 void (*ndo_uninit)(struct net_device *dev);
1140 int (*ndo_open)(struct net_device *dev);
1141 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001142 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1143 struct net_device *dev);
1144 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1145 struct net_device *dev,
1146 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001147 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001148 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001149 void *accel_priv,
1150 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001151 void (*ndo_change_rx_flags)(struct net_device *dev,
1152 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001153 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001154 int (*ndo_set_mac_address)(struct net_device *dev,
1155 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001156 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001157 int (*ndo_do_ioctl)(struct net_device *dev,
1158 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001159 int (*ndo_set_config)(struct net_device *dev,
1160 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001161 int (*ndo_change_mtu)(struct net_device *dev,
1162 int new_mtu);
1163 int (*ndo_neigh_setup)(struct net_device *dev,
1164 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001165 void (*ndo_tx_timeout) (struct net_device *dev);
1166
Eric Dumazet28172732010-07-07 14:58:56 -07001167 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1168 struct rtnl_link_stats64 *storage);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001169 bool (*ndo_has_offload_stats)(int attr_id);
1170 int (*ndo_get_offload_stats)(int attr_id,
1171 const struct net_device *dev,
1172 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001173 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174
Jiri Pirko8e586132011-12-08 19:52:37 -05001175 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001176 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001177 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001178 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001179#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001180 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001181 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001182 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001183 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001184#endif
Cong Wange0d10952013-08-01 11:10:25 +08001185#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001186 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001187#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001188 int (*ndo_set_vf_mac)(struct net_device *dev,
1189 int queue, u8 *mac);
1190 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001191 int queue, u16 vlan,
1192 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001193 int (*ndo_set_vf_rate)(struct net_device *dev,
1194 int vf, int min_tx_rate,
1195 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001196 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1197 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001198 int (*ndo_set_vf_trust)(struct net_device *dev,
1199 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001200 int (*ndo_get_vf_config)(struct net_device *dev,
1201 int vf,
1202 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001203 int (*ndo_set_vf_link_state)(struct net_device *dev,
1204 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001205 int (*ndo_get_vf_stats)(struct net_device *dev,
1206 int vf,
1207 struct ifla_vf_stats
1208 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001209 int (*ndo_set_vf_port)(struct net_device *dev,
1210 int vf,
1211 struct nlattr *port[]);
1212 int (*ndo_get_vf_port)(struct net_device *dev,
1213 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001214 int (*ndo_set_vf_guid)(struct net_device *dev,
1215 int vf, u64 guid,
1216 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001217 int (*ndo_set_vf_rss_query_en)(
1218 struct net_device *dev,
1219 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001220 int (*ndo_setup_tc)(struct net_device *dev,
1221 u32 handle,
1222 __be16 protocol,
1223 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001224#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001225 int (*ndo_fcoe_enable)(struct net_device *dev);
1226 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001227 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1228 u16 xid,
1229 struct scatterlist *sgl,
1230 unsigned int sgc);
1231 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1232 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001233 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1234 u16 xid,
1235 struct scatterlist *sgl,
1236 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001237 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1238 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001239#endif
1240
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001241#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001242#define NETDEV_FCOE_WWNN 0
1243#define NETDEV_FCOE_WWPN 1
1244 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1245 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001246#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001247
Ben Hutchingsc4454772011-01-19 11:03:53 +00001248#ifdef CONFIG_RFS_ACCEL
1249 int (*ndo_rx_flow_steer)(struct net_device *dev,
1250 const struct sk_buff *skb,
1251 u16 rxq_index,
1252 u32 flow_id);
1253#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001254 int (*ndo_add_slave)(struct net_device *dev,
1255 struct net_device *slave_dev);
1256 int (*ndo_del_slave)(struct net_device *dev,
1257 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001258 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1259 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001260 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001261 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001262 int (*ndo_neigh_construct)(struct net_device *dev,
1263 struct neighbour *n);
1264 void (*ndo_neigh_destroy)(struct net_device *dev,
1265 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001266
1267 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001268 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001269 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001270 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001271 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001272 u16 flags);
1273 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001274 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001275 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001276 const unsigned char *addr,
1277 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001278 int (*ndo_fdb_dump)(struct sk_buff *skb,
1279 struct netlink_callback *cb,
1280 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001281 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001282 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001283
1284 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001285 struct nlmsghdr *nlh,
1286 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001287 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1288 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001289 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001290 u32 filter_mask,
1291 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001292 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001293 struct nlmsghdr *nlh,
1294 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001295 int (*ndo_change_carrier)(struct net_device *dev,
1296 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001297 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001298 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001299 int (*ndo_get_phys_port_name)(struct net_device *dev,
1300 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001301 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1302 struct udp_tunnel_info *ti);
1303 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1304 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001305 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1306 struct net_device *dev);
1307 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1308 void *priv);
1309
1310 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1311 struct net_device *dev,
1312 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001313 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001314 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1315 int queue_index,
1316 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001317 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001318 int (*ndo_change_proto_down)(struct net_device *dev,
1319 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001320 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1321 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001322 void (*ndo_set_rx_headroom)(struct net_device *dev,
1323 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001324 int (*ndo_xdp)(struct net_device *dev,
1325 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001326};
1327
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001328/**
1329 * enum net_device_priv_flags - &struct net_device priv_flags
1330 *
1331 * These are the &struct net_device, they are only set internally
1332 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001333 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001334 * during any kernel release.
1335 *
1336 * You should have a pretty good reason to be extending these flags.
1337 *
1338 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1339 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001340 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001341 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001342 * @IFF_WAN_HDLC: WAN HDLC device
1343 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1344 * release skb->dst
1345 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1346 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1347 * @IFF_MACVLAN_PORT: device used as macvlan port
1348 * @IFF_BRIDGE_PORT: device used as bridge port
1349 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1350 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1351 * @IFF_UNICAST_FLT: Supports unicast filtering
1352 * @IFF_TEAM_PORT: device used as team port
1353 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1354 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1355 * change when it's running
1356 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001357 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1358 * underlying stacked devices
1359 * @IFF_IPVLAN_MASTER: IPvlan master device
1360 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001361 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001362 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001363 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001364 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001365 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001366 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001367 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1368 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001369 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001370 */
1371enum netdev_priv_flags {
1372 IFF_802_1Q_VLAN = 1<<0,
1373 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001374 IFF_BONDING = 1<<2,
1375 IFF_ISATAP = 1<<3,
1376 IFF_WAN_HDLC = 1<<4,
1377 IFF_XMIT_DST_RELEASE = 1<<5,
1378 IFF_DONT_BRIDGE = 1<<6,
1379 IFF_DISABLE_NETPOLL = 1<<7,
1380 IFF_MACVLAN_PORT = 1<<8,
1381 IFF_BRIDGE_PORT = 1<<9,
1382 IFF_OVS_DATAPATH = 1<<10,
1383 IFF_TX_SKB_SHARING = 1<<11,
1384 IFF_UNICAST_FLT = 1<<12,
1385 IFF_TEAM_PORT = 1<<13,
1386 IFF_SUPP_NOFCS = 1<<14,
1387 IFF_LIVE_ADDR_CHANGE = 1<<15,
1388 IFF_MACVLAN = 1<<16,
1389 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1390 IFF_IPVLAN_MASTER = 1<<18,
1391 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001392 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001393 IFF_NO_QUEUE = 1<<21,
1394 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001395 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001396 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001397 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001398 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001399 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001400};
1401
1402#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1403#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001404#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001405#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001406#define IFF_WAN_HDLC IFF_WAN_HDLC
1407#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1408#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1409#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1410#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1411#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1412#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1413#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1414#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1415#define IFF_TEAM_PORT IFF_TEAM_PORT
1416#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1417#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1418#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001419#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001420#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1421#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001422#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001423#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001424#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001425#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001426#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001427#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001428#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001429
Karoly Kemeny536721b2014-07-30 20:27:36 +02001430/**
1431 * struct net_device - The DEVICE structure.
1432 * Actually, this whole structure is a big mistake. It mixes I/O
1433 * data with strictly "high-level" data, and it has to know about
1434 * almost every data structure used in the INET module.
1435 *
1436 * @name: This is the first field of the "visible" part of this structure
1437 * (i.e. as seen by users in the "Space.c" file). It is the name
1438 * of the interface.
1439 *
1440 * @name_hlist: Device name hash chain, please keep it close to name[]
1441 * @ifalias: SNMP alias
1442 * @mem_end: Shared memory end
1443 * @mem_start: Shared memory start
1444 * @base_addr: Device I/O address
1445 * @irq: Device IRQ number
1446 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001447 * @carrier_changes: Stats to monitor carrier on<->off transitions
1448 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001449 * @state: Generic network queuing layer state, see netdev_state_t
1450 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001451 * @napi_list: List entry used for polling NAPI devices
1452 * @unreg_list: List entry when we are unregistering the
1453 * device; see the function unregister_netdev
1454 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001455 * @ptype_all: Device-specific packet handlers for all protocols
1456 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001457 *
1458 * @adj_list: Directly linked devices, like slaves for bonding
1459 * @all_adj_list: All linked devices, *including* neighbours
1460 * @features: Currently active device features
1461 * @hw_features: User-changeable features
1462 *
1463 * @wanted_features: User-requested features
1464 * @vlan_features: Mask of features inheritable by VLAN devices
1465 *
1466 * @hw_enc_features: Mask of features inherited by encapsulating devices
1467 * This field indicates what encapsulation
1468 * offloads the hardware is capable of doing,
1469 * and drivers will need to set them appropriately.
1470 *
1471 * @mpls_features: Mask of features inheritable by MPLS
1472 *
1473 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001474 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001475 *
1476 * @stats: Statistics struct, which was left as a legacy, use
1477 * rtnl_link_stats64 instead
1478 *
1479 * @rx_dropped: Dropped packets by core network,
1480 * do not use this in drivers
1481 * @tx_dropped: Dropped packets by core network,
1482 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001483 * @rx_nohandler: nohandler dropped packets by core network on
1484 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001485 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001486 * @wireless_handlers: List of functions to handle Wireless Extensions,
1487 * instead of ioctl,
1488 * see <net/iw_handler.h> for details.
1489 * @wireless_data: Instance data managed by the core of wireless extensions
1490 *
1491 * @netdev_ops: Includes several pointers to callbacks,
1492 * if one wants to override the ndo_*() functions
1493 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001494 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1495 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001496 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001497 * of Layer 2 headers.
1498 *
1499 * @flags: Interface flags (a la BSD)
1500 * @priv_flags: Like 'flags' but invisible to userspace,
1501 * see if.h for the definitions
1502 * @gflags: Global flags ( kept as legacy )
1503 * @padded: How much padding added by alloc_netdev()
1504 * @operstate: RFC2863 operstate
1505 * @link_mode: Mapping policy to operstate
1506 * @if_port: Selectable AUI, TP, ...
1507 * @dma: DMA channel
1508 * @mtu: Interface MTU value
1509 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001510 * @hard_header_len: Maximum hardware header length.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001511 *
1512 * @needed_headroom: Extra headroom the hardware may need, but not in all
1513 * cases can this be guaranteed
1514 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1515 * cases can this be guaranteed. Some cases also use
1516 * LL_MAX_HEADER instead to allocate the skb
1517 *
1518 * interface address info:
1519 *
1520 * @perm_addr: Permanent hw address
1521 * @addr_assign_type: Hw address assignment type
1522 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001523 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001524 * @dev_id: Used to differentiate devices that share
1525 * the same link layer address
1526 * @dev_port: Used to differentiate devices that share
1527 * the same function
1528 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001529 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001530 * has been enabled due to the need to listen to
1531 * additional unicast addresses in a device that
1532 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001533 * @uc: unicast mac addresses
1534 * @mc: multicast mac addresses
1535 * @dev_addrs: list of device hw addresses
1536 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001537 * @promiscuity: Number of times the NIC is told to work in
1538 * promiscuous mode; if it becomes 0 the NIC will
1539 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001540 * @allmulti: Counter, enables or disables allmulticast mode
1541 *
1542 * @vlan_info: VLAN info
1543 * @dsa_ptr: dsa specific data
1544 * @tipc_ptr: TIPC specific data
1545 * @atalk_ptr: AppleTalk link
1546 * @ip_ptr: IPv4 specific data
1547 * @dn_ptr: DECnet specific data
1548 * @ip6_ptr: IPv6 specific data
1549 * @ax25_ptr: AX.25 specific data
1550 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1551 *
1552 * @last_rx: Time of last Rx
1553 * @dev_addr: Hw address (before bcast,
1554 * because most packets are unicast)
1555 *
1556 * @_rx: Array of RX queues
1557 * @num_rx_queues: Number of RX queues
1558 * allocated at register_netdev() time
1559 * @real_num_rx_queues: Number of RX queues currently active in device
1560 *
1561 * @rx_handler: handler for received packets
1562 * @rx_handler_data: XXX: need comments on this one
1563 * @ingress_queue: XXX: need comments on this one
1564 * @broadcast: hw bcast address
1565 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001566 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1567 * indexed by RX queue number. Assigned by driver.
1568 * This must only be set if the ndo_rx_flow_steer
1569 * operation is defined
1570 * @index_hlist: Device index hash chain
1571 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001572 * @_tx: Array of TX queues
1573 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1574 * @real_num_tx_queues: Number of TX queues currently active in device
1575 * @qdisc: Root qdisc from userspace point of view
1576 * @tx_queue_len: Max frames per queue allowed
1577 * @tx_global_lock: XXX: need comments on this one
1578 *
1579 * @xps_maps: XXX: need comments on this one
1580 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001581 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001582 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001583 * @watchdog_timer: List of timers
1584 *
1585 * @pcpu_refcnt: Number of references to this device
1586 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001587 * @link_watch_list: XXX: need comments on this one
1588 *
1589 * @reg_state: Register/unregister state machine
1590 * @dismantle: Device is going to be freed
1591 * @rtnl_link_state: This enum represents the phases of creating
1592 * a new link
1593 *
1594 * @destructor: Called from unregister,
1595 * can be used to call free_netdev
1596 * @npinfo: XXX: need comments on this one
1597 * @nd_net: Network namespace this network device is inside
1598 *
1599 * @ml_priv: Mid-layer private
1600 * @lstats: Loopback statistics
1601 * @tstats: Tunnel statistics
1602 * @dstats: Dummy statistics
1603 * @vstats: Virtual ethernet statistics
1604 *
1605 * @garp_port: GARP
1606 * @mrp_port: MRP
1607 *
1608 * @dev: Class/net/name entry
1609 * @sysfs_groups: Space for optional device, statistics and wireless
1610 * sysfs groups
1611 *
1612 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1613 * @rtnl_link_ops: Rtnl_link_ops
1614 *
1615 * @gso_max_size: Maximum size of generic segmentation offload
1616 * @gso_max_segs: Maximum number of segments that can be passed to the
1617 * NIC for GSO
1618 *
1619 * @dcbnl_ops: Data Center Bridging netlink ops
1620 * @num_tc: Number of traffic classes in the net device
1621 * @tc_to_txq: XXX: need comments on this one
1622 * @prio_tc_map XXX: need comments on this one
1623 *
1624 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1625 *
1626 * @priomap: XXX: need comments on this one
1627 * @phydev: Physical device may attach itself
1628 * for hardware timestamping
1629 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001630 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1631 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001632 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001633 * @proto_down: protocol port state information can be sent to the
1634 * switch driver and used to set the phys state of the
1635 * switch port.
1636 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 * FIXME: cleanup struct net_device such that network protocol info
1638 * moves out.
1639 */
1640
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001641struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001643 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001644 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 /*
1646 * I/O specific fields
1647 * FIXME: Merge these and struct ifmap into one
1648 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001649 unsigned long mem_end;
1650 unsigned long mem_start;
1651 unsigned long base_addr;
1652 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001654 atomic_t carrier_changes;
1655
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001657 * Some hardware also needs these fields (state,dev_list,
1658 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 * part of the usual set specified in Space.c.
1660 */
1661
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 unsigned long state;
1663
Pavel Emelianov7562f872007-05-03 15:13:45 -07001664 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001665 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001666 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001667 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001668 struct list_head ptype_all;
1669 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001670
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001671 struct {
1672 struct list_head upper;
1673 struct list_head lower;
1674 } adj_list;
1675
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001676 struct {
1677 struct list_head upper;
1678 struct list_head lower;
1679 } all_adj_list;
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001680
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001681 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001682 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001683 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001684 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001685 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001686 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001687 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001688
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001690 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691
Rusty Russellc45d2862007-03-28 14:29:08 -07001692 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001693
Eric Dumazet015f0682014-03-27 08:45:56 -07001694 atomic_long_t rx_dropped;
1695 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001696 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
Johannes Bergb86e0282007-04-26 20:48:23 -07001698#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001699 const struct iw_handler_def *wireless_handlers;
1700 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001701#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001702 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001703 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001704#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001705 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001706#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001707#ifdef CONFIG_NET_L3_MASTER_DEV
1708 const struct l3mdev_ops *l3mdev_ops;
1709#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001710#if IS_ENABLED(CONFIG_IPV6)
1711 const struct ndisc_ops *ndisc_ops;
1712#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001714 const struct header_ops *header_ops;
1715
Karoly Kemeny536721b2014-07-30 20:27:36 +02001716 unsigned int flags;
1717 unsigned int priv_flags;
1718
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001720 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721
Karoly Kemeny536721b2014-07-30 20:27:36 +02001722 unsigned char operstate;
1723 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001724
Karoly Kemeny536721b2014-07-30 20:27:36 +02001725 unsigned char if_port;
1726 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001727
Karoly Kemeny536721b2014-07-30 20:27:36 +02001728 unsigned int mtu;
1729 unsigned short type;
1730 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
Johannes Bergf5184d22008-05-12 20:48:31 -07001732 unsigned short needed_headroom;
1733 unsigned short needed_tailroom;
1734
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001736 unsigned char perm_addr[MAX_ADDR_LEN];
1737 unsigned char addr_assign_type;
1738 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001739 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001740 unsigned short dev_id;
1741 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001742 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001743 unsigned char name_assign_type;
1744 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001745 struct netdev_hw_addr_list uc;
1746 struct netdev_hw_addr_list mc;
1747 struct netdev_hw_addr_list dev_addrs;
1748
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001749#ifdef CONFIG_SYSFS
1750 struct kset *queues_kset;
1751#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001752 unsigned int promiscuity;
1753 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001756 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001757
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001758#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001759 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001760#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001761#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001762 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001763#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001764#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001765 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001766#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001767 void *atalk_ptr;
1768 struct in_device __rcu *ip_ptr;
1769 struct dn_dev __rcu *dn_ptr;
1770 struct inet6_dev __rcu *ip6_ptr;
1771 void *ax25_ptr;
1772 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001773 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001774#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1775 struct mpls_dev __rcu *mpls_ptr;
1776#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001778/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001779 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001780 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001781 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001782
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001783 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001784 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001785
Michael Daltona953be52014-01-16 22:23:28 -08001786#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001787 struct netdev_rx_queue *_rx;
1788
Tom Herbert0a9627f2010-03-16 08:03:29 +00001789 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001790 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001791#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001792
Eric Dumazet3b47d302014-11-06 21:09:44 -08001793 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001794 rx_handler_func_t __rcu *rx_handler;
1795 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001796
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001797#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001798 struct tcf_proto __rcu *ingress_cl_list;
1799#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001800 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001801#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001802 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001803#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001804
Karoly Kemeny536721b2014-07-30 20:27:36 +02001805 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001806#ifdef CONFIG_RFS_ACCEL
1807 struct cpu_rmap *rx_cpu_rmap;
1808#endif
1809 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001810
1811/*
1812 * Cache lines mostly used on transmit path
1813 */
David S. Millere8a04642008-07-17 00:34:19 -07001814 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1815 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001816 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001817 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001818#ifdef CONFIG_NET_SCHED
1819 DECLARE_HASHTABLE (qdisc_hash, 4);
1820#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001821 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001822 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001823 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001824
Tom Herbertbf264142010-11-26 08:36:09 +00001825#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001826 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001827#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001828#ifdef CONFIG_NET_CLS_ACT
1829 struct tcf_proto __rcu *egress_cl_list;
1830#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001831
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001832 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001833 struct timer_list watchdog_timer;
1834
Eric Dumazet29b44332010-10-11 10:22:12 +00001835 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837
Eric Dumazete014deb2009-11-17 05:59:21 +00001838 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001839
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001841 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 NETREG_UNREGISTERING, /* called unregister_netdevice */
1843 NETREG_UNREGISTERED, /* completed unregister todo */
1844 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001845 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001846 } reg_state:8;
1847
Karoly Kemeny536721b2014-07-30 20:27:36 +02001848 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001849
1850 enum {
1851 RTNL_LINK_INITIALIZED,
1852 RTNL_LINK_INITIALIZING,
1853 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001855 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001858 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859#endif
David S. Millereae792b2008-07-15 03:03:33 -07001860
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001861 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001862
David S. Miller49517042008-05-12 03:29:11 -07001863 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001864 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001865 void *ml_priv;
1866 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001867 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001868 struct pcpu_dstats __percpu *dstats;
1869 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001870 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001871
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001872 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001873 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001875 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001876 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001877 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001878
Patrick McHardy38f7b872007-06-13 12:03:51 -07001879 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001880
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001881 /* for setting kernel sock attribute on TCP connection setup */
1882#define GSO_MAX_SIZE 65536
1883 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001884#define GSO_MAX_SEGS 65535
1885 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001886
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001887#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001888 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001889#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001890 u8 num_tc;
1891 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1892 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001893
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001894#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001895 unsigned int fcoe_ddp_xid;
1896#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001897#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001898 struct netprio_map __rcu *priomap;
1899#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001900 struct phy_device *phydev;
1901 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001902 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001903 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001905#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906
1907#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908
David S. Millere8a04642008-07-17 00:34:19 -07001909static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001910int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1911{
1912 return dev->prio_tc_map[prio & TC_BITMASK];
1913}
1914
1915static inline
1916int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1917{
1918 if (tc >= dev->num_tc)
1919 return -EINVAL;
1920
1921 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1922 return 0;
1923}
1924
1925static inline
1926void netdev_reset_tc(struct net_device *dev)
1927{
1928 dev->num_tc = 0;
1929 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1930 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1931}
1932
1933static inline
1934int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1935{
1936 if (tc >= dev->num_tc)
1937 return -EINVAL;
1938
1939 dev->tc_to_txq[tc].count = count;
1940 dev->tc_to_txq[tc].offset = offset;
1941 return 0;
1942}
1943
1944static inline
1945int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1946{
1947 if (num_tc > TC_MAX_QUEUE)
1948 return -EINVAL;
1949
1950 dev->num_tc = num_tc;
1951 return 0;
1952}
1953
1954static inline
1955int netdev_get_num_tc(struct net_device *dev)
1956{
1957 return dev->num_tc;
1958}
1959
1960static inline
David S. Millere8a04642008-07-17 00:34:19 -07001961struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1962 unsigned int index)
1963{
1964 return &dev->_tx[index];
1965}
1966
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001967static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1968 const struct sk_buff *skb)
1969{
1970 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1971}
1972
David S. Millere8a04642008-07-17 00:34:19 -07001973static inline void netdev_for_each_tx_queue(struct net_device *dev,
1974 void (*f)(struct net_device *,
1975 struct netdev_queue *,
1976 void *),
1977 void *arg)
1978{
1979 unsigned int i;
1980
1981 for (i = 0; i < dev->num_tx_queues; i++)
1982 f(dev, &dev->_tx[i], arg);
1983}
1984
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001985#define netdev_lockdep_set_classes(dev) \
1986{ \
1987 static struct lock_class_key qdisc_tx_busylock_key; \
1988 static struct lock_class_key qdisc_running_key; \
1989 static struct lock_class_key qdisc_xmit_lock_key; \
1990 static struct lock_class_key dev_addr_list_lock_key; \
1991 unsigned int i; \
1992 \
1993 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1994 (dev)->qdisc_running_key = &qdisc_running_key; \
1995 lockdep_set_class(&(dev)->addr_list_lock, \
1996 &dev_addr_list_lock_key); \
1997 for (i = 0; i < (dev)->num_tx_queues; i++) \
1998 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1999 &qdisc_xmit_lock_key); \
2000}
2001
Joe Perchesf629d202013-09-26 14:48:15 -07002002struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08002003 struct sk_buff *skb,
2004 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00002005
Paolo Abeni871b6422016-02-26 10:45:37 +01002006/* returns the headroom that the master device needs to take in account
2007 * when forwarding to this dev
2008 */
2009static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2010{
2011 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2012}
2013
2014static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2015{
2016 if (dev->netdev_ops->ndo_set_rx_headroom)
2017 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2018}
2019
2020/* set the device rx headroom to the dev's default */
2021static inline void netdev_reset_rx_headroom(struct net_device *dev)
2022{
2023 netdev_set_rx_headroom(dev, -1);
2024}
2025
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002026/*
2027 * Net namespace inlines
2028 */
2029static inline
2030struct net *dev_net(const struct net_device *dev)
2031{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002032 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002033}
2034
2035static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002036void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002037{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002038 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002039}
2040
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07002041static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002042{
Alexander Duyck3fc88672014-09-12 23:12:46 -07002043#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002044 if (dev->dsa_ptr != NULL)
2045 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002046#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002047 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00002048}
2049
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002050/**
2051 * netdev_priv - access network device private data
2052 * @dev: network device
2053 *
2054 * Get network device private data
2055 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002056static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002058 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059}
2060
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061/* Set the sysfs physical device reference for the network logical device
2062 * if set prior to registration will cause a symlink during initialization.
2063 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002064#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065
Marcel Holtmann384912e2009-08-31 21:08:19 +00002066/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002067 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002068 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002069 */
2070#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2071
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002072/* Default NAPI poll() weight
2073 * Device drivers are strongly advised to not use bigger value
2074 */
2075#define NAPI_POLL_WEIGHT 64
2076
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002077/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002078 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002079 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002080 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002081 * @poll: polling function
2082 * @weight: default weight
2083 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002084 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2085 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002086 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002087void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2088 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002089
Alexander Duyckd8156532008-07-08 15:13:05 -07002090/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002091 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002092 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002093 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002094 * @poll: polling function
2095 * @weight: default weight
2096 *
2097 * This variant of netif_napi_add() should be used from drivers using NAPI
2098 * to exclusively poll a TX queue.
2099 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2100 */
2101static inline void netif_tx_napi_add(struct net_device *dev,
2102 struct napi_struct *napi,
2103 int (*poll)(struct napi_struct *, int),
2104 int weight)
2105{
2106 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2107 netif_napi_add(dev, napi, poll, weight);
2108}
2109
2110/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002111 * netif_napi_del - remove a NAPI context
2112 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002113 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002114 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002115 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002116void netif_napi_del(struct napi_struct *napi);
2117
2118struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002119 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002120 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002121
Herbert Xu74895942009-05-26 18:50:27 +00002122 /* Length of frag0. */
2123 unsigned int frag0_len;
2124
Herbert Xu86911732009-01-29 14:19:50 +00002125 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002126 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002127
Herbert Xud565b0a2008-12-15 23:38:52 -08002128 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002129 u16 flush;
2130
2131 /* Save the IP ID here and check when we get to the transport layer */
2132 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002133
2134 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002135 u16 count;
2136
Tom Herbert15e23962015-02-10 16:30:31 -08002137 /* Start offset for remote checksum offload */
2138 u16 gro_remcsum_start;
2139
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002140 /* jiffies when first packet was created/queued */
2141 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002142
Tom Herbertafe93322014-09-17 12:25:57 -07002143 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002144 u16 proto;
2145
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002146 /* This is non-zero if the packet may be of the same flow. */
2147 u8 same_flow:1;
2148
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002149 /* Used in tunnel GRO receive */
2150 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002151
2152 /* GRO checksum is valid */
2153 u8 csum_valid:1;
2154
Tom Herbert662880f2014-08-27 21:26:56 -07002155 /* Number of checksums via CHECKSUM_UNNECESSARY */
2156 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002157
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002158 /* Free the skb? */
2159 u8 free:2;
2160#define NAPI_GRO_FREE 1
2161#define NAPI_GRO_FREE_STOLEN_HEAD 2
2162
Tom Herbertefc98d02014-10-03 15:48:08 -07002163 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2164 u8 is_ipv6:1;
2165
Alexander Duycka0ca1532016-04-05 09:13:39 -07002166 /* Used in GRE, set in fou/gue_gro_receive */
2167 u8 is_fou:1;
2168
Alexander Duyck15305452016-04-10 21:44:57 -04002169 /* Used to determine if flush_id can be ignored */
2170 u8 is_atomic:1;
2171
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002172 /* Number of gro_receive callbacks this packet already went through */
2173 u8 recursion_counter:4;
2174
2175 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002176
Jerry Chubf5a7552014-01-07 10:23:19 -08002177 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2178 __wsum csum;
2179
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002180 /* used in skb_gro_receive() slow path */
2181 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002182};
2183
2184#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002185
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002186#define GRO_RECURSION_LIMIT 15
2187static inline int gro_recursion_inc_test(struct sk_buff *skb)
2188{
2189 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2190}
2191
2192typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2193static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2194 struct sk_buff **head,
2195 struct sk_buff *skb)
2196{
2197 if (unlikely(gro_recursion_inc_test(skb))) {
2198 NAPI_GRO_CB(skb)->flush |= 1;
2199 return NULL;
2200 }
2201
2202 return cb(head, skb);
2203}
2204
2205typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2206 struct sk_buff *);
2207static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2208 struct sock *sk,
2209 struct sk_buff **head,
2210 struct sk_buff *skb)
2211{
2212 if (unlikely(gro_recursion_inc_test(skb))) {
2213 NAPI_GRO_CB(skb)->flush |= 1;
2214 return NULL;
2215 }
2216
2217 return cb(sk, head, skb);
2218}
2219
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002221 __be16 type; /* This is really htons(ether_type). */
2222 struct net_device *dev; /* NULL is wildcarded here */
2223 int (*func) (struct sk_buff *,
2224 struct net_device *,
2225 struct packet_type *,
2226 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002227 bool (*id_match)(struct packet_type *ptype,
2228 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 void *af_packet_priv;
2230 struct list_head list;
2231};
2232
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002233struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2235 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002237 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002238 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002239};
2240
2241struct packet_offload {
2242 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002243 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002244 struct offload_callbacks callbacks;
2245 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246};
2247
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002248/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002249struct pcpu_sw_netstats {
2250 u64 rx_packets;
2251 u64 rx_bytes;
2252 u64 tx_packets;
2253 u64 tx_bytes;
2254 struct u64_stats_sync syncp;
2255};
2256
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002257#define __netdev_alloc_pcpu_stats(type, gfp) \
2258({ \
2259 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2260 if (pcpu_stats) { \
2261 int __cpu; \
2262 for_each_possible_cpu(__cpu) { \
2263 typeof(type) *stat; \
2264 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2265 u64_stats_init(&stat->syncp); \
2266 } \
2267 } \
2268 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002269})
2270
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002271#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002272 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002273
Jiri Pirko764f5e52015-12-03 12:12:12 +01002274enum netdev_lag_tx_type {
2275 NETDEV_LAG_TX_TYPE_UNKNOWN,
2276 NETDEV_LAG_TX_TYPE_RANDOM,
2277 NETDEV_LAG_TX_TYPE_BROADCAST,
2278 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2279 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2280 NETDEV_LAG_TX_TYPE_HASH,
2281};
2282
2283struct netdev_lag_upper_info {
2284 enum netdev_lag_tx_type tx_type;
2285};
2286
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002287struct netdev_lag_lower_state_info {
2288 u8 link_up : 1,
2289 tx_enabled : 1;
2290};
2291
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292#include <linux/notifier.h>
2293
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002294/* netdevice notifier chain. Please remember to update the rtnetlink
2295 * notification exclusion list in rtnetlink_event() when adding new
2296 * types.
2297 */
2298#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2299#define NETDEV_DOWN 0x0002
2300#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2301 detected a hardware crash and restarted
2302 - we can use this eg to kick tcp sessions
2303 once done */
2304#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2305#define NETDEV_REGISTER 0x0005
2306#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002307#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002308#define NETDEV_CHANGEADDR 0x0008
2309#define NETDEV_GOING_DOWN 0x0009
2310#define NETDEV_CHANGENAME 0x000A
2311#define NETDEV_FEAT_CHANGE 0x000B
2312#define NETDEV_BONDING_FAILOVER 0x000C
2313#define NETDEV_PRE_UP 0x000D
2314#define NETDEV_PRE_TYPE_CHANGE 0x000E
2315#define NETDEV_POST_TYPE_CHANGE 0x000F
2316#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002317#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002318#define NETDEV_RELEASE 0x0012
2319#define NETDEV_NOTIFY_PEERS 0x0013
2320#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002321#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002322#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002323#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002324#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002325#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002326#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002327#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002328#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002329#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002330
Joe Perchesf629d202013-09-26 14:48:15 -07002331int register_netdevice_notifier(struct notifier_block *nb);
2332int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002333
2334struct netdev_notifier_info {
2335 struct net_device *dev;
2336};
2337
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002338struct netdev_notifier_change_info {
2339 struct netdev_notifier_info info; /* must be first */
2340 unsigned int flags_changed;
2341};
2342
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002343struct netdev_notifier_changeupper_info {
2344 struct netdev_notifier_info info; /* must be first */
2345 struct net_device *upper_dev; /* new upper dev */
2346 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002347 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002348 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002349};
2350
Jiri Pirko04d48262015-12-03 12:12:15 +01002351struct netdev_notifier_changelowerstate_info {
2352 struct netdev_notifier_info info; /* must be first */
2353 void *lower_state_info; /* is lower dev state */
2354};
2355
Cong Wang75538c22013-05-29 11:30:50 +08002356static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2357 struct net_device *dev)
2358{
2359 info->dev = dev;
2360}
2361
Jiri Pirko351638e2013-05-28 01:30:21 +00002362static inline struct net_device *
2363netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2364{
2365 return info->dev;
2366}
2367
Joe Perchesf629d202013-09-26 14:48:15 -07002368int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002369
2370
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371extern rwlock_t dev_base_lock; /* Device list lock */
2372
Eric W. Biederman881d9662007-09-17 11:56:21 -07002373#define for_each_netdev(net, d) \
2374 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002375#define for_each_netdev_reverse(net, d) \
2376 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002377#define for_each_netdev_rcu(net, d) \
2378 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002379#define for_each_netdev_safe(net, d, n) \
2380 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2381#define for_each_netdev_continue(net, d) \
2382 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002383#define for_each_netdev_continue_rcu(net, d) \
2384 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002385#define for_each_netdev_in_bond_rcu(bond, slave) \
2386 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002387 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002388#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2389
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002390static inline struct net_device *next_net_device(struct net_device *dev)
2391{
2392 struct list_head *lh;
2393 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002394
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002395 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002396 lh = dev->dev_list.next;
2397 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2398}
2399
Eric Dumazetce81b762009-11-11 17:34:30 +00002400static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2401{
2402 struct list_head *lh;
2403 struct net *net;
2404
2405 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002406 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002407 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2408}
2409
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002410static inline struct net_device *first_net_device(struct net *net)
2411{
2412 return list_empty(&net->dev_base_head) ? NULL :
2413 net_device_entry(net->dev_base_head.next);
2414}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002415
Eric Dumazetccf43432011-01-26 18:08:02 +00002416static inline struct net_device *first_net_device_rcu(struct net *net)
2417{
2418 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2419
2420 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2421}
2422
Joe Perchesf629d202013-09-26 14:48:15 -07002423int netdev_boot_setup_check(struct net_device *dev);
2424unsigned long netdev_boot_base(const char *prefix, int unit);
2425struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2426 const char *hwaddr);
2427struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2428struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2429void dev_add_pack(struct packet_type *pt);
2430void dev_remove_pack(struct packet_type *pt);
2431void __dev_remove_pack(struct packet_type *pt);
2432void dev_add_offload(struct packet_offload *po);
2433void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002435int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002436int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002437struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2438 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002439struct net_device *dev_get_by_name(struct net *net, const char *name);
2440struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2441struct net_device *__dev_get_by_name(struct net *net, const char *name);
2442int dev_alloc_name(struct net_device *dev, const char *name);
2443int dev_open(struct net_device *dev);
2444int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002445int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002446void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002447int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002448int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002449int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002450int register_netdevice(struct net_device *dev);
2451void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2452void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002453static inline void unregister_netdevice(struct net_device *dev)
2454{
2455 unregister_netdevice_queue(dev, NULL);
2456}
2457
Joe Perchesf629d202013-09-26 14:48:15 -07002458int netdev_refcnt_read(const struct net_device *dev);
2459void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002460void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002461void synchronize_net(void);
2462int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002463
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002464DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002465#define XMIT_RECURSION_LIMIT 10
2466
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002467static inline int dev_recursion_level(void)
2468{
2469 return this_cpu_read(xmit_recursion);
2470}
2471
Joe Perchesf629d202013-09-26 14:48:15 -07002472struct net_device *dev_get_by_index(struct net *net, int ifindex);
2473struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2474struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2475int netdev_get_name(struct net *net, char *name, int ifindex);
2476int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002477int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002478
2479static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2480{
2481 return NAPI_GRO_CB(skb)->data_offset;
2482}
2483
2484static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2485{
2486 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2487}
2488
2489static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2490{
2491 NAPI_GRO_CB(skb)->data_offset += len;
2492}
2493
Herbert Xua5b1cf22009-05-26 18:50:28 +00002494static inline void *skb_gro_header_fast(struct sk_buff *skb,
2495 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002496{
Herbert Xu78a478d2009-05-26 18:50:21 +00002497 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002498}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002499
Herbert Xua5b1cf22009-05-26 18:50:28 +00002500static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2501{
2502 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2503}
2504
2505static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2506 unsigned int offset)
2507{
Herbert Xu17dd7592011-07-27 06:16:28 -07002508 if (!pskb_may_pull(skb, hlen))
2509 return NULL;
2510
Herbert Xua5b1cf22009-05-26 18:50:28 +00002511 NAPI_GRO_CB(skb)->frag0 = NULL;
2512 NAPI_GRO_CB(skb)->frag0_len = 0;
Herbert Xu17dd7592011-07-27 06:16:28 -07002513 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002514}
2515
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002516static inline void *skb_gro_network_header(struct sk_buff *skb)
2517{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002518 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2519 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002520}
2521
Jerry Chubf5a7552014-01-07 10:23:19 -08002522static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2523 const void *start, unsigned int len)
2524{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002525 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002526 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2527 csum_partial(start, len, 0));
2528}
2529
Tom Herbert573e8fc2014-08-22 13:33:47 -07002530/* GRO checksum functions. These are logical equivalents of the normal
2531 * checksum functions (in skbuff.h) except that they operate on the GRO
2532 * offsets and fields in sk_buff.
2533 */
2534
2535__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2536
Tom Herbert15e23962015-02-10 16:30:31 -08002537static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2538{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002539 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002540}
2541
Tom Herbert573e8fc2014-08-22 13:33:47 -07002542static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2543 bool zero_okay,
2544 __sum16 check)
2545{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002546 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2547 skb_checksum_start_offset(skb) <
2548 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002549 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002550 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002551 (!zero_okay || check));
2552}
2553
2554static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2555 __wsum psum)
2556{
2557 if (NAPI_GRO_CB(skb)->csum_valid &&
2558 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2559 return 0;
2560
2561 NAPI_GRO_CB(skb)->csum = psum;
2562
2563 return __skb_gro_checksum_complete(skb);
2564}
2565
Tom Herbert573e8fc2014-08-22 13:33:47 -07002566static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2567{
Tom Herbert662880f2014-08-27 21:26:56 -07002568 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2569 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2570 NAPI_GRO_CB(skb)->csum_cnt--;
2571 } else {
2572 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2573 * verified a new top level checksum or an encapsulated one
2574 * during GRO. This saves work if we fallback to normal path.
2575 */
2576 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002577 }
2578}
2579
2580#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2581 compute_pseudo) \
2582({ \
2583 __sum16 __ret = 0; \
2584 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2585 __ret = __skb_gro_checksum_validate_complete(skb, \
2586 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002587 if (__ret) \
2588 __skb_mark_checksum_bad(skb); \
2589 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002590 skb_gro_incr_csum_unnecessary(skb); \
2591 __ret; \
2592})
2593
2594#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2595 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2596
2597#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2598 compute_pseudo) \
2599 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2600
2601#define skb_gro_checksum_simple_validate(skb) \
2602 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2603
Tom Herbertd96535a2014-08-31 15:12:42 -07002604static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2605{
2606 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2607 !NAPI_GRO_CB(skb)->csum_valid);
2608}
2609
2610static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2611 __sum16 check, __wsum pseudo)
2612{
2613 NAPI_GRO_CB(skb)->csum = ~pseudo;
2614 NAPI_GRO_CB(skb)->csum_valid = 1;
2615}
2616
2617#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2618do { \
2619 if (__skb_gro_checksum_convert_check(skb)) \
2620 __skb_gro_checksum_convert(skb, check, \
2621 compute_pseudo(skb, proto)); \
2622} while (0)
2623
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002624struct gro_remcsum {
2625 int offset;
2626 __wsum delta;
2627};
2628
2629static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2630{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002631 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002632 grc->delta = 0;
2633}
2634
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002635static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2636 unsigned int off, size_t hdrlen,
2637 int start, int offset,
2638 struct gro_remcsum *grc,
2639 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002640{
2641 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002642 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002643
2644 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2645
Tom Herbert15e23962015-02-10 16:30:31 -08002646 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002647 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2648 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002649 }
2650
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002651 ptr = skb_gro_header_fast(skb, off);
2652 if (skb_gro_header_hard(skb, off + plen)) {
2653 ptr = skb_gro_header_slow(skb, off + plen, off);
2654 if (!ptr)
2655 return NULL;
2656 }
2657
2658 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2659 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002660
2661 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002662 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002663
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002664 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002665 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002666
2667 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002668}
2669
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002670static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2671 struct gro_remcsum *grc)
2672{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002673 void *ptr;
2674 size_t plen = grc->offset + sizeof(u16);
2675
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002676 if (!grc->delta)
2677 return;
2678
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002679 ptr = skb_gro_header_fast(skb, grc->offset);
2680 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2681 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2682 if (!ptr)
2683 return;
2684 }
2685
2686 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002687}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002688
Tom Herbert6ae23ad2015-12-14 11:19:46 -08002689struct skb_csum_offl_spec {
2690 __u16 ipv4_okay:1,
2691 ipv6_okay:1,
2692 encap_okay:1,
2693 ip_options_okay:1,
2694 ext_hdrs_okay:1,
2695 tcp_okay:1,
2696 udp_okay:1,
2697 sctp_okay:1,
2698 vlan_okay:1,
2699 no_encapped_ipv6:1,
2700 no_not_encapped:1;
2701};
2702
2703bool __skb_csum_offload_chk(struct sk_buff *skb,
2704 const struct skb_csum_offl_spec *spec,
2705 bool *csum_encapped,
2706 bool csum_help);
2707
2708static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2709 const struct skb_csum_offl_spec *spec,
2710 bool *csum_encapped,
2711 bool csum_help)
2712{
2713 if (skb->ip_summed != CHECKSUM_PARTIAL)
2714 return false;
2715
2716 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2717}
2718
2719static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2720 const struct skb_csum_offl_spec *spec)
2721{
2722 bool csum_encapped;
2723
2724 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2725}
2726
2727static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2728{
2729 static const struct skb_csum_offl_spec csum_offl_spec = {
2730 .ipv4_okay = 1,
2731 .ip_options_okay = 1,
2732 .ipv6_okay = 1,
2733 .vlan_okay = 1,
2734 .tcp_okay = 1,
2735 .udp_okay = 1,
2736 };
2737
2738 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2739}
2740
2741static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2742{
2743 static const struct skb_csum_offl_spec csum_offl_spec = {
2744 .ipv4_okay = 1,
2745 .ip_options_okay = 1,
2746 .tcp_okay = 1,
2747 .udp_okay = 1,
2748 .vlan_okay = 1,
2749 };
2750
2751 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2752}
2753
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002754static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2755 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002756 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002757 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002758{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002759 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002760 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002761
2762 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002763}
2764
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002765static inline int dev_parse_header(const struct sk_buff *skb,
2766 unsigned char *haddr)
2767{
2768 const struct net_device *dev = skb->dev;
2769
Patrick McHardy1b833362007-10-18 05:09:28 -07002770 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002771 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002772 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002773}
2774
Willem de Bruijn2793a232016-03-09 21:58:32 -05002775/* ll_header must have at least hard_header_len allocated */
2776static inline bool dev_validate_header(const struct net_device *dev,
2777 char *ll_header, int len)
2778{
2779 if (likely(len >= dev->hard_header_len))
2780 return true;
2781
2782 if (capable(CAP_SYS_RAWIO)) {
2783 memset(ll_header + len, 0, dev->hard_header_len - len);
2784 return true;
2785 }
2786
2787 if (dev->header_ops && dev->header_ops->validate)
2788 return dev->header_ops->validate(ll_header, len);
2789
2790 return false;
2791}
2792
Linus Torvalds1da177e2005-04-16 15:20:36 -07002793typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002794int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002795static inline int unregister_gifconf(unsigned int family)
2796{
2797 return register_gifconf(family, NULL);
2798}
2799
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002800#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002801#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002802struct sd_flow_limit {
2803 u64 count;
2804 unsigned int num_buckets;
2805 unsigned int history_head;
2806 u16 history[FLOW_LIMIT_HISTORY];
2807 u8 buckets[];
2808};
2809
2810extern int netdev_flow_limit_table_len;
2811#endif /* CONFIG_NET_FLOW_LIMIT */
2812
Linus Torvalds1da177e2005-04-16 15:20:36 -07002813/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002814 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002815 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002816struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002817 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002818 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002819
Changli Gaodee42872010-05-02 05:42:16 +00002820 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002821 unsigned int processed;
2822 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002823 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002824#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002825 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002826#endif
2827#ifdef CONFIG_NET_FLOW_LIMIT
2828 struct sd_flow_limit __rcu *flow_limit;
2829#endif
2830 struct Qdisc *output_queue;
2831 struct Qdisc **output_queue_tailp;
2832 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002833
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002834#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002835 /* input_queue_head should be written by cpu owning this struct,
2836 * and only read by other cpus. Worth using a cache line.
2837 */
2838 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2839
2840 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002841 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002842 struct softnet_data *rps_ipi_next;
2843 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002844 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002845#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002846 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002847 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002848 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002849
Linus Torvalds1da177e2005-04-16 15:20:36 -07002850};
2851
Tom Herbert76cc8b12010-05-20 18:37:59 +00002852static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002853{
2854#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002855 sd->input_queue_head++;
2856#endif
2857}
2858
2859static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2860 unsigned int *qtail)
2861{
2862#ifdef CONFIG_RPS
2863 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002864#endif
2865}
2866
Tom Herbert0a9627f2010-03-16 08:03:29 +00002867DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002868
Joe Perchesf629d202013-09-26 14:48:15 -07002869void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002870void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002871
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002872static inline void netif_tx_schedule_all(struct net_device *dev)
2873{
2874 unsigned int i;
2875
2876 for (i = 0; i < dev->num_tx_queues; i++)
2877 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2878}
2879
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002880static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002881{
Tom Herbert734664982011-11-28 16:32:44 +00002882 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002883}
2884
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002885/**
2886 * netif_start_queue - allow transmit
2887 * @dev: network device
2888 *
2889 * Allow upper layers to call the device hard_start_xmit routine.
2890 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002891static inline void netif_start_queue(struct net_device *dev)
2892{
David S. Millere8a04642008-07-17 00:34:19 -07002893 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002894}
2895
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002896static inline void netif_tx_start_all_queues(struct net_device *dev)
2897{
2898 unsigned int i;
2899
2900 for (i = 0; i < dev->num_tx_queues; i++) {
2901 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2902 netif_tx_start_queue(txq);
2903 }
2904}
2905
John Fastabend46e5da40a2014-09-12 20:04:52 -07002906void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002907
Dave Jonesd29f7492008-07-22 14:09:06 -07002908/**
2909 * netif_wake_queue - restart transmit
2910 * @dev: network device
2911 *
2912 * Allow upper layers to call the device hard_start_xmit routine.
2913 * Used for flow control when transmit resources are available.
2914 */
David S. Miller79d16382008-07-08 23:14:46 -07002915static inline void netif_wake_queue(struct net_device *dev)
2916{
David S. Millere8a04642008-07-17 00:34:19 -07002917 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002918}
2919
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002920static inline void netif_tx_wake_all_queues(struct net_device *dev)
2921{
2922 unsigned int i;
2923
2924 for (i = 0; i < dev->num_tx_queues; i++) {
2925 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2926 netif_tx_wake_queue(txq);
2927 }
2928}
2929
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002930static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002931{
Tom Herbert734664982011-11-28 16:32:44 +00002932 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002933}
2934
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002935/**
2936 * netif_stop_queue - stop transmitted packets
2937 * @dev: network device
2938 *
2939 * Stop upper layers calling the device hard_start_xmit routine.
2940 * Used for flow control when transmit resources are unavailable.
2941 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002942static inline void netif_stop_queue(struct net_device *dev)
2943{
David S. Millere8a04642008-07-17 00:34:19 -07002944 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002945}
2946
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002947void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002948
David S. Miller4d295152012-03-07 21:02:35 -05002949static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002950{
Tom Herbert734664982011-11-28 16:32:44 +00002951 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002952}
2953
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002954/**
2955 * netif_queue_stopped - test if transmit queue is flowblocked
2956 * @dev: network device
2957 *
2958 * Test if transmit queue on device is currently unable to send.
2959 */
David S. Miller4d295152012-03-07 21:02:35 -05002960static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002961{
David S. Millere8a04642008-07-17 00:34:19 -07002962 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002963}
2964
David S. Miller4d295152012-03-07 21:02:35 -05002965static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002966{
Tom Herbert734664982011-11-28 16:32:44 +00002967 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2968}
2969
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002970static inline bool
2971netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002972{
2973 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2974}
2975
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002976static inline bool
2977netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2978{
2979 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2980}
2981
Eric Dumazet53511452014-10-08 08:19:27 -07002982/**
2983 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2984 * @dev_queue: pointer to transmit queue
2985 *
2986 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002987 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002988 */
2989static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2990{
2991#ifdef CONFIG_BQL
2992 prefetchw(&dev_queue->dql.num_queued);
2993#endif
2994}
2995
2996/**
2997 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2998 * @dev_queue: pointer to transmit queue
2999 *
3000 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003001 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07003002 */
3003static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3004{
3005#ifdef CONFIG_BQL
3006 prefetchw(&dev_queue->dql.limit);
3007#endif
3008}
3009
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003010static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3011 unsigned int bytes)
3012{
Tom Herbert114cf582011-11-28 16:33:09 +00003013#ifdef CONFIG_BQL
3014 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00003015
3016 if (likely(dql_avail(&dev_queue->dql) >= 0))
3017 return;
3018
3019 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3020
3021 /*
3022 * The XOFF flag must be set before checking the dql_avail below,
3023 * because in netdev_tx_completed_queue we update the dql_completed
3024 * before checking the XOFF flag.
3025 */
3026 smp_mb();
3027
3028 /* check again in case another CPU has just made room avail */
3029 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3030 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003031#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003032}
3033
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003034/**
3035 * netdev_sent_queue - report the number of bytes queued to hardware
3036 * @dev: network device
3037 * @bytes: number of bytes queued to the hardware device queue
3038 *
3039 * Report the number of bytes queued for sending/completion to the network
3040 * device hardware queue. @bytes should be a good approximation and should
3041 * exactly match netdev_completed_queue() @bytes
3042 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003043static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3044{
3045 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3046}
3047
3048static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00003049 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003050{
Tom Herbert114cf582011-11-28 16:33:09 +00003051#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00003052 if (unlikely(!bytes))
3053 return;
3054
3055 dql_completed(&dev_queue->dql, bytes);
3056
3057 /*
3058 * Without the memory barrier there is a small possiblity that
3059 * netdev_tx_sent_queue will miss the update and cause the queue to
3060 * be stopped forever
3061 */
3062 smp_mb();
3063
3064 if (dql_avail(&dev_queue->dql) < 0)
3065 return;
3066
3067 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3068 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00003069#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003070}
3071
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003072/**
3073 * netdev_completed_queue - report bytes and packets completed by device
3074 * @dev: network device
3075 * @pkts: actual number of packets sent over the medium
3076 * @bytes: actual number of bytes sent over the medium
3077 *
3078 * Report the number of bytes and packets transmitted by the network device
3079 * hardware queue over the physical medium, @bytes must exactly match the
3080 * @bytes amount passed to netdev_sent_queue()
3081 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003082static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003083 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003084{
3085 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3086}
3087
3088static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3089{
Tom Herbert114cf582011-11-28 16:33:09 +00003090#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003091 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003092 dql_reset(&q->dql);
3093#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003094}
3095
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003096/**
3097 * netdev_reset_queue - reset the packets and bytes count of a network device
3098 * @dev_queue: network device
3099 *
3100 * Reset the bytes and packet count of a network device and clear the
3101 * software flow control OFF bit for this network device
3102 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003103static inline void netdev_reset_queue(struct net_device *dev_queue)
3104{
3105 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003106}
3107
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003108/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003109 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3110 * @dev: network device
3111 * @queue_index: given tx queue index
3112 *
3113 * Returns 0 if given tx queue index >= number of device tx queues,
3114 * otherwise returns the originally passed tx queue index.
3115 */
3116static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3117{
3118 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3119 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3120 dev->name, queue_index,
3121 dev->real_num_tx_queues);
3122 return 0;
3123 }
3124
3125 return queue_index;
3126}
3127
3128/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003129 * netif_running - test if up
3130 * @dev: network device
3131 *
3132 * Test if the device has been brought up.
3133 */
David S. Miller4d295152012-03-07 21:02:35 -05003134static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003135{
3136 return test_bit(__LINK_STATE_START, &dev->state);
3137}
3138
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003139/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003140 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003141 * stop, and a check if it's stopped. All other device management is
3142 * done at the overall netdevice level.
3143 * Also test the device if we're multiqueue.
3144 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003145
3146/**
3147 * netif_start_subqueue - allow sending packets on subqueue
3148 * @dev: network device
3149 * @queue_index: sub queue index
3150 *
3151 * Start individual transmit queue of a device with multiple transmit queues.
3152 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003153static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3154{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003155 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003156
3157 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003158}
3159
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003160/**
3161 * netif_stop_subqueue - stop sending packets on subqueue
3162 * @dev: network device
3163 * @queue_index: sub queue index
3164 *
3165 * Stop individual transmit queue of a device with multiple transmit queues.
3166 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003167static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3168{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003169 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003170 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003171}
3172
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003173/**
3174 * netif_subqueue_stopped - test status of subqueue
3175 * @dev: network device
3176 * @queue_index: sub queue index
3177 *
3178 * Check individual transmit queue of a device with multiple transmit queues.
3179 */
David S. Miller4d295152012-03-07 21:02:35 -05003180static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3181 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003182{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003183 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003184
3185 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003186}
3187
David S. Miller4d295152012-03-07 21:02:35 -05003188static inline bool netif_subqueue_stopped(const struct net_device *dev,
3189 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003190{
3191 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3192}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003193
John Fastabend46e5da40a2014-09-12 20:04:52 -07003194void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003195
Alexander Duyck537c00d2013-01-10 08:57:02 +00003196#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003197int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003198 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003199#else
3200static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003201 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003202 u16 index)
3203{
3204 return 0;
3205}
3206#endif
3207
Jiri Pirko5605c762015-05-12 14:56:12 +02003208u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3209 unsigned int num_tx_queues);
3210
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003211/*
3212 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3213 * as a distribution range limit for the returned value.
3214 */
3215static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003216 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003217{
3218 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3219}
3220
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003221/**
3222 * netif_is_multiqueue - test if device has multiple transmit queues
3223 * @dev: network device
3224 *
3225 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003226 */
David S. Miller4d295152012-03-07 21:02:35 -05003227static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003228{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003229 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003230}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003231
Joe Perchesf629d202013-09-26 14:48:15 -07003232int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003233
Michael Daltona953be52014-01-16 22:23:28 -08003234#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003235int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003236#else
3237static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3238 unsigned int rxq)
3239{
3240 return 0;
3241}
3242#endif
3243
Michael Daltona953be52014-01-16 22:23:28 -08003244#ifdef CONFIG_SYSFS
3245static inline unsigned int get_netdev_rx_queue_index(
3246 struct netdev_rx_queue *queue)
3247{
3248 struct net_device *dev = queue->dev;
3249 int index = queue - dev->_rx;
3250
3251 BUG_ON(index >= dev->num_rx_queues);
3252 return index;
3253}
3254#endif
3255
Yuval Mintz16917b82012-07-01 03:18:50 +00003256#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003257int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003258
Eric Dumazete6247022013-12-05 04:45:08 -08003259enum skb_free_reason {
3260 SKB_REASON_CONSUMED,
3261 SKB_REASON_DROPPED,
3262};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003263
Eric Dumazete6247022013-12-05 04:45:08 -08003264void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3265void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3266
3267/*
3268 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3269 * interrupt context or with hardware interrupts being disabled.
3270 * (in_irq() || irqs_disabled())
3271 *
3272 * We provide four helpers that can be used in following contexts :
3273 *
3274 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3275 * replacing kfree_skb(skb)
3276 *
3277 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3278 * Typically used in place of consume_skb(skb) in TX completion path
3279 *
3280 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3281 * replacing kfree_skb(skb)
3282 *
3283 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3284 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003285 */
Eric Dumazete6247022013-12-05 04:45:08 -08003286static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3287{
3288 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3289}
3290
3291static inline void dev_consume_skb_irq(struct sk_buff *skb)
3292{
3293 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3294}
3295
3296static inline void dev_kfree_skb_any(struct sk_buff *skb)
3297{
3298 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3299}
3300
3301static inline void dev_consume_skb_any(struct sk_buff *skb)
3302{
3303 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3304}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003305
Joe Perchesf629d202013-09-26 14:48:15 -07003306int netif_rx(struct sk_buff *skb);
3307int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003308int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003309gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3310void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3311struct sk_buff *napi_get_frags(struct napi_struct *napi);
3312gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003313struct packet_offload *gro_find_receive_by_type(__be16 type);
3314struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003315
3316static inline void napi_free_frags(struct napi_struct *napi)
3317{
3318 kfree_skb(napi->skb);
3319 napi->skb = NULL;
3320}
3321
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003322bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003323int netdev_rx_handler_register(struct net_device *dev,
3324 rx_handler_func_t *rx_handler,
3325 void *rx_handler_data);
3326void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003327
Joe Perchesf629d202013-09-26 14:48:15 -07003328bool dev_valid_name(const char *name);
3329int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3330int dev_ethtool(struct net *net, struct ifreq *);
3331unsigned int dev_get_flags(const struct net_device *);
3332int __dev_change_flags(struct net_device *, unsigned int flags);
3333int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003334void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3335 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003336int dev_change_name(struct net_device *, const char *);
3337int dev_set_alias(struct net_device *, const char *, size_t);
3338int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3339int dev_set_mtu(struct net_device *, int);
3340void dev_set_group(struct net_device *, int);
3341int dev_set_mac_address(struct net_device *, struct sockaddr *);
3342int dev_change_carrier(struct net_device *, bool new_carrier);
3343int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003344 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003345int dev_get_phys_port_name(struct net_device *dev,
3346 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003347int dev_change_proto_down(struct net_device *dev, bool proto_down);
Brenden Blancoa7862b42016-07-19 12:16:48 -07003348int dev_change_xdp_fd(struct net_device *dev, int fd);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003349struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003350struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3351 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003352int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003353int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003354bool is_skb_forwardable(const struct net_device *dev,
3355 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003356
David Ahern74b20582016-05-10 11:19:50 -07003357void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3358
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003359extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003360
3361/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003362void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003363
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003364/**
3365 * dev_put - release reference to device
3366 * @dev: network device
3367 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003368 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003369 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003370static inline void dev_put(struct net_device *dev)
3371{
Christoph Lameter933393f2011-12-22 11:58:51 -06003372 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003373}
3374
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003375/**
3376 * dev_hold - get reference to device
3377 * @dev: network device
3378 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003379 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003380 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003381static inline void dev_hold(struct net_device *dev)
3382{
Christoph Lameter933393f2011-12-22 11:58:51 -06003383 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003384}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003385
3386/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3387 * and _off may be called from IRQ context, but it is caller
3388 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003389 *
3390 * The name carrier is inappropriate, these functions should really be
3391 * called netif_lowerlayer_*() because they represent the state of any
3392 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003393 */
3394
Joe Perchesf629d202013-09-26 14:48:15 -07003395void linkwatch_init_dev(struct net_device *dev);
3396void linkwatch_fire_event(struct net_device *dev);
3397void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003398
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003399/**
3400 * netif_carrier_ok - test if carrier present
3401 * @dev: network device
3402 *
3403 * Check if carrier is present on device
3404 */
David S. Miller4d295152012-03-07 21:02:35 -05003405static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003406{
3407 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3408}
3409
Joe Perchesf629d202013-09-26 14:48:15 -07003410unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003411
Joe Perchesf629d202013-09-26 14:48:15 -07003412void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003413
Joe Perchesf629d202013-09-26 14:48:15 -07003414void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003415
Joe Perchesf629d202013-09-26 14:48:15 -07003416void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003417
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003418/**
3419 * netif_dormant_on - mark device as dormant.
3420 * @dev: network device
3421 *
3422 * Mark device as dormant (as per RFC2863).
3423 *
3424 * The dormant state indicates that the relevant interface is not
3425 * actually in a condition to pass packets (i.e., it is not 'up') but is
3426 * in a "pending" state, waiting for some external event. For "on-
3427 * demand" interfaces, this new state identifies the situation where the
3428 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003429 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003430static inline void netif_dormant_on(struct net_device *dev)
3431{
3432 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3433 linkwatch_fire_event(dev);
3434}
3435
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003436/**
3437 * netif_dormant_off - set device as not dormant.
3438 * @dev: network device
3439 *
3440 * Device is not in dormant state.
3441 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003442static inline void netif_dormant_off(struct net_device *dev)
3443{
3444 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3445 linkwatch_fire_event(dev);
3446}
3447
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003448/**
3449 * netif_dormant - test if carrier present
3450 * @dev: network device
3451 *
3452 * Check if carrier is present on device
3453 */
David S. Miller4d295152012-03-07 21:02:35 -05003454static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003455{
3456 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3457}
3458
3459
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003460/**
3461 * netif_oper_up - test if device is operational
3462 * @dev: network device
3463 *
3464 * Check if carrier is operational
3465 */
David S. Miller4d295152012-03-07 21:02:35 -05003466static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003467{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003468 return (dev->operstate == IF_OPER_UP ||
3469 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3470}
3471
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003472/**
3473 * netif_device_present - is device available or removed
3474 * @dev: network device
3475 *
3476 * Check if device has not been removed from system.
3477 */
David S. Miller4d295152012-03-07 21:02:35 -05003478static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003479{
3480 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3481}
3482
Joe Perchesf629d202013-09-26 14:48:15 -07003483void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003484
Joe Perchesf629d202013-09-26 14:48:15 -07003485void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003486
3487/*
3488 * Network interface message level settings
3489 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003490
3491enum {
3492 NETIF_MSG_DRV = 0x0001,
3493 NETIF_MSG_PROBE = 0x0002,
3494 NETIF_MSG_LINK = 0x0004,
3495 NETIF_MSG_TIMER = 0x0008,
3496 NETIF_MSG_IFDOWN = 0x0010,
3497 NETIF_MSG_IFUP = 0x0020,
3498 NETIF_MSG_RX_ERR = 0x0040,
3499 NETIF_MSG_TX_ERR = 0x0080,
3500 NETIF_MSG_TX_QUEUED = 0x0100,
3501 NETIF_MSG_INTR = 0x0200,
3502 NETIF_MSG_TX_DONE = 0x0400,
3503 NETIF_MSG_RX_STATUS = 0x0800,
3504 NETIF_MSG_PKTDATA = 0x1000,
3505 NETIF_MSG_HW = 0x2000,
3506 NETIF_MSG_WOL = 0x4000,
3507};
3508
3509#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3510#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3511#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3512#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3513#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3514#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3515#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3516#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3517#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3518#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3519#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3520#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3521#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3522#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3523#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3524
3525static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3526{
3527 /* use default */
3528 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3529 return default_msg_enable_bits;
3530 if (debug_value == 0) /* no output */
3531 return 0;
3532 /* set low N bits */
3533 return (1 << debug_value) - 1;
3534}
3535
David S. Millerc773e842008-07-08 23:13:53 -07003536static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003537{
David S. Millerc773e842008-07-08 23:13:53 -07003538 spin_lock(&txq->_xmit_lock);
3539 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003540}
3541
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003542static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3543{
3544 spin_lock_bh(&txq->_xmit_lock);
3545 txq->xmit_lock_owner = smp_processor_id();
3546}
3547
David S. Miller4d295152012-03-07 21:02:35 -05003548static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003549{
David S. Miller4d295152012-03-07 21:02:35 -05003550 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003551 if (likely(ok))
3552 txq->xmit_lock_owner = smp_processor_id();
3553 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003554}
3555
David S. Millerc773e842008-07-08 23:13:53 -07003556static inline void __netif_tx_unlock(struct netdev_queue *txq)
3557{
3558 txq->xmit_lock_owner = -1;
3559 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003560}
3561
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003562static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3563{
3564 txq->xmit_lock_owner = -1;
3565 spin_unlock_bh(&txq->_xmit_lock);
3566}
3567
Eric Dumazet08baf562009-05-25 22:58:01 -07003568static inline void txq_trans_update(struct netdev_queue *txq)
3569{
3570 if (txq->xmit_lock_owner != -1)
3571 txq->trans_start = jiffies;
3572}
3573
Florian Westphalba162f82016-05-03 16:31:00 +02003574/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3575static inline void netif_trans_update(struct net_device *dev)
3576{
Florian Westphal9b366272016-05-03 16:33:14 +02003577 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3578
3579 if (txq->trans_start != jiffies)
3580 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003581}
3582
David S. Millerc3f26a22008-07-31 16:58:50 -07003583/**
3584 * netif_tx_lock - grab network device transmit lock
3585 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003586 *
3587 * Get network device transmit lock
3588 */
3589static inline void netif_tx_lock(struct net_device *dev)
3590{
3591 unsigned int i;
3592 int cpu;
3593
3594 spin_lock(&dev->tx_global_lock);
3595 cpu = smp_processor_id();
3596 for (i = 0; i < dev->num_tx_queues; i++) {
3597 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3598
3599 /* We are the only thread of execution doing a
3600 * freeze, but we have to grab the _xmit_lock in
3601 * order to synchronize with threads which are in
3602 * the ->hard_start_xmit() handler and already
3603 * checked the frozen bit.
3604 */
3605 __netif_tx_lock(txq, cpu);
3606 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3607 __netif_tx_unlock(txq);
3608 }
3609}
3610
3611static inline void netif_tx_lock_bh(struct net_device *dev)
3612{
3613 local_bh_disable();
3614 netif_tx_lock(dev);
3615}
3616
Herbert Xu932ff272006-06-09 12:20:56 -07003617static inline void netif_tx_unlock(struct net_device *dev)
3618{
David S. Millere8a04642008-07-17 00:34:19 -07003619 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003620
David S. Millere8a04642008-07-17 00:34:19 -07003621 for (i = 0; i < dev->num_tx_queues; i++) {
3622 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003623
David S. Millerc3f26a22008-07-31 16:58:50 -07003624 /* No need to grab the _xmit_lock here. If the
3625 * queue is not stopped for another reason, we
3626 * force a schedule.
3627 */
3628 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003629 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003630 }
3631 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003632}
3633
3634static inline void netif_tx_unlock_bh(struct net_device *dev)
3635{
David S. Millere8a04642008-07-17 00:34:19 -07003636 netif_tx_unlock(dev);
3637 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003638}
3639
David S. Millerc773e842008-07-08 23:13:53 -07003640#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003641 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003642 __netif_tx_lock(txq, cpu); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003643 } \
3644}
3645
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003646#define HARD_TX_TRYLOCK(dev, txq) \
3647 (((dev->features & NETIF_F_LLTX) == 0) ? \
3648 __netif_tx_trylock(txq) : \
3649 true )
3650
David S. Millerc773e842008-07-08 23:13:53 -07003651#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003652 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003653 __netif_tx_unlock(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003654 } \
3655}
3656
Linus Torvalds1da177e2005-04-16 15:20:36 -07003657static inline void netif_tx_disable(struct net_device *dev)
3658{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003659 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003660 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003661
David S. Millerc3f26a22008-07-31 16:58:50 -07003662 local_bh_disable();
3663 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003664 for (i = 0; i < dev->num_tx_queues; i++) {
3665 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003666
3667 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003668 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003669 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003670 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003671 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003672}
3673
David S. Millere308a5d2008-07-15 00:13:44 -07003674static inline void netif_addr_lock(struct net_device *dev)
3675{
3676 spin_lock(&dev->addr_list_lock);
3677}
3678
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003679static inline void netif_addr_lock_nested(struct net_device *dev)
3680{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003681 int subclass = SINGLE_DEPTH_NESTING;
3682
3683 if (dev->netdev_ops->ndo_get_lock_subclass)
3684 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3685
3686 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003687}
3688
David S. Millere308a5d2008-07-15 00:13:44 -07003689static inline void netif_addr_lock_bh(struct net_device *dev)
3690{
3691 spin_lock_bh(&dev->addr_list_lock);
3692}
3693
3694static inline void netif_addr_unlock(struct net_device *dev)
3695{
3696 spin_unlock(&dev->addr_list_lock);
3697}
3698
3699static inline void netif_addr_unlock_bh(struct net_device *dev)
3700{
3701 spin_unlock_bh(&dev->addr_list_lock);
3702}
3703
Jiri Pirkof001fde2009-05-05 02:48:28 +00003704/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003705 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003706 * rcu_read_lock held.
3707 */
3708#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003709 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003710
Linus Torvalds1da177e2005-04-16 15:20:36 -07003711/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3712
Joe Perchesf629d202013-09-26 14:48:15 -07003713void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003714
3715/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003716struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003717 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003718 void (*setup)(struct net_device *),
3719 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003720#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3721 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003722
Tom Gundersenc835a672014-07-14 16:37:24 +02003723#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3724 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3725 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003726
Joe Perchesf629d202013-09-26 14:48:15 -07003727int register_netdev(struct net_device *dev);
3728void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003729
Jiri Pirko22bedad32010-04-01 21:22:57 +00003730/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003731int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3732 struct netdev_hw_addr_list *from_list, int addr_len);
3733void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3734 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003735int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3736 struct net_device *dev,
3737 int (*sync)(struct net_device *, const unsigned char *),
3738 int (*unsync)(struct net_device *,
3739 const unsigned char *));
3740void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3741 struct net_device *dev,
3742 int (*unsync)(struct net_device *,
3743 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003744void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003745
Jiri Pirkof001fde2009-05-05 02:48:28 +00003746/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003747int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3748 unsigned char addr_type);
3749int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3750 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003751void dev_addr_flush(struct net_device *dev);
3752int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003753
3754/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003755int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3756int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3757int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3758int dev_uc_sync(struct net_device *to, struct net_device *from);
3759int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3760void dev_uc_unsync(struct net_device *to, struct net_device *from);
3761void dev_uc_flush(struct net_device *dev);
3762void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003763
Alexander Duyck670e5b82014-05-28 18:44:46 -07003764/**
3765 * __dev_uc_sync - Synchonize device's unicast list
3766 * @dev: device to sync
3767 * @sync: function to call if address should be added
3768 * @unsync: function to call if address should be removed
3769 *
3770 * Add newly added addresses to the interface, and release
3771 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003772 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003773static inline int __dev_uc_sync(struct net_device *dev,
3774 int (*sync)(struct net_device *,
3775 const unsigned char *),
3776 int (*unsync)(struct net_device *,
3777 const unsigned char *))
3778{
3779 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3780}
3781
3782/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003783 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003784 * @dev: device to sync
3785 * @unsync: function to call if address should be removed
3786 *
3787 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003788 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003789static inline void __dev_uc_unsync(struct net_device *dev,
3790 int (*unsync)(struct net_device *,
3791 const unsigned char *))
3792{
3793 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3794}
3795
Jiri Pirko22bedad32010-04-01 21:22:57 +00003796/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003797int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3798int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3799int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3800int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3801int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3802int dev_mc_sync(struct net_device *to, struct net_device *from);
3803int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3804void dev_mc_unsync(struct net_device *to, struct net_device *from);
3805void dev_mc_flush(struct net_device *dev);
3806void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003807
Alexander Duyck670e5b82014-05-28 18:44:46 -07003808/**
3809 * __dev_mc_sync - Synchonize device's multicast list
3810 * @dev: device to sync
3811 * @sync: function to call if address should be added
3812 * @unsync: function to call if address should be removed
3813 *
3814 * Add newly added addresses to the interface, and release
3815 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003816 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003817static inline int __dev_mc_sync(struct net_device *dev,
3818 int (*sync)(struct net_device *,
3819 const unsigned char *),
3820 int (*unsync)(struct net_device *,
3821 const unsigned char *))
3822{
3823 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3824}
3825
3826/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003827 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003828 * @dev: device to sync
3829 * @unsync: function to call if address should be removed
3830 *
3831 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003832 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003833static inline void __dev_mc_unsync(struct net_device *dev,
3834 int (*unsync)(struct net_device *,
3835 const unsigned char *))
3836{
3837 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3838}
3839
Linus Torvalds1da177e2005-04-16 15:20:36 -07003840/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003841void dev_set_rx_mode(struct net_device *dev);
3842void __dev_set_rx_mode(struct net_device *dev);
3843int dev_set_promiscuity(struct net_device *dev, int inc);
3844int dev_set_allmulti(struct net_device *dev, int inc);
3845void netdev_state_change(struct net_device *dev);
3846void netdev_notify_peers(struct net_device *dev);
3847void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003848/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003849void dev_load(struct net *net, const char *name);
3850struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3851 struct rtnl_link_stats64 *storage);
3852void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3853 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003854
3855extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003856extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003857extern int weight_p;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003858
Joe Perchesf629d202013-09-26 14:48:15 -07003859bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003860struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3861 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003862struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3863 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003864
3865/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003866#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3867 for (iter = &(dev)->adj_list.upper, \
3868 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3869 updev; \
3870 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3871
3872/* iterate through upper list, must be called under RCU read lock */
Veaceslav Falico2f268f12013-09-25 09:20:07 +02003873#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3874 for (iter = &(dev)->all_adj_list.upper, \
3875 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3876 updev; \
3877 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003878
Joe Perchesf629d202013-09-26 14:48:15 -07003879void *netdev_lower_get_next_private(struct net_device *dev,
3880 struct list_head **iter);
3881void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3882 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003883
3884#define netdev_for_each_lower_private(dev, priv, iter) \
3885 for (iter = (dev)->adj_list.lower.next, \
3886 priv = netdev_lower_get_next_private(dev, &(iter)); \
3887 priv; \
3888 priv = netdev_lower_get_next_private(dev, &(iter)))
3889
3890#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3891 for (iter = &(dev)->adj_list.lower, \
3892 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3893 priv; \
3894 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3895
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003896void *netdev_lower_get_next(struct net_device *dev,
3897 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003898
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003899#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003900 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003901 ldev = netdev_lower_get_next(dev, &(iter)); \
3902 ldev; \
3903 ldev = netdev_lower_get_next(dev, &(iter)))
3904
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003905struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3906 struct list_head **iter);
3907struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3908 struct list_head **iter);
3909
3910#define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3911 for (iter = (dev)->all_adj_list.lower.next, \
3912 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3913 ldev; \
3914 ldev = netdev_all_lower_get_next(dev, &(iter)))
3915
3916#define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
Ido Schimmele4961b02016-10-19 16:57:08 +03003917 for (iter = &(dev)->all_adj_list.lower, \
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003918 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3919 ldev; \
3920 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3921
Joe Perchesf629d202013-09-26 14:48:15 -07003922void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003923void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003924struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3925struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3926int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3927int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003928 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003929 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003930void netdev_upper_dev_unlink(struct net_device *dev,
3931 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003932void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003933void *netdev_lower_dev_get_private(struct net_device *dev,
3934 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003935void netdev_lower_state_changed(struct net_device *lower_dev,
3936 void *lower_state_info);
Jiri Pirko18bfb9242016-07-05 11:27:38 +02003937int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3938 struct neighbour *n);
3939void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3940 struct neighbour *n);
Eric Dumazet960fb622014-11-16 06:23:05 -08003941
3942/* RSS keys are 40 or 52 bytes long */
3943#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003944extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003945void netdev_rss_key_fill(void *buffer, size_t len);
3946
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003947int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003948int skb_checksum_help(struct sk_buff *skb);
3949struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3950 netdev_features_t features, bool tx_path);
3951struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3952 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003953
Moni Shoua61bd3852015-02-03 16:48:29 +02003954struct netdev_bonding_info {
3955 ifslave slave;
3956 ifbond master;
3957};
3958
3959struct netdev_notifier_bonding_info {
3960 struct netdev_notifier_info info; /* must be first */
3961 struct netdev_bonding_info bonding_info;
3962};
3963
3964void netdev_bonding_info_change(struct net_device *dev,
3965 struct netdev_bonding_info *bonding_info);
3966
Cong Wang12b00042013-02-05 16:36:38 +00003967static inline
3968struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3969{
3970 return __skb_gso_segment(skb, features, true);
3971}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003972__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003973
3974static inline bool can_checksum_protocol(netdev_features_t features,
3975 __be16 protocol)
3976{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003977 if (protocol == htons(ETH_P_FCOE))
3978 return !!(features & NETIF_F_FCOE_CRC);
3979
3980 /* Assume this is an IP checksum (not SCTP CRC) */
3981
3982 if (features & NETIF_F_HW_CSUM) {
3983 /* Can checksum everything */
3984 return true;
3985 }
3986
3987 switch (protocol) {
3988 case htons(ETH_P_IP):
3989 return !!(features & NETIF_F_IP_CSUM);
3990 case htons(ETH_P_IPV6):
3991 return !!(features & NETIF_F_IPV6_CSUM);
3992 default:
3993 return false;
3994 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003995}
Cong Wang12b00042013-02-05 16:36:38 +00003996
Tom Herbert6ae23ad2015-12-14 11:19:46 -08003997/* Map an ethertype into IP protocol if possible */
3998static inline int eproto_to_ipproto(int eproto)
3999{
4000 switch (eproto) {
4001 case htons(ETH_P_IP):
4002 return IPPROTO_IP;
4003 case htons(ETH_P_IPV6):
4004 return IPPROTO_IPV6;
4005 default:
4006 return -1;
4007 }
4008}
4009
Herbert Xufb286bb2005-11-10 13:01:24 -08004010#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07004011void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08004012#else
4013static inline void netdev_rx_csum_fault(struct net_device *dev)
4014{
4015}
4016#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07004017/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07004018void net_enable_timestamp(void);
4019void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004020
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004021#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07004022int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00004023#else
4024#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004025#endif
4026
David S. Miller47982482014-08-22 16:21:53 -07004027static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004028 struct sk_buff *skb, struct net_device *dev,
4029 bool more)
David S. Miller47982482014-08-22 16:21:53 -07004030{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004031 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07004032 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07004033}
4034
David S. Miller10b3ad82014-08-29 21:07:24 -07004035static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004036 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07004037{
4038 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07004039 int rc;
David S. Miller47982482014-08-22 16:21:53 -07004040
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004041 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07004042 if (rc == NETDEV_TX_OK)
4043 txq_trans_update(txq);
4044
4045 return rc;
David S. Miller47982482014-08-22 16:21:53 -07004046}
4047
Linus Torvalds42a2d922013-11-13 17:40:34 +09004048int netdev_class_create_file_ns(struct class_attribute *class_attr,
4049 const void *ns);
4050void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4051 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04004052
4053static inline int netdev_class_create_file(struct class_attribute *class_attr)
4054{
4055 return netdev_class_create_file_ns(class_attr, NULL);
4056}
4057
4058static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4059{
4060 netdev_class_remove_file_ns(class_attr, NULL);
4061}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004062
Johannes Berg04600792010-08-05 17:45:15 +02004063extern struct kobj_ns_type_operations net_ns_type_operations;
4064
Joe Perchesf629d202013-09-26 14:48:15 -07004065const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004066
Joe Perchesf629d202013-09-26 14:48:15 -07004067void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004068
Michal Kubečekda081432014-05-20 08:29:25 +02004069static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4070 netdev_features_t f2)
4071{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004072 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4073 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004074 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004075 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004076 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004077 }
Michal Kubečekda081432014-05-20 08:29:25 +02004078
Tom Herbertc8cd0982015-12-14 11:19:44 -08004079 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004080}
4081
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004082static inline netdev_features_t netdev_get_wanted_features(
4083 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004084{
4085 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4086}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004087netdev_features_t netdev_increment_features(netdev_features_t all,
4088 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004089
4090/* Allow TSO being used on stacked device :
4091 * Performing the GSO segmentation before last device
4092 * is a performance improvement.
4093 */
4094static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4095 netdev_features_t mask)
4096{
4097 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4098}
4099
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004100int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004101void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004102void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004103
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004104void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4105 struct net_device *dev);
4106
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004107netdev_features_t passthru_features_check(struct sk_buff *skb,
4108 struct net_device *dev,
4109 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004110netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004111
David S. Miller4d295152012-03-07 21:02:35 -05004112static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004113{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004114 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004115
4116 /* check flags correspondence */
4117 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4118 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4119 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4120 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004121 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004122 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4123 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004124 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4125 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004126 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4127 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004128 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4129 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
Alexander Duyck802ab552016-04-10 21:45:03 -04004130 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004131 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004132 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004133
Herbert Xubcd76112006-06-30 13:36:35 -07004134 return (features & feature) == feature;
4135}
4136
David S. Miller4d295152012-03-07 21:02:35 -05004137static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004138{
Herbert Xu278b2512009-06-03 21:20:51 -07004139 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004140 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004141}
4142
Johannes Berg8b86a612015-04-17 15:45:04 +02004143static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004144 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004145{
Jesse Grossfc741212011-01-09 06:23:32 +00004146 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004147 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4148 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004149}
4150
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004151static inline void netif_set_gso_max_size(struct net_device *dev,
4152 unsigned int size)
4153{
4154 dev->gso_max_size = size;
4155}
4156
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004157static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4158 int pulled_hlen, u16 mac_offset,
4159 int mac_len)
4160{
4161 skb->protocol = protocol;
4162 skb->encapsulation = 1;
4163 skb_push(skb, pulled_hlen);
4164 skb_reset_transport_header(skb);
4165 skb->mac_header = mac_offset;
4166 skb->network_header = skb->mac_header + mac_len;
4167 skb->mac_len = mac_len;
4168}
4169
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004170static inline bool netif_is_macsec(const struct net_device *dev)
4171{
4172 return dev->priv_flags & IFF_MACSEC;
4173}
4174
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004175static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004176{
4177 return dev->priv_flags & IFF_MACVLAN;
4178}
4179
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004180static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004181{
4182 return dev->priv_flags & IFF_MACVLAN_PORT;
4183}
4184
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004185static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004186{
4187 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4188}
4189
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004190static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004191{
4192 return dev->priv_flags & IFF_IPVLAN_MASTER;
4193}
4194
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004195static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004196{
4197 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4198}
4199
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004200static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004201{
4202 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4203}
4204
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004205static inline bool netif_supports_nofcs(struct net_device *dev)
4206{
4207 return dev->priv_flags & IFF_SUPP_NOFCS;
4208}
4209
David Ahern007979e2015-09-29 20:07:10 -07004210static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004211{
David Ahern007979e2015-09-29 20:07:10 -07004212 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004213}
4214
David Ahernfee6d4c2015-10-05 08:51:24 -07004215static inline bool netif_is_l3_slave(const struct net_device *dev)
4216{
4217 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4218}
4219
Jiri Pirko0894ae32015-08-27 09:31:19 +02004220static inline bool netif_is_bridge_master(const struct net_device *dev)
4221{
4222 return dev->priv_flags & IFF_EBRIDGE;
4223}
4224
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004225static inline bool netif_is_bridge_port(const struct net_device *dev)
4226{
4227 return dev->priv_flags & IFF_BRIDGE_PORT;
4228}
4229
Jiri Pirko35d4e172015-08-27 09:31:20 +02004230static inline bool netif_is_ovs_master(const struct net_device *dev)
4231{
4232 return dev->priv_flags & IFF_OPENVSWITCH;
4233}
4234
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004235static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004236{
4237 return dev->priv_flags & IFF_TEAM;
4238}
4239
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004240static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004241{
4242 return dev->priv_flags & IFF_TEAM_PORT;
4243}
4244
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004245static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004246{
4247 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4248}
4249
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004250static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004251{
4252 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4253}
4254
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004255static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4256{
4257 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4258}
4259
Eric Dumazet02875872014-10-05 18:38:35 -07004260/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4261static inline void netif_keep_dst(struct net_device *dev)
4262{
4263 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4264}
4265
Paolo Abeni18d3df32016-07-14 18:00:10 +02004266/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4267static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4268{
4269 /* TODO: reserve and use an additional IFF bit, if we get more users */
4270 return dev->priv_flags & IFF_MACSEC;
4271}
4272
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004273extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004274
Joe Perches571ba422010-02-09 11:49:47 +00004275/* Logging, debugging and troubleshooting/diagnostic helpers. */
4276
4277/* netdev_printk helpers, similar to dev_printk */
4278
4279static inline const char *netdev_name(const struct net_device *dev)
4280{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004281 if (!dev->name[0] || strchr(dev->name, '%'))
4282 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004283 return dev->name;
4284}
4285
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004286static inline const char *netdev_reg_state(const struct net_device *dev)
4287{
4288 switch (dev->reg_state) {
4289 case NETREG_UNINITIALIZED: return " (uninitialized)";
4290 case NETREG_REGISTERED: return "";
4291 case NETREG_UNREGISTERING: return " (unregistering)";
4292 case NETREG_UNREGISTERED: return " (unregistered)";
4293 case NETREG_RELEASED: return " (released)";
4294 case NETREG_DUMMY: return " (dummy)";
4295 }
4296
4297 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4298 return " (unknown)";
4299}
4300
Joe Perchesf629d202013-09-26 14:48:15 -07004301__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004302void netdev_printk(const char *level, const struct net_device *dev,
4303 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004304__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004305void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004306__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004307void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004308__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004309void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004310__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004311void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004312__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004313void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004314__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004315void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004316__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004317void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004318
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004319#define MODULE_ALIAS_NETDEV(device) \
4320 MODULE_ALIAS("netdev-" device)
4321
Jim Cromieb558c962011-12-19 17:11:18 -05004322#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004323#define netdev_dbg(__dev, format, args...) \
4324do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004325 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004326} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004327#elif defined(DEBUG)
4328#define netdev_dbg(__dev, format, args...) \
4329 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004330#else
4331#define netdev_dbg(__dev, format, args...) \
4332({ \
4333 if (0) \
4334 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004335})
4336#endif
4337
4338#if defined(VERBOSE_DEBUG)
4339#define netdev_vdbg netdev_dbg
4340#else
4341
4342#define netdev_vdbg(dev, format, args...) \
4343({ \
4344 if (0) \
4345 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4346 0; \
4347})
4348#endif
4349
4350/*
4351 * netdev_WARN() acts like dev_printk(), but with the key difference
4352 * of using a WARN/WARN_ON to get the message out, including the
4353 * file/line information and a backtrace.
4354 */
4355#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004356 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4357 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004358
Joe Perchesb3d95c52010-02-09 11:49:49 +00004359/* netif printk helpers, similar to netdev_printk */
4360
4361#define netif_printk(priv, type, level, dev, fmt, args...) \
4362do { \
4363 if (netif_msg_##type(priv)) \
4364 netdev_printk(level, (dev), fmt, ##args); \
4365} while (0)
4366
Joe Perchesf45f4322010-06-27 01:02:36 +00004367#define netif_level(level, priv, type, dev, fmt, args...) \
4368do { \
4369 if (netif_msg_##type(priv)) \
4370 netdev_##level(dev, fmt, ##args); \
4371} while (0)
4372
Joe Perchesb3d95c52010-02-09 11:49:49 +00004373#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004374 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004375#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004376 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004377#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004378 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004379#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004380 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004381#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004382 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004383#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004384 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004385#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004386 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004387
Joe Perches0053ea92012-05-30 07:43:34 +00004388#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004389#define netif_dbg(priv, type, netdev, format, args...) \
4390do { \
4391 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004392 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004393} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004394#elif defined(DEBUG)
4395#define netif_dbg(priv, type, dev, format, args...) \
4396 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004397#else
4398#define netif_dbg(priv, type, dev, format, args...) \
4399({ \
4400 if (0) \
4401 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4402 0; \
4403})
4404#endif
4405
4406#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004407#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004408#else
4409#define netif_vdbg(priv, type, dev, format, args...) \
4410({ \
4411 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004412 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004413 0; \
4414})
4415#endif
Joe Perches571ba422010-02-09 11:49:47 +00004416
Cong Wang900ff8c2013-02-18 19:20:33 +00004417/*
4418 * The list of packet types we will receive (as opposed to discard)
4419 * and the routines to invoke.
4420 *
4421 * Why 16. Because with 16 the only overlap we get on a hash of the
4422 * low nibble of the protocol value is RARP/SNAP/X.25.
4423 *
4424 * NOTE: That is no longer true with the addition of VLAN tags. Not
4425 * sure which should go first, but I bet it won't make much
4426 * difference if we are running VLANs. The good news is that
4427 * this protocol won't be in the list unless compiled in, so
4428 * the average user (w/out VLANs) will not be adversely affected.
4429 * --BLG
4430 *
4431 * 0800 IP
4432 * 8100 802.1Q VLAN
4433 * 0001 802.3
4434 * 0002 AX.25
4435 * 0004 802.2
4436 * 8035 RARP
4437 * 0005 SNAP
4438 * 0805 X.25
4439 * 0806 ARP
4440 * 8137 IPX
4441 * 0009 Localtalk
4442 * 86DD IPv6
4443 */
4444#define PTYPE_HASH_SIZE (16)
4445#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4446
Jiri Pirko385a1542009-05-27 15:48:07 -07004447#endif /* _LINUX_NETDEVICE_H */