blob: 2095b6ab3661f4bf87d6d0d2999aad3d687b9646 [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 Pirko22bedad2010-04-01 21:22:57 +0000209#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad2010-04-01 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 Pirko22bedad2010-04-01 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 Pirko22bedad2010-04-01 21:22:57 +0000230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800231
Jiri Pirko22bedad2010-04-01 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 Pirko22bedad2010-04-01 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. Millerf6b72b62011-07-14 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 Horman2d8bff12015-09-23 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 Fastabend46e5da42014-09-12 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,
John Fastabend16e5cc62016-02-16 21:16:43 -0800792};
793
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800794struct tc_cls_u32_offload;
795
John Fastabend16e5cc62016-02-16 21:16:43 -0800796struct tc_to_netdev {
797 unsigned int type;
798 union {
799 u8 tc;
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800800 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200801 struct tc_cls_flower_offload *cls_flower;
Yotam Gigib87f7932016-07-21 12:03:12 +0200802 struct tc_cls_matchall_offload *cls_mall;
John Fastabend16e5cc62016-02-16 21:16:43 -0800803 };
804};
805
Brenden Blancoa7862b42016-07-19 12:16:48 -0700806/* These structures hold the attributes of xdp state that are being passed
807 * to the netdevice through the xdp op.
808 */
809enum xdp_netdev_command {
810 /* Set or clear a bpf program used in the earliest stages of packet
811 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
812 * is responsible for calling bpf_prog_put on any old progs that are
813 * stored. In case of error, the callee need not release the new prog
814 * reference, but on success it takes ownership and must bpf_prog_put
815 * when it is no longer used.
816 */
817 XDP_SETUP_PROG,
818 /* Check if a bpf program is set on the device. The callee should
819 * return true if a program is currently attached and running.
820 */
821 XDP_QUERY_PROG,
822};
823
824struct netdev_xdp {
825 enum xdp_netdev_command command;
826 union {
827 /* XDP_SETUP_PROG */
828 struct bpf_prog *prog;
829 /* XDP_QUERY_PROG */
830 bool prog_attached;
831 };
832};
John Fastabend16e5cc62016-02-16 21:16:43 -0800833
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800834/*
835 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800836 * The following hooks can be defined; unless noted otherwise, they are
837 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800838 *
839 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500840 * This function is called once when a network device is registered.
841 * The network device can use this for any late stage initialization
842 * or semantic validation. It can fail with an error code which will
843 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800844 *
845 * void (*ndo_uninit)(struct net_device *dev);
846 * This function is called when device is unregistered or when registration
847 * fails. It is not called if init fails.
848 *
849 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500850 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800851 * state.
852 *
853 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500854 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800855 * state.
856 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000857 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
858 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800859 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030860 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
861 * the queue before that can happen; it's for obsolete devices and weird
862 * corner cases, but the stack really does a non-trivial amount
863 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500864 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800865 *
Eric Dumazetcdba7562016-01-06 06:53:50 -0800866 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
867 * netdev_features_t features);
868 * Adjusts the requested feature flags according to device-specific
869 * constraints, and returns the resulting flags. Must not modify
870 * the device state.
871 *
Jason Wangf663dd92014-01-10 16:18:26 +0800872 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100873 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500874 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800875 * transmit queues.
876 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800877 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
878 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500879 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800880 *
881 * void (*ndo_set_rx_mode)(struct net_device *dev);
882 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000883 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500884 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800885 *
886 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
887 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700888 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500889 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800890 *
891 * int (*ndo_validate_addr)(struct net_device *dev);
892 * Test if Media Access Control address is valid for the device.
893 *
894 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500895 * Called when a user requests an ioctl which can't be handled by
896 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800897 * not supported error code.
898 *
899 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
900 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500901 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800902 * interface (PCI) for low level management.
903 *
904 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
905 * Called when a user wants to change the Maximum Transfer Unit
906 * of a device. If not defined, any request to change MTU will
907 * will return an error.
908 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800909 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500910 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800911 * for dev->watchdog ticks.
912 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000913 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700914 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700915 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800916 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000917 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000918 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
919 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700920 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000921 * (which should normally be dev->stats) and return a pointer to
922 * it. The structure may be changed asynchronously only if each
923 * field is written atomically.
924 * 3. Update dev->stats asynchronously and atomically, and define
925 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800926 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530927 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500928 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000929 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800930 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530931 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500932 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000933 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800934 *
935 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000936 *
937 * SR-IOV management functions.
938 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
939 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400940 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
941 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000942 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000943 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000944 * int (*ndo_get_vf_config)(struct net_device *dev,
945 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300946 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700947 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
948 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300949 *
950 * Enable or disable the VF ability to query its RSS Redirection Table and
951 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500952 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300953 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700954 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000955 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
956 * Called to setup 'tc' number of traffic classes in the net device. This
957 * is always called from the stack with the rtnl lock held and netif tx
958 * queues stopped. This allows the netdevice to perform queue management
959 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000960 *
Yi Zoue9bce842011-03-09 08:48:03 +0000961 * Fiber Channel over Ethernet (FCoE) offload functions.
962 * int (*ndo_fcoe_enable)(struct net_device *dev);
963 * Called when the FCoE protocol stack wants to start using LLD for FCoE
964 * so the underlying device can perform whatever needed configuration or
965 * initialization to support acceleration of FCoE traffic.
966 *
967 * int (*ndo_fcoe_disable)(struct net_device *dev);
968 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
969 * so the underlying device can perform whatever needed clean-ups to
970 * stop supporting acceleration of FCoE traffic.
971 *
972 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
973 * struct scatterlist *sgl, unsigned int sgc);
974 * Called when the FCoE Initiator wants to initialize an I/O that
975 * is a possible candidate for Direct Data Placement (DDP). The LLD can
976 * perform necessary setup and returns 1 to indicate the device is set up
977 * successfully to perform DDP on this I/O, otherwise this returns 0.
978 *
979 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
980 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
981 * indicated by the FC exchange id 'xid', so the underlying device can
982 * clean up and reuse resources for later DDP requests.
983 *
984 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
985 * struct scatterlist *sgl, unsigned int sgc);
986 * Called when the FCoE Target wants to initialize an I/O that
987 * is a possible candidate for Direct Data Placement (DDP). The LLD can
988 * perform necessary setup and returns 1 to indicate the device is set up
989 * successfully to perform DDP on this I/O, otherwise this returns 0.
990 *
Neerav Parikh68bad942012-01-04 20:23:39 +0000991 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
992 * struct netdev_fcoe_hbainfo *hbainfo);
993 * Called when the FCoE Protocol stack wants information on the underlying
994 * device. This information is utilized by the FCoE protocol stack to
995 * register attributes with Fiber Channel management service as per the
996 * FC-GS Fabric Device Management Information(FDMI) specification.
997 *
Yi Zoue9bce842011-03-09 08:48:03 +0000998 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
999 * Called when the underlying device wants to override default World Wide
1000 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1001 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1002 * protocol stack to use.
1003 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001004 * RFS acceleration.
1005 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1006 * u16 rxq_index, u32 flow_id);
1007 * Set hardware filter for RFS. rxq_index is the target queue index;
1008 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1009 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001010 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001011 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001012 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1013 * Called to make another netdev an underling.
1014 *
1015 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1016 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001017 *
1018 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001019 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001020 * Called to update device configuration to new features. Passed
1021 * feature set might be less than what was returned by ndo_fix_features()).
1022 * Must return >0 or -errno if it changed dev->features itself.
1023 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001024 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1025 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001026 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001027 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001028 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1029 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001030 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001031 * Deletes the FDB entry from dev coresponding to addr.
1032 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001033 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001034 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001035 * Used to add FDB entries to dump requests. Implementers should add
1036 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001037 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001038 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1039 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001040 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001041 * struct net_device *dev, u32 filter_mask,
1042 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001043 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1044 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001045 *
1046 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1047 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1048 * which do not represent real hardware may define this to allow their
1049 * userspace components to manage their virtual carrier state. Devices
1050 * that determine carrier state from physical hardware properties (eg
1051 * network cables) or protocol-dependent mechanisms (eg
1052 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001053 *
1054 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001055 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001056 * Called to get ID of physical port of this device. If driver does
1057 * not implement this, it is assumed that the hw is not able to have
1058 * multiple net devices on single physical port.
Joseph Gasparakis53cf5272013-09-04 02:13:38 -07001059 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001060 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1061 * struct udp_tunnel_info *ti);
1062 * Called by UDP tunnel to notify a driver about the UDP port and socket
1063 * address family that a UDP tunnel is listnening to. It is called only
1064 * when a new port starts listening. The operation is protected by the
1065 * RTNL.
1066 *
1067 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1068 * struct udp_tunnel_info *ti);
1069 * Called by UDP tunnel to notify the driver about a UDP port and socket
1070 * address family that the UDP tunnel is not listening to anymore. The
1071 * operation is protected by the RTNL.
1072 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001073 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1074 * struct net_device *dev)
1075 * Called by upper layer devices to accelerate switching or other
1076 * station functionality into hardware. 'pdev is the lowerdev
1077 * to use for the offload and 'dev' is the net device that will
1078 * back the offload. Returns a pointer to the private structure
1079 * the upper layer will maintain.
1080 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1081 * Called by upper layer device to delete the station created
1082 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1083 * the station and priv is the structure returned by the add
1084 * operation.
1085 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1086 * struct net_device *dev,
1087 * void *priv);
1088 * Callback to use for xmit over the accelerated station. This
1089 * is used in place of ndo_start_xmit on accelerated net
1090 * devices.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001091 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1092 * struct net_device *dev
1093 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001094 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001095 * performing offload operations on a given packet. This is to give
1096 * the device an opportunity to implement any restrictions that cannot
1097 * be otherwise expressed by feature flags. The check is called with
1098 * the set of features that the stack has calculated and it returns
1099 * those the driver believes to be appropriate.
John Fastabend822b3b22015-03-18 14:57:33 +02001100 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1101 * int queue_index, u32 maxrate);
1102 * Called when a user wants to set a max-rate limitation of specific
1103 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001104 * int (*ndo_get_iflink)(const struct net_device *dev);
1105 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001106 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001107 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001108 * This function is used to pass protocol port error state information
1109 * to the switch driver. The switch driver can react to the proto_down
1110 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001111 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1112 * This function is used to get egress tunnel information for given skb.
1113 * This is useful for retrieving outer tunnel header parameters while
1114 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001115 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1116 * This function is used to specify the headroom that the skb must
1117 * consider when allocation skb during packet reception. Setting
1118 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001119 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001120 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001121 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1122 * This function is used to set or query state related to XDP on the
1123 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001124 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001125 */
1126struct net_device_ops {
1127 int (*ndo_init)(struct net_device *dev);
1128 void (*ndo_uninit)(struct net_device *dev);
1129 int (*ndo_open)(struct net_device *dev);
1130 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001131 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1132 struct net_device *dev);
1133 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1134 struct net_device *dev,
1135 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001136 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001137 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001138 void *accel_priv,
1139 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001140 void (*ndo_change_rx_flags)(struct net_device *dev,
1141 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001142 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001143 int (*ndo_set_mac_address)(struct net_device *dev,
1144 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001145 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001146 int (*ndo_do_ioctl)(struct net_device *dev,
1147 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001148 int (*ndo_set_config)(struct net_device *dev,
1149 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001150 int (*ndo_change_mtu)(struct net_device *dev,
1151 int new_mtu);
1152 int (*ndo_neigh_setup)(struct net_device *dev,
1153 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001154 void (*ndo_tx_timeout) (struct net_device *dev);
1155
Eric Dumazet28172732010-07-07 14:58:56 -07001156 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1157 struct rtnl_link_stats64 *storage);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001158 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1159
Jiri Pirko8e586132011-12-08 19:52:37 -05001160 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001161 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001162 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001163 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001164#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001165 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001166 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001167 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001168 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001169#endif
Cong Wange0d10952013-08-01 11:10:25 +08001170#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001171 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001172#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001173 int (*ndo_set_vf_mac)(struct net_device *dev,
1174 int queue, u8 *mac);
1175 int (*ndo_set_vf_vlan)(struct net_device *dev,
1176 int queue, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001177 int (*ndo_set_vf_rate)(struct net_device *dev,
1178 int vf, int min_tx_rate,
1179 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001180 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1181 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001182 int (*ndo_set_vf_trust)(struct net_device *dev,
1183 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001184 int (*ndo_get_vf_config)(struct net_device *dev,
1185 int vf,
1186 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001187 int (*ndo_set_vf_link_state)(struct net_device *dev,
1188 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001189 int (*ndo_get_vf_stats)(struct net_device *dev,
1190 int vf,
1191 struct ifla_vf_stats
1192 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001193 int (*ndo_set_vf_port)(struct net_device *dev,
1194 int vf,
1195 struct nlattr *port[]);
1196 int (*ndo_get_vf_port)(struct net_device *dev,
1197 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001198 int (*ndo_set_vf_guid)(struct net_device *dev,
1199 int vf, u64 guid,
1200 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001201 int (*ndo_set_vf_rss_query_en)(
1202 struct net_device *dev,
1203 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001204 int (*ndo_setup_tc)(struct net_device *dev,
1205 u32 handle,
1206 __be16 protocol,
1207 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001208#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001209 int (*ndo_fcoe_enable)(struct net_device *dev);
1210 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001211 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1212 u16 xid,
1213 struct scatterlist *sgl,
1214 unsigned int sgc);
1215 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1216 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001217 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1218 u16 xid,
1219 struct scatterlist *sgl,
1220 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001221 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1222 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001223#endif
1224
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001225#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001226#define NETDEV_FCOE_WWNN 0
1227#define NETDEV_FCOE_WWPN 1
1228 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1229 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001230#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001231
Ben Hutchingsc4454772011-01-19 11:03:53 +00001232#ifdef CONFIG_RFS_ACCEL
1233 int (*ndo_rx_flow_steer)(struct net_device *dev,
1234 const struct sk_buff *skb,
1235 u16 rxq_index,
1236 u32 flow_id);
1237#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001238 int (*ndo_add_slave)(struct net_device *dev,
1239 struct net_device *slave_dev);
1240 int (*ndo_del_slave)(struct net_device *dev,
1241 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001242 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1243 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001244 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001245 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001246 int (*ndo_neigh_construct)(struct net_device *dev,
1247 struct neighbour *n);
1248 void (*ndo_neigh_destroy)(struct net_device *dev,
1249 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001250
1251 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001252 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001253 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001254 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001255 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001256 u16 flags);
1257 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001258 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001259 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001260 const unsigned char *addr,
1261 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001262 int (*ndo_fdb_dump)(struct sk_buff *skb,
1263 struct netlink_callback *cb,
1264 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001265 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001266 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001267
1268 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001269 struct nlmsghdr *nlh,
1270 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001271 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1272 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001273 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001274 u32 filter_mask,
1275 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001276 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001277 struct nlmsghdr *nlh,
1278 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001279 int (*ndo_change_carrier)(struct net_device *dev,
1280 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001281 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001282 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001283 int (*ndo_get_phys_port_name)(struct net_device *dev,
1284 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001285 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1286 struct udp_tunnel_info *ti);
1287 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1288 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001289 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1290 struct net_device *dev);
1291 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1292 void *priv);
1293
1294 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1295 struct net_device *dev,
1296 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001297 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001298 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1299 int queue_index,
1300 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001301 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001302 int (*ndo_change_proto_down)(struct net_device *dev,
1303 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001304 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1305 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001306 void (*ndo_set_rx_headroom)(struct net_device *dev,
1307 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001308 int (*ndo_xdp)(struct net_device *dev,
1309 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001310};
1311
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001312/**
1313 * enum net_device_priv_flags - &struct net_device priv_flags
1314 *
1315 * These are the &struct net_device, they are only set internally
1316 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001317 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001318 * during any kernel release.
1319 *
1320 * You should have a pretty good reason to be extending these flags.
1321 *
1322 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1323 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001324 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001325 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001326 * @IFF_WAN_HDLC: WAN HDLC device
1327 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1328 * release skb->dst
1329 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1330 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1331 * @IFF_MACVLAN_PORT: device used as macvlan port
1332 * @IFF_BRIDGE_PORT: device used as bridge port
1333 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1334 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1335 * @IFF_UNICAST_FLT: Supports unicast filtering
1336 * @IFF_TEAM_PORT: device used as team port
1337 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1338 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1339 * change when it's running
1340 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001341 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1342 * underlying stacked devices
1343 * @IFF_IPVLAN_MASTER: IPvlan master device
1344 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001345 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001346 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001347 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001348 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001349 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001350 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001351 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1352 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001353 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001354 */
1355enum netdev_priv_flags {
1356 IFF_802_1Q_VLAN = 1<<0,
1357 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001358 IFF_BONDING = 1<<2,
1359 IFF_ISATAP = 1<<3,
1360 IFF_WAN_HDLC = 1<<4,
1361 IFF_XMIT_DST_RELEASE = 1<<5,
1362 IFF_DONT_BRIDGE = 1<<6,
1363 IFF_DISABLE_NETPOLL = 1<<7,
1364 IFF_MACVLAN_PORT = 1<<8,
1365 IFF_BRIDGE_PORT = 1<<9,
1366 IFF_OVS_DATAPATH = 1<<10,
1367 IFF_TX_SKB_SHARING = 1<<11,
1368 IFF_UNICAST_FLT = 1<<12,
1369 IFF_TEAM_PORT = 1<<13,
1370 IFF_SUPP_NOFCS = 1<<14,
1371 IFF_LIVE_ADDR_CHANGE = 1<<15,
1372 IFF_MACVLAN = 1<<16,
1373 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1374 IFF_IPVLAN_MASTER = 1<<18,
1375 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001376 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001377 IFF_NO_QUEUE = 1<<21,
1378 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001379 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001380 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001381 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001382 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001383 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001384};
1385
1386#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1387#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001388#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001389#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001390#define IFF_WAN_HDLC IFF_WAN_HDLC
1391#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1392#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1393#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1394#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1395#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1396#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1397#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1398#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1399#define IFF_TEAM_PORT IFF_TEAM_PORT
1400#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1401#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1402#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001403#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001404#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1405#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001406#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001407#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001408#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001409#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001410#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001411#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001412#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001413
Karoly Kemeny536721b2014-07-30 20:27:36 +02001414/**
1415 * struct net_device - The DEVICE structure.
1416 * Actually, this whole structure is a big mistake. It mixes I/O
1417 * data with strictly "high-level" data, and it has to know about
1418 * almost every data structure used in the INET module.
1419 *
1420 * @name: This is the first field of the "visible" part of this structure
1421 * (i.e. as seen by users in the "Space.c" file). It is the name
1422 * of the interface.
1423 *
1424 * @name_hlist: Device name hash chain, please keep it close to name[]
1425 * @ifalias: SNMP alias
1426 * @mem_end: Shared memory end
1427 * @mem_start: Shared memory start
1428 * @base_addr: Device I/O address
1429 * @irq: Device IRQ number
1430 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001431 * @carrier_changes: Stats to monitor carrier on<->off transitions
1432 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001433 * @state: Generic network queuing layer state, see netdev_state_t
1434 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001435 * @napi_list: List entry used for polling NAPI devices
1436 * @unreg_list: List entry when we are unregistering the
1437 * device; see the function unregister_netdev
1438 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001439 * @ptype_all: Device-specific packet handlers for all protocols
1440 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001441 *
1442 * @adj_list: Directly linked devices, like slaves for bonding
1443 * @all_adj_list: All linked devices, *including* neighbours
1444 * @features: Currently active device features
1445 * @hw_features: User-changeable features
1446 *
1447 * @wanted_features: User-requested features
1448 * @vlan_features: Mask of features inheritable by VLAN devices
1449 *
1450 * @hw_enc_features: Mask of features inherited by encapsulating devices
1451 * This field indicates what encapsulation
1452 * offloads the hardware is capable of doing,
1453 * and drivers will need to set them appropriately.
1454 *
1455 * @mpls_features: Mask of features inheritable by MPLS
1456 *
1457 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001458 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001459 *
1460 * @stats: Statistics struct, which was left as a legacy, use
1461 * rtnl_link_stats64 instead
1462 *
1463 * @rx_dropped: Dropped packets by core network,
1464 * do not use this in drivers
1465 * @tx_dropped: Dropped packets by core network,
1466 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001467 * @rx_nohandler: nohandler dropped packets by core network on
1468 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001469 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001470 * @wireless_handlers: List of functions to handle Wireless Extensions,
1471 * instead of ioctl,
1472 * see <net/iw_handler.h> for details.
1473 * @wireless_data: Instance data managed by the core of wireless extensions
1474 *
1475 * @netdev_ops: Includes several pointers to callbacks,
1476 * if one wants to override the ndo_*() functions
1477 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001478 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1479 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001480 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001481 * of Layer 2 headers.
1482 *
1483 * @flags: Interface flags (a la BSD)
1484 * @priv_flags: Like 'flags' but invisible to userspace,
1485 * see if.h for the definitions
1486 * @gflags: Global flags ( kept as legacy )
1487 * @padded: How much padding added by alloc_netdev()
1488 * @operstate: RFC2863 operstate
1489 * @link_mode: Mapping policy to operstate
1490 * @if_port: Selectable AUI, TP, ...
1491 * @dma: DMA channel
1492 * @mtu: Interface MTU value
1493 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001494 * @hard_header_len: Maximum hardware header length.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001495 *
1496 * @needed_headroom: Extra headroom the hardware may need, but not in all
1497 * cases can this be guaranteed
1498 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1499 * cases can this be guaranteed. Some cases also use
1500 * LL_MAX_HEADER instead to allocate the skb
1501 *
1502 * interface address info:
1503 *
1504 * @perm_addr: Permanent hw address
1505 * @addr_assign_type: Hw address assignment type
1506 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001507 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001508 * @dev_id: Used to differentiate devices that share
1509 * the same link layer address
1510 * @dev_port: Used to differentiate devices that share
1511 * the same function
1512 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001513 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001514 * has been enabled due to the need to listen to
1515 * additional unicast addresses in a device that
1516 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001517 * @uc: unicast mac addresses
1518 * @mc: multicast mac addresses
1519 * @dev_addrs: list of device hw addresses
1520 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001521 * @promiscuity: Number of times the NIC is told to work in
1522 * promiscuous mode; if it becomes 0 the NIC will
1523 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001524 * @allmulti: Counter, enables or disables allmulticast mode
1525 *
1526 * @vlan_info: VLAN info
1527 * @dsa_ptr: dsa specific data
1528 * @tipc_ptr: TIPC specific data
1529 * @atalk_ptr: AppleTalk link
1530 * @ip_ptr: IPv4 specific data
1531 * @dn_ptr: DECnet specific data
1532 * @ip6_ptr: IPv6 specific data
1533 * @ax25_ptr: AX.25 specific data
1534 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1535 *
1536 * @last_rx: Time of last Rx
1537 * @dev_addr: Hw address (before bcast,
1538 * because most packets are unicast)
1539 *
1540 * @_rx: Array of RX queues
1541 * @num_rx_queues: Number of RX queues
1542 * allocated at register_netdev() time
1543 * @real_num_rx_queues: Number of RX queues currently active in device
1544 *
1545 * @rx_handler: handler for received packets
1546 * @rx_handler_data: XXX: need comments on this one
1547 * @ingress_queue: XXX: need comments on this one
1548 * @broadcast: hw bcast address
1549 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001550 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1551 * indexed by RX queue number. Assigned by driver.
1552 * This must only be set if the ndo_rx_flow_steer
1553 * operation is defined
1554 * @index_hlist: Device index hash chain
1555 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001556 * @_tx: Array of TX queues
1557 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1558 * @real_num_tx_queues: Number of TX queues currently active in device
1559 * @qdisc: Root qdisc from userspace point of view
1560 * @tx_queue_len: Max frames per queue allowed
1561 * @tx_global_lock: XXX: need comments on this one
1562 *
1563 * @xps_maps: XXX: need comments on this one
1564 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001565 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001566 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001567 * @watchdog_timer: List of timers
1568 *
1569 * @pcpu_refcnt: Number of references to this device
1570 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001571 * @link_watch_list: XXX: need comments on this one
1572 *
1573 * @reg_state: Register/unregister state machine
1574 * @dismantle: Device is going to be freed
1575 * @rtnl_link_state: This enum represents the phases of creating
1576 * a new link
1577 *
1578 * @destructor: Called from unregister,
1579 * can be used to call free_netdev
1580 * @npinfo: XXX: need comments on this one
1581 * @nd_net: Network namespace this network device is inside
1582 *
1583 * @ml_priv: Mid-layer private
1584 * @lstats: Loopback statistics
1585 * @tstats: Tunnel statistics
1586 * @dstats: Dummy statistics
1587 * @vstats: Virtual ethernet statistics
1588 *
1589 * @garp_port: GARP
1590 * @mrp_port: MRP
1591 *
1592 * @dev: Class/net/name entry
1593 * @sysfs_groups: Space for optional device, statistics and wireless
1594 * sysfs groups
1595 *
1596 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1597 * @rtnl_link_ops: Rtnl_link_ops
1598 *
1599 * @gso_max_size: Maximum size of generic segmentation offload
1600 * @gso_max_segs: Maximum number of segments that can be passed to the
1601 * NIC for GSO
1602 *
1603 * @dcbnl_ops: Data Center Bridging netlink ops
1604 * @num_tc: Number of traffic classes in the net device
1605 * @tc_to_txq: XXX: need comments on this one
1606 * @prio_tc_map XXX: need comments on this one
1607 *
1608 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1609 *
1610 * @priomap: XXX: need comments on this one
1611 * @phydev: Physical device may attach itself
1612 * for hardware timestamping
1613 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001614 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1615 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001616 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001617 * @proto_down: protocol port state information can be sent to the
1618 * switch driver and used to set the phys state of the
1619 * switch port.
1620 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 * FIXME: cleanup struct net_device such that network protocol info
1622 * moves out.
1623 */
1624
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001625struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001627 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001628 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 /*
1630 * I/O specific fields
1631 * FIXME: Merge these and struct ifmap into one
1632 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001633 unsigned long mem_end;
1634 unsigned long mem_start;
1635 unsigned long base_addr;
1636 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001638 atomic_t carrier_changes;
1639
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001641 * Some hardware also needs these fields (state,dev_list,
1642 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 * part of the usual set specified in Space.c.
1644 */
1645
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 unsigned long state;
1647
Pavel Emelianov7562f872007-05-03 15:13:45 -07001648 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001649 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001650 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001651 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001652 struct list_head ptype_all;
1653 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001654
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001655 struct {
1656 struct list_head upper;
1657 struct list_head lower;
1658 } adj_list;
1659
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001660 struct {
1661 struct list_head upper;
1662 struct list_head lower;
1663 } all_adj_list;
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001664
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001665 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001666 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001667 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001668 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001669 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001670 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001671 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001672
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001674 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675
Rusty Russellc45d2862007-03-28 14:29:08 -07001676 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001677
Eric Dumazet015f0682014-03-27 08:45:56 -07001678 atomic_long_t rx_dropped;
1679 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001680 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681
Johannes Bergb86e0282007-04-26 20:48:23 -07001682#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001683 const struct iw_handler_def *wireless_handlers;
1684 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001685#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001686 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001687 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001688#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001689 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001690#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001691#ifdef CONFIG_NET_L3_MASTER_DEV
1692 const struct l3mdev_ops *l3mdev_ops;
1693#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001694#if IS_ENABLED(CONFIG_IPV6)
1695 const struct ndisc_ops *ndisc_ops;
1696#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001698 const struct header_ops *header_ops;
1699
Karoly Kemeny536721b2014-07-30 20:27:36 +02001700 unsigned int flags;
1701 unsigned int priv_flags;
1702
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001704 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
Karoly Kemeny536721b2014-07-30 20:27:36 +02001706 unsigned char operstate;
1707 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001708
Karoly Kemeny536721b2014-07-30 20:27:36 +02001709 unsigned char if_port;
1710 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001711
Karoly Kemeny536721b2014-07-30 20:27:36 +02001712 unsigned int mtu;
1713 unsigned short type;
1714 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715
Johannes Bergf5184d22008-05-12 20:48:31 -07001716 unsigned short needed_headroom;
1717 unsigned short needed_tailroom;
1718
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001720 unsigned char perm_addr[MAX_ADDR_LEN];
1721 unsigned char addr_assign_type;
1722 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001723 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001724 unsigned short dev_id;
1725 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001726 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001727 unsigned char name_assign_type;
1728 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001729 struct netdev_hw_addr_list uc;
1730 struct netdev_hw_addr_list mc;
1731 struct netdev_hw_addr_list dev_addrs;
1732
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001733#ifdef CONFIG_SYSFS
1734 struct kset *queues_kset;
1735#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001736 unsigned int promiscuity;
1737 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001740 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001741
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001742#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001743 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001744#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001745#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001746 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001747#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001748#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001749 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001750#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001751 void *atalk_ptr;
1752 struct in_device __rcu *ip_ptr;
1753 struct dn_dev __rcu *dn_ptr;
1754 struct inet6_dev __rcu *ip6_ptr;
1755 void *ax25_ptr;
1756 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001757 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001758#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1759 struct mpls_dev __rcu *mpls_ptr;
1760#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001762/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001763 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001764 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001765 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001766
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001767 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001768 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001769
Michael Daltona953be52014-01-16 22:23:28 -08001770#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001771 struct netdev_rx_queue *_rx;
1772
Tom Herbert0a9627f2010-03-16 08:03:29 +00001773 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001774 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001775#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001776
Eric Dumazet3b47d302014-11-06 21:09:44 -08001777 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001778 rx_handler_func_t __rcu *rx_handler;
1779 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001780
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001781#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001782 struct tcf_proto __rcu *ingress_cl_list;
1783#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001784 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001785#ifdef CONFIG_NETFILTER_INGRESS
1786 struct list_head nf_hooks_ingress;
1787#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001788
Karoly Kemeny536721b2014-07-30 20:27:36 +02001789 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001790#ifdef CONFIG_RFS_ACCEL
1791 struct cpu_rmap *rx_cpu_rmap;
1792#endif
1793 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001794
1795/*
1796 * Cache lines mostly used on transmit path
1797 */
David S. Millere8a04642008-07-17 00:34:19 -07001798 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1799 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001800 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001801 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001802#ifdef CONFIG_NET_SCHED
1803 DECLARE_HASHTABLE (qdisc_hash, 4);
1804#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001805 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001806 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001807 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001808
Tom Herbertbf264142010-11-26 08:36:09 +00001809#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001810 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001811#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001812#ifdef CONFIG_NET_CLS_ACT
1813 struct tcf_proto __rcu *egress_cl_list;
1814#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001815
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001816 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001817 struct timer_list watchdog_timer;
1818
Eric Dumazet29b44332010-10-11 10:22:12 +00001819 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821
Eric Dumazete014deb2009-11-17 05:59:21 +00001822 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001823
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001825 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 NETREG_UNREGISTERING, /* called unregister_netdevice */
1827 NETREG_UNREGISTERED, /* completed unregister todo */
1828 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001829 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001830 } reg_state:8;
1831
Karoly Kemeny536721b2014-07-30 20:27:36 +02001832 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001833
1834 enum {
1835 RTNL_LINK_INITIALIZED,
1836 RTNL_LINK_INITIALIZING,
1837 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001839 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001842 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843#endif
David S. Millereae792b2008-07-15 03:03:33 -07001844
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001845 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001846
David S. Miller49517042008-05-12 03:29:11 -07001847 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001848 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001849 void *ml_priv;
1850 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001851 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001852 struct pcpu_dstats __percpu *dstats;
1853 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001854 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001855
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001856 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001857 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001859 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001860 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001861 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001862
Patrick McHardy38f7b872007-06-13 12:03:51 -07001863 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001864
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001865 /* for setting kernel sock attribute on TCP connection setup */
1866#define GSO_MAX_SIZE 65536
1867 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001868#define GSO_MAX_SEGS 65535
1869 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001870
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001871#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001872 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001873#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001874 u8 num_tc;
1875 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1876 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001877
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001878#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001879 unsigned int fcoe_ddp_xid;
1880#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001881#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001882 struct netprio_map __rcu *priomap;
1883#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001884 struct phy_device *phydev;
1885 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001886 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001887 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001889#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890
1891#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892
David S. Millere8a04642008-07-17 00:34:19 -07001893static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001894int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1895{
1896 return dev->prio_tc_map[prio & TC_BITMASK];
1897}
1898
1899static inline
1900int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1901{
1902 if (tc >= dev->num_tc)
1903 return -EINVAL;
1904
1905 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1906 return 0;
1907}
1908
1909static inline
1910void netdev_reset_tc(struct net_device *dev)
1911{
1912 dev->num_tc = 0;
1913 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1914 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1915}
1916
1917static inline
1918int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1919{
1920 if (tc >= dev->num_tc)
1921 return -EINVAL;
1922
1923 dev->tc_to_txq[tc].count = count;
1924 dev->tc_to_txq[tc].offset = offset;
1925 return 0;
1926}
1927
1928static inline
1929int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1930{
1931 if (num_tc > TC_MAX_QUEUE)
1932 return -EINVAL;
1933
1934 dev->num_tc = num_tc;
1935 return 0;
1936}
1937
1938static inline
1939int netdev_get_num_tc(struct net_device *dev)
1940{
1941 return dev->num_tc;
1942}
1943
1944static inline
David S. Millere8a04642008-07-17 00:34:19 -07001945struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1946 unsigned int index)
1947{
1948 return &dev->_tx[index];
1949}
1950
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001951static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1952 const struct sk_buff *skb)
1953{
1954 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1955}
1956
David S. Millere8a04642008-07-17 00:34:19 -07001957static inline void netdev_for_each_tx_queue(struct net_device *dev,
1958 void (*f)(struct net_device *,
1959 struct netdev_queue *,
1960 void *),
1961 void *arg)
1962{
1963 unsigned int i;
1964
1965 for (i = 0; i < dev->num_tx_queues; i++)
1966 f(dev, &dev->_tx[i], arg);
1967}
1968
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001969#define netdev_lockdep_set_classes(dev) \
1970{ \
1971 static struct lock_class_key qdisc_tx_busylock_key; \
1972 static struct lock_class_key qdisc_running_key; \
1973 static struct lock_class_key qdisc_xmit_lock_key; \
1974 static struct lock_class_key dev_addr_list_lock_key; \
1975 unsigned int i; \
1976 \
1977 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1978 (dev)->qdisc_running_key = &qdisc_running_key; \
1979 lockdep_set_class(&(dev)->addr_list_lock, \
1980 &dev_addr_list_lock_key); \
1981 for (i = 0; i < (dev)->num_tx_queues; i++) \
1982 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1983 &qdisc_xmit_lock_key); \
1984}
1985
Joe Perchesf629d202013-09-26 14:48:15 -07001986struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001987 struct sk_buff *skb,
1988 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001989
Paolo Abeni871b6422016-02-26 10:45:37 +01001990/* returns the headroom that the master device needs to take in account
1991 * when forwarding to this dev
1992 */
1993static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1994{
1995 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1996}
1997
1998static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1999{
2000 if (dev->netdev_ops->ndo_set_rx_headroom)
2001 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2002}
2003
2004/* set the device rx headroom to the dev's default */
2005static inline void netdev_reset_rx_headroom(struct net_device *dev)
2006{
2007 netdev_set_rx_headroom(dev, -1);
2008}
2009
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002010/*
2011 * Net namespace inlines
2012 */
2013static inline
2014struct net *dev_net(const struct net_device *dev)
2015{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002016 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002017}
2018
2019static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002020void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002021{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002022 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002023}
2024
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07002025static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002026{
Alexander Duyck3fc88672014-09-12 23:12:46 -07002027#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002028 if (dev->dsa_ptr != NULL)
2029 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00002030#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07002031 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00002032}
2033
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002034/**
2035 * netdev_priv - access network device private data
2036 * @dev: network device
2037 *
2038 * Get network device private data
2039 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002040static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002042 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043}
2044
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045/* Set the sysfs physical device reference for the network logical device
2046 * if set prior to registration will cause a symlink during initialization.
2047 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002048#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049
Marcel Holtmann384912e2009-08-31 21:08:19 +00002050/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002051 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002052 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002053 */
2054#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2055
Eric Dumazet82dc3c62013-03-05 15:57:22 +00002056/* Default NAPI poll() weight
2057 * Device drivers are strongly advised to not use bigger value
2058 */
2059#define NAPI_POLL_WEIGHT 64
2060
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002061/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002062 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002063 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002064 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002065 * @poll: polling function
2066 * @weight: default weight
2067 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002068 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2069 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002070 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002071void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2072 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002073
Alexander Duyckd8156532008-07-08 15:13:05 -07002074/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002075 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002076 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002077 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002078 * @poll: polling function
2079 * @weight: default weight
2080 *
2081 * This variant of netif_napi_add() should be used from drivers using NAPI
2082 * to exclusively poll a TX queue.
2083 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2084 */
2085static inline void netif_tx_napi_add(struct net_device *dev,
2086 struct napi_struct *napi,
2087 int (*poll)(struct napi_struct *, int),
2088 int weight)
2089{
2090 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2091 netif_napi_add(dev, napi, poll, weight);
2092}
2093
2094/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002095 * netif_napi_del - remove a NAPI context
2096 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002097 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002098 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002099 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002100void netif_napi_del(struct napi_struct *napi);
2101
2102struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002103 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002104 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002105
Herbert Xu74895942009-05-26 18:50:27 +00002106 /* Length of frag0. */
2107 unsigned int frag0_len;
2108
Herbert Xu86911732009-01-29 14:19:50 +00002109 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002110 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002111
Herbert Xud565b0a2008-12-15 23:38:52 -08002112 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002113 u16 flush;
2114
2115 /* Save the IP ID here and check when we get to the transport layer */
2116 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002117
2118 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002119 u16 count;
2120
Tom Herbert15e23962015-02-10 16:30:31 -08002121 /* Start offset for remote checksum offload */
2122 u16 gro_remcsum_start;
2123
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002124 /* jiffies when first packet was created/queued */
2125 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002126
Tom Herbertafe93322014-09-17 12:25:57 -07002127 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002128 u16 proto;
2129
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002130 /* This is non-zero if the packet may be of the same flow. */
2131 u8 same_flow:1;
2132
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002133 /* Used in tunnel GRO receive */
2134 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002135
2136 /* GRO checksum is valid */
2137 u8 csum_valid:1;
2138
Tom Herbert662880f2014-08-27 21:26:56 -07002139 /* Number of checksums via CHECKSUM_UNNECESSARY */
2140 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002141
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002142 /* Free the skb? */
2143 u8 free:2;
2144#define NAPI_GRO_FREE 1
2145#define NAPI_GRO_FREE_STOLEN_HEAD 2
2146
Tom Herbertefc98d02014-10-03 15:48:08 -07002147 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2148 u8 is_ipv6:1;
2149
Alexander Duycka0ca1532016-04-05 09:13:39 -07002150 /* Used in GRE, set in fou/gue_gro_receive */
2151 u8 is_fou:1;
2152
Alexander Duyck15305452016-04-10 21:44:57 -04002153 /* Used to determine if flush_id can be ignored */
2154 u8 is_atomic:1;
2155
2156 /* 5 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002157
Jerry Chubf5a7552014-01-07 10:23:19 -08002158 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2159 __wsum csum;
2160
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002161 /* used in skb_gro_receive() slow path */
2162 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002163};
2164
2165#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002166
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002168 __be16 type; /* This is really htons(ether_type). */
2169 struct net_device *dev; /* NULL is wildcarded here */
2170 int (*func) (struct sk_buff *,
2171 struct net_device *,
2172 struct packet_type *,
2173 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002174 bool (*id_match)(struct packet_type *ptype,
2175 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176 void *af_packet_priv;
2177 struct list_head list;
2178};
2179
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002180struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2182 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002184 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002185 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002186};
2187
2188struct packet_offload {
2189 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002190 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002191 struct offload_callbacks callbacks;
2192 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193};
2194
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002195/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002196struct pcpu_sw_netstats {
2197 u64 rx_packets;
2198 u64 rx_bytes;
2199 u64 tx_packets;
2200 u64 tx_bytes;
2201 struct u64_stats_sync syncp;
2202};
2203
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002204#define __netdev_alloc_pcpu_stats(type, gfp) \
2205({ \
2206 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2207 if (pcpu_stats) { \
2208 int __cpu; \
2209 for_each_possible_cpu(__cpu) { \
2210 typeof(type) *stat; \
2211 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2212 u64_stats_init(&stat->syncp); \
2213 } \
2214 } \
2215 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002216})
2217
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002218#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002219 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002220
Jiri Pirko764f5e52015-12-03 12:12:12 +01002221enum netdev_lag_tx_type {
2222 NETDEV_LAG_TX_TYPE_UNKNOWN,
2223 NETDEV_LAG_TX_TYPE_RANDOM,
2224 NETDEV_LAG_TX_TYPE_BROADCAST,
2225 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2226 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2227 NETDEV_LAG_TX_TYPE_HASH,
2228};
2229
2230struct netdev_lag_upper_info {
2231 enum netdev_lag_tx_type tx_type;
2232};
2233
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002234struct netdev_lag_lower_state_info {
2235 u8 link_up : 1,
2236 tx_enabled : 1;
2237};
2238
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239#include <linux/notifier.h>
2240
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002241/* netdevice notifier chain. Please remember to update the rtnetlink
2242 * notification exclusion list in rtnetlink_event() when adding new
2243 * types.
2244 */
2245#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2246#define NETDEV_DOWN 0x0002
2247#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2248 detected a hardware crash and restarted
2249 - we can use this eg to kick tcp sessions
2250 once done */
2251#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2252#define NETDEV_REGISTER 0x0005
2253#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002254#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002255#define NETDEV_CHANGEADDR 0x0008
2256#define NETDEV_GOING_DOWN 0x0009
2257#define NETDEV_CHANGENAME 0x000A
2258#define NETDEV_FEAT_CHANGE 0x000B
2259#define NETDEV_BONDING_FAILOVER 0x000C
2260#define NETDEV_PRE_UP 0x000D
2261#define NETDEV_PRE_TYPE_CHANGE 0x000E
2262#define NETDEV_POST_TYPE_CHANGE 0x000F
2263#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002264#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002265#define NETDEV_RELEASE 0x0012
2266#define NETDEV_NOTIFY_PEERS 0x0013
2267#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002268#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002269#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002270#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002271#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002272#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002273#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002274#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002275#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002276#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002277
Joe Perchesf629d202013-09-26 14:48:15 -07002278int register_netdevice_notifier(struct notifier_block *nb);
2279int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002280
2281struct netdev_notifier_info {
2282 struct net_device *dev;
2283};
2284
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002285struct netdev_notifier_change_info {
2286 struct netdev_notifier_info info; /* must be first */
2287 unsigned int flags_changed;
2288};
2289
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002290struct netdev_notifier_changeupper_info {
2291 struct netdev_notifier_info info; /* must be first */
2292 struct net_device *upper_dev; /* new upper dev */
2293 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002294 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002295 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002296};
2297
Jiri Pirko04d48262015-12-03 12:12:15 +01002298struct netdev_notifier_changelowerstate_info {
2299 struct netdev_notifier_info info; /* must be first */
2300 void *lower_state_info; /* is lower dev state */
2301};
2302
Cong Wang75538c22013-05-29 11:30:50 +08002303static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2304 struct net_device *dev)
2305{
2306 info->dev = dev;
2307}
2308
Jiri Pirko351638e2013-05-28 01:30:21 +00002309static inline struct net_device *
2310netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2311{
2312 return info->dev;
2313}
2314
Joe Perchesf629d202013-09-26 14:48:15 -07002315int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002316
2317
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318extern rwlock_t dev_base_lock; /* Device list lock */
2319
Eric W. Biederman881d9662007-09-17 11:56:21 -07002320#define for_each_netdev(net, d) \
2321 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002322#define for_each_netdev_reverse(net, d) \
2323 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002324#define for_each_netdev_rcu(net, d) \
2325 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002326#define for_each_netdev_safe(net, d, n) \
2327 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2328#define for_each_netdev_continue(net, d) \
2329 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002330#define for_each_netdev_continue_rcu(net, d) \
2331 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002332#define for_each_netdev_in_bond_rcu(bond, slave) \
2333 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002334 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002335#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2336
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002337static inline struct net_device *next_net_device(struct net_device *dev)
2338{
2339 struct list_head *lh;
2340 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002341
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002342 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002343 lh = dev->dev_list.next;
2344 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2345}
2346
Eric Dumazetce81b762009-11-11 17:34:30 +00002347static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2348{
2349 struct list_head *lh;
2350 struct net *net;
2351
2352 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002353 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002354 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2355}
2356
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002357static inline struct net_device *first_net_device(struct net *net)
2358{
2359 return list_empty(&net->dev_base_head) ? NULL :
2360 net_device_entry(net->dev_base_head.next);
2361}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002362
Eric Dumazetccf43432011-01-26 18:08:02 +00002363static inline struct net_device *first_net_device_rcu(struct net *net)
2364{
2365 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2366
2367 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2368}
2369
Joe Perchesf629d202013-09-26 14:48:15 -07002370int netdev_boot_setup_check(struct net_device *dev);
2371unsigned long netdev_boot_base(const char *prefix, int unit);
2372struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2373 const char *hwaddr);
2374struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2375struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2376void dev_add_pack(struct packet_type *pt);
2377void dev_remove_pack(struct packet_type *pt);
2378void __dev_remove_pack(struct packet_type *pt);
2379void dev_add_offload(struct packet_offload *po);
2380void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002382int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002383int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002384struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2385 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002386struct net_device *dev_get_by_name(struct net *net, const char *name);
2387struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2388struct net_device *__dev_get_by_name(struct net *net, const char *name);
2389int dev_alloc_name(struct net_device *dev, const char *name);
2390int dev_open(struct net_device *dev);
2391int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002392int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002393void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002394int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002395int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002396int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002397int register_netdevice(struct net_device *dev);
2398void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2399void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002400static inline void unregister_netdevice(struct net_device *dev)
2401{
2402 unregister_netdevice_queue(dev, NULL);
2403}
2404
Joe Perchesf629d202013-09-26 14:48:15 -07002405int netdev_refcnt_read(const struct net_device *dev);
2406void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002407void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002408void synchronize_net(void);
2409int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002410
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002411DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002412#define XMIT_RECURSION_LIMIT 10
2413
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002414static inline int dev_recursion_level(void)
2415{
2416 return this_cpu_read(xmit_recursion);
2417}
2418
Joe Perchesf629d202013-09-26 14:48:15 -07002419struct net_device *dev_get_by_index(struct net *net, int ifindex);
2420struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2421struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2422int netdev_get_name(struct net *net, char *name, int ifindex);
2423int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002424int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002425
2426static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2427{
2428 return NAPI_GRO_CB(skb)->data_offset;
2429}
2430
2431static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2432{
2433 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2434}
2435
2436static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2437{
2438 NAPI_GRO_CB(skb)->data_offset += len;
2439}
2440
Herbert Xua5b1cf22009-05-26 18:50:28 +00002441static inline void *skb_gro_header_fast(struct sk_buff *skb,
2442 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002443{
Herbert Xu78a478d2009-05-26 18:50:21 +00002444 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002445}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446
Herbert Xua5b1cf22009-05-26 18:50:28 +00002447static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2448{
2449 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2450}
2451
2452static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2453 unsigned int offset)
2454{
Herbert Xu17dd7592011-07-27 06:16:28 -07002455 if (!pskb_may_pull(skb, hlen))
2456 return NULL;
2457
Herbert Xua5b1cf22009-05-26 18:50:28 +00002458 NAPI_GRO_CB(skb)->frag0 = NULL;
2459 NAPI_GRO_CB(skb)->frag0_len = 0;
Herbert Xu17dd7592011-07-27 06:16:28 -07002460 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002461}
2462
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002463static inline void *skb_gro_network_header(struct sk_buff *skb)
2464{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002465 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2466 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002467}
2468
Jerry Chubf5a7552014-01-07 10:23:19 -08002469static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2470 const void *start, unsigned int len)
2471{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002472 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002473 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2474 csum_partial(start, len, 0));
2475}
2476
Tom Herbert573e8fc2014-08-22 13:33:47 -07002477/* GRO checksum functions. These are logical equivalents of the normal
2478 * checksum functions (in skbuff.h) except that they operate on the GRO
2479 * offsets and fields in sk_buff.
2480 */
2481
2482__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2483
Tom Herbert15e23962015-02-10 16:30:31 -08002484static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2485{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002486 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002487}
2488
Tom Herbert573e8fc2014-08-22 13:33:47 -07002489static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2490 bool zero_okay,
2491 __sum16 check)
2492{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002493 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2494 skb_checksum_start_offset(skb) <
2495 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002496 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002497 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002498 (!zero_okay || check));
2499}
2500
2501static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2502 __wsum psum)
2503{
2504 if (NAPI_GRO_CB(skb)->csum_valid &&
2505 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2506 return 0;
2507
2508 NAPI_GRO_CB(skb)->csum = psum;
2509
2510 return __skb_gro_checksum_complete(skb);
2511}
2512
Tom Herbert573e8fc2014-08-22 13:33:47 -07002513static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2514{
Tom Herbert662880f2014-08-27 21:26:56 -07002515 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2516 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2517 NAPI_GRO_CB(skb)->csum_cnt--;
2518 } else {
2519 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2520 * verified a new top level checksum or an encapsulated one
2521 * during GRO. This saves work if we fallback to normal path.
2522 */
2523 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002524 }
2525}
2526
2527#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2528 compute_pseudo) \
2529({ \
2530 __sum16 __ret = 0; \
2531 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2532 __ret = __skb_gro_checksum_validate_complete(skb, \
2533 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002534 if (__ret) \
2535 __skb_mark_checksum_bad(skb); \
2536 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002537 skb_gro_incr_csum_unnecessary(skb); \
2538 __ret; \
2539})
2540
2541#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2542 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2543
2544#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2545 compute_pseudo) \
2546 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2547
2548#define skb_gro_checksum_simple_validate(skb) \
2549 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2550
Tom Herbertd96535a2014-08-31 15:12:42 -07002551static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2552{
2553 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2554 !NAPI_GRO_CB(skb)->csum_valid);
2555}
2556
2557static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2558 __sum16 check, __wsum pseudo)
2559{
2560 NAPI_GRO_CB(skb)->csum = ~pseudo;
2561 NAPI_GRO_CB(skb)->csum_valid = 1;
2562}
2563
2564#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2565do { \
2566 if (__skb_gro_checksum_convert_check(skb)) \
2567 __skb_gro_checksum_convert(skb, check, \
2568 compute_pseudo(skb, proto)); \
2569} while (0)
2570
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002571struct gro_remcsum {
2572 int offset;
2573 __wsum delta;
2574};
2575
2576static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2577{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002578 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002579 grc->delta = 0;
2580}
2581
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002582static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2583 unsigned int off, size_t hdrlen,
2584 int start, int offset,
2585 struct gro_remcsum *grc,
2586 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002587{
2588 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002589 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002590
2591 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2592
Tom Herbert15e23962015-02-10 16:30:31 -08002593 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002594 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2595 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002596 }
2597
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002598 ptr = skb_gro_header_fast(skb, off);
2599 if (skb_gro_header_hard(skb, off + plen)) {
2600 ptr = skb_gro_header_slow(skb, off + plen, off);
2601 if (!ptr)
2602 return NULL;
2603 }
2604
2605 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2606 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002607
2608 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002609 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002610
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002611 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002612 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002613
2614 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002615}
2616
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002617static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2618 struct gro_remcsum *grc)
2619{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002620 void *ptr;
2621 size_t plen = grc->offset + sizeof(u16);
2622
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002623 if (!grc->delta)
2624 return;
2625
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002626 ptr = skb_gro_header_fast(skb, grc->offset);
2627 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2628 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2629 if (!ptr)
2630 return;
2631 }
2632
2633 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002634}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002635
Tom Herbert6ae23ad2015-12-14 11:19:46 -08002636struct skb_csum_offl_spec {
2637 __u16 ipv4_okay:1,
2638 ipv6_okay:1,
2639 encap_okay:1,
2640 ip_options_okay:1,
2641 ext_hdrs_okay:1,
2642 tcp_okay:1,
2643 udp_okay:1,
2644 sctp_okay:1,
2645 vlan_okay:1,
2646 no_encapped_ipv6:1,
2647 no_not_encapped:1;
2648};
2649
2650bool __skb_csum_offload_chk(struct sk_buff *skb,
2651 const struct skb_csum_offl_spec *spec,
2652 bool *csum_encapped,
2653 bool csum_help);
2654
2655static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2656 const struct skb_csum_offl_spec *spec,
2657 bool *csum_encapped,
2658 bool csum_help)
2659{
2660 if (skb->ip_summed != CHECKSUM_PARTIAL)
2661 return false;
2662
2663 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2664}
2665
2666static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2667 const struct skb_csum_offl_spec *spec)
2668{
2669 bool csum_encapped;
2670
2671 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2672}
2673
2674static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2675{
2676 static const struct skb_csum_offl_spec csum_offl_spec = {
2677 .ipv4_okay = 1,
2678 .ip_options_okay = 1,
2679 .ipv6_okay = 1,
2680 .vlan_okay = 1,
2681 .tcp_okay = 1,
2682 .udp_okay = 1,
2683 };
2684
2685 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2686}
2687
2688static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2689{
2690 static const struct skb_csum_offl_spec csum_offl_spec = {
2691 .ipv4_okay = 1,
2692 .ip_options_okay = 1,
2693 .tcp_okay = 1,
2694 .udp_okay = 1,
2695 .vlan_okay = 1,
2696 };
2697
2698 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2699}
2700
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002701static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2702 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002703 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002704 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002705{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002706 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002707 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002708
2709 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002710}
2711
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002712static inline int dev_parse_header(const struct sk_buff *skb,
2713 unsigned char *haddr)
2714{
2715 const struct net_device *dev = skb->dev;
2716
Patrick McHardy1b833362007-10-18 05:09:28 -07002717 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002718 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002719 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002720}
2721
Willem de Bruijn2793a232016-03-09 21:58:32 -05002722/* ll_header must have at least hard_header_len allocated */
2723static inline bool dev_validate_header(const struct net_device *dev,
2724 char *ll_header, int len)
2725{
2726 if (likely(len >= dev->hard_header_len))
2727 return true;
2728
2729 if (capable(CAP_SYS_RAWIO)) {
2730 memset(ll_header + len, 0, dev->hard_header_len - len);
2731 return true;
2732 }
2733
2734 if (dev->header_ops && dev->header_ops->validate)
2735 return dev->header_ops->validate(ll_header, len);
2736
2737 return false;
2738}
2739
Linus Torvalds1da177e2005-04-16 15:20:36 -07002740typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002741int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002742static inline int unregister_gifconf(unsigned int family)
2743{
2744 return register_gifconf(family, NULL);
2745}
2746
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002747#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002748#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002749struct sd_flow_limit {
2750 u64 count;
2751 unsigned int num_buckets;
2752 unsigned int history_head;
2753 u16 history[FLOW_LIMIT_HISTORY];
2754 u8 buckets[];
2755};
2756
2757extern int netdev_flow_limit_table_len;
2758#endif /* CONFIG_NET_FLOW_LIMIT */
2759
Linus Torvalds1da177e2005-04-16 15:20:36 -07002760/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002761 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002762 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002763struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002765 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766
Changli Gaodee42872010-05-02 05:42:16 +00002767 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002768 unsigned int processed;
2769 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002770 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002771#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002772 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002773#endif
2774#ifdef CONFIG_NET_FLOW_LIMIT
2775 struct sd_flow_limit __rcu *flow_limit;
2776#endif
2777 struct Qdisc *output_queue;
2778 struct Qdisc **output_queue_tailp;
2779 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002780
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002781#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002782 /* input_queue_head should be written by cpu owning this struct,
2783 * and only read by other cpus. Worth using a cache line.
2784 */
2785 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2786
2787 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002788 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002789 struct softnet_data *rps_ipi_next;
2790 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002791 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002792#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002793 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002794 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002795 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002796
Linus Torvalds1da177e2005-04-16 15:20:36 -07002797};
2798
Tom Herbert76cc8b12010-05-20 18:37:59 +00002799static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002800{
2801#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002802 sd->input_queue_head++;
2803#endif
2804}
2805
2806static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2807 unsigned int *qtail)
2808{
2809#ifdef CONFIG_RPS
2810 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002811#endif
2812}
2813
Tom Herbert0a9627f2010-03-16 08:03:29 +00002814DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002815
Joe Perchesf629d202013-09-26 14:48:15 -07002816void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da42014-09-12 20:04:52 -07002817void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002818
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002819static inline void netif_tx_schedule_all(struct net_device *dev)
2820{
2821 unsigned int i;
2822
2823 for (i = 0; i < dev->num_tx_queues; i++)
2824 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2825}
2826
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002827static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002828{
Tom Herbert734664982011-11-28 16:32:44 +00002829 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002830}
2831
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002832/**
2833 * netif_start_queue - allow transmit
2834 * @dev: network device
2835 *
2836 * Allow upper layers to call the device hard_start_xmit routine.
2837 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002838static inline void netif_start_queue(struct net_device *dev)
2839{
David S. Millere8a04642008-07-17 00:34:19 -07002840 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002841}
2842
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002843static inline void netif_tx_start_all_queues(struct net_device *dev)
2844{
2845 unsigned int i;
2846
2847 for (i = 0; i < dev->num_tx_queues; i++) {
2848 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2849 netif_tx_start_queue(txq);
2850 }
2851}
2852
John Fastabend46e5da42014-09-12 20:04:52 -07002853void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002854
Dave Jonesd29f7492008-07-22 14:09:06 -07002855/**
2856 * netif_wake_queue - restart transmit
2857 * @dev: network device
2858 *
2859 * Allow upper layers to call the device hard_start_xmit routine.
2860 * Used for flow control when transmit resources are available.
2861 */
David S. Miller79d16382008-07-08 23:14:46 -07002862static inline void netif_wake_queue(struct net_device *dev)
2863{
David S. Millere8a04642008-07-17 00:34:19 -07002864 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865}
2866
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002867static inline void netif_tx_wake_all_queues(struct net_device *dev)
2868{
2869 unsigned int i;
2870
2871 for (i = 0; i < dev->num_tx_queues; i++) {
2872 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2873 netif_tx_wake_queue(txq);
2874 }
2875}
2876
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002877static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002878{
Tom Herbert734664982011-11-28 16:32:44 +00002879 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002880}
2881
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002882/**
2883 * netif_stop_queue - stop transmitted packets
2884 * @dev: network device
2885 *
2886 * Stop upper layers calling the device hard_start_xmit routine.
2887 * Used for flow control when transmit resources are unavailable.
2888 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002889static inline void netif_stop_queue(struct net_device *dev)
2890{
David S. Millere8a04642008-07-17 00:34:19 -07002891 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002892}
2893
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002894void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002895
David S. Miller4d295152012-03-07 21:02:35 -05002896static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002897{
Tom Herbert734664982011-11-28 16:32:44 +00002898 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002899}
2900
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002901/**
2902 * netif_queue_stopped - test if transmit queue is flowblocked
2903 * @dev: network device
2904 *
2905 * Test if transmit queue on device is currently unable to send.
2906 */
David S. Miller4d295152012-03-07 21:02:35 -05002907static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002908{
David S. Millere8a04642008-07-17 00:34:19 -07002909 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002910}
2911
David S. Miller4d295152012-03-07 21:02:35 -05002912static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002913{
Tom Herbert734664982011-11-28 16:32:44 +00002914 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2915}
2916
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002917static inline bool
2918netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002919{
2920 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2921}
2922
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002923static inline bool
2924netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2925{
2926 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2927}
2928
Eric Dumazet53511452014-10-08 08:19:27 -07002929/**
2930 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2931 * @dev_queue: pointer to transmit queue
2932 *
2933 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002934 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002935 */
2936static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2937{
2938#ifdef CONFIG_BQL
2939 prefetchw(&dev_queue->dql.num_queued);
2940#endif
2941}
2942
2943/**
2944 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2945 * @dev_queue: pointer to transmit queue
2946 *
2947 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002948 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002949 */
2950static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2951{
2952#ifdef CONFIG_BQL
2953 prefetchw(&dev_queue->dql.limit);
2954#endif
2955}
2956
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002957static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2958 unsigned int bytes)
2959{
Tom Herbert114cf582011-11-28 16:33:09 +00002960#ifdef CONFIG_BQL
2961 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002962
2963 if (likely(dql_avail(&dev_queue->dql) >= 0))
2964 return;
2965
2966 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2967
2968 /*
2969 * The XOFF flag must be set before checking the dql_avail below,
2970 * because in netdev_tx_completed_queue we update the dql_completed
2971 * before checking the XOFF flag.
2972 */
2973 smp_mb();
2974
2975 /* check again in case another CPU has just made room avail */
2976 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2977 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002978#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002979}
2980
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002981/**
2982 * netdev_sent_queue - report the number of bytes queued to hardware
2983 * @dev: network device
2984 * @bytes: number of bytes queued to the hardware device queue
2985 *
2986 * Report the number of bytes queued for sending/completion to the network
2987 * device hardware queue. @bytes should be a good approximation and should
2988 * exactly match netdev_completed_queue() @bytes
2989 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002990static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2991{
2992 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2993}
2994
2995static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002996 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002997{
Tom Herbert114cf582011-11-28 16:33:09 +00002998#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002999 if (unlikely(!bytes))
3000 return;
3001
3002 dql_completed(&dev_queue->dql, bytes);
3003
3004 /*
3005 * Without the memory barrier there is a small possiblity that
3006 * netdev_tx_sent_queue will miss the update and cause the queue to
3007 * be stopped forever
3008 */
3009 smp_mb();
3010
3011 if (dql_avail(&dev_queue->dql) < 0)
3012 return;
3013
3014 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3015 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00003016#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003017}
3018
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003019/**
3020 * netdev_completed_queue - report bytes and packets completed by device
3021 * @dev: network device
3022 * @pkts: actual number of packets sent over the medium
3023 * @bytes: actual number of bytes sent over the medium
3024 *
3025 * Report the number of bytes and packets transmitted by the network device
3026 * hardware queue over the physical medium, @bytes must exactly match the
3027 * @bytes amount passed to netdev_sent_queue()
3028 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003029static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003030 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003031{
3032 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3033}
3034
3035static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3036{
Tom Herbert114cf582011-11-28 16:33:09 +00003037#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003038 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003039 dql_reset(&q->dql);
3040#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003041}
3042
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003043/**
3044 * netdev_reset_queue - reset the packets and bytes count of a network device
3045 * @dev_queue: network device
3046 *
3047 * Reset the bytes and packet count of a network device and clear the
3048 * software flow control OFF bit for this network device
3049 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003050static inline void netdev_reset_queue(struct net_device *dev_queue)
3051{
3052 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003053}
3054
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003055/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003056 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3057 * @dev: network device
3058 * @queue_index: given tx queue index
3059 *
3060 * Returns 0 if given tx queue index >= number of device tx queues,
3061 * otherwise returns the originally passed tx queue index.
3062 */
3063static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3064{
3065 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3066 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3067 dev->name, queue_index,
3068 dev->real_num_tx_queues);
3069 return 0;
3070 }
3071
3072 return queue_index;
3073}
3074
3075/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003076 * netif_running - test if up
3077 * @dev: network device
3078 *
3079 * Test if the device has been brought up.
3080 */
David S. Miller4d295152012-03-07 21:02:35 -05003081static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003082{
3083 return test_bit(__LINK_STATE_START, &dev->state);
3084}
3085
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003086/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003087 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003088 * stop, and a check if it's stopped. All other device management is
3089 * done at the overall netdevice level.
3090 * Also test the device if we're multiqueue.
3091 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003092
3093/**
3094 * netif_start_subqueue - allow sending packets on subqueue
3095 * @dev: network device
3096 * @queue_index: sub queue index
3097 *
3098 * Start individual transmit queue of a device with multiple transmit queues.
3099 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003100static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3101{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003102 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003103
3104 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003105}
3106
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003107/**
3108 * netif_stop_subqueue - stop sending packets on subqueue
3109 * @dev: network device
3110 * @queue_index: sub queue index
3111 *
3112 * Stop individual transmit queue of a device with multiple transmit queues.
3113 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003114static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3115{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003116 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003117 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003118}
3119
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003120/**
3121 * netif_subqueue_stopped - test status of subqueue
3122 * @dev: network device
3123 * @queue_index: sub queue index
3124 *
3125 * Check individual transmit queue of a device with multiple transmit queues.
3126 */
David S. Miller4d295152012-03-07 21:02:35 -05003127static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3128 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003129{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003130 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003131
3132 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003133}
3134
David S. Miller4d295152012-03-07 21:02:35 -05003135static inline bool netif_subqueue_stopped(const struct net_device *dev,
3136 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003137{
3138 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3139}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003140
John Fastabend46e5da42014-09-12 20:04:52 -07003141void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003142
Alexander Duyck537c00d2013-01-10 08:57:02 +00003143#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003144int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003145 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003146#else
3147static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003148 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003149 u16 index)
3150{
3151 return 0;
3152}
3153#endif
3154
Jiri Pirko5605c762015-05-12 14:56:12 +02003155u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3156 unsigned int num_tx_queues);
3157
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003158/*
3159 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3160 * as a distribution range limit for the returned value.
3161 */
3162static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003163 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003164{
3165 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3166}
3167
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003168/**
3169 * netif_is_multiqueue - test if device has multiple transmit queues
3170 * @dev: network device
3171 *
3172 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003173 */
David S. Miller4d295152012-03-07 21:02:35 -05003174static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003175{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003176 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003177}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003178
Joe Perchesf629d202013-09-26 14:48:15 -07003179int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003180
Michael Daltona953be52014-01-16 22:23:28 -08003181#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003182int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003183#else
3184static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3185 unsigned int rxq)
3186{
3187 return 0;
3188}
3189#endif
3190
Michael Daltona953be52014-01-16 22:23:28 -08003191#ifdef CONFIG_SYSFS
3192static inline unsigned int get_netdev_rx_queue_index(
3193 struct netdev_rx_queue *queue)
3194{
3195 struct net_device *dev = queue->dev;
3196 int index = queue - dev->_rx;
3197
3198 BUG_ON(index >= dev->num_rx_queues);
3199 return index;
3200}
3201#endif
3202
Yuval Mintz16917b82012-07-01 03:18:50 +00003203#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003204int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003205
Eric Dumazete6247022013-12-05 04:45:08 -08003206enum skb_free_reason {
3207 SKB_REASON_CONSUMED,
3208 SKB_REASON_DROPPED,
3209};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003210
Eric Dumazete6247022013-12-05 04:45:08 -08003211void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3212void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3213
3214/*
3215 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3216 * interrupt context or with hardware interrupts being disabled.
3217 * (in_irq() || irqs_disabled())
3218 *
3219 * We provide four helpers that can be used in following contexts :
3220 *
3221 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3222 * replacing kfree_skb(skb)
3223 *
3224 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3225 * Typically used in place of consume_skb(skb) in TX completion path
3226 *
3227 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3228 * replacing kfree_skb(skb)
3229 *
3230 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3231 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003232 */
Eric Dumazete6247022013-12-05 04:45:08 -08003233static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3234{
3235 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3236}
3237
3238static inline void dev_consume_skb_irq(struct sk_buff *skb)
3239{
3240 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3241}
3242
3243static inline void dev_kfree_skb_any(struct sk_buff *skb)
3244{
3245 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3246}
3247
3248static inline void dev_consume_skb_any(struct sk_buff *skb)
3249{
3250 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3251}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003252
Joe Perchesf629d202013-09-26 14:48:15 -07003253int netif_rx(struct sk_buff *skb);
3254int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003255int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003256gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3257void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3258struct sk_buff *napi_get_frags(struct napi_struct *napi);
3259gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003260struct packet_offload *gro_find_receive_by_type(__be16 type);
3261struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003262
3263static inline void napi_free_frags(struct napi_struct *napi)
3264{
3265 kfree_skb(napi->skb);
3266 napi->skb = NULL;
3267}
3268
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003269bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003270int netdev_rx_handler_register(struct net_device *dev,
3271 rx_handler_func_t *rx_handler,
3272 void *rx_handler_data);
3273void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003274
Joe Perchesf629d202013-09-26 14:48:15 -07003275bool dev_valid_name(const char *name);
3276int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3277int dev_ethtool(struct net *net, struct ifreq *);
3278unsigned int dev_get_flags(const struct net_device *);
3279int __dev_change_flags(struct net_device *, unsigned int flags);
3280int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003281void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3282 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003283int dev_change_name(struct net_device *, const char *);
3284int dev_set_alias(struct net_device *, const char *, size_t);
3285int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3286int dev_set_mtu(struct net_device *, int);
3287void dev_set_group(struct net_device *, int);
3288int dev_set_mac_address(struct net_device *, struct sockaddr *);
3289int dev_change_carrier(struct net_device *, bool new_carrier);
3290int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003291 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003292int dev_get_phys_port_name(struct net_device *dev,
3293 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003294int dev_change_proto_down(struct net_device *dev, bool proto_down);
Brenden Blancoa7862b42016-07-19 12:16:48 -07003295int dev_change_xdp_fd(struct net_device *dev, int fd);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003296struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003297struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3298 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003299int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003300int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003301bool is_skb_forwardable(const struct net_device *dev,
3302 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003303
David Ahern74b20582016-05-10 11:19:50 -07003304void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3305
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003306extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003307
3308/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003309void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003310
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003311/**
3312 * dev_put - release reference to device
3313 * @dev: network device
3314 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003315 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003316 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003317static inline void dev_put(struct net_device *dev)
3318{
Christoph Lameter933393f2011-12-22 11:58:51 -06003319 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003320}
3321
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003322/**
3323 * dev_hold - get reference to device
3324 * @dev: network device
3325 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003326 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003327 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003328static inline void dev_hold(struct net_device *dev)
3329{
Christoph Lameter933393f2011-12-22 11:58:51 -06003330 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003331}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003332
3333/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3334 * and _off may be called from IRQ context, but it is caller
3335 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003336 *
3337 * The name carrier is inappropriate, these functions should really be
3338 * called netif_lowerlayer_*() because they represent the state of any
3339 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003340 */
3341
Joe Perchesf629d202013-09-26 14:48:15 -07003342void linkwatch_init_dev(struct net_device *dev);
3343void linkwatch_fire_event(struct net_device *dev);
3344void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003345
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003346/**
3347 * netif_carrier_ok - test if carrier present
3348 * @dev: network device
3349 *
3350 * Check if carrier is present on device
3351 */
David S. Miller4d295152012-03-07 21:02:35 -05003352static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003353{
3354 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3355}
3356
Joe Perchesf629d202013-09-26 14:48:15 -07003357unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003358
Joe Perchesf629d202013-09-26 14:48:15 -07003359void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003360
Joe Perchesf629d202013-09-26 14:48:15 -07003361void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003362
Joe Perchesf629d202013-09-26 14:48:15 -07003363void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003364
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003365/**
3366 * netif_dormant_on - mark device as dormant.
3367 * @dev: network device
3368 *
3369 * Mark device as dormant (as per RFC2863).
3370 *
3371 * The dormant state indicates that the relevant interface is not
3372 * actually in a condition to pass packets (i.e., it is not 'up') but is
3373 * in a "pending" state, waiting for some external event. For "on-
3374 * demand" interfaces, this new state identifies the situation where the
3375 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003376 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003377static inline void netif_dormant_on(struct net_device *dev)
3378{
3379 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3380 linkwatch_fire_event(dev);
3381}
3382
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003383/**
3384 * netif_dormant_off - set device as not dormant.
3385 * @dev: network device
3386 *
3387 * Device is not in dormant state.
3388 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003389static inline void netif_dormant_off(struct net_device *dev)
3390{
3391 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3392 linkwatch_fire_event(dev);
3393}
3394
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003395/**
3396 * netif_dormant - test if carrier present
3397 * @dev: network device
3398 *
3399 * Check if carrier is present on device
3400 */
David S. Miller4d295152012-03-07 21:02:35 -05003401static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003402{
3403 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3404}
3405
3406
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003407/**
3408 * netif_oper_up - test if device is operational
3409 * @dev: network device
3410 *
3411 * Check if carrier is operational
3412 */
David S. Miller4d295152012-03-07 21:02:35 -05003413static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003414{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003415 return (dev->operstate == IF_OPER_UP ||
3416 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3417}
3418
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003419/**
3420 * netif_device_present - is device available or removed
3421 * @dev: network device
3422 *
3423 * Check if device has not been removed from system.
3424 */
David S. Miller4d295152012-03-07 21:02:35 -05003425static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003426{
3427 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3428}
3429
Joe Perchesf629d202013-09-26 14:48:15 -07003430void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003431
Joe Perchesf629d202013-09-26 14:48:15 -07003432void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003433
3434/*
3435 * Network interface message level settings
3436 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003437
3438enum {
3439 NETIF_MSG_DRV = 0x0001,
3440 NETIF_MSG_PROBE = 0x0002,
3441 NETIF_MSG_LINK = 0x0004,
3442 NETIF_MSG_TIMER = 0x0008,
3443 NETIF_MSG_IFDOWN = 0x0010,
3444 NETIF_MSG_IFUP = 0x0020,
3445 NETIF_MSG_RX_ERR = 0x0040,
3446 NETIF_MSG_TX_ERR = 0x0080,
3447 NETIF_MSG_TX_QUEUED = 0x0100,
3448 NETIF_MSG_INTR = 0x0200,
3449 NETIF_MSG_TX_DONE = 0x0400,
3450 NETIF_MSG_RX_STATUS = 0x0800,
3451 NETIF_MSG_PKTDATA = 0x1000,
3452 NETIF_MSG_HW = 0x2000,
3453 NETIF_MSG_WOL = 0x4000,
3454};
3455
3456#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3457#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3458#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3459#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3460#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3461#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3462#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3463#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3464#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3465#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3466#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3467#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3468#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3469#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3470#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3471
3472static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3473{
3474 /* use default */
3475 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3476 return default_msg_enable_bits;
3477 if (debug_value == 0) /* no output */
3478 return 0;
3479 /* set low N bits */
3480 return (1 << debug_value) - 1;
3481}
3482
David S. Millerc773e842008-07-08 23:13:53 -07003483static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003484{
David S. Millerc773e842008-07-08 23:13:53 -07003485 spin_lock(&txq->_xmit_lock);
3486 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003487}
3488
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003489static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3490{
3491 spin_lock_bh(&txq->_xmit_lock);
3492 txq->xmit_lock_owner = smp_processor_id();
3493}
3494
David S. Miller4d295152012-03-07 21:02:35 -05003495static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003496{
David S. Miller4d295152012-03-07 21:02:35 -05003497 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003498 if (likely(ok))
3499 txq->xmit_lock_owner = smp_processor_id();
3500 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003501}
3502
David S. Millerc773e842008-07-08 23:13:53 -07003503static inline void __netif_tx_unlock(struct netdev_queue *txq)
3504{
3505 txq->xmit_lock_owner = -1;
3506 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003507}
3508
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003509static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3510{
3511 txq->xmit_lock_owner = -1;
3512 spin_unlock_bh(&txq->_xmit_lock);
3513}
3514
Eric Dumazet08baf562009-05-25 22:58:01 -07003515static inline void txq_trans_update(struct netdev_queue *txq)
3516{
3517 if (txq->xmit_lock_owner != -1)
3518 txq->trans_start = jiffies;
3519}
3520
Florian Westphalba162f82016-05-03 16:31:00 +02003521/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3522static inline void netif_trans_update(struct net_device *dev)
3523{
Florian Westphal9b366272016-05-03 16:33:14 +02003524 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3525
3526 if (txq->trans_start != jiffies)
3527 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003528}
3529
David S. Millerc3f26a22008-07-31 16:58:50 -07003530/**
3531 * netif_tx_lock - grab network device transmit lock
3532 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003533 *
3534 * Get network device transmit lock
3535 */
3536static inline void netif_tx_lock(struct net_device *dev)
3537{
3538 unsigned int i;
3539 int cpu;
3540
3541 spin_lock(&dev->tx_global_lock);
3542 cpu = smp_processor_id();
3543 for (i = 0; i < dev->num_tx_queues; i++) {
3544 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3545
3546 /* We are the only thread of execution doing a
3547 * freeze, but we have to grab the _xmit_lock in
3548 * order to synchronize with threads which are in
3549 * the ->hard_start_xmit() handler and already
3550 * checked the frozen bit.
3551 */
3552 __netif_tx_lock(txq, cpu);
3553 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3554 __netif_tx_unlock(txq);
3555 }
3556}
3557
3558static inline void netif_tx_lock_bh(struct net_device *dev)
3559{
3560 local_bh_disable();
3561 netif_tx_lock(dev);
3562}
3563
Herbert Xu932ff272006-06-09 12:20:56 -07003564static inline void netif_tx_unlock(struct net_device *dev)
3565{
David S. Millere8a04642008-07-17 00:34:19 -07003566 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003567
David S. Millere8a04642008-07-17 00:34:19 -07003568 for (i = 0; i < dev->num_tx_queues; i++) {
3569 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003570
David S. Millerc3f26a22008-07-31 16:58:50 -07003571 /* No need to grab the _xmit_lock here. If the
3572 * queue is not stopped for another reason, we
3573 * force a schedule.
3574 */
3575 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003576 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003577 }
3578 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003579}
3580
3581static inline void netif_tx_unlock_bh(struct net_device *dev)
3582{
David S. Millere8a04642008-07-17 00:34:19 -07003583 netif_tx_unlock(dev);
3584 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003585}
3586
David S. Millerc773e842008-07-08 23:13:53 -07003587#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003588 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003589 __netif_tx_lock(txq, cpu); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003590 } \
3591}
3592
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003593#define HARD_TX_TRYLOCK(dev, txq) \
3594 (((dev->features & NETIF_F_LLTX) == 0) ? \
3595 __netif_tx_trylock(txq) : \
3596 true )
3597
David S. Millerc773e842008-07-08 23:13:53 -07003598#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003599 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003600 __netif_tx_unlock(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003601 } \
3602}
3603
Linus Torvalds1da177e2005-04-16 15:20:36 -07003604static inline void netif_tx_disable(struct net_device *dev)
3605{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003606 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003607 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003608
David S. Millerc3f26a22008-07-31 16:58:50 -07003609 local_bh_disable();
3610 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003611 for (i = 0; i < dev->num_tx_queues; i++) {
3612 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003613
3614 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003615 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003616 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003617 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003618 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003619}
3620
David S. Millere308a5d2008-07-15 00:13:44 -07003621static inline void netif_addr_lock(struct net_device *dev)
3622{
3623 spin_lock(&dev->addr_list_lock);
3624}
3625
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003626static inline void netif_addr_lock_nested(struct net_device *dev)
3627{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003628 int subclass = SINGLE_DEPTH_NESTING;
3629
3630 if (dev->netdev_ops->ndo_get_lock_subclass)
3631 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3632
3633 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003634}
3635
David S. Millere308a5d2008-07-15 00:13:44 -07003636static inline void netif_addr_lock_bh(struct net_device *dev)
3637{
3638 spin_lock_bh(&dev->addr_list_lock);
3639}
3640
3641static inline void netif_addr_unlock(struct net_device *dev)
3642{
3643 spin_unlock(&dev->addr_list_lock);
3644}
3645
3646static inline void netif_addr_unlock_bh(struct net_device *dev)
3647{
3648 spin_unlock_bh(&dev->addr_list_lock);
3649}
3650
Jiri Pirkof001fde2009-05-05 02:48:28 +00003651/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003652 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003653 * rcu_read_lock held.
3654 */
3655#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003656 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003657
Linus Torvalds1da177e2005-04-16 15:20:36 -07003658/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3659
Joe Perchesf629d202013-09-26 14:48:15 -07003660void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003661
3662/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003663struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003664 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003665 void (*setup)(struct net_device *),
3666 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003667#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3668 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003669
Tom Gundersenc835a672014-07-14 16:37:24 +02003670#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3671 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3672 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003673
Joe Perchesf629d202013-09-26 14:48:15 -07003674int register_netdev(struct net_device *dev);
3675void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003676
Jiri Pirko22bedad2010-04-01 21:22:57 +00003677/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003678int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3679 struct netdev_hw_addr_list *from_list, int addr_len);
3680void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3681 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003682int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3683 struct net_device *dev,
3684 int (*sync)(struct net_device *, const unsigned char *),
3685 int (*unsync)(struct net_device *,
3686 const unsigned char *));
3687void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3688 struct net_device *dev,
3689 int (*unsync)(struct net_device *,
3690 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003691void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad2010-04-01 21:22:57 +00003692
Jiri Pirkof001fde2009-05-05 02:48:28 +00003693/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003694int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3695 unsigned char addr_type);
3696int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3697 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003698void dev_addr_flush(struct net_device *dev);
3699int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003700
3701/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003702int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3703int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3704int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3705int dev_uc_sync(struct net_device *to, struct net_device *from);
3706int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3707void dev_uc_unsync(struct net_device *to, struct net_device *from);
3708void dev_uc_flush(struct net_device *dev);
3709void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003710
Alexander Duyck670e5b82014-05-28 18:44:46 -07003711/**
3712 * __dev_uc_sync - Synchonize device's unicast list
3713 * @dev: device to sync
3714 * @sync: function to call if address should be added
3715 * @unsync: function to call if address should be removed
3716 *
3717 * Add newly added addresses to the interface, and release
3718 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003719 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003720static inline int __dev_uc_sync(struct net_device *dev,
3721 int (*sync)(struct net_device *,
3722 const unsigned char *),
3723 int (*unsync)(struct net_device *,
3724 const unsigned char *))
3725{
3726 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3727}
3728
3729/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003730 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003731 * @dev: device to sync
3732 * @unsync: function to call if address should be removed
3733 *
3734 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003735 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003736static inline void __dev_uc_unsync(struct net_device *dev,
3737 int (*unsync)(struct net_device *,
3738 const unsigned char *))
3739{
3740 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3741}
3742
Jiri Pirko22bedad2010-04-01 21:22:57 +00003743/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003744int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3745int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3746int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3747int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3748int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3749int dev_mc_sync(struct net_device *to, struct net_device *from);
3750int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3751void dev_mc_unsync(struct net_device *to, struct net_device *from);
3752void dev_mc_flush(struct net_device *dev);
3753void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003754
Alexander Duyck670e5b82014-05-28 18:44:46 -07003755/**
3756 * __dev_mc_sync - Synchonize device's multicast list
3757 * @dev: device to sync
3758 * @sync: function to call if address should be added
3759 * @unsync: function to call if address should be removed
3760 *
3761 * Add newly added addresses to the interface, and release
3762 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003763 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003764static inline int __dev_mc_sync(struct net_device *dev,
3765 int (*sync)(struct net_device *,
3766 const unsigned char *),
3767 int (*unsync)(struct net_device *,
3768 const unsigned char *))
3769{
3770 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3771}
3772
3773/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003774 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003775 * @dev: device to sync
3776 * @unsync: function to call if address should be removed
3777 *
3778 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003779 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003780static inline void __dev_mc_unsync(struct net_device *dev,
3781 int (*unsync)(struct net_device *,
3782 const unsigned char *))
3783{
3784 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3785}
3786
Linus Torvalds1da177e2005-04-16 15:20:36 -07003787/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003788void dev_set_rx_mode(struct net_device *dev);
3789void __dev_set_rx_mode(struct net_device *dev);
3790int dev_set_promiscuity(struct net_device *dev, int inc);
3791int dev_set_allmulti(struct net_device *dev, int inc);
3792void netdev_state_change(struct net_device *dev);
3793void netdev_notify_peers(struct net_device *dev);
3794void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003795/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003796void dev_load(struct net *net, const char *name);
3797struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3798 struct rtnl_link_stats64 *storage);
3799void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3800 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003801
3802extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003803extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003804extern int weight_p;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003805
Joe Perchesf629d202013-09-26 14:48:15 -07003806bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003807struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3808 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003809struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3810 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003811
3812/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003813#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3814 for (iter = &(dev)->adj_list.upper, \
3815 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3816 updev; \
3817 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3818
3819/* iterate through upper list, must be called under RCU read lock */
Veaceslav Falico2f268f12013-09-25 09:20:07 +02003820#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3821 for (iter = &(dev)->all_adj_list.upper, \
3822 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3823 updev; \
3824 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003825
Joe Perchesf629d202013-09-26 14:48:15 -07003826void *netdev_lower_get_next_private(struct net_device *dev,
3827 struct list_head **iter);
3828void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3829 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003830
3831#define netdev_for_each_lower_private(dev, priv, iter) \
3832 for (iter = (dev)->adj_list.lower.next, \
3833 priv = netdev_lower_get_next_private(dev, &(iter)); \
3834 priv; \
3835 priv = netdev_lower_get_next_private(dev, &(iter)))
3836
3837#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3838 for (iter = &(dev)->adj_list.lower, \
3839 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3840 priv; \
3841 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3842
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003843void *netdev_lower_get_next(struct net_device *dev,
3844 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003845
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003846#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003847 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003848 ldev = netdev_lower_get_next(dev, &(iter)); \
3849 ldev; \
3850 ldev = netdev_lower_get_next(dev, &(iter)))
3851
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003852struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3853 struct list_head **iter);
3854struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3855 struct list_head **iter);
3856
3857#define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3858 for (iter = (dev)->all_adj_list.lower.next, \
3859 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3860 ldev; \
3861 ldev = netdev_all_lower_get_next(dev, &(iter)))
3862
3863#define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3864 for (iter = (dev)->all_adj_list.lower.next, \
3865 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3866 ldev; \
3867 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3868
Joe Perchesf629d202013-09-26 14:48:15 -07003869void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003870void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003871struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3872struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3873int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3874int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003875 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003876 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003877void netdev_upper_dev_unlink(struct net_device *dev,
3878 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003879void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003880void *netdev_lower_dev_get_private(struct net_device *dev,
3881 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003882void netdev_lower_state_changed(struct net_device *lower_dev,
3883 void *lower_state_info);
Jiri Pirko18bfb922016-07-05 11:27:38 +02003884int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3885 struct neighbour *n);
3886void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3887 struct neighbour *n);
Eric Dumazet960fb622014-11-16 06:23:05 -08003888
3889/* RSS keys are 40 or 52 bytes long */
3890#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003891extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003892void netdev_rss_key_fill(void *buffer, size_t len);
3893
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003894int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003895int skb_checksum_help(struct sk_buff *skb);
3896struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3897 netdev_features_t features, bool tx_path);
3898struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3899 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003900
Moni Shoua61bd3852015-02-03 16:48:29 +02003901struct netdev_bonding_info {
3902 ifslave slave;
3903 ifbond master;
3904};
3905
3906struct netdev_notifier_bonding_info {
3907 struct netdev_notifier_info info; /* must be first */
3908 struct netdev_bonding_info bonding_info;
3909};
3910
3911void netdev_bonding_info_change(struct net_device *dev,
3912 struct netdev_bonding_info *bonding_info);
3913
Cong Wang12b00042013-02-05 16:36:38 +00003914static inline
3915struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3916{
3917 return __skb_gso_segment(skb, features, true);
3918}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003919__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003920
3921static inline bool can_checksum_protocol(netdev_features_t features,
3922 __be16 protocol)
3923{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003924 if (protocol == htons(ETH_P_FCOE))
3925 return !!(features & NETIF_F_FCOE_CRC);
3926
3927 /* Assume this is an IP checksum (not SCTP CRC) */
3928
3929 if (features & NETIF_F_HW_CSUM) {
3930 /* Can checksum everything */
3931 return true;
3932 }
3933
3934 switch (protocol) {
3935 case htons(ETH_P_IP):
3936 return !!(features & NETIF_F_IP_CSUM);
3937 case htons(ETH_P_IPV6):
3938 return !!(features & NETIF_F_IPV6_CSUM);
3939 default:
3940 return false;
3941 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003942}
Cong Wang12b00042013-02-05 16:36:38 +00003943
Tom Herbert6ae23ad2015-12-14 11:19:46 -08003944/* Map an ethertype into IP protocol if possible */
3945static inline int eproto_to_ipproto(int eproto)
3946{
3947 switch (eproto) {
3948 case htons(ETH_P_IP):
3949 return IPPROTO_IP;
3950 case htons(ETH_P_IPV6):
3951 return IPPROTO_IPV6;
3952 default:
3953 return -1;
3954 }
3955}
3956
Herbert Xufb286bb2005-11-10 13:01:24 -08003957#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003958void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003959#else
3960static inline void netdev_rx_csum_fault(struct net_device *dev)
3961{
3962}
3963#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003964/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003965void net_enable_timestamp(void);
3966void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003967
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003968#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003969int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003970#else
3971#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003972#endif
3973
David S. Miller47982482014-08-22 16:21:53 -07003974static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003975 struct sk_buff *skb, struct net_device *dev,
3976 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003977{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003978 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003979 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003980}
3981
David S. Miller10b3ad82014-08-29 21:07:24 -07003982static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003983 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003984{
3985 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003986 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003987
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003988 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003989 if (rc == NETDEV_TX_OK)
3990 txq_trans_update(txq);
3991
3992 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003993}
3994
Linus Torvalds42a2d922013-11-13 17:40:34 +09003995int netdev_class_create_file_ns(struct class_attribute *class_attr,
3996 const void *ns);
3997void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3998 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003999
4000static inline int netdev_class_create_file(struct class_attribute *class_attr)
4001{
4002 return netdev_class_create_file_ns(class_attr, NULL);
4003}
4004
4005static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4006{
4007 netdev_class_remove_file_ns(class_attr, NULL);
4008}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004009
Johannes Berg04600792010-08-05 17:45:15 +02004010extern struct kobj_ns_type_operations net_ns_type_operations;
4011
Joe Perchesf629d202013-09-26 14:48:15 -07004012const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004013
Joe Perchesf629d202013-09-26 14:48:15 -07004014void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004015
Michal Kubečekda081432014-05-20 08:29:25 +02004016static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4017 netdev_features_t f2)
4018{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004019 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4020 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004021 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004022 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004023 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004024 }
Michal Kubečekda081432014-05-20 08:29:25 +02004025
Tom Herbertc8cd0982015-12-14 11:19:44 -08004026 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004027}
4028
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004029static inline netdev_features_t netdev_get_wanted_features(
4030 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004031{
4032 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4033}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004034netdev_features_t netdev_increment_features(netdev_features_t all,
4035 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004036
4037/* Allow TSO being used on stacked device :
4038 * Performing the GSO segmentation before last device
4039 * is a performance improvement.
4040 */
4041static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4042 netdev_features_t mask)
4043{
4044 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4045}
4046
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004047int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004048void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004049void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004050
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004051void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4052 struct net_device *dev);
4053
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004054netdev_features_t passthru_features_check(struct sk_buff *skb,
4055 struct net_device *dev,
4056 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004057netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004058
David S. Miller4d295152012-03-07 21:02:35 -05004059static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004060{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004061 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004062
4063 /* check flags correspondence */
4064 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4065 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4066 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4067 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004068 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004069 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4070 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004071 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4072 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004073 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4074 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004075 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4076 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 -04004077 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004078 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004079 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004080
Herbert Xubcd76112006-06-30 13:36:35 -07004081 return (features & feature) == feature;
4082}
4083
David S. Miller4d295152012-03-07 21:02:35 -05004084static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004085{
Herbert Xu278b2512009-06-03 21:20:51 -07004086 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004087 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004088}
4089
Johannes Berg8b86a612015-04-17 15:45:04 +02004090static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004091 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004092{
Jesse Grossfc741212011-01-09 06:23:32 +00004093 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004094 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4095 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004096}
4097
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004098static inline void netif_set_gso_max_size(struct net_device *dev,
4099 unsigned int size)
4100{
4101 dev->gso_max_size = size;
4102}
4103
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004104static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4105 int pulled_hlen, u16 mac_offset,
4106 int mac_len)
4107{
4108 skb->protocol = protocol;
4109 skb->encapsulation = 1;
4110 skb_push(skb, pulled_hlen);
4111 skb_reset_transport_header(skb);
4112 skb->mac_header = mac_offset;
4113 skb->network_header = skb->mac_header + mac_len;
4114 skb->mac_len = mac_len;
4115}
4116
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004117static inline bool netif_is_macsec(const struct net_device *dev)
4118{
4119 return dev->priv_flags & IFF_MACSEC;
4120}
4121
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004122static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004123{
4124 return dev->priv_flags & IFF_MACVLAN;
4125}
4126
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004127static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004128{
4129 return dev->priv_flags & IFF_MACVLAN_PORT;
4130}
4131
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004132static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004133{
4134 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4135}
4136
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004137static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004138{
4139 return dev->priv_flags & IFF_IPVLAN_MASTER;
4140}
4141
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004142static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004143{
4144 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4145}
4146
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004147static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004148{
4149 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4150}
4151
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004152static inline bool netif_supports_nofcs(struct net_device *dev)
4153{
4154 return dev->priv_flags & IFF_SUPP_NOFCS;
4155}
4156
David Ahern007979e2015-09-29 20:07:10 -07004157static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004158{
David Ahern007979e2015-09-29 20:07:10 -07004159 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004160}
4161
David Ahernfee6d4c2015-10-05 08:51:24 -07004162static inline bool netif_is_l3_slave(const struct net_device *dev)
4163{
4164 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4165}
4166
Jiri Pirko0894ae32015-08-27 09:31:19 +02004167static inline bool netif_is_bridge_master(const struct net_device *dev)
4168{
4169 return dev->priv_flags & IFF_EBRIDGE;
4170}
4171
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004172static inline bool netif_is_bridge_port(const struct net_device *dev)
4173{
4174 return dev->priv_flags & IFF_BRIDGE_PORT;
4175}
4176
Jiri Pirko35d4e172015-08-27 09:31:20 +02004177static inline bool netif_is_ovs_master(const struct net_device *dev)
4178{
4179 return dev->priv_flags & IFF_OPENVSWITCH;
4180}
4181
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004182static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004183{
4184 return dev->priv_flags & IFF_TEAM;
4185}
4186
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004187static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004188{
4189 return dev->priv_flags & IFF_TEAM_PORT;
4190}
4191
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004192static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004193{
4194 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4195}
4196
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004197static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004198{
4199 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4200}
4201
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004202static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4203{
4204 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4205}
4206
Eric Dumazet02875872014-10-05 18:38:35 -07004207/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4208static inline void netif_keep_dst(struct net_device *dev)
4209{
4210 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4211}
4212
Paolo Abeni18d3df32016-07-14 18:00:10 +02004213/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4214static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4215{
4216 /* TODO: reserve and use an additional IFF bit, if we get more users */
4217 return dev->priv_flags & IFF_MACSEC;
4218}
4219
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004220extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004221
Joe Perches571ba422010-02-09 11:49:47 +00004222/* Logging, debugging and troubleshooting/diagnostic helpers. */
4223
4224/* netdev_printk helpers, similar to dev_printk */
4225
4226static inline const char *netdev_name(const struct net_device *dev)
4227{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004228 if (!dev->name[0] || strchr(dev->name, '%'))
4229 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004230 return dev->name;
4231}
4232
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004233static inline const char *netdev_reg_state(const struct net_device *dev)
4234{
4235 switch (dev->reg_state) {
4236 case NETREG_UNINITIALIZED: return " (uninitialized)";
4237 case NETREG_REGISTERED: return "";
4238 case NETREG_UNREGISTERING: return " (unregistering)";
4239 case NETREG_UNREGISTERED: return " (unregistered)";
4240 case NETREG_RELEASED: return " (released)";
4241 case NETREG_DUMMY: return " (dummy)";
4242 }
4243
4244 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4245 return " (unknown)";
4246}
4247
Joe Perchesf629d202013-09-26 14:48:15 -07004248__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004249void netdev_printk(const char *level, const struct net_device *dev,
4250 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004251__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004252void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004253__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004254void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004255__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004256void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004257__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004258void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004259__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004260void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004261__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004262void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004263__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004264void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004265
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004266#define MODULE_ALIAS_NETDEV(device) \
4267 MODULE_ALIAS("netdev-" device)
4268
Jim Cromieb558c962011-12-19 17:11:18 -05004269#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004270#define netdev_dbg(__dev, format, args...) \
4271do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004272 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004273} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004274#elif defined(DEBUG)
4275#define netdev_dbg(__dev, format, args...) \
4276 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004277#else
4278#define netdev_dbg(__dev, format, args...) \
4279({ \
4280 if (0) \
4281 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004282})
4283#endif
4284
4285#if defined(VERBOSE_DEBUG)
4286#define netdev_vdbg netdev_dbg
4287#else
4288
4289#define netdev_vdbg(dev, format, args...) \
4290({ \
4291 if (0) \
4292 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4293 0; \
4294})
4295#endif
4296
4297/*
4298 * netdev_WARN() acts like dev_printk(), but with the key difference
4299 * of using a WARN/WARN_ON to get the message out, including the
4300 * file/line information and a backtrace.
4301 */
4302#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004303 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4304 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004305
Joe Perchesb3d95c52010-02-09 11:49:49 +00004306/* netif printk helpers, similar to netdev_printk */
4307
4308#define netif_printk(priv, type, level, dev, fmt, args...) \
4309do { \
4310 if (netif_msg_##type(priv)) \
4311 netdev_printk(level, (dev), fmt, ##args); \
4312} while (0)
4313
Joe Perchesf45f4322010-06-27 01:02:36 +00004314#define netif_level(level, priv, type, dev, fmt, args...) \
4315do { \
4316 if (netif_msg_##type(priv)) \
4317 netdev_##level(dev, fmt, ##args); \
4318} while (0)
4319
Joe Perchesb3d95c52010-02-09 11:49:49 +00004320#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004321 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004322#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004323 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004324#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004325 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004326#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004327 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004328#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004329 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004330#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004331 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004332#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004333 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004334
Joe Perches0053ea92012-05-30 07:43:34 +00004335#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004336#define netif_dbg(priv, type, netdev, format, args...) \
4337do { \
4338 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004339 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004340} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004341#elif defined(DEBUG)
4342#define netif_dbg(priv, type, dev, format, args...) \
4343 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004344#else
4345#define netif_dbg(priv, type, dev, format, args...) \
4346({ \
4347 if (0) \
4348 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4349 0; \
4350})
4351#endif
4352
4353#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004354#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004355#else
4356#define netif_vdbg(priv, type, dev, format, args...) \
4357({ \
4358 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004359 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004360 0; \
4361})
4362#endif
Joe Perches571ba422010-02-09 11:49:47 +00004363
Cong Wang900ff8c2013-02-18 19:20:33 +00004364/*
4365 * The list of packet types we will receive (as opposed to discard)
4366 * and the routines to invoke.
4367 *
4368 * Why 16. Because with 16 the only overlap we get on a hash of the
4369 * low nibble of the protocol value is RARP/SNAP/X.25.
4370 *
4371 * NOTE: That is no longer true with the addition of VLAN tags. Not
4372 * sure which should go first, but I bet it won't make much
4373 * difference if we are running VLANs. The good news is that
4374 * this protocol won't be in the list unless compiled in, so
4375 * the average user (w/out VLANs) will not be adversely affected.
4376 * --BLG
4377 *
4378 * 0800 IP
4379 * 8100 802.1Q VLAN
4380 * 0001 802.3
4381 * 0002 AX.25
4382 * 0004 802.2
4383 * 8035 RARP
4384 * 0005 SNAP
4385 * 0805 X.25
4386 * 0806 ARP
4387 * 8137 IPX
4388 * 0009 Localtalk
4389 * 86DD IPv6
4390 */
4391#define PTYPE_HASH_SIZE (16)
4392#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4393
Jiri Pirko385a1542009-05-27 15:48:07 -07004394#endif /* _LINUX_NETDEVICE_H */