blob: 9c23bd2efb56393e7c616d42d427101c3f5f51d2 [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>
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -080044#ifdef CONFIG_DCB
Alexander Duyck2f90b862008-11-20 20:52:10 -080045#include <net/dcbnl.h>
46#endif
Neil Horman5bc14212011-11-22 05:10:51 +000047#include <net/netprio_cgroup.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020048
Michał Mirosława59e2ec2011-11-15 15:29:55 +000049#include <linux/netdev_features.h>
John Fastabend77162022012-04-15 06:43:56 +000050#include <linux/neighbour.h>
David Howells607ca462012-10-13 10:46:48 +010051#include <uapi/linux/netdevice.h>
Moni Shoua61bd3852015-02-03 16:48:29 +020052#include <uapi/linux/if_bonding.h>
John Fastabende4c67342016-02-16 21:16:15 -080053#include <uapi/linux/pkt_cls.h>
Jiri Kosina59cc1f62016-08-10 11:05:15 +020054#include <linux/hashtable.h>
Michał Mirosława59e2ec2011-11-15 15:29:55 +000055
Jeff Moyer115c1d62005-06-22 22:05:31 -070056struct netpoll_info;
Paul Gortmaker313162d2012-01-30 11:46:54 -050057struct device;
Richard Cochranc1f19b52010-07-17 08:49:36 +000058struct phy_device;
Andrew Lunnc6e970a2017-03-28 23:45:06 +020059struct dsa_switch_tree;
60
Johannes Berg704232c2007-04-23 12:20:05 -070061/* 802.11 specific */
62struct wireless_dev;
Alexander Aring98a18b62014-11-02 06:44:54 +010063/* 802.15.4 specific */
64struct wpan_dev;
Robert Shearman03c57742015-04-22 11:14:37 +010065struct mpls_dev;
Alexander Duyck7c46a642016-06-16 12:21:00 -070066/* UDP Tunnel offloads */
67struct udp_tunnel_info;
Brenden Blancoa7862b42016-07-19 12:16:48 -070068struct bpf_prog;
Linus Torvalds1da177e2005-04-16 15:20:36 -070069
Joe Perchesf629d202013-09-26 14:48:15 -070070void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000072
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000073/* Backlog congestion levels */
74#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75#define NET_RX_DROP 1 /* packet dropped */
76
Patrick McHardy572a9d72009-11-10 06:14:14 +000077/*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -050089 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
Patrick McHardy572a9d72009-11-10 06:14:14 +000091 * others are propagated to higher layers.
92 */
93
94/* qdisc ->enqueue() return codes. */
95#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000096#define NET_XMIT_DROP 0x01 /* skb dropped */
97#define NET_XMIT_CN 0x02 /* congestion notification */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000098#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -0200100/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +0000103#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000106/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000107#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000108
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000109enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000113};
114typedef enum netdev_tx netdev_tx_t;
115
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000116/*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120static inline bool dev_xmit_complete(int rc)
121{
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132}
133
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500135 * Compute the worst-case header length according to the protocols
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 * used.
137 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800138
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800139#if defined(CONFIG_HYPERV_NET)
140# define LL_MAX_HEADER 128
141#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700142# if defined(CONFIG_MAC80211_MESH)
143# define LL_MAX_HEADER 128
144# else
145# define LL_MAX_HEADER 96
146# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700148# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149#endif
150
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000151#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153#define MAX_HEADER LL_MAX_HEADER
154#else
155#define MAX_HEADER (LL_MAX_HEADER + 48)
156#endif
157
158/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800162
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800163struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187};
188
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189
190#include <linux/cache.h>
191#include <linux/skbuff.h>
192
Eric Dumazetadc93002011-11-17 03:13:26 +0000193#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100194#include <linux/static_key.h>
195extern struct static_key rps_needed;
Eric Dumazet13bfff22016-12-07 08:29:10 -0800196extern struct static_key rfs_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000197#endif
198
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199struct neighbour;
200struct neigh_parms;
201struct sk_buff;
202
Jiri Pirkof001fde2009-05-05 02:48:28 +0000203struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000207#define NETDEV_HW_ADDR_T_LAN 1
208#define NETDEV_HW_ADDR_T_SAN 2
209#define NETDEV_HW_ADDR_T_SLAVE 3
210#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000211#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000212 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000213 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700214 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400215 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000216 struct rcu_head rcu_head;
217};
218
Jiri Pirko31278e72009-06-17 01:12:19 +0000219struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222};
223
Jiri Pirko22bedad32010-04-01 21:22:57 +0000224#define netdev_hw_addr_list_count(l) ((l)->count)
225#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226#define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
228
229#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800231#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800233
Jiri Pirko22bedad32010-04-01 21:22:57 +0000234#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700236#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece2010-02-04 10:22:25 -0800238
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800239struct hh_cache {
Alexey Dobriyan5b3dc2f2017-04-10 11:11:17 +0300240 unsigned int hh_len;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800241 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
243 /* cached hardware header; allow for machine alignment needs. */
244#define HH_DATA_MOD 16
245#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247#define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250};
251
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500252/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700265struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000268 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500274 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700275};
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277/* These flag bits are private to the generic network queueing
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500278 * layer; they may not be explicitly referenced by any other
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 * code.
280 */
281
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800282enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288};
289
290
291/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500292 * This structure holds boot-time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800293 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 */
295struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298};
299#define NETDEV_BOOT_SETUP_MAX 8
300
Joe Perchesf629d202013-09-26 14:48:15 -0700301int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
303/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500310 * to the per-CPU poll_list, and whoever clears that bit
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000317 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700318 int (*poll)(struct napi_struct *, int);
319#ifdef CONFIG_NETPOLL
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700320 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 */
Eric Dumazet39e6c822017-02-28 10:34:50 -0800333 NAPI_STATE_MISSED, /* reschedule a napi */
David S. Millera0a46192008-01-07 20:35:07 -0800334 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Eric Dumazet217f6972016-11-15 10:15:11 -0800338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339};
340
341enum {
Eric Dumazet39e6c822017-02-28 10:34:50 -0800342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700349};
350
Ben Hutchings5b252f02009-10-29 07:17:09 +0000351enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700352 GRO_MERGED,
353 GRO_MERGED_FREE,
354 GRO_HELD,
355 GRO_NORMAL,
356 GRO_DROP,
Steffen Klassert25393d32017-02-15 09:39:44 +0100357 GRO_CONSUMED,
Herbert Xud1c76af2009-03-16 10:50:02 -0700358};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000359typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700360
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000361/*
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364 * further.
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000369 *
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
372 *
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
375 *
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
379 *
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381 * do with the skb.
382 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500383 * If the rx_handler consumed the skb in some way, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500385 * the skb to be delivered in some other way.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000386 *
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
390 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500391 * If the rx_handler decides the skb should be ignored, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400393 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000394 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000396 * delivered, it should return RX_HANDLER_PASS.
397 *
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
400 */
401
402enum rx_handler_result {
403 RX_HANDLER_CONSUMED,
404 RX_HANDLER_ANOTHER,
405 RX_HANDLER_EXACT,
406 RX_HANDLER_PASS,
407};
408typedef enum rx_handler_result rx_handler_result_t;
409typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000410
Joe Perchesf629d202013-09-26 14:48:15 -0700411void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700412void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700413
David S. Miller4d295152012-03-07 21:02:35 -0500414static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800415{
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
417}
418
Eric Dumazet39e6c822017-02-28 10:34:50 -0800419bool napi_schedule_prep(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700420
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 Dumazet364b6052016-11-15 10:15:13 -0800456bool napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700457/**
458 * napi_complete - NAPI processing complete
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500459 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700460 *
461 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800462 * Consider using napi_complete_done() instead.
Eric Dumazet364b6052016-11-15 10:15:13 -0800463 * Return false if device should avoid rearming interrupts.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700464 */
Eric Dumazet364b6052016-11-15 10:15:13 -0800465static inline bool napi_complete(struct napi_struct *n)
Eric Dumazet3b47d302014-11-06 21:09:44 -0800466{
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_del - remove a NAPI from global table
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 * Warning: caller must observe RCU grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800475 * before freeing memory containing @napi, if
476 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800477 * Note: core networking stack automatically calls it
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500478 * from netif_napi_del().
Eric Dumazet93d05d42015-11-18 06:31:03 -0800479 * Drivers might want to call this helper to combine all
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500480 * the needed RCU grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300481 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800482bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300483
484/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700485 * napi_disable - prevent NAPI from scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500486 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700487 *
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
490 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800491void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700492
493/**
494 * napi_enable - enable NAPI scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500495 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500static inline void napi_enable(struct napi_struct *n)
501{
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100503 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700504 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400505 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700506}
507
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700508/**
509 * napi_synchronize - wait until NAPI is not running
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500510 * @n: NAPI context
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516static inline void napi_synchronize(const struct napi_struct *n)
517{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700523}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700524
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800525enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700528 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700529};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200530
531#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
534
535#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 QUEUE_STATE_FROZEN)
538#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 QUEUE_STATE_FROZEN)
540
Tom Herbert734664982011-11-28 16:32:44 +0000541/*
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
549 */
David S. Miller79d16382008-07-08 23:14:46 -0700550
David S. Millerbb949fb2008-07-08 16:55:56 -0700551struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700552/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500553 * read-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700554 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700555 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700556 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700557 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000558#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000559 struct kobject kobj;
560#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000561#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 int numa_node;
563#endif
Florian Westphalc0ef0792016-05-03 03:29:09 +0200564 unsigned long tx_maxrate;
565 /*
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
568 */
569 unsigned long trans_timeout;
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700570/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500571 * write-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700575 /*
Florian Westphal9b366272016-05-03 16:33:14 +0200576 * Time (in jiffies) of last Tx
Eric Dumazet9d214932009-05-17 20:55:16 -0700577 */
578 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000579
Tom Herbert114cf582011-11-28 16:33:09 +0000580 unsigned long state;
581
582#ifdef CONFIG_BQL
583 struct dql dql;
584#endif
David S. Millere8a04642008-07-17 00:34:19 -0700585} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700586
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000587static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588{
589#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 return q->numa_node;
591#else
Changli Gaob236da62010-12-14 03:09:15 +0000592 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000593#endif
594}
595
596static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 q->numa_node = node;
600#endif
601}
602
Eric Dumazetdf334542010-03-24 19:13:54 +0000603#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000604/*
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
607 */
608struct rps_map {
609 unsigned int len;
610 struct rcu_head rcu;
611 u16 cpus[0];
612};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000613#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000614
Tom Herbertfec5e652010-04-16 16:01:27 -0700615/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700619 */
620struct rps_dev_flow {
621 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000622 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700623 unsigned int last_qtail;
624};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000625#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700626
627/*
628 * The rps_dev_flow_table structure contains a table of flow mappings.
629 */
630struct rps_dev_flow_table {
631 unsigned int mask;
632 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700633 struct rps_dev_flow flows[0];
634};
635#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000636 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700637
638/*
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
Eric Dumazet567e4b72015-02-06 12:59:01 -0800643 * rps_cpu_mask is used to partition the space, depending on number of
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
Eric Dumazet567e4b72015-02-06 12:59:01 -0800646 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700647 */
648struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800649 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800650
651 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700652};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800653#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700654
655#define RPS_NO_CPU 0xffff
656
Eric Dumazet567e4b72015-02-06 12:59:01 -0800657extern u32 rps_cpu_mask;
658extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659
Tom Herbertfec5e652010-04-16 16:01:27 -0700660static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 u32 hash)
662{
663 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700666
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500667 /* We only give a hint, preemption can change CPU under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800668 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700669
Eric Dumazet567e4b72015-02-06 12:59:01 -0800670 if (table->ents[index] != val)
671 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700672 }
673}
674
Ben Hutchingsc4454772011-01-19 11:03:53 +0000675#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700676bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000678#endif
Michael Daltona953be52014-01-16 22:23:28 -0800679#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000680
Tom Herbert0a9627f2010-03-16 08:03:29 +0000681/* This structure contains an instance of an RX queue. */
682struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800683#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800686#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000687 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000688 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000689} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800690
691/*
692 * RX queue sysfs structures and functions.
693 */
694struct rx_queue_attribute {
695 struct attribute attr;
696 ssize_t (*show)(struct netdev_rx_queue *queue,
697 struct rx_queue_attribute *attr, char *buf);
698 ssize_t (*store)(struct netdev_rx_queue *queue,
699 struct rx_queue_attribute *attr, const char *buf, size_t len);
700};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800701
Tom Herbertbf264142010-11-26 08:36:09 +0000702#ifdef CONFIG_XPS
703/*
704 * This structure holds an XPS map which can be of variable length. The
705 * map is an array of queues.
706 */
707struct xps_map {
708 unsigned int len;
709 unsigned int alloc_len;
710 struct rcu_head rcu;
711 u16 queues[0];
712};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000713#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100714#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000716
717/*
718 * This structure holds all XPS maps for device. Maps are indexed by CPU.
719 */
720struct xps_dev_maps {
721 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000722 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000723};
Alexander Duyck184c4492016-10-28 11:50:13 -0400724#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
Tom Herbertbf264142010-11-26 08:36:09 +0000726#endif /* CONFIG_XPS */
727
John Fastabend4f57c082011-01-17 08:06:04 +0000728#define TC_MAX_QUEUE 16
729#define TC_BITMASK 15
730/* HW offloaded queuing disciplines txq count and offset maps */
731struct netdev_tc_txq {
732 u16 count;
733 u16 offset;
734};
735
Neerav Parikh68bad942012-01-04 20:23:39 +0000736#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
737/*
738 * This structure is to hold information about the device
739 * configured to run FCoE protocol stack.
740 */
741struct netdev_fcoe_hbainfo {
742 char manufacturer[64];
743 char serial_number[64];
744 char hardware_version[64];
745 char driver_version[64];
746 char optionrom_version[64];
747 char firmware_version[64];
748 char model[256];
749 char model_description[256];
750};
751#endif
752
Jiri Pirko02637fc2014-11-28 14:34:16 +0100753#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200754
Jiri Pirko02637fc2014-11-28 14:34:16 +0100755/* This structure holds a unique identifier to identify some
756 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200757 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100758struct netdev_phys_item_id {
759 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200760 unsigned char id_len;
761};
762
Scott Feldmand754f982015-07-18 18:24:49 -0700763static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
764 struct netdev_phys_item_id *b)
765{
766 return a->id_len == b->id_len &&
767 memcmp(a->id, b->id, a->id_len) == 0;
768}
769
Daniel Borkmann99932d42014-02-16 15:55:20 +0100770typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
771 struct sk_buff *skb);
772
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800773/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800774 * that are being passed to the netdevice through the setup_tc op.
775 */
776enum {
777 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800778 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200779 TC_SETUP_CLSFLOWER,
Yotam Gigib87f7932016-07-21 12:03:12 +0200780 TC_SETUP_MATCHALL,
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100781 TC_SETUP_CLSBPF,
John Fastabend16e5cc62016-02-16 21:16:43 -0800782};
783
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800784struct tc_cls_u32_offload;
785
John Fastabend16e5cc62016-02-16 21:16:43 -0800786struct tc_to_netdev {
787 unsigned int type;
788 union {
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800789 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200790 struct tc_cls_flower_offload *cls_flower;
Yotam Gigib87f7932016-07-21 12:03:12 +0200791 struct tc_cls_matchall_offload *cls_mall;
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100792 struct tc_cls_bpf_offload *cls_bpf;
Amritha Nambiar56f36ac2017-03-15 10:39:25 -0700793 struct tc_mqprio_qopt *mqprio;
John Fastabend16e5cc62016-02-16 21:16:43 -0800794 };
Hadar Hen Zion7091d8c2016-12-01 14:06:37 +0200795 bool egress_dev;
John Fastabend16e5cc62016-02-16 21:16:43 -0800796};
797
Brenden Blancoa7862b42016-07-19 12:16:48 -0700798/* These structures hold the attributes of xdp state that are being passed
799 * to the netdevice through the xdp op.
800 */
801enum xdp_netdev_command {
802 /* Set or clear a bpf program used in the earliest stages of packet
803 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
804 * is responsible for calling bpf_prog_put on any old progs that are
805 * stored. In case of error, the callee need not release the new prog
806 * reference, but on success it takes ownership and must bpf_prog_put
807 * when it is no longer used.
808 */
809 XDP_SETUP_PROG,
810 /* Check if a bpf program is set on the device. The callee should
811 * return true if a program is currently attached and running.
812 */
813 XDP_QUERY_PROG,
814};
815
Jakub Kicinskiddf9f972017-04-30 21:46:46 -0700816struct netlink_ext_ack;
817
Brenden Blancoa7862b42016-07-19 12:16:48 -0700818struct netdev_xdp {
819 enum xdp_netdev_command command;
820 union {
821 /* XDP_SETUP_PROG */
Jakub Kicinskiddf9f972017-04-30 21:46:46 -0700822 struct {
823 struct bpf_prog *prog;
824 struct netlink_ext_ack *extack;
825 };
Brenden Blancoa7862b42016-07-19 12:16:48 -0700826 /* XDP_QUERY_PROG */
827 bool prog_attached;
828 };
829};
John Fastabend16e5cc62016-02-16 21:16:43 -0800830
Steffen Klassertd77e38e2017-04-14 10:06:10 +0200831#ifdef CONFIG_XFRM_OFFLOAD
832struct xfrmdev_ops {
833 int (*xdo_dev_state_add) (struct xfrm_state *x);
834 void (*xdo_dev_state_delete) (struct xfrm_state *x);
835 void (*xdo_dev_state_free) (struct xfrm_state *x);
836 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
837 struct xfrm_state *x);
838};
839#endif
840
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800841/*
842 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800843 * The following hooks can be defined; unless noted otherwise, they are
844 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800845 *
846 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500847 * This function is called once when a network device is registered.
848 * The network device can use this for any late stage initialization
849 * or semantic validation. It can fail with an error code which will
850 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800851 *
852 * void (*ndo_uninit)(struct net_device *dev);
853 * This function is called when device is unregistered or when registration
854 * fails. It is not called if init fails.
855 *
856 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500857 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800858 * state.
859 *
860 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500861 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800862 * state.
863 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000864 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
865 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800866 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030867 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
868 * the queue before that can happen; it's for obsolete devices and weird
869 * corner cases, but the stack really does a non-trivial amount
870 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500871 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800872 *
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -0800873 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
874 * struct net_device *dev
875 * netdev_features_t features);
876 * Called by core transmit path to determine if device is capable of
877 * performing offload operations on a given packet. This is to give
878 * the device an opportunity to implement any restrictions that cannot
879 * be otherwise expressed by feature flags. The check is called with
880 * the set of features that the stack has calculated and it returns
881 * those the driver believes to be appropriate.
Eric Dumazetcdba7562016-01-06 06:53:50 -0800882 *
Jason Wangf663dd92014-01-10 16:18:26 +0800883 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100884 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500885 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800886 * transmit queues.
887 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800888 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
889 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500890 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800891 *
892 * void (*ndo_set_rx_mode)(struct net_device *dev);
893 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000894 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500895 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800896 *
897 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
898 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700899 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500900 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800901 *
902 * int (*ndo_validate_addr)(struct net_device *dev);
903 * Test if Media Access Control address is valid for the device.
904 *
905 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500906 * Called when a user requests an ioctl which can't be handled by
907 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800908 * not supported error code.
909 *
910 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
911 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500912 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800913 * interface (PCI) for low level management.
914 *
915 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
916 * Called when a user wants to change the Maximum Transfer Unit
917 * of a device. If not defined, any request to change MTU will
918 * will return an error.
919 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800920 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500921 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800922 * for dev->watchdog ticks.
923 *
stephen hemmingerbc1f4472017-01-06 19:12:52 -0800924 * void (*ndo_get_stats64)(struct net_device *dev,
925 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700926 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800927 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000928 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000929 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
930 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700931 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000932 * (which should normally be dev->stats) and return a pointer to
933 * it. The structure may be changed asynchronously only if each
934 * field is written atomically.
935 * 3. Update dev->stats asynchronously and atomically, and define
936 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800937 *
Or Gerlitz3df5b3c2016-11-22 23:09:54 +0200938 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200939 * Return true if this device supports offload stats of this attr_id.
940 *
941 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
942 * void *attr_data)
943 * Get statistics for offload operations by attr_id. Write it into the
944 * attr_data pointer.
945 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530946 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500947 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000948 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800949 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530950 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500951 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000952 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800953 *
954 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000955 *
956 * SR-IOV management functions.
957 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300958 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
959 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400960 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
961 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000962 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000963 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000964 * int (*ndo_get_vf_config)(struct net_device *dev,
965 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300966 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700967 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
968 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300969 *
970 * Enable or disable the VF ability to query its RSS Redirection Table and
971 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500972 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300973 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700974 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
Florian Fainelli6a4bc2b2017-01-26 14:44:17 -0800975 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
976 * __be16 protocol, struct tc_to_netdev *tc);
977 * Called to setup any 'tc' scheduler, classifier or action on @dev.
978 * This is always called from the stack with the rtnl lock held and netif
979 * tx queues stopped. This allows the netdevice to perform queue
980 * management safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000981 *
Yi Zoue9bce842011-03-09 08:48:03 +0000982 * Fiber Channel over Ethernet (FCoE) offload functions.
983 * int (*ndo_fcoe_enable)(struct net_device *dev);
984 * Called when the FCoE protocol stack wants to start using LLD for FCoE
985 * so the underlying device can perform whatever needed configuration or
986 * initialization to support acceleration of FCoE traffic.
987 *
988 * int (*ndo_fcoe_disable)(struct net_device *dev);
989 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
990 * so the underlying device can perform whatever needed clean-ups to
991 * stop supporting acceleration of FCoE traffic.
992 *
993 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
994 * struct scatterlist *sgl, unsigned int sgc);
995 * Called when the FCoE Initiator wants to initialize an I/O that
996 * is a possible candidate for Direct Data Placement (DDP). The LLD can
997 * perform necessary setup and returns 1 to indicate the device is set up
998 * successfully to perform DDP on this I/O, otherwise this returns 0.
999 *
1000 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1001 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1002 * indicated by the FC exchange id 'xid', so the underlying device can
1003 * clean up and reuse resources for later DDP requests.
1004 *
1005 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1006 * struct scatterlist *sgl, unsigned int sgc);
1007 * Called when the FCoE Target wants to initialize an I/O that
1008 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1009 * perform necessary setup and returns 1 to indicate the device is set up
1010 * successfully to perform DDP on this I/O, otherwise this returns 0.
1011 *
Neerav Parikh68bad942012-01-04 20:23:39 +00001012 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1013 * struct netdev_fcoe_hbainfo *hbainfo);
1014 * Called when the FCoE Protocol stack wants information on the underlying
1015 * device. This information is utilized by the FCoE protocol stack to
1016 * register attributes with Fiber Channel management service as per the
1017 * FC-GS Fabric Device Management Information(FDMI) specification.
1018 *
Yi Zoue9bce842011-03-09 08:48:03 +00001019 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1020 * Called when the underlying device wants to override default World Wide
1021 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1022 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1023 * protocol stack to use.
1024 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001025 * RFS acceleration.
1026 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1027 * u16 rxq_index, u32 flow_id);
1028 * Set hardware filter for RFS. rxq_index is the target queue index;
1029 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1030 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001031 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001032 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001033 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1034 * Called to make another netdev an underling.
1035 *
1036 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1037 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001038 *
1039 * Feature/offload setting functions.
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -08001040 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1041 * netdev_features_t features);
1042 * Adjusts the requested feature flags according to device-specific
1043 * constraints, and returns the resulting flags. Must not modify
1044 * the device state.
1045 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001046 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001047 * Called to update device configuration to new features. Passed
1048 * feature set might be less than what was returned by ndo_fix_features()).
1049 * Must return >0 or -errno if it changed dev->features itself.
1050 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001051 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1052 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001053 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001054 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001055 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1056 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001057 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001058 * Deletes the FDB entry from dev coresponding to addr.
1059 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001060 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001061 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001062 * Used to add FDB entries to dump requests. Implementers should add
1063 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001064 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001065 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1066 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001067 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001068 * struct net_device *dev, u32 filter_mask,
1069 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001070 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1071 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001072 *
1073 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1074 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1075 * which do not represent real hardware may define this to allow their
1076 * userspace components to manage their virtual carrier state. Devices
1077 * that determine carrier state from physical hardware properties (eg
1078 * network cables) or protocol-dependent mechanisms (eg
1079 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001080 *
1081 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001082 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001083 * Called to get ID of physical port of this device. If driver does
1084 * not implement this, it is assumed that the hw is not able to have
1085 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001086 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001087 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1088 * struct udp_tunnel_info *ti);
1089 * Called by UDP tunnel to notify a driver about the UDP port and socket
1090 * address family that a UDP tunnel is listnening to. It is called only
1091 * when a new port starts listening. The operation is protected by the
1092 * RTNL.
1093 *
1094 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1095 * struct udp_tunnel_info *ti);
1096 * Called by UDP tunnel to notify the driver about a UDP port and socket
1097 * address family that the UDP tunnel is not listening to anymore. The
1098 * operation is protected by the RTNL.
1099 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001100 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1101 * struct net_device *dev)
1102 * Called by upper layer devices to accelerate switching or other
1103 * station functionality into hardware. 'pdev is the lowerdev
1104 * to use for the offload and 'dev' is the net device that will
1105 * back the offload. Returns a pointer to the private structure
1106 * the upper layer will maintain.
1107 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1108 * Called by upper layer device to delete the station created
1109 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1110 * the station and priv is the structure returned by the add
1111 * operation.
1112 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1113 * struct net_device *dev,
1114 * void *priv);
1115 * Callback to use for xmit over the accelerated station. This
1116 * is used in place of ndo_start_xmit on accelerated net
1117 * devices.
John Fastabend822b3b22015-03-18 14:57:33 +02001118 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1119 * int queue_index, u32 maxrate);
1120 * Called when a user wants to set a max-rate limitation of specific
1121 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001122 * int (*ndo_get_iflink)(const struct net_device *dev);
1123 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001124 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001125 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001126 * This function is used to pass protocol port error state information
1127 * to the switch driver. The switch driver can react to the proto_down
1128 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001129 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1130 * This function is used to get egress tunnel information for given skb.
1131 * This is useful for retrieving outer tunnel header parameters while
1132 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001133 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1134 * This function is used to specify the headroom that the skb must
1135 * consider when allocation skb during packet reception. Setting
1136 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001137 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001138 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001139 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1140 * This function is used to set or query state related to XDP on the
1141 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001142 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001143 */
1144struct net_device_ops {
1145 int (*ndo_init)(struct net_device *dev);
1146 void (*ndo_uninit)(struct net_device *dev);
1147 int (*ndo_open)(struct net_device *dev);
1148 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001149 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1150 struct net_device *dev);
1151 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1152 struct net_device *dev,
1153 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001154 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001155 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001156 void *accel_priv,
1157 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001158 void (*ndo_change_rx_flags)(struct net_device *dev,
1159 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001160 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001161 int (*ndo_set_mac_address)(struct net_device *dev,
1162 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001163 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001164 int (*ndo_do_ioctl)(struct net_device *dev,
1165 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001166 int (*ndo_set_config)(struct net_device *dev,
1167 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001168 int (*ndo_change_mtu)(struct net_device *dev,
1169 int new_mtu);
1170 int (*ndo_neigh_setup)(struct net_device *dev,
1171 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001172 void (*ndo_tx_timeout) (struct net_device *dev);
1173
stephen hemmingerbc1f4472017-01-06 19:12:52 -08001174 void (*ndo_get_stats64)(struct net_device *dev,
1175 struct rtnl_link_stats64 *storage);
Or Gerlitz3df5b3c2016-11-22 23:09:54 +02001176 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001177 int (*ndo_get_offload_stats)(int attr_id,
1178 const struct net_device *dev,
1179 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001180 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1181
Jiri Pirko8e586132011-12-08 19:52:37 -05001182 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001183 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001184 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001185 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001186#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001187 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001188 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001189 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001190 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001191#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001192 int (*ndo_set_vf_mac)(struct net_device *dev,
1193 int queue, u8 *mac);
1194 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001195 int queue, u16 vlan,
1196 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001197 int (*ndo_set_vf_rate)(struct net_device *dev,
1198 int vf, int min_tx_rate,
1199 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001200 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1201 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001202 int (*ndo_set_vf_trust)(struct net_device *dev,
1203 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001204 int (*ndo_get_vf_config)(struct net_device *dev,
1205 int vf,
1206 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001207 int (*ndo_set_vf_link_state)(struct net_device *dev,
1208 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001209 int (*ndo_get_vf_stats)(struct net_device *dev,
1210 int vf,
1211 struct ifla_vf_stats
1212 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001213 int (*ndo_set_vf_port)(struct net_device *dev,
1214 int vf,
1215 struct nlattr *port[]);
1216 int (*ndo_get_vf_port)(struct net_device *dev,
1217 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001218 int (*ndo_set_vf_guid)(struct net_device *dev,
1219 int vf, u64 guid,
1220 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001221 int (*ndo_set_vf_rss_query_en)(
1222 struct net_device *dev,
1223 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001224 int (*ndo_setup_tc)(struct net_device *dev,
1225 u32 handle,
1226 __be16 protocol,
1227 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001228#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001229 int (*ndo_fcoe_enable)(struct net_device *dev);
1230 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001231 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1232 u16 xid,
1233 struct scatterlist *sgl,
1234 unsigned int sgc);
1235 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1236 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001237 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1238 u16 xid,
1239 struct scatterlist *sgl,
1240 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001241 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1242 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001243#endif
1244
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001245#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001246#define NETDEV_FCOE_WWNN 0
1247#define NETDEV_FCOE_WWPN 1
1248 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1249 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001250#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001251
Ben Hutchingsc4454772011-01-19 11:03:53 +00001252#ifdef CONFIG_RFS_ACCEL
1253 int (*ndo_rx_flow_steer)(struct net_device *dev,
1254 const struct sk_buff *skb,
1255 u16 rxq_index,
1256 u32 flow_id);
1257#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001258 int (*ndo_add_slave)(struct net_device *dev,
1259 struct net_device *slave_dev);
1260 int (*ndo_del_slave)(struct net_device *dev,
1261 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001262 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1263 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001264 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001265 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001266 int (*ndo_neigh_construct)(struct net_device *dev,
1267 struct neighbour *n);
1268 void (*ndo_neigh_destroy)(struct net_device *dev,
1269 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001270
1271 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001272 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001273 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001274 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001275 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001276 u16 flags);
1277 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001278 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001279 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001280 const unsigned char *addr,
1281 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001282 int (*ndo_fdb_dump)(struct sk_buff *skb,
1283 struct netlink_callback *cb,
1284 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001285 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001286 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001287
1288 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001289 struct nlmsghdr *nlh,
1290 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001291 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1292 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001293 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001294 u32 filter_mask,
1295 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001296 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001297 struct nlmsghdr *nlh,
1298 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001299 int (*ndo_change_carrier)(struct net_device *dev,
1300 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001301 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001302 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001303 int (*ndo_get_phys_port_name)(struct net_device *dev,
1304 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001305 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1306 struct udp_tunnel_info *ti);
1307 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1308 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001309 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1310 struct net_device *dev);
1311 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1312 void *priv);
1313
1314 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1315 struct net_device *dev,
1316 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001317 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001318 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1319 int queue_index,
1320 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001321 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001322 int (*ndo_change_proto_down)(struct net_device *dev,
1323 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001324 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1325 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001326 void (*ndo_set_rx_headroom)(struct net_device *dev,
1327 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001328 int (*ndo_xdp)(struct net_device *dev,
1329 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001330};
1331
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001332/**
1333 * enum net_device_priv_flags - &struct net_device priv_flags
1334 *
1335 * These are the &struct net_device, they are only set internally
1336 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001337 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001338 * during any kernel release.
1339 *
1340 * You should have a pretty good reason to be extending these flags.
1341 *
1342 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1343 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001344 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001345 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001346 * @IFF_WAN_HDLC: WAN HDLC device
1347 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1348 * release skb->dst
1349 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1350 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1351 * @IFF_MACVLAN_PORT: device used as macvlan port
1352 * @IFF_BRIDGE_PORT: device used as bridge port
1353 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1354 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1355 * @IFF_UNICAST_FLT: Supports unicast filtering
1356 * @IFF_TEAM_PORT: device used as team port
1357 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1358 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1359 * change when it's running
1360 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001361 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1362 * underlying stacked devices
1363 * @IFF_IPVLAN_MASTER: IPvlan master device
1364 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001365 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001366 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001367 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001368 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001369 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001370 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001371 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1372 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001373 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001374 */
1375enum netdev_priv_flags {
1376 IFF_802_1Q_VLAN = 1<<0,
1377 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001378 IFF_BONDING = 1<<2,
1379 IFF_ISATAP = 1<<3,
1380 IFF_WAN_HDLC = 1<<4,
1381 IFF_XMIT_DST_RELEASE = 1<<5,
1382 IFF_DONT_BRIDGE = 1<<6,
1383 IFF_DISABLE_NETPOLL = 1<<7,
1384 IFF_MACVLAN_PORT = 1<<8,
1385 IFF_BRIDGE_PORT = 1<<9,
1386 IFF_OVS_DATAPATH = 1<<10,
1387 IFF_TX_SKB_SHARING = 1<<11,
1388 IFF_UNICAST_FLT = 1<<12,
1389 IFF_TEAM_PORT = 1<<13,
1390 IFF_SUPP_NOFCS = 1<<14,
1391 IFF_LIVE_ADDR_CHANGE = 1<<15,
1392 IFF_MACVLAN = 1<<16,
1393 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1394 IFF_IPVLAN_MASTER = 1<<18,
1395 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001396 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001397 IFF_NO_QUEUE = 1<<21,
1398 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001399 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001400 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001401 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001402 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001403 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001404};
1405
1406#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1407#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001408#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001409#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001410#define IFF_WAN_HDLC IFF_WAN_HDLC
1411#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1412#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1413#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1414#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1415#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1416#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1417#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1418#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1419#define IFF_TEAM_PORT IFF_TEAM_PORT
1420#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1421#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1422#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001423#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001424#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1425#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001426#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001427#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001428#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001429#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001430#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001431#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001432#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001433
Karoly Kemeny536721b2014-07-30 20:27:36 +02001434/**
1435 * struct net_device - The DEVICE structure.
1436 * Actually, this whole structure is a big mistake. It mixes I/O
1437 * data with strictly "high-level" data, and it has to know about
1438 * almost every data structure used in the INET module.
1439 *
1440 * @name: This is the first field of the "visible" part of this structure
1441 * (i.e. as seen by users in the "Space.c" file). It is the name
1442 * of the interface.
1443 *
1444 * @name_hlist: Device name hash chain, please keep it close to name[]
1445 * @ifalias: SNMP alias
1446 * @mem_end: Shared memory end
1447 * @mem_start: Shared memory start
1448 * @base_addr: Device I/O address
1449 * @irq: Device IRQ number
1450 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001451 * @carrier_changes: Stats to monitor carrier on<->off transitions
1452 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001453 * @state: Generic network queuing layer state, see netdev_state_t
1454 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001455 * @napi_list: List entry used for polling NAPI devices
1456 * @unreg_list: List entry when we are unregistering the
1457 * device; see the function unregister_netdev
1458 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001459 * @ptype_all: Device-specific packet handlers for all protocols
1460 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001461 *
1462 * @adj_list: Directly linked devices, like slaves for bonding
Karoly Kemeny536721b2014-07-30 20:27:36 +02001463 * @features: Currently active device features
1464 * @hw_features: User-changeable features
1465 *
1466 * @wanted_features: User-requested features
1467 * @vlan_features: Mask of features inheritable by VLAN devices
1468 *
1469 * @hw_enc_features: Mask of features inherited by encapsulating devices
1470 * This field indicates what encapsulation
1471 * offloads the hardware is capable of doing,
1472 * and drivers will need to set them appropriately.
1473 *
1474 * @mpls_features: Mask of features inheritable by MPLS
1475 *
1476 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001477 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001478 *
1479 * @stats: Statistics struct, which was left as a legacy, use
1480 * rtnl_link_stats64 instead
1481 *
1482 * @rx_dropped: Dropped packets by core network,
1483 * do not use this in drivers
1484 * @tx_dropped: Dropped packets by core network,
1485 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001486 * @rx_nohandler: nohandler dropped packets by core network on
1487 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001488 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001489 * @wireless_handlers: List of functions to handle Wireless Extensions,
1490 * instead of ioctl,
1491 * see <net/iw_handler.h> for details.
1492 * @wireless_data: Instance data managed by the core of wireless extensions
1493 *
1494 * @netdev_ops: Includes several pointers to callbacks,
1495 * if one wants to override the ndo_*() functions
1496 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001497 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1498 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001499 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001500 * of Layer 2 headers.
1501 *
1502 * @flags: Interface flags (a la BSD)
1503 * @priv_flags: Like 'flags' but invisible to userspace,
1504 * see if.h for the definitions
1505 * @gflags: Global flags ( kept as legacy )
1506 * @padded: How much padding added by alloc_netdev()
1507 * @operstate: RFC2863 operstate
1508 * @link_mode: Mapping policy to operstate
1509 * @if_port: Selectable AUI, TP, ...
1510 * @dma: DMA channel
1511 * @mtu: Interface MTU value
Jarod Wilson61e84622016-10-07 22:04:33 -04001512 * @min_mtu: Interface Minimum MTU value
1513 * @max_mtu: Interface Maximum MTU value
Karoly Kemeny536721b2014-07-30 20:27:36 +02001514 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001515 * @hard_header_len: Maximum hardware header length.
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05001516 * @min_header_len: Minimum hardware header length
Karoly Kemeny536721b2014-07-30 20:27:36 +02001517 *
1518 * @needed_headroom: Extra headroom the hardware may need, but not in all
1519 * cases can this be guaranteed
1520 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1521 * cases can this be guaranteed. Some cases also use
1522 * LL_MAX_HEADER instead to allocate the skb
1523 *
1524 * interface address info:
1525 *
1526 * @perm_addr: Permanent hw address
1527 * @addr_assign_type: Hw address assignment type
1528 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001529 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001530 * @dev_id: Used to differentiate devices that share
1531 * the same link layer address
1532 * @dev_port: Used to differentiate devices that share
1533 * the same function
1534 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001535 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001536 * has been enabled due to the need to listen to
1537 * additional unicast addresses in a device that
1538 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001539 * @uc: unicast mac addresses
1540 * @mc: multicast mac addresses
1541 * @dev_addrs: list of device hw addresses
1542 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001543 * @promiscuity: Number of times the NIC is told to work in
1544 * promiscuous mode; if it becomes 0 the NIC will
1545 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001546 * @allmulti: Counter, enables or disables allmulticast mode
1547 *
1548 * @vlan_info: VLAN info
1549 * @dsa_ptr: dsa specific data
1550 * @tipc_ptr: TIPC specific data
1551 * @atalk_ptr: AppleTalk link
1552 * @ip_ptr: IPv4 specific data
1553 * @dn_ptr: DECnet specific data
1554 * @ip6_ptr: IPv6 specific data
1555 * @ax25_ptr: AX.25 specific data
1556 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1557 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001558 * @dev_addr: Hw address (before bcast,
1559 * because most packets are unicast)
1560 *
1561 * @_rx: Array of RX queues
1562 * @num_rx_queues: Number of RX queues
1563 * allocated at register_netdev() time
1564 * @real_num_rx_queues: Number of RX queues currently active in device
1565 *
1566 * @rx_handler: handler for received packets
1567 * @rx_handler_data: XXX: need comments on this one
1568 * @ingress_queue: XXX: need comments on this one
1569 * @broadcast: hw bcast address
1570 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001571 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1572 * indexed by RX queue number. Assigned by driver.
1573 * This must only be set if the ndo_rx_flow_steer
1574 * operation is defined
1575 * @index_hlist: Device index hash chain
1576 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001577 * @_tx: Array of TX queues
1578 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1579 * @real_num_tx_queues: Number of TX queues currently active in device
1580 * @qdisc: Root qdisc from userspace point of view
1581 * @tx_queue_len: Max frames per queue allowed
1582 * @tx_global_lock: XXX: need comments on this one
1583 *
1584 * @xps_maps: XXX: need comments on this one
1585 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001586 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001587 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001588 * @watchdog_timer: List of timers
1589 *
1590 * @pcpu_refcnt: Number of references to this device
1591 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001592 * @link_watch_list: XXX: need comments on this one
1593 *
1594 * @reg_state: Register/unregister state machine
1595 * @dismantle: Device is going to be freed
1596 * @rtnl_link_state: This enum represents the phases of creating
1597 * a new link
1598 *
1599 * @destructor: Called from unregister,
1600 * can be used to call free_netdev
1601 * @npinfo: XXX: need comments on this one
1602 * @nd_net: Network namespace this network device is inside
1603 *
1604 * @ml_priv: Mid-layer private
1605 * @lstats: Loopback statistics
1606 * @tstats: Tunnel statistics
1607 * @dstats: Dummy statistics
1608 * @vstats: Virtual ethernet statistics
1609 *
1610 * @garp_port: GARP
1611 * @mrp_port: MRP
1612 *
1613 * @dev: Class/net/name entry
1614 * @sysfs_groups: Space for optional device, statistics and wireless
1615 * sysfs groups
1616 *
1617 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1618 * @rtnl_link_ops: Rtnl_link_ops
1619 *
1620 * @gso_max_size: Maximum size of generic segmentation offload
1621 * @gso_max_segs: Maximum number of segments that can be passed to the
1622 * NIC for GSO
1623 *
1624 * @dcbnl_ops: Data Center Bridging netlink ops
1625 * @num_tc: Number of traffic classes in the net device
1626 * @tc_to_txq: XXX: need comments on this one
Randy Dunlap920c1cd2016-11-21 18:28:36 -08001627 * @prio_tc_map: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001628 *
1629 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1630 *
1631 * @priomap: XXX: need comments on this one
1632 * @phydev: Physical device may attach itself
1633 * for hardware timestamping
1634 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001635 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1636 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001637 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001638 * @proto_down: protocol port state information can be sent to the
1639 * switch driver and used to set the phys state of the
1640 * switch port.
1641 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 * FIXME: cleanup struct net_device such that network protocol info
1643 * moves out.
1644 */
1645
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001646struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001648 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001649 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 /*
1651 * I/O specific fields
1652 * FIXME: Merge these and struct ifmap into one
1653 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001654 unsigned long mem_end;
1655 unsigned long mem_start;
1656 unsigned long base_addr;
1657 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001659 atomic_t carrier_changes;
1660
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001662 * Some hardware also needs these fields (state,dev_list,
1663 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 * part of the usual set specified in Space.c.
1665 */
1666
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 unsigned long state;
1668
Pavel Emelianov7562f872007-05-03 15:13:45 -07001669 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001670 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001671 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001672 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001673 struct list_head ptype_all;
1674 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001675
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001676 struct {
1677 struct list_head upper;
1678 struct list_head lower;
1679 } adj_list;
1680
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001681 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001682 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001683 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001684 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001685 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001686 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001687 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001688
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001690 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691
Rusty Russellc45d2862007-03-28 14:29:08 -07001692 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001693
Eric Dumazet015f0682014-03-27 08:45:56 -07001694 atomic_long_t rx_dropped;
1695 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001696 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
Johannes Bergb86e0282007-04-26 20:48:23 -07001698#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001699 const struct iw_handler_def *wireless_handlers;
1700 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001701#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001702 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001703 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001704#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001705 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001706#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001707#ifdef CONFIG_NET_L3_MASTER_DEV
1708 const struct l3mdev_ops *l3mdev_ops;
1709#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001710#if IS_ENABLED(CONFIG_IPV6)
1711 const struct ndisc_ops *ndisc_ops;
1712#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
Steffen Klassertd77e38e2017-04-14 10:06:10 +02001714#ifdef CONFIG_XFRM
1715 const struct xfrmdev_ops *xfrmdev_ops;
1716#endif
1717
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001718 const struct header_ops *header_ops;
1719
Karoly Kemeny536721b2014-07-30 20:27:36 +02001720 unsigned int flags;
1721 unsigned int priv_flags;
1722
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001724 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725
Karoly Kemeny536721b2014-07-30 20:27:36 +02001726 unsigned char operstate;
1727 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001728
Karoly Kemeny536721b2014-07-30 20:27:36 +02001729 unsigned char if_port;
1730 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001731
Karoly Kemeny536721b2014-07-30 20:27:36 +02001732 unsigned int mtu;
Jarod Wilson61e84622016-10-07 22:04:33 -04001733 unsigned int min_mtu;
1734 unsigned int max_mtu;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001735 unsigned short type;
1736 unsigned short hard_header_len;
Alexey Dobriyand92be7a2017-04-10 11:25:26 +03001737 unsigned char min_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738
Johannes Bergf5184d22008-05-12 20:48:31 -07001739 unsigned short needed_headroom;
1740 unsigned short needed_tailroom;
1741
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001743 unsigned char perm_addr[MAX_ADDR_LEN];
1744 unsigned char addr_assign_type;
1745 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001746 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001747 unsigned short dev_id;
1748 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001749 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001750 unsigned char name_assign_type;
1751 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001752 struct netdev_hw_addr_list uc;
1753 struct netdev_hw_addr_list mc;
1754 struct netdev_hw_addr_list dev_addrs;
1755
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001756#ifdef CONFIG_SYSFS
1757 struct kset *queues_kset;
1758#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001759 unsigned int promiscuity;
1760 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001763 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001764
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001765#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001766 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001767#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001768#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001769 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001770#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001771#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001772 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001773#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001774 void *atalk_ptr;
1775 struct in_device __rcu *ip_ptr;
1776 struct dn_dev __rcu *dn_ptr;
1777 struct inet6_dev __rcu *ip6_ptr;
1778 void *ax25_ptr;
1779 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001780 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001781#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1782 struct mpls_dev __rcu *mpls_ptr;
1783#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001785/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001786 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001787 */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001788 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001789 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001790
Michael Daltona953be52014-01-16 22:23:28 -08001791#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001792 struct netdev_rx_queue *_rx;
1793
Tom Herbert0a9627f2010-03-16 08:03:29 +00001794 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001795 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001796#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001797
Eric Dumazet7acedaf2017-04-25 11:36:52 -07001798 struct bpf_prog __rcu *xdp_prog;
Eric Dumazet3b47d302014-11-06 21:09:44 -08001799 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001800 rx_handler_func_t __rcu *rx_handler;
1801 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001802
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001803#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001804 struct tcf_proto __rcu *ingress_cl_list;
1805#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001806 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001807#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001808 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001809#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001810
Karoly Kemeny536721b2014-07-30 20:27:36 +02001811 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001812#ifdef CONFIG_RFS_ACCEL
1813 struct cpu_rmap *rx_cpu_rmap;
1814#endif
1815 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001816
1817/*
1818 * Cache lines mostly used on transmit path
1819 */
David S. Millere8a04642008-07-17 00:34:19 -07001820 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1821 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001822 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001823 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001824#ifdef CONFIG_NET_SCHED
1825 DECLARE_HASHTABLE (qdisc_hash, 4);
1826#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001827 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001828 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001829 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001830
Tom Herbertbf264142010-11-26 08:36:09 +00001831#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001832 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001833#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001834#ifdef CONFIG_NET_CLS_ACT
1835 struct tcf_proto __rcu *egress_cl_list;
1836#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001837
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001838 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001839 struct timer_list watchdog_timer;
1840
Eric Dumazet29b44332010-10-11 10:22:12 +00001841 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843
Eric Dumazete014deb2009-11-17 05:59:21 +00001844 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001845
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001847 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 NETREG_UNREGISTERING, /* called unregister_netdevice */
1849 NETREG_UNREGISTERED, /* completed unregister todo */
1850 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001851 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001852 } reg_state:8;
1853
Karoly Kemeny536721b2014-07-30 20:27:36 +02001854 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001855
1856 enum {
1857 RTNL_LINK_INITIALIZED,
1858 RTNL_LINK_INITIALIZING,
1859 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001861 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001864 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865#endif
David S. Millereae792b2008-07-15 03:03:33 -07001866
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001867 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001868
David S. Miller49517042008-05-12 03:29:11 -07001869 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001870 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001871 void *ml_priv;
1872 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001873 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001874 struct pcpu_dstats __percpu *dstats;
1875 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001876 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001877
Tobias Klauserfb585b42017-02-10 16:43:50 +01001878#if IS_ENABLED(CONFIG_GARP)
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001879 struct garp_port __rcu *garp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001880#endif
1881#if IS_ENABLED(CONFIG_MRP)
David Wardfebf0182013-02-08 17:17:06 +00001882 struct mrp_port __rcu *mrp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001883#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001885 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001886 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001887 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001888
Patrick McHardy38f7b872007-06-13 12:03:51 -07001889 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001890
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001891 /* for setting kernel sock attribute on TCP connection setup */
1892#define GSO_MAX_SIZE 65536
1893 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001894#define GSO_MAX_SEGS 65535
1895 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001896
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001897#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001898 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001899#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001900 u8 num_tc;
1901 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1902 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001903
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001904#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001905 unsigned int fcoe_ddp_xid;
1906#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001907#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001908 struct netprio_map __rcu *priomap;
1909#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001910 struct phy_device *phydev;
1911 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001912 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001913 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001915#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916
David S. Millerb5cdae32017-04-18 15:36:58 -04001917static inline bool netif_elide_gro(const struct net_device *dev)
1918{
1919 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1920 return true;
1921 return false;
1922}
1923
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925
David S. Millere8a04642008-07-17 00:34:19 -07001926static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001927int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1928{
1929 return dev->prio_tc_map[prio & TC_BITMASK];
1930}
1931
1932static inline
1933int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1934{
1935 if (tc >= dev->num_tc)
1936 return -EINVAL;
1937
1938 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1939 return 0;
1940}
1941
Alexander Duyck8d059b02016-10-28 11:43:49 -04001942int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
Alexander Duyck9cf1f6a2016-10-28 11:43:20 -04001943void netdev_reset_tc(struct net_device *dev);
1944int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1945int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
John Fastabend4f57c082011-01-17 08:06:04 +00001946
1947static inline
1948int netdev_get_num_tc(struct net_device *dev)
1949{
1950 return dev->num_tc;
1951}
1952
1953static inline
David S. Millere8a04642008-07-17 00:34:19 -07001954struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1955 unsigned int index)
1956{
1957 return &dev->_tx[index];
1958}
1959
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001960static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1961 const struct sk_buff *skb)
1962{
1963 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1964}
1965
David S. Millere8a04642008-07-17 00:34:19 -07001966static inline void netdev_for_each_tx_queue(struct net_device *dev,
1967 void (*f)(struct net_device *,
1968 struct netdev_queue *,
1969 void *),
1970 void *arg)
1971{
1972 unsigned int i;
1973
1974 for (i = 0; i < dev->num_tx_queues; i++)
1975 f(dev, &dev->_tx[i], arg);
1976}
1977
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001978#define netdev_lockdep_set_classes(dev) \
1979{ \
1980 static struct lock_class_key qdisc_tx_busylock_key; \
1981 static struct lock_class_key qdisc_running_key; \
1982 static struct lock_class_key qdisc_xmit_lock_key; \
1983 static struct lock_class_key dev_addr_list_lock_key; \
1984 unsigned int i; \
1985 \
1986 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1987 (dev)->qdisc_running_key = &qdisc_running_key; \
1988 lockdep_set_class(&(dev)->addr_list_lock, \
1989 &dev_addr_list_lock_key); \
1990 for (i = 0; i < (dev)->num_tx_queues; i++) \
1991 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1992 &qdisc_xmit_lock_key); \
1993}
1994
Joe Perchesf629d202013-09-26 14:48:15 -07001995struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001996 struct sk_buff *skb,
1997 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001998
Paolo Abeni871b6422016-02-26 10:45:37 +01001999/* returns the headroom that the master device needs to take in account
2000 * when forwarding to this dev
2001 */
2002static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2003{
2004 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2005}
2006
2007static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2008{
2009 if (dev->netdev_ops->ndo_set_rx_headroom)
2010 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2011}
2012
2013/* set the device rx headroom to the dev's default */
2014static inline void netdev_reset_rx_headroom(struct net_device *dev)
2015{
2016 netdev_set_rx_headroom(dev, -1);
2017}
2018
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002019/*
2020 * Net namespace inlines
2021 */
2022static inline
2023struct net *dev_net(const struct net_device *dev)
2024{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002025 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002026}
2027
2028static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002029void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002030{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002031 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002032}
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 Dumazet82dc3c63c2013-03-05 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
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002156 /* Number of gro_receive callbacks this packet already went through */
2157 u8 recursion_counter:4;
2158
2159 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002160
Jerry Chubf5a7552014-01-07 10:23:19 -08002161 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2162 __wsum csum;
2163
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002164 /* used in skb_gro_receive() slow path */
2165 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002166};
2167
2168#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002169
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002170#define GRO_RECURSION_LIMIT 15
2171static inline int gro_recursion_inc_test(struct sk_buff *skb)
2172{
2173 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2174}
2175
2176typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2177static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2178 struct sk_buff **head,
2179 struct sk_buff *skb)
2180{
2181 if (unlikely(gro_recursion_inc_test(skb))) {
2182 NAPI_GRO_CB(skb)->flush |= 1;
2183 return NULL;
2184 }
2185
2186 return cb(head, skb);
2187}
2188
2189typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2190 struct sk_buff *);
2191static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2192 struct sock *sk,
2193 struct sk_buff **head,
2194 struct sk_buff *skb)
2195{
2196 if (unlikely(gro_recursion_inc_test(skb))) {
2197 NAPI_GRO_CB(skb)->flush |= 1;
2198 return NULL;
2199 }
2200
2201 return cb(sk, head, skb);
2202}
2203
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002205 __be16 type; /* This is really htons(ether_type). */
2206 struct net_device *dev; /* NULL is wildcarded here */
2207 int (*func) (struct sk_buff *,
2208 struct net_device *,
2209 struct packet_type *,
2210 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002211 bool (*id_match)(struct packet_type *ptype,
2212 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 void *af_packet_priv;
2214 struct list_head list;
2215};
2216
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002217struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2219 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002221 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002222 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002223};
2224
2225struct packet_offload {
2226 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002227 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002228 struct offload_callbacks callbacks;
2229 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230};
2231
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002232/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002233struct pcpu_sw_netstats {
2234 u64 rx_packets;
2235 u64 rx_bytes;
2236 u64 tx_packets;
2237 u64 tx_bytes;
2238 struct u64_stats_sync syncp;
2239};
2240
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002241#define __netdev_alloc_pcpu_stats(type, gfp) \
2242({ \
2243 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2244 if (pcpu_stats) { \
2245 int __cpu; \
2246 for_each_possible_cpu(__cpu) { \
2247 typeof(type) *stat; \
2248 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2249 u64_stats_init(&stat->syncp); \
2250 } \
2251 } \
2252 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002253})
2254
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002255#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002256 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002257
Jiri Pirko764f5e52015-12-03 12:12:12 +01002258enum netdev_lag_tx_type {
2259 NETDEV_LAG_TX_TYPE_UNKNOWN,
2260 NETDEV_LAG_TX_TYPE_RANDOM,
2261 NETDEV_LAG_TX_TYPE_BROADCAST,
2262 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2263 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2264 NETDEV_LAG_TX_TYPE_HASH,
2265};
2266
2267struct netdev_lag_upper_info {
2268 enum netdev_lag_tx_type tx_type;
2269};
2270
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002271struct netdev_lag_lower_state_info {
2272 u8 link_up : 1,
2273 tx_enabled : 1;
2274};
2275
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276#include <linux/notifier.h>
2277
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002278/* netdevice notifier chain. Please remember to update the rtnetlink
2279 * notification exclusion list in rtnetlink_event() when adding new
2280 * types.
2281 */
2282#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2283#define NETDEV_DOWN 0x0002
2284#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2285 detected a hardware crash and restarted
2286 - we can use this eg to kick tcp sessions
2287 once done */
2288#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2289#define NETDEV_REGISTER 0x0005
2290#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002291#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002292#define NETDEV_CHANGEADDR 0x0008
2293#define NETDEV_GOING_DOWN 0x0009
2294#define NETDEV_CHANGENAME 0x000A
2295#define NETDEV_FEAT_CHANGE 0x000B
2296#define NETDEV_BONDING_FAILOVER 0x000C
2297#define NETDEV_PRE_UP 0x000D
2298#define NETDEV_PRE_TYPE_CHANGE 0x000E
2299#define NETDEV_POST_TYPE_CHANGE 0x000F
2300#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002301#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002302#define NETDEV_RELEASE 0x0012
2303#define NETDEV_NOTIFY_PEERS 0x0013
2304#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002305#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002306#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002307#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002308#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002309#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002310#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002311#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002312#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002313#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002314
Joe Perchesf629d202013-09-26 14:48:15 -07002315int register_netdevice_notifier(struct notifier_block *nb);
2316int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002317
2318struct netdev_notifier_info {
2319 struct net_device *dev;
2320};
2321
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002322struct netdev_notifier_change_info {
2323 struct netdev_notifier_info info; /* must be first */
2324 unsigned int flags_changed;
2325};
2326
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002327struct netdev_notifier_changeupper_info {
2328 struct netdev_notifier_info info; /* must be first */
2329 struct net_device *upper_dev; /* new upper dev */
2330 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002331 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002332 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002333};
2334
Jiri Pirko04d48262015-12-03 12:12:15 +01002335struct netdev_notifier_changelowerstate_info {
2336 struct netdev_notifier_info info; /* must be first */
2337 void *lower_state_info; /* is lower dev state */
2338};
2339
Cong Wang75538c22013-05-29 11:30:50 +08002340static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2341 struct net_device *dev)
2342{
2343 info->dev = dev;
2344}
2345
Jiri Pirko351638e2013-05-28 01:30:21 +00002346static inline struct net_device *
2347netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2348{
2349 return info->dev;
2350}
2351
Joe Perchesf629d202013-09-26 14:48:15 -07002352int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002353
2354
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355extern rwlock_t dev_base_lock; /* Device list lock */
2356
Eric W. Biederman881d9662007-09-17 11:56:21 -07002357#define for_each_netdev(net, d) \
2358 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002359#define for_each_netdev_reverse(net, d) \
2360 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002361#define for_each_netdev_rcu(net, d) \
2362 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002363#define for_each_netdev_safe(net, d, n) \
2364 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2365#define for_each_netdev_continue(net, d) \
2366 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002367#define for_each_netdev_continue_rcu(net, d) \
2368 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002369#define for_each_netdev_in_bond_rcu(bond, slave) \
2370 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002371 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002372#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2373
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002374static inline struct net_device *next_net_device(struct net_device *dev)
2375{
2376 struct list_head *lh;
2377 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002378
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002379 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002380 lh = dev->dev_list.next;
2381 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2382}
2383
Eric Dumazetce81b762009-11-11 17:34:30 +00002384static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2385{
2386 struct list_head *lh;
2387 struct net *net;
2388
2389 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002390 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002391 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2392}
2393
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002394static inline struct net_device *first_net_device(struct net *net)
2395{
2396 return list_empty(&net->dev_base_head) ? NULL :
2397 net_device_entry(net->dev_base_head.next);
2398}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002399
Eric Dumazetccf43432011-01-26 18:08:02 +00002400static inline struct net_device *first_net_device_rcu(struct net *net)
2401{
2402 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2403
2404 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2405}
2406
Joe Perchesf629d202013-09-26 14:48:15 -07002407int netdev_boot_setup_check(struct net_device *dev);
2408unsigned long netdev_boot_base(const char *prefix, int unit);
2409struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2410 const char *hwaddr);
2411struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2412struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2413void dev_add_pack(struct packet_type *pt);
2414void dev_remove_pack(struct packet_type *pt);
2415void __dev_remove_pack(struct packet_type *pt);
2416void dev_add_offload(struct packet_offload *po);
2417void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002419int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002420int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002421struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2422 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002423struct net_device *dev_get_by_name(struct net *net, const char *name);
2424struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2425struct net_device *__dev_get_by_name(struct net *net, const char *name);
2426int dev_alloc_name(struct net_device *dev, const char *name);
2427int dev_open(struct net_device *dev);
2428int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002429int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002430void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002431int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002432int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002433int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002434int register_netdevice(struct net_device *dev);
2435void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2436void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002437static inline void unregister_netdevice(struct net_device *dev)
2438{
2439 unregister_netdevice_queue(dev, NULL);
2440}
2441
Joe Perchesf629d202013-09-26 14:48:15 -07002442int netdev_refcnt_read(const struct net_device *dev);
2443void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002444void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002445void synchronize_net(void);
2446int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002447
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002448DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002449#define XMIT_RECURSION_LIMIT 10
2450
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002451static inline int dev_recursion_level(void)
2452{
2453 return this_cpu_read(xmit_recursion);
2454}
2455
Joe Perchesf629d202013-09-26 14:48:15 -07002456struct net_device *dev_get_by_index(struct net *net, int ifindex);
2457struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2458struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2459int netdev_get_name(struct net *net, char *name, int ifindex);
2460int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002461int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002462
2463static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2464{
2465 return NAPI_GRO_CB(skb)->data_offset;
2466}
2467
2468static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2469{
2470 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2471}
2472
2473static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2474{
2475 NAPI_GRO_CB(skb)->data_offset += len;
2476}
2477
Herbert Xua5b1cf22009-05-26 18:50:28 +00002478static inline void *skb_gro_header_fast(struct sk_buff *skb,
2479 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002480{
Herbert Xu78a478d2009-05-26 18:50:21 +00002481 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002482}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483
Herbert Xua5b1cf22009-05-26 18:50:28 +00002484static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2485{
2486 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2487}
2488
Herbert Xu57ea52a2017-01-10 12:24:15 -08002489static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2490{
2491 NAPI_GRO_CB(skb)->frag0 = NULL;
2492 NAPI_GRO_CB(skb)->frag0_len = 0;
2493}
2494
Herbert Xua5b1cf22009-05-26 18:50:28 +00002495static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2496 unsigned int offset)
2497{
Herbert Xu17dd7592011-07-27 06:16:28 -07002498 if (!pskb_may_pull(skb, hlen))
2499 return NULL;
2500
Herbert Xu57ea52a2017-01-10 12:24:15 -08002501 skb_gro_frag0_invalidate(skb);
Herbert Xu17dd7592011-07-27 06:16:28 -07002502 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002503}
2504
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002505static inline void *skb_gro_network_header(struct sk_buff *skb)
2506{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002507 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2508 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002509}
2510
Jerry Chubf5a7552014-01-07 10:23:19 -08002511static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2512 const void *start, unsigned int len)
2513{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002514 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002515 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2516 csum_partial(start, len, 0));
2517}
2518
Tom Herbert573e8fc2014-08-22 13:33:47 -07002519/* GRO checksum functions. These are logical equivalents of the normal
2520 * checksum functions (in skbuff.h) except that they operate on the GRO
2521 * offsets and fields in sk_buff.
2522 */
2523
2524__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2525
Tom Herbert15e23962015-02-10 16:30:31 -08002526static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2527{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002528 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002529}
2530
Tom Herbert573e8fc2014-08-22 13:33:47 -07002531static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2532 bool zero_okay,
2533 __sum16 check)
2534{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002535 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2536 skb_checksum_start_offset(skb) <
2537 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002538 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002539 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002540 (!zero_okay || check));
2541}
2542
2543static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2544 __wsum psum)
2545{
2546 if (NAPI_GRO_CB(skb)->csum_valid &&
2547 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2548 return 0;
2549
2550 NAPI_GRO_CB(skb)->csum = psum;
2551
2552 return __skb_gro_checksum_complete(skb);
2553}
2554
Tom Herbert573e8fc2014-08-22 13:33:47 -07002555static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2556{
Tom Herbert662880f2014-08-27 21:26:56 -07002557 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2558 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2559 NAPI_GRO_CB(skb)->csum_cnt--;
2560 } else {
2561 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2562 * verified a new top level checksum or an encapsulated one
2563 * during GRO. This saves work if we fallback to normal path.
2564 */
2565 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002566 }
2567}
2568
2569#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2570 compute_pseudo) \
2571({ \
2572 __sum16 __ret = 0; \
2573 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2574 __ret = __skb_gro_checksum_validate_complete(skb, \
2575 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002576 if (__ret) \
2577 __skb_mark_checksum_bad(skb); \
2578 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002579 skb_gro_incr_csum_unnecessary(skb); \
2580 __ret; \
2581})
2582
2583#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2584 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2585
2586#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2587 compute_pseudo) \
2588 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2589
2590#define skb_gro_checksum_simple_validate(skb) \
2591 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2592
Tom Herbertd96535a2014-08-31 15:12:42 -07002593static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2594{
2595 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2596 !NAPI_GRO_CB(skb)->csum_valid);
2597}
2598
2599static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2600 __sum16 check, __wsum pseudo)
2601{
2602 NAPI_GRO_CB(skb)->csum = ~pseudo;
2603 NAPI_GRO_CB(skb)->csum_valid = 1;
2604}
2605
2606#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2607do { \
2608 if (__skb_gro_checksum_convert_check(skb)) \
2609 __skb_gro_checksum_convert(skb, check, \
2610 compute_pseudo(skb, proto)); \
2611} while (0)
2612
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002613struct gro_remcsum {
2614 int offset;
2615 __wsum delta;
2616};
2617
2618static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2619{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002620 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002621 grc->delta = 0;
2622}
2623
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002624static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2625 unsigned int off, size_t hdrlen,
2626 int start, int offset,
2627 struct gro_remcsum *grc,
2628 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002629{
2630 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002631 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002632
2633 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2634
Tom Herbert15e23962015-02-10 16:30:31 -08002635 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002636 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2637 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002638 }
2639
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002640 ptr = skb_gro_header_fast(skb, off);
2641 if (skb_gro_header_hard(skb, off + plen)) {
2642 ptr = skb_gro_header_slow(skb, off + plen, off);
2643 if (!ptr)
2644 return NULL;
2645 }
2646
2647 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2648 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002649
2650 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002651 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002652
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002653 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002654 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002655
2656 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002657}
2658
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002659static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2660 struct gro_remcsum *grc)
2661{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002662 void *ptr;
2663 size_t plen = grc->offset + sizeof(u16);
2664
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002665 if (!grc->delta)
2666 return;
2667
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002668 ptr = skb_gro_header_fast(skb, grc->offset);
2669 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2670 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2671 if (!ptr)
2672 return;
2673 }
2674
2675 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002676}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002677
Steffen Klassert25393d32017-02-15 09:39:44 +01002678#ifdef CONFIG_XFRM_OFFLOAD
2679static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2680{
2681 if (PTR_ERR(pp) != -EINPROGRESS)
2682 NAPI_GRO_CB(skb)->flush |= flush;
2683}
2684#else
Steffen Klassert5f114162017-02-15 09:39:39 +01002685static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2686{
2687 NAPI_GRO_CB(skb)->flush |= flush;
2688}
Steffen Klassert25393d32017-02-15 09:39:44 +01002689#endif
Steffen Klassert5f114162017-02-15 09:39:39 +01002690
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002691static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2692 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002693 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002694 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002695{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002696 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002697 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002698
2699 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002700}
2701
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002702static inline int dev_parse_header(const struct sk_buff *skb,
2703 unsigned char *haddr)
2704{
2705 const struct net_device *dev = skb->dev;
2706
Patrick McHardy1b833362007-10-18 05:09:28 -07002707 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002708 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002709 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002710}
2711
Willem de Bruijn2793a232016-03-09 21:58:32 -05002712/* ll_header must have at least hard_header_len allocated */
2713static inline bool dev_validate_header(const struct net_device *dev,
2714 char *ll_header, int len)
2715{
2716 if (likely(len >= dev->hard_header_len))
2717 return true;
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05002718 if (len < dev->min_header_len)
2719 return false;
Willem de Bruijn2793a232016-03-09 21:58:32 -05002720
2721 if (capable(CAP_SYS_RAWIO)) {
2722 memset(ll_header + len, 0, dev->hard_header_len - len);
2723 return true;
2724 }
2725
2726 if (dev->header_ops && dev->header_ops->validate)
2727 return dev->header_ops->validate(ll_header, len);
2728
2729 return false;
2730}
2731
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002733int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002734static inline int unregister_gifconf(unsigned int family)
2735{
2736 return register_gifconf(family, NULL);
2737}
2738
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002739#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002740#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002741struct sd_flow_limit {
2742 u64 count;
2743 unsigned int num_buckets;
2744 unsigned int history_head;
2745 u16 history[FLOW_LIMIT_HISTORY];
2746 u8 buckets[];
2747};
2748
2749extern int netdev_flow_limit_table_len;
2750#endif /* CONFIG_NET_FLOW_LIMIT */
2751
Linus Torvalds1da177e2005-04-16 15:20:36 -07002752/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002753 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002754 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002755struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002756 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002757 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002758
Changli Gaodee42872010-05-02 05:42:16 +00002759 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002760 unsigned int processed;
2761 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002762 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002763#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002764 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002765#endif
2766#ifdef CONFIG_NET_FLOW_LIMIT
2767 struct sd_flow_limit __rcu *flow_limit;
2768#endif
2769 struct Qdisc *output_queue;
2770 struct Qdisc **output_queue_tailp;
2771 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002772
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002773#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002774 /* input_queue_head should be written by cpu owning this struct,
2775 * and only read by other cpus. Worth using a cache line.
2776 */
2777 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2778
2779 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002780 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002781 struct softnet_data *rps_ipi_next;
2782 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002783 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002784#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002785 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002786 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002787 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002788
Linus Torvalds1da177e2005-04-16 15:20:36 -07002789};
2790
Tom Herbert76cc8b12010-05-20 18:37:59 +00002791static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002792{
2793#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002794 sd->input_queue_head++;
2795#endif
2796}
2797
2798static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2799 unsigned int *qtail)
2800{
2801#ifdef CONFIG_RPS
2802 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002803#endif
2804}
2805
Tom Herbert0a9627f2010-03-16 08:03:29 +00002806DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002807
Joe Perchesf629d202013-09-26 14:48:15 -07002808void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002809void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002810
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002811static inline void netif_tx_schedule_all(struct net_device *dev)
2812{
2813 unsigned int i;
2814
2815 for (i = 0; i < dev->num_tx_queues; i++)
2816 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2817}
2818
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002819static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002820{
Tom Herbert734664982011-11-28 16:32:44 +00002821 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002822}
2823
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002824/**
2825 * netif_start_queue - allow transmit
2826 * @dev: network device
2827 *
2828 * Allow upper layers to call the device hard_start_xmit routine.
2829 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002830static inline void netif_start_queue(struct net_device *dev)
2831{
David S. Millere8a04642008-07-17 00:34:19 -07002832 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002833}
2834
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002835static inline void netif_tx_start_all_queues(struct net_device *dev)
2836{
2837 unsigned int i;
2838
2839 for (i = 0; i < dev->num_tx_queues; i++) {
2840 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2841 netif_tx_start_queue(txq);
2842 }
2843}
2844
John Fastabend46e5da40a2014-09-12 20:04:52 -07002845void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002846
Dave Jonesd29f7492008-07-22 14:09:06 -07002847/**
2848 * netif_wake_queue - restart transmit
2849 * @dev: network device
2850 *
2851 * Allow upper layers to call the device hard_start_xmit routine.
2852 * Used for flow control when transmit resources are available.
2853 */
David S. Miller79d16382008-07-08 23:14:46 -07002854static inline void netif_wake_queue(struct net_device *dev)
2855{
David S. Millere8a04642008-07-17 00:34:19 -07002856 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002857}
2858
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002859static inline void netif_tx_wake_all_queues(struct net_device *dev)
2860{
2861 unsigned int i;
2862
2863 for (i = 0; i < dev->num_tx_queues; i++) {
2864 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2865 netif_tx_wake_queue(txq);
2866 }
2867}
2868
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002869static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002870{
Tom Herbert734664982011-11-28 16:32:44 +00002871 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002872}
2873
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002874/**
2875 * netif_stop_queue - stop transmitted packets
2876 * @dev: network device
2877 *
2878 * Stop upper layers calling the device hard_start_xmit routine.
2879 * Used for flow control when transmit resources are unavailable.
2880 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002881static inline void netif_stop_queue(struct net_device *dev)
2882{
David S. Millere8a04642008-07-17 00:34:19 -07002883 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002884}
2885
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002886void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002887
David S. Miller4d295152012-03-07 21:02:35 -05002888static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002889{
Tom Herbert734664982011-11-28 16:32:44 +00002890 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002891}
2892
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002893/**
2894 * netif_queue_stopped - test if transmit queue is flowblocked
2895 * @dev: network device
2896 *
2897 * Test if transmit queue on device is currently unable to send.
2898 */
David S. Miller4d295152012-03-07 21:02:35 -05002899static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002900{
David S. Millere8a04642008-07-17 00:34:19 -07002901 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002902}
2903
David S. Miller4d295152012-03-07 21:02:35 -05002904static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002905{
Tom Herbert734664982011-11-28 16:32:44 +00002906 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2907}
2908
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002909static inline bool
2910netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002911{
2912 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2913}
2914
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002915static inline bool
2916netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2917{
2918 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2919}
2920
Eric Dumazet53511452014-10-08 08:19:27 -07002921/**
2922 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2923 * @dev_queue: pointer to transmit queue
2924 *
2925 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002926 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002927 */
2928static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2929{
2930#ifdef CONFIG_BQL
2931 prefetchw(&dev_queue->dql.num_queued);
2932#endif
2933}
2934
2935/**
2936 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2937 * @dev_queue: pointer to transmit queue
2938 *
2939 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002940 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002941 */
2942static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2943{
2944#ifdef CONFIG_BQL
2945 prefetchw(&dev_queue->dql.limit);
2946#endif
2947}
2948
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002949static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2950 unsigned int bytes)
2951{
Tom Herbert114cf582011-11-28 16:33:09 +00002952#ifdef CONFIG_BQL
2953 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002954
2955 if (likely(dql_avail(&dev_queue->dql) >= 0))
2956 return;
2957
2958 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2959
2960 /*
2961 * The XOFF flag must be set before checking the dql_avail below,
2962 * because in netdev_tx_completed_queue we update the dql_completed
2963 * before checking the XOFF flag.
2964 */
2965 smp_mb();
2966
2967 /* check again in case another CPU has just made room avail */
2968 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2969 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002970#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002971}
2972
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002973/**
2974 * netdev_sent_queue - report the number of bytes queued to hardware
2975 * @dev: network device
2976 * @bytes: number of bytes queued to the hardware device queue
2977 *
2978 * Report the number of bytes queued for sending/completion to the network
2979 * device hardware queue. @bytes should be a good approximation and should
2980 * exactly match netdev_completed_queue() @bytes
2981 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002982static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2983{
2984 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2985}
2986
2987static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002988 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002989{
Tom Herbert114cf582011-11-28 16:33:09 +00002990#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002991 if (unlikely(!bytes))
2992 return;
2993
2994 dql_completed(&dev_queue->dql, bytes);
2995
2996 /*
2997 * Without the memory barrier there is a small possiblity that
2998 * netdev_tx_sent_queue will miss the update and cause the queue to
2999 * be stopped forever
3000 */
3001 smp_mb();
3002
3003 if (dql_avail(&dev_queue->dql) < 0)
3004 return;
3005
3006 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3007 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00003008#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003009}
3010
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003011/**
3012 * netdev_completed_queue - report bytes and packets completed by device
3013 * @dev: network device
3014 * @pkts: actual number of packets sent over the medium
3015 * @bytes: actual number of bytes sent over the medium
3016 *
3017 * Report the number of bytes and packets transmitted by the network device
3018 * hardware queue over the physical medium, @bytes must exactly match the
3019 * @bytes amount passed to netdev_sent_queue()
3020 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003021static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003022 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003023{
3024 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3025}
3026
3027static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3028{
Tom Herbert114cf582011-11-28 16:33:09 +00003029#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003030 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003031 dql_reset(&q->dql);
3032#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003033}
3034
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003035/**
3036 * netdev_reset_queue - reset the packets and bytes count of a network device
3037 * @dev_queue: network device
3038 *
3039 * Reset the bytes and packet count of a network device and clear the
3040 * software flow control OFF bit for this network device
3041 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003042static inline void netdev_reset_queue(struct net_device *dev_queue)
3043{
3044 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003045}
3046
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003047/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003048 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3049 * @dev: network device
3050 * @queue_index: given tx queue index
3051 *
3052 * Returns 0 if given tx queue index >= number of device tx queues,
3053 * otherwise returns the originally passed tx queue index.
3054 */
3055static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3056{
3057 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3058 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3059 dev->name, queue_index,
3060 dev->real_num_tx_queues);
3061 return 0;
3062 }
3063
3064 return queue_index;
3065}
3066
3067/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003068 * netif_running - test if up
3069 * @dev: network device
3070 *
3071 * Test if the device has been brought up.
3072 */
David S. Miller4d295152012-03-07 21:02:35 -05003073static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003074{
3075 return test_bit(__LINK_STATE_START, &dev->state);
3076}
3077
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003078/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003079 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003080 * stop, and a check if it's stopped. All other device management is
3081 * done at the overall netdevice level.
3082 * Also test the device if we're multiqueue.
3083 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003084
3085/**
3086 * netif_start_subqueue - allow sending packets on subqueue
3087 * @dev: network device
3088 * @queue_index: sub queue index
3089 *
3090 * Start individual transmit queue of a device with multiple transmit queues.
3091 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003092static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3093{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003094 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003095
3096 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003097}
3098
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003099/**
3100 * netif_stop_subqueue - stop sending packets on subqueue
3101 * @dev: network device
3102 * @queue_index: sub queue index
3103 *
3104 * Stop individual transmit queue of a device with multiple transmit queues.
3105 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003106static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3107{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003108 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003109 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003110}
3111
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003112/**
3113 * netif_subqueue_stopped - test status of subqueue
3114 * @dev: network device
3115 * @queue_index: sub queue index
3116 *
3117 * Check individual transmit queue of a device with multiple transmit queues.
3118 */
David S. Miller4d295152012-03-07 21:02:35 -05003119static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3120 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003121{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003122 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003123
3124 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003125}
3126
David S. Miller4d295152012-03-07 21:02:35 -05003127static inline bool netif_subqueue_stopped(const struct net_device *dev,
3128 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003129{
3130 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3131}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003132
Florian Fainelli738b35c2017-01-11 21:13:02 -08003133/**
3134 * netif_wake_subqueue - allow sending packets on subqueue
3135 * @dev: network device
3136 * @queue_index: sub queue index
3137 *
3138 * Resume individual transmit queue of a device with multiple transmit queues.
3139 */
3140static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3141{
3142 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3143
3144 netif_tx_wake_queue(txq);
3145}
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003146
Alexander Duyck537c00d2013-01-10 08:57:02 +00003147#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003148int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003149 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003150#else
3151static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003152 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003153 u16 index)
3154{
3155 return 0;
3156}
3157#endif
3158
Jiri Pirko5605c762015-05-12 14:56:12 +02003159u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3160 unsigned int num_tx_queues);
3161
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003162/*
3163 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3164 * as a distribution range limit for the returned value.
3165 */
3166static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003167 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003168{
3169 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3170}
3171
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003172/**
3173 * netif_is_multiqueue - test if device has multiple transmit queues
3174 * @dev: network device
3175 *
3176 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003177 */
David S. Miller4d295152012-03-07 21:02:35 -05003178static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003179{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003180 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003181}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003182
Joe Perchesf629d202013-09-26 14:48:15 -07003183int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003184
Michael Daltona953be52014-01-16 22:23:28 -08003185#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003186int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003187#else
3188static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3189 unsigned int rxq)
3190{
3191 return 0;
3192}
3193#endif
3194
Michael Daltona953be52014-01-16 22:23:28 -08003195#ifdef CONFIG_SYSFS
3196static inline unsigned int get_netdev_rx_queue_index(
3197 struct netdev_rx_queue *queue)
3198{
3199 struct net_device *dev = queue->dev;
3200 int index = queue - dev->_rx;
3201
3202 BUG_ON(index >= dev->num_rx_queues);
3203 return index;
3204}
3205#endif
3206
Yuval Mintz16917b82012-07-01 03:18:50 +00003207#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003208int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003209
Eric Dumazete6247022013-12-05 04:45:08 -08003210enum skb_free_reason {
3211 SKB_REASON_CONSUMED,
3212 SKB_REASON_DROPPED,
3213};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003214
Eric Dumazete6247022013-12-05 04:45:08 -08003215void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3216void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3217
3218/*
3219 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3220 * interrupt context or with hardware interrupts being disabled.
3221 * (in_irq() || irqs_disabled())
3222 *
3223 * We provide four helpers that can be used in following contexts :
3224 *
3225 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3226 * replacing kfree_skb(skb)
3227 *
3228 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3229 * Typically used in place of consume_skb(skb) in TX completion path
3230 *
3231 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3232 * replacing kfree_skb(skb)
3233 *
3234 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3235 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003236 */
Eric Dumazete6247022013-12-05 04:45:08 -08003237static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3238{
3239 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3240}
3241
3242static inline void dev_consume_skb_irq(struct sk_buff *skb)
3243{
3244 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3245}
3246
3247static inline void dev_kfree_skb_any(struct sk_buff *skb)
3248{
3249 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3250}
3251
3252static inline void dev_consume_skb_any(struct sk_buff *skb)
3253{
3254 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3255}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003256
Joe Perchesf629d202013-09-26 14:48:15 -07003257int netif_rx(struct sk_buff *skb);
3258int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003259int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003260gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3261void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3262struct sk_buff *napi_get_frags(struct napi_struct *napi);
3263gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003264struct packet_offload *gro_find_receive_by_type(__be16 type);
3265struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003266
3267static inline void napi_free_frags(struct napi_struct *napi)
3268{
3269 kfree_skb(napi->skb);
3270 napi->skb = NULL;
3271}
3272
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003273bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003274int netdev_rx_handler_register(struct net_device *dev,
3275 rx_handler_func_t *rx_handler,
3276 void *rx_handler_data);
3277void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003278
Joe Perchesf629d202013-09-26 14:48:15 -07003279bool dev_valid_name(const char *name);
3280int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3281int dev_ethtool(struct net *net, struct ifreq *);
3282unsigned int dev_get_flags(const struct net_device *);
3283int __dev_change_flags(struct net_device *, unsigned int flags);
3284int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003285void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3286 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003287int dev_change_name(struct net_device *, const char *);
3288int dev_set_alias(struct net_device *, const char *, size_t);
3289int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3290int dev_set_mtu(struct net_device *, int);
3291void dev_set_group(struct net_device *, int);
3292int dev_set_mac_address(struct net_device *, struct sockaddr *);
3293int dev_change_carrier(struct net_device *, bool new_carrier);
3294int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003295 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003296int dev_get_phys_port_name(struct net_device *dev,
3297 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003298int dev_change_proto_down(struct net_device *dev, bool proto_down);
Jakub Kicinskiddf9f972017-04-30 21:46:46 -07003299int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3300 int fd, u32 flags);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003301struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003302struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3303 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003304int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003305int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003306bool is_skb_forwardable(const struct net_device *dev,
3307 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003308
Martin KaFai Lau4e3264d2016-11-09 15:36:33 -08003309static __always_inline int ____dev_forward_skb(struct net_device *dev,
3310 struct sk_buff *skb)
3311{
3312 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3313 unlikely(!is_skb_forwardable(dev, skb))) {
3314 atomic_long_inc(&dev->rx_dropped);
3315 kfree_skb(skb);
3316 return NET_RX_DROP;
3317 }
3318
3319 skb_scrub_packet(skb, true);
3320 skb->priority = 0;
3321 return 0;
3322}
3323
David Ahern74b20582016-05-10 11:19:50 -07003324void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3325
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003326extern int netdev_budget;
Matthew Whitehead7acf8a12017-04-19 12:37:10 -04003327extern unsigned int netdev_budget_usecs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003328
3329/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003330void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003331
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003332/**
3333 * dev_put - release reference to device
3334 * @dev: network device
3335 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003336 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003337 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003338static inline void dev_put(struct net_device *dev)
3339{
Christoph Lameter933393f2011-12-22 11:58:51 -06003340 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003341}
3342
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003343/**
3344 * dev_hold - get reference to device
3345 * @dev: network device
3346 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003347 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003348 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003349static inline void dev_hold(struct net_device *dev)
3350{
Christoph Lameter933393f2011-12-22 11:58:51 -06003351 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003352}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003353
3354/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3355 * and _off may be called from IRQ context, but it is caller
3356 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003357 *
3358 * The name carrier is inappropriate, these functions should really be
3359 * called netif_lowerlayer_*() because they represent the state of any
3360 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003361 */
3362
Joe Perchesf629d202013-09-26 14:48:15 -07003363void linkwatch_init_dev(struct net_device *dev);
3364void linkwatch_fire_event(struct net_device *dev);
3365void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003366
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003367/**
3368 * netif_carrier_ok - test if carrier present
3369 * @dev: network device
3370 *
3371 * Check if carrier is present on device
3372 */
David S. Miller4d295152012-03-07 21:02:35 -05003373static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003374{
3375 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3376}
3377
Joe Perchesf629d202013-09-26 14:48:15 -07003378unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003379
Joe Perchesf629d202013-09-26 14:48:15 -07003380void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003381
Joe Perchesf629d202013-09-26 14:48:15 -07003382void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003383
Joe Perchesf629d202013-09-26 14:48:15 -07003384void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003385
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003386/**
3387 * netif_dormant_on - mark device as dormant.
3388 * @dev: network device
3389 *
3390 * Mark device as dormant (as per RFC2863).
3391 *
3392 * The dormant state indicates that the relevant interface is not
3393 * actually in a condition to pass packets (i.e., it is not 'up') but is
3394 * in a "pending" state, waiting for some external event. For "on-
3395 * demand" interfaces, this new state identifies the situation where the
3396 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003397 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003398static inline void netif_dormant_on(struct net_device *dev)
3399{
3400 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3401 linkwatch_fire_event(dev);
3402}
3403
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003404/**
3405 * netif_dormant_off - set device as not dormant.
3406 * @dev: network device
3407 *
3408 * Device is not in dormant state.
3409 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003410static inline void netif_dormant_off(struct net_device *dev)
3411{
3412 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3413 linkwatch_fire_event(dev);
3414}
3415
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003416/**
Zhang Shengju8ecbc402017-04-26 11:05:12 +08003417 * netif_dormant - test if device is dormant
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003418 * @dev: network device
3419 *
Zhang Shengju8ecbc402017-04-26 11:05:12 +08003420 * Check if device is dormant.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003421 */
David S. Miller4d295152012-03-07 21:02:35 -05003422static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003423{
3424 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3425}
3426
3427
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003428/**
3429 * netif_oper_up - test if device is operational
3430 * @dev: network device
3431 *
3432 * Check if carrier is operational
3433 */
David S. Miller4d295152012-03-07 21:02:35 -05003434static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003435{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003436 return (dev->operstate == IF_OPER_UP ||
3437 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3438}
3439
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003440/**
3441 * netif_device_present - is device available or removed
3442 * @dev: network device
3443 *
3444 * Check if device has not been removed from system.
3445 */
David S. Miller4d295152012-03-07 21:02:35 -05003446static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003447{
3448 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3449}
3450
Joe Perchesf629d202013-09-26 14:48:15 -07003451void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003452
Joe Perchesf629d202013-09-26 14:48:15 -07003453void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003454
3455/*
3456 * Network interface message level settings
3457 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003458
3459enum {
3460 NETIF_MSG_DRV = 0x0001,
3461 NETIF_MSG_PROBE = 0x0002,
3462 NETIF_MSG_LINK = 0x0004,
3463 NETIF_MSG_TIMER = 0x0008,
3464 NETIF_MSG_IFDOWN = 0x0010,
3465 NETIF_MSG_IFUP = 0x0020,
3466 NETIF_MSG_RX_ERR = 0x0040,
3467 NETIF_MSG_TX_ERR = 0x0080,
3468 NETIF_MSG_TX_QUEUED = 0x0100,
3469 NETIF_MSG_INTR = 0x0200,
3470 NETIF_MSG_TX_DONE = 0x0400,
3471 NETIF_MSG_RX_STATUS = 0x0800,
3472 NETIF_MSG_PKTDATA = 0x1000,
3473 NETIF_MSG_HW = 0x2000,
3474 NETIF_MSG_WOL = 0x4000,
3475};
3476
3477#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3478#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3479#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3480#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3481#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3482#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3483#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3484#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3485#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3486#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3487#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3488#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3489#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3490#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3491#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3492
3493static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3494{
3495 /* use default */
3496 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3497 return default_msg_enable_bits;
3498 if (debug_value == 0) /* no output */
3499 return 0;
3500 /* set low N bits */
3501 return (1 << debug_value) - 1;
3502}
3503
David S. Millerc773e842008-07-08 23:13:53 -07003504static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003505{
David S. Millerc773e842008-07-08 23:13:53 -07003506 spin_lock(&txq->_xmit_lock);
3507 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003508}
3509
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003510static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3511{
3512 __acquire(&txq->_xmit_lock);
3513 return true;
3514}
3515
3516static inline void __netif_tx_release(struct netdev_queue *txq)
3517{
3518 __release(&txq->_xmit_lock);
3519}
3520
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003521static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3522{
3523 spin_lock_bh(&txq->_xmit_lock);
3524 txq->xmit_lock_owner = smp_processor_id();
3525}
3526
David S. Miller4d295152012-03-07 21:02:35 -05003527static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003528{
David S. Miller4d295152012-03-07 21:02:35 -05003529 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003530 if (likely(ok))
3531 txq->xmit_lock_owner = smp_processor_id();
3532 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003533}
3534
David S. Millerc773e842008-07-08 23:13:53 -07003535static inline void __netif_tx_unlock(struct netdev_queue *txq)
3536{
3537 txq->xmit_lock_owner = -1;
3538 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003539}
3540
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003541static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3542{
3543 txq->xmit_lock_owner = -1;
3544 spin_unlock_bh(&txq->_xmit_lock);
3545}
3546
Eric Dumazet08baf562009-05-25 22:58:01 -07003547static inline void txq_trans_update(struct netdev_queue *txq)
3548{
3549 if (txq->xmit_lock_owner != -1)
3550 txq->trans_start = jiffies;
3551}
3552
Florian Westphalba162f82016-05-03 16:31:00 +02003553/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3554static inline void netif_trans_update(struct net_device *dev)
3555{
Florian Westphal9b366272016-05-03 16:33:14 +02003556 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3557
3558 if (txq->trans_start != jiffies)
3559 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003560}
3561
David S. Millerc3f26a22008-07-31 16:58:50 -07003562/**
3563 * netif_tx_lock - grab network device transmit lock
3564 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003565 *
3566 * Get network device transmit lock
3567 */
3568static inline void netif_tx_lock(struct net_device *dev)
3569{
3570 unsigned int i;
3571 int cpu;
3572
3573 spin_lock(&dev->tx_global_lock);
3574 cpu = smp_processor_id();
3575 for (i = 0; i < dev->num_tx_queues; i++) {
3576 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3577
3578 /* We are the only thread of execution doing a
3579 * freeze, but we have to grab the _xmit_lock in
3580 * order to synchronize with threads which are in
3581 * the ->hard_start_xmit() handler and already
3582 * checked the frozen bit.
3583 */
3584 __netif_tx_lock(txq, cpu);
3585 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3586 __netif_tx_unlock(txq);
3587 }
3588}
3589
3590static inline void netif_tx_lock_bh(struct net_device *dev)
3591{
3592 local_bh_disable();
3593 netif_tx_lock(dev);
3594}
3595
Herbert Xu932ff272006-06-09 12:20:56 -07003596static inline void netif_tx_unlock(struct net_device *dev)
3597{
David S. Millere8a04642008-07-17 00:34:19 -07003598 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003599
David S. Millere8a04642008-07-17 00:34:19 -07003600 for (i = 0; i < dev->num_tx_queues; i++) {
3601 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003602
David S. Millerc3f26a22008-07-31 16:58:50 -07003603 /* No need to grab the _xmit_lock here. If the
3604 * queue is not stopped for another reason, we
3605 * force a schedule.
3606 */
3607 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003608 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003609 }
3610 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003611}
3612
3613static inline void netif_tx_unlock_bh(struct net_device *dev)
3614{
David S. Millere8a04642008-07-17 00:34:19 -07003615 netif_tx_unlock(dev);
3616 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003617}
3618
David S. Millerc773e842008-07-08 23:13:53 -07003619#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003620 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003621 __netif_tx_lock(txq, cpu); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003622 } else { \
3623 __netif_tx_acquire(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003624 } \
3625}
3626
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003627#define HARD_TX_TRYLOCK(dev, txq) \
3628 (((dev->features & NETIF_F_LLTX) == 0) ? \
3629 __netif_tx_trylock(txq) : \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003630 __netif_tx_acquire(txq))
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003631
David S. Millerc773e842008-07-08 23:13:53 -07003632#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003633 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003634 __netif_tx_unlock(txq); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003635 } else { \
3636 __netif_tx_release(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003637 } \
3638}
3639
Linus Torvalds1da177e2005-04-16 15:20:36 -07003640static inline void netif_tx_disable(struct net_device *dev)
3641{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003642 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003643 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003644
David S. Millerc3f26a22008-07-31 16:58:50 -07003645 local_bh_disable();
3646 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003647 for (i = 0; i < dev->num_tx_queues; i++) {
3648 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003649
3650 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003651 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003652 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003653 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003654 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003655}
3656
David S. Millere308a5d2008-07-15 00:13:44 -07003657static inline void netif_addr_lock(struct net_device *dev)
3658{
3659 spin_lock(&dev->addr_list_lock);
3660}
3661
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003662static inline void netif_addr_lock_nested(struct net_device *dev)
3663{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003664 int subclass = SINGLE_DEPTH_NESTING;
3665
3666 if (dev->netdev_ops->ndo_get_lock_subclass)
3667 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3668
3669 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003670}
3671
David S. Millere308a5d2008-07-15 00:13:44 -07003672static inline void netif_addr_lock_bh(struct net_device *dev)
3673{
3674 spin_lock_bh(&dev->addr_list_lock);
3675}
3676
3677static inline void netif_addr_unlock(struct net_device *dev)
3678{
3679 spin_unlock(&dev->addr_list_lock);
3680}
3681
3682static inline void netif_addr_unlock_bh(struct net_device *dev)
3683{
3684 spin_unlock_bh(&dev->addr_list_lock);
3685}
3686
Jiri Pirkof001fde2009-05-05 02:48:28 +00003687/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003688 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003689 * rcu_read_lock held.
3690 */
3691#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003692 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003693
Linus Torvalds1da177e2005-04-16 15:20:36 -07003694/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3695
Joe Perchesf629d202013-09-26 14:48:15 -07003696void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003697
3698/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003699struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003700 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003701 void (*setup)(struct net_device *),
3702 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003703#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3704 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003705
Tom Gundersenc835a672014-07-14 16:37:24 +02003706#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3707 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3708 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003709
Joe Perchesf629d202013-09-26 14:48:15 -07003710int register_netdev(struct net_device *dev);
3711void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003712
Jiri Pirko22bedad32010-04-01 21:22:57 +00003713/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003714int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3715 struct netdev_hw_addr_list *from_list, int addr_len);
3716void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3717 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003718int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3719 struct net_device *dev,
3720 int (*sync)(struct net_device *, const unsigned char *),
3721 int (*unsync)(struct net_device *,
3722 const unsigned char *));
3723void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3724 struct net_device *dev,
3725 int (*unsync)(struct net_device *,
3726 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003727void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003728
Jiri Pirkof001fde2009-05-05 02:48:28 +00003729/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003730int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3731 unsigned char addr_type);
3732int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3733 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003734void dev_addr_flush(struct net_device *dev);
3735int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003736
3737/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003738int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3739int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3740int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3741int dev_uc_sync(struct net_device *to, struct net_device *from);
3742int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3743void dev_uc_unsync(struct net_device *to, struct net_device *from);
3744void dev_uc_flush(struct net_device *dev);
3745void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003746
Alexander Duyck670e5b82014-05-28 18:44:46 -07003747/**
3748 * __dev_uc_sync - Synchonize device's unicast list
3749 * @dev: device to sync
3750 * @sync: function to call if address should be added
3751 * @unsync: function to call if address should be removed
3752 *
3753 * Add newly added addresses to the interface, and release
3754 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003755 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003756static inline int __dev_uc_sync(struct net_device *dev,
3757 int (*sync)(struct net_device *,
3758 const unsigned char *),
3759 int (*unsync)(struct net_device *,
3760 const unsigned char *))
3761{
3762 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3763}
3764
3765/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003766 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003767 * @dev: device to sync
3768 * @unsync: function to call if address should be removed
3769 *
3770 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003771 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003772static inline void __dev_uc_unsync(struct net_device *dev,
3773 int (*unsync)(struct net_device *,
3774 const unsigned char *))
3775{
3776 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3777}
3778
Jiri Pirko22bedad32010-04-01 21:22:57 +00003779/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003780int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3781int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3782int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3783int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3784int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3785int dev_mc_sync(struct net_device *to, struct net_device *from);
3786int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3787void dev_mc_unsync(struct net_device *to, struct net_device *from);
3788void dev_mc_flush(struct net_device *dev);
3789void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003790
Alexander Duyck670e5b82014-05-28 18:44:46 -07003791/**
3792 * __dev_mc_sync - Synchonize device's multicast list
3793 * @dev: device to sync
3794 * @sync: function to call if address should be added
3795 * @unsync: function to call if address should be removed
3796 *
3797 * Add newly added addresses to the interface, and release
3798 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003799 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003800static inline int __dev_mc_sync(struct net_device *dev,
3801 int (*sync)(struct net_device *,
3802 const unsigned char *),
3803 int (*unsync)(struct net_device *,
3804 const unsigned char *))
3805{
3806 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3807}
3808
3809/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003810 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003811 * @dev: device to sync
3812 * @unsync: function to call if address should be removed
3813 *
3814 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003815 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003816static inline void __dev_mc_unsync(struct net_device *dev,
3817 int (*unsync)(struct net_device *,
3818 const unsigned char *))
3819{
3820 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3821}
3822
Linus Torvalds1da177e2005-04-16 15:20:36 -07003823/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003824void dev_set_rx_mode(struct net_device *dev);
3825void __dev_set_rx_mode(struct net_device *dev);
3826int dev_set_promiscuity(struct net_device *dev, int inc);
3827int dev_set_allmulti(struct net_device *dev, int inc);
3828void netdev_state_change(struct net_device *dev);
3829void netdev_notify_peers(struct net_device *dev);
3830void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003831/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003832void dev_load(struct net *net, const char *name);
3833struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3834 struct rtnl_link_stats64 *storage);
3835void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3836 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003837
3838extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003839extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003840extern int weight_p;
Matthias Tafelmeier3d48b532016-12-29 21:37:21 +01003841extern int dev_weight_rx_bias;
3842extern int dev_weight_tx_bias;
3843extern int dev_rx_weight;
3844extern int dev_tx_weight;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003845
Joe Perchesf629d202013-09-26 14:48:15 -07003846bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003847struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3848 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003849struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3850 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003851
3852/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003853#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3854 for (iter = &(dev)->adj_list.upper, \
3855 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3856 updev; \
3857 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3858
David Ahern1a3f0602016-10-17 19:15:44 -07003859int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3860 int (*fn)(struct net_device *upper_dev,
3861 void *data),
3862 void *data);
3863
3864bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3865 struct net_device *upper_dev);
3866
Joe Perchesf629d202013-09-26 14:48:15 -07003867void *netdev_lower_get_next_private(struct net_device *dev,
3868 struct list_head **iter);
3869void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3870 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003871
3872#define netdev_for_each_lower_private(dev, priv, iter) \
3873 for (iter = (dev)->adj_list.lower.next, \
3874 priv = netdev_lower_get_next_private(dev, &(iter)); \
3875 priv; \
3876 priv = netdev_lower_get_next_private(dev, &(iter)))
3877
3878#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3879 for (iter = &(dev)->adj_list.lower, \
3880 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3881 priv; \
3882 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3883
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003884void *netdev_lower_get_next(struct net_device *dev,
3885 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003886
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003887#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003888 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003889 ldev = netdev_lower_get_next(dev, &(iter)); \
3890 ldev; \
3891 ldev = netdev_lower_get_next(dev, &(iter)))
3892
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003893struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3894 struct list_head **iter);
3895struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3896 struct list_head **iter);
3897
David Ahern1a3f0602016-10-17 19:15:44 -07003898int netdev_walk_all_lower_dev(struct net_device *dev,
3899 int (*fn)(struct net_device *lower_dev,
3900 void *data),
3901 void *data);
3902int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3903 int (*fn)(struct net_device *lower_dev,
3904 void *data),
3905 void *data);
3906
Joe Perchesf629d202013-09-26 14:48:15 -07003907void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003908void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003909struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3910struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3911int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3912int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003913 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003914 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003915void netdev_upper_dev_unlink(struct net_device *dev,
3916 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003917void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003918void *netdev_lower_dev_get_private(struct net_device *dev,
3919 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003920void netdev_lower_state_changed(struct net_device *lower_dev,
3921 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003922
3923/* RSS keys are 40 or 52 bytes long */
3924#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003925extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003926void netdev_rss_key_fill(void *buffer, size_t len);
3927
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003928int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003929int skb_checksum_help(struct sk_buff *skb);
3930struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3931 netdev_features_t features, bool tx_path);
3932struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3933 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003934
Moni Shoua61bd3852015-02-03 16:48:29 +02003935struct netdev_bonding_info {
3936 ifslave slave;
3937 ifbond master;
3938};
3939
3940struct netdev_notifier_bonding_info {
3941 struct netdev_notifier_info info; /* must be first */
3942 struct netdev_bonding_info bonding_info;
3943};
3944
3945void netdev_bonding_info_change(struct net_device *dev,
3946 struct netdev_bonding_info *bonding_info);
3947
Cong Wang12b00042013-02-05 16:36:38 +00003948static inline
3949struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3950{
3951 return __skb_gso_segment(skb, features, true);
3952}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003953__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003954
3955static inline bool can_checksum_protocol(netdev_features_t features,
3956 __be16 protocol)
3957{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003958 if (protocol == htons(ETH_P_FCOE))
3959 return !!(features & NETIF_F_FCOE_CRC);
3960
3961 /* Assume this is an IP checksum (not SCTP CRC) */
3962
3963 if (features & NETIF_F_HW_CSUM) {
3964 /* Can checksum everything */
3965 return true;
3966 }
3967
3968 switch (protocol) {
3969 case htons(ETH_P_IP):
3970 return !!(features & NETIF_F_IP_CSUM);
3971 case htons(ETH_P_IPV6):
3972 return !!(features & NETIF_F_IPV6_CSUM);
3973 default:
3974 return false;
3975 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003976}
Cong Wang12b00042013-02-05 16:36:38 +00003977
Herbert Xufb286bb2005-11-10 13:01:24 -08003978#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003979void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003980#else
3981static inline void netdev_rx_csum_fault(struct net_device *dev)
3982{
3983}
3984#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003985/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003986void net_enable_timestamp(void);
3987void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003988
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003989#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003990int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003991#else
3992#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003993#endif
3994
David S. Miller47982482014-08-22 16:21:53 -07003995static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003996 struct sk_buff *skb, struct net_device *dev,
3997 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003998{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003999 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07004000 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07004001}
4002
David S. Miller10b3ad82014-08-29 21:07:24 -07004003static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004004 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07004005{
4006 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07004007 int rc;
David S. Miller47982482014-08-22 16:21:53 -07004008
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004009 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07004010 if (rc == NETDEV_TX_OK)
4011 txq_trans_update(txq);
4012
4013 return rc;
David S. Miller47982482014-08-22 16:21:53 -07004014}
4015
Linus Torvalds42a2d922013-11-13 17:40:34 +09004016int netdev_class_create_file_ns(struct class_attribute *class_attr,
4017 const void *ns);
4018void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4019 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04004020
4021static inline int netdev_class_create_file(struct class_attribute *class_attr)
4022{
4023 return netdev_class_create_file_ns(class_attr, NULL);
4024}
4025
4026static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4027{
4028 netdev_class_remove_file_ns(class_attr, NULL);
4029}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004030
Johannes Berg04600792010-08-05 17:45:15 +02004031extern struct kobj_ns_type_operations net_ns_type_operations;
4032
Joe Perchesf629d202013-09-26 14:48:15 -07004033const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004034
Joe Perchesf629d202013-09-26 14:48:15 -07004035void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004036
Michal Kubečekda081432014-05-20 08:29:25 +02004037static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4038 netdev_features_t f2)
4039{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004040 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4041 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004042 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004043 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004044 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004045 }
Michal Kubečekda081432014-05-20 08:29:25 +02004046
Tom Herbertc8cd0982015-12-14 11:19:44 -08004047 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004048}
4049
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004050static inline netdev_features_t netdev_get_wanted_features(
4051 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004052{
4053 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4054}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004055netdev_features_t netdev_increment_features(netdev_features_t all,
4056 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004057
4058/* Allow TSO being used on stacked device :
4059 * Performing the GSO segmentation before last device
4060 * is a performance improvement.
4061 */
4062static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4063 netdev_features_t mask)
4064{
4065 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4066}
4067
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004068int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004069void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004070void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004071
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004072void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4073 struct net_device *dev);
4074
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004075netdev_features_t passthru_features_check(struct sk_buff *skb,
4076 struct net_device *dev,
4077 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004078netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004079
David S. Miller4d295152012-03-07 21:02:35 -05004080static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004081{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004082 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004083
4084 /* check flags correspondence */
4085 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4086 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4087 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4088 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004089 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004090 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4091 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004092 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4093 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004094 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4095 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004096 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4097 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 -04004098 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004099 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004100 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Steffen Klassertc7ef8f02017-04-14 10:05:36 +02004101 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004102
Herbert Xubcd76112006-06-30 13:36:35 -07004103 return (features & feature) == feature;
4104}
4105
David S. Miller4d295152012-03-07 21:02:35 -05004106static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004107{
Herbert Xu278b2512009-06-03 21:20:51 -07004108 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004109 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004110}
4111
Johannes Berg8b86a612015-04-17 15:45:04 +02004112static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004113 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004114{
Jesse Grossfc741212011-01-09 06:23:32 +00004115 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004116 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4117 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004118}
4119
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004120static inline void netif_set_gso_max_size(struct net_device *dev,
4121 unsigned int size)
4122{
4123 dev->gso_max_size = size;
4124}
4125
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004126static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4127 int pulled_hlen, u16 mac_offset,
4128 int mac_len)
4129{
4130 skb->protocol = protocol;
4131 skb->encapsulation = 1;
4132 skb_push(skb, pulled_hlen);
4133 skb_reset_transport_header(skb);
4134 skb->mac_header = mac_offset;
4135 skb->network_header = skb->mac_header + mac_len;
4136 skb->mac_len = mac_len;
4137}
4138
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004139static inline bool netif_is_macsec(const struct net_device *dev)
4140{
4141 return dev->priv_flags & IFF_MACSEC;
4142}
4143
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004144static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004145{
4146 return dev->priv_flags & IFF_MACVLAN;
4147}
4148
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004149static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004150{
4151 return dev->priv_flags & IFF_MACVLAN_PORT;
4152}
4153
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004154static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004155{
4156 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4157}
4158
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004159static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004160{
4161 return dev->priv_flags & IFF_IPVLAN_MASTER;
4162}
4163
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004164static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004165{
4166 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4167}
4168
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004169static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004170{
4171 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4172}
4173
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004174static inline bool netif_supports_nofcs(struct net_device *dev)
4175{
4176 return dev->priv_flags & IFF_SUPP_NOFCS;
4177}
4178
David Ahern007979e2015-09-29 20:07:10 -07004179static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004180{
David Ahern007979e2015-09-29 20:07:10 -07004181 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004182}
4183
David Ahernfee6d4c2015-10-05 08:51:24 -07004184static inline bool netif_is_l3_slave(const struct net_device *dev)
4185{
4186 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4187}
4188
Jiri Pirko0894ae32015-08-27 09:31:19 +02004189static inline bool netif_is_bridge_master(const struct net_device *dev)
4190{
4191 return dev->priv_flags & IFF_EBRIDGE;
4192}
4193
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004194static inline bool netif_is_bridge_port(const struct net_device *dev)
4195{
4196 return dev->priv_flags & IFF_BRIDGE_PORT;
4197}
4198
Jiri Pirko35d4e172015-08-27 09:31:20 +02004199static inline bool netif_is_ovs_master(const struct net_device *dev)
4200{
4201 return dev->priv_flags & IFF_OPENVSWITCH;
4202}
4203
Jiri Pirko5be66142017-04-18 16:55:36 +02004204static inline bool netif_is_ovs_port(const struct net_device *dev)
4205{
4206 return dev->priv_flags & IFF_OVS_DATAPATH;
4207}
4208
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004209static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004210{
4211 return dev->priv_flags & IFF_TEAM;
4212}
4213
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004214static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004215{
4216 return dev->priv_flags & IFF_TEAM_PORT;
4217}
4218
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004219static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004220{
4221 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4222}
4223
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004224static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004225{
4226 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4227}
4228
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004229static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4230{
4231 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4232}
4233
Eric Dumazet02875872014-10-05 18:38:35 -07004234/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4235static inline void netif_keep_dst(struct net_device *dev)
4236{
4237 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4238}
4239
Paolo Abeni18d3df32016-07-14 18:00:10 +02004240/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4241static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4242{
4243 /* TODO: reserve and use an additional IFF bit, if we get more users */
4244 return dev->priv_flags & IFF_MACSEC;
4245}
4246
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004247extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004248
Joe Perches571ba422010-02-09 11:49:47 +00004249/* Logging, debugging and troubleshooting/diagnostic helpers. */
4250
4251/* netdev_printk helpers, similar to dev_printk */
4252
4253static inline const char *netdev_name(const struct net_device *dev)
4254{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004255 if (!dev->name[0] || strchr(dev->name, '%'))
4256 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004257 return dev->name;
4258}
4259
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004260static inline const char *netdev_reg_state(const struct net_device *dev)
4261{
4262 switch (dev->reg_state) {
4263 case NETREG_UNINITIALIZED: return " (uninitialized)";
4264 case NETREG_REGISTERED: return "";
4265 case NETREG_UNREGISTERING: return " (unregistering)";
4266 case NETREG_UNREGISTERED: return " (unregistered)";
4267 case NETREG_RELEASED: return " (released)";
4268 case NETREG_DUMMY: return " (dummy)";
4269 }
4270
4271 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4272 return " (unknown)";
4273}
4274
Joe Perchesf629d202013-09-26 14:48:15 -07004275__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004276void netdev_printk(const char *level, const struct net_device *dev,
4277 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004278__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004279void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004280__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004281void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004282__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004283void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004284__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004285void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004286__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004287void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004288__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004289void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004290__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004291void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004292
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004293#define MODULE_ALIAS_NETDEV(device) \
4294 MODULE_ALIAS("netdev-" device)
4295
Jim Cromieb558c962011-12-19 17:11:18 -05004296#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004297#define netdev_dbg(__dev, format, args...) \
4298do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004299 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004300} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004301#elif defined(DEBUG)
4302#define netdev_dbg(__dev, format, args...) \
4303 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004304#else
4305#define netdev_dbg(__dev, format, args...) \
4306({ \
4307 if (0) \
4308 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004309})
4310#endif
4311
4312#if defined(VERBOSE_DEBUG)
4313#define netdev_vdbg netdev_dbg
4314#else
4315
4316#define netdev_vdbg(dev, format, args...) \
4317({ \
4318 if (0) \
4319 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4320 0; \
4321})
4322#endif
4323
4324/*
4325 * netdev_WARN() acts like dev_printk(), but with the key difference
4326 * of using a WARN/WARN_ON to get the message out, including the
4327 * file/line information and a backtrace.
4328 */
4329#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004330 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4331 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004332
Joe Perchesb3d95c52010-02-09 11:49:49 +00004333/* netif printk helpers, similar to netdev_printk */
4334
4335#define netif_printk(priv, type, level, dev, fmt, args...) \
4336do { \
4337 if (netif_msg_##type(priv)) \
4338 netdev_printk(level, (dev), fmt, ##args); \
4339} while (0)
4340
Joe Perchesf45f4322010-06-27 01:02:36 +00004341#define netif_level(level, priv, type, dev, fmt, args...) \
4342do { \
4343 if (netif_msg_##type(priv)) \
4344 netdev_##level(dev, fmt, ##args); \
4345} while (0)
4346
Joe Perchesb3d95c52010-02-09 11:49:49 +00004347#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004348 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004349#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004350 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004351#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004352 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004353#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004354 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004355#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004356 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004357#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004358 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004359#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004360 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004361
Joe Perches0053ea92012-05-30 07:43:34 +00004362#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004363#define netif_dbg(priv, type, netdev, format, args...) \
4364do { \
4365 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004366 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004367} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004368#elif defined(DEBUG)
4369#define netif_dbg(priv, type, dev, format, args...) \
4370 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004371#else
4372#define netif_dbg(priv, type, dev, format, args...) \
4373({ \
4374 if (0) \
4375 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4376 0; \
4377})
4378#endif
4379
Edward Creef617f272017-01-27 15:02:26 +00004380/* if @cond then downgrade to debug, else print at @level */
4381#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4382 do { \
4383 if (cond) \
4384 netif_dbg(priv, type, netdev, fmt, ##args); \
4385 else \
4386 netif_ ## level(priv, type, netdev, fmt, ##args); \
4387 } while (0)
4388
Joe Perchesb3d95c52010-02-09 11:49:49 +00004389#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004390#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004391#else
4392#define netif_vdbg(priv, type, dev, format, args...) \
4393({ \
4394 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004395 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004396 0; \
4397})
4398#endif
Joe Perches571ba422010-02-09 11:49:47 +00004399
Cong Wang900ff8c2013-02-18 19:20:33 +00004400/*
4401 * The list of packet types we will receive (as opposed to discard)
4402 * and the routines to invoke.
4403 *
4404 * Why 16. Because with 16 the only overlap we get on a hash of the
4405 * low nibble of the protocol value is RARP/SNAP/X.25.
4406 *
4407 * NOTE: That is no longer true with the addition of VLAN tags. Not
4408 * sure which should go first, but I bet it won't make much
4409 * difference if we are running VLANs. The good news is that
4410 * this protocol won't be in the list unless compiled in, so
4411 * the average user (w/out VLANs) will not be adversely affected.
4412 * --BLG
4413 *
4414 * 0800 IP
4415 * 8100 802.1Q VLAN
4416 * 0001 802.3
4417 * 0002 AX.25
4418 * 0004 802.2
4419 * 8035 RARP
4420 * 0005 SNAP
4421 * 0805 X.25
4422 * 0806 ARP
4423 * 8137 IPX
4424 * 0009 Localtalk
4425 * 86DD IPv6
4426 */
4427#define PTYPE_HASH_SIZE (16)
4428#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4429
Jiri Pirko385a1542009-05-27 15:48:07 -07004430#endif /* _LINUX_NETDEVICE_H */