blob: 7c7118b3bd692831e1c58e22c017d6985169b301 [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 */
Martin KaFai Lau58038692017-06-15 17:29:09 -0700827 struct {
828 bool prog_attached;
829 u32 prog_id;
830 };
Brenden Blancoa7862b42016-07-19 12:16:48 -0700831 };
832};
John Fastabend16e5cc62016-02-16 21:16:43 -0800833
Steffen Klassertd77e38e2017-04-14 10:06:10 +0200834#ifdef CONFIG_XFRM_OFFLOAD
835struct xfrmdev_ops {
836 int (*xdo_dev_state_add) (struct xfrm_state *x);
837 void (*xdo_dev_state_delete) (struct xfrm_state *x);
838 void (*xdo_dev_state_free) (struct xfrm_state *x);
839 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
840 struct xfrm_state *x);
841};
842#endif
843
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800844/*
845 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800846 * The following hooks can be defined; unless noted otherwise, they are
847 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800848 *
849 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500850 * This function is called once when a network device is registered.
851 * The network device can use this for any late stage initialization
852 * or semantic validation. It can fail with an error code which will
853 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800854 *
855 * void (*ndo_uninit)(struct net_device *dev);
856 * This function is called when device is unregistered or when registration
857 * fails. It is not called if init fails.
858 *
859 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500860 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800861 * state.
862 *
863 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500864 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800865 * state.
866 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000867 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
868 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800869 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030870 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
871 * the queue before that can happen; it's for obsolete devices and weird
872 * corner cases, but the stack really does a non-trivial amount
873 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500874 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800875 *
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -0800876 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
877 * struct net_device *dev
878 * netdev_features_t features);
879 * Called by core transmit path to determine if device is capable of
880 * performing offload operations on a given packet. This is to give
881 * the device an opportunity to implement any restrictions that cannot
882 * be otherwise expressed by feature flags. The check is called with
883 * the set of features that the stack has calculated and it returns
884 * those the driver believes to be appropriate.
Eric Dumazetcdba7562016-01-06 06:53:50 -0800885 *
Jason Wangf663dd92014-01-10 16:18:26 +0800886 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100887 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500888 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800889 * transmit queues.
890 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800891 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
892 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500893 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800894 *
895 * void (*ndo_set_rx_mode)(struct net_device *dev);
896 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000897 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500898 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800899 *
900 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
901 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700902 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500903 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800904 *
905 * int (*ndo_validate_addr)(struct net_device *dev);
906 * Test if Media Access Control address is valid for the device.
907 *
908 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500909 * Called when a user requests an ioctl which can't be handled by
910 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800911 * not supported error code.
912 *
913 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
914 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500915 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800916 * interface (PCI) for low level management.
917 *
918 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
919 * Called when a user wants to change the Maximum Transfer Unit
Magnus Dammdb46a0e2017-06-14 16:15:24 +0900920 * of a device.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800921 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800922 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500923 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800924 * for dev->watchdog ticks.
925 *
stephen hemmingerbc1f4472017-01-06 19:12:52 -0800926 * void (*ndo_get_stats64)(struct net_device *dev,
927 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700928 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800929 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000930 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000931 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
932 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700933 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000934 * (which should normally be dev->stats) and return a pointer to
935 * it. The structure may be changed asynchronously only if each
936 * field is written atomically.
937 * 3. Update dev->stats asynchronously and atomically, and define
938 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800939 *
Or Gerlitz3df5b3c2016-11-22 23:09:54 +0200940 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200941 * Return true if this device supports offload stats of this attr_id.
942 *
943 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
944 * void *attr_data)
945 * Get statistics for offload operations by attr_id. Write it into the
946 * attr_data pointer.
947 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530948 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500949 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000950 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800951 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530952 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500953 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000954 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800955 *
956 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000957 *
958 * SR-IOV management functions.
959 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300960 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
961 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400962 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
963 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000964 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000965 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000966 * int (*ndo_get_vf_config)(struct net_device *dev,
967 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300968 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700969 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
970 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300971 *
972 * Enable or disable the VF ability to query its RSS Redirection Table and
973 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500974 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300975 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700976 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
Jiri Pirkoa5fcf8a2017-06-06 17:00:16 +0200977 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, u32 chain_index,
Florian Fainelli6a4bc2b2017-01-26 14:44:17 -0800978 * __be16 protocol, struct tc_to_netdev *tc);
979 * Called to setup any 'tc' scheduler, classifier or action on @dev.
980 * This is always called from the stack with the rtnl lock held and netif
981 * tx queues stopped. This allows the netdevice to perform queue
982 * management safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000983 *
Yi Zoue9bce842011-03-09 08:48:03 +0000984 * Fiber Channel over Ethernet (FCoE) offload functions.
985 * int (*ndo_fcoe_enable)(struct net_device *dev);
986 * Called when the FCoE protocol stack wants to start using LLD for FCoE
987 * so the underlying device can perform whatever needed configuration or
988 * initialization to support acceleration of FCoE traffic.
989 *
990 * int (*ndo_fcoe_disable)(struct net_device *dev);
991 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
992 * so the underlying device can perform whatever needed clean-ups to
993 * stop supporting acceleration of FCoE traffic.
994 *
995 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
996 * struct scatterlist *sgl, unsigned int sgc);
997 * Called when the FCoE Initiator wants to initialize an I/O that
998 * is a possible candidate for Direct Data Placement (DDP). The LLD can
999 * perform necessary setup and returns 1 to indicate the device is set up
1000 * successfully to perform DDP on this I/O, otherwise this returns 0.
1001 *
1002 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1003 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1004 * indicated by the FC exchange id 'xid', so the underlying device can
1005 * clean up and reuse resources for later DDP requests.
1006 *
1007 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1008 * struct scatterlist *sgl, unsigned int sgc);
1009 * Called when the FCoE Target wants to initialize an I/O that
1010 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1011 * perform necessary setup and returns 1 to indicate the device is set up
1012 * successfully to perform DDP on this I/O, otherwise this returns 0.
1013 *
Neerav Parikh68bad942012-01-04 20:23:39 +00001014 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1015 * struct netdev_fcoe_hbainfo *hbainfo);
1016 * Called when the FCoE Protocol stack wants information on the underlying
1017 * device. This information is utilized by the FCoE protocol stack to
1018 * register attributes with Fiber Channel management service as per the
1019 * FC-GS Fabric Device Management Information(FDMI) specification.
1020 *
Yi Zoue9bce842011-03-09 08:48:03 +00001021 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1022 * Called when the underlying device wants to override default World Wide
1023 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1024 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1025 * protocol stack to use.
1026 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001027 * RFS acceleration.
1028 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1029 * u16 rxq_index, u32 flow_id);
1030 * Set hardware filter for RFS. rxq_index is the target queue index;
1031 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1032 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001033 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001034 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001035 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1036 * Called to make another netdev an underling.
1037 *
1038 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1039 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001040 *
1041 * Feature/offload setting functions.
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -08001042 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1043 * netdev_features_t features);
1044 * Adjusts the requested feature flags according to device-specific
1045 * constraints, and returns the resulting flags. Must not modify
1046 * the device state.
1047 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001048 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001049 * Called to update device configuration to new features. Passed
1050 * feature set might be less than what was returned by ndo_fix_features()).
1051 * Must return >0 or -errno if it changed dev->features itself.
1052 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001053 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1054 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001055 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001056 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001057 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1058 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001059 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001060 * Deletes the FDB entry from dev coresponding to addr.
1061 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001062 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001063 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001064 * Used to add FDB entries to dump requests. Implementers should add
1065 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001066 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001067 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1068 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001069 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001070 * struct net_device *dev, u32 filter_mask,
1071 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001072 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1073 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001074 *
1075 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1076 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1077 * which do not represent real hardware may define this to allow their
1078 * userspace components to manage their virtual carrier state. Devices
1079 * that determine carrier state from physical hardware properties (eg
1080 * network cables) or protocol-dependent mechanisms (eg
1081 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001082 *
1083 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001084 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001085 * Called to get ID of physical port of this device. If driver does
1086 * not implement this, it is assumed that the hw is not able to have
1087 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001088 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001089 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1090 * struct udp_tunnel_info *ti);
1091 * Called by UDP tunnel to notify a driver about the UDP port and socket
1092 * address family that a UDP tunnel is listnening to. It is called only
1093 * when a new port starts listening. The operation is protected by the
1094 * RTNL.
1095 *
1096 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1097 * struct udp_tunnel_info *ti);
1098 * Called by UDP tunnel to notify the driver about a UDP port and socket
1099 * address family that the UDP tunnel is not listening to anymore. The
1100 * operation is protected by the RTNL.
1101 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001102 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1103 * struct net_device *dev)
1104 * Called by upper layer devices to accelerate switching or other
1105 * station functionality into hardware. 'pdev is the lowerdev
1106 * to use for the offload and 'dev' is the net device that will
1107 * back the offload. Returns a pointer to the private structure
1108 * the upper layer will maintain.
1109 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1110 * Called by upper layer device to delete the station created
1111 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1112 * the station and priv is the structure returned by the add
1113 * operation.
1114 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1115 * struct net_device *dev,
1116 * void *priv);
1117 * Callback to use for xmit over the accelerated station. This
1118 * is used in place of ndo_start_xmit on accelerated net
1119 * devices.
John Fastabend822b3b22015-03-18 14:57:33 +02001120 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1121 * int queue_index, u32 maxrate);
1122 * Called when a user wants to set a max-rate limitation of specific
1123 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001124 * int (*ndo_get_iflink)(const struct net_device *dev);
1125 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001126 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001127 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001128 * This function is used to pass protocol port error state information
1129 * to the switch driver. The switch driver can react to the proto_down
1130 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001131 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1132 * This function is used to get egress tunnel information for given skb.
1133 * This is useful for retrieving outer tunnel header parameters while
1134 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001135 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1136 * This function is used to specify the headroom that the skb must
1137 * consider when allocation skb during packet reception. Setting
1138 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001139 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001140 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001141 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1142 * This function is used to set or query state related to XDP on the
1143 * netdevice. See definition of enum xdp_netdev_command for details.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001144 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001145 */
1146struct net_device_ops {
1147 int (*ndo_init)(struct net_device *dev);
1148 void (*ndo_uninit)(struct net_device *dev);
1149 int (*ndo_open)(struct net_device *dev);
1150 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001151 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1152 struct net_device *dev);
1153 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1154 struct net_device *dev,
1155 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001156 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001157 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001158 void *accel_priv,
1159 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001160 void (*ndo_change_rx_flags)(struct net_device *dev,
1161 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001162 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001163 int (*ndo_set_mac_address)(struct net_device *dev,
1164 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001165 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001166 int (*ndo_do_ioctl)(struct net_device *dev,
1167 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001168 int (*ndo_set_config)(struct net_device *dev,
1169 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001170 int (*ndo_change_mtu)(struct net_device *dev,
1171 int new_mtu);
1172 int (*ndo_neigh_setup)(struct net_device *dev,
1173 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001174 void (*ndo_tx_timeout) (struct net_device *dev);
1175
stephen hemmingerbc1f4472017-01-06 19:12:52 -08001176 void (*ndo_get_stats64)(struct net_device *dev,
1177 struct rtnl_link_stats64 *storage);
Or Gerlitz3df5b3c2016-11-22 23:09:54 +02001178 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001179 int (*ndo_get_offload_stats)(int attr_id,
1180 const struct net_device *dev,
1181 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001182 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1183
Jiri Pirko8e586132011-12-08 19:52:37 -05001184 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001185 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001186 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001187 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001188#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001189 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001190 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001191 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001192 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001193#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001194 int (*ndo_set_vf_mac)(struct net_device *dev,
1195 int queue, u8 *mac);
1196 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001197 int queue, u16 vlan,
1198 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001199 int (*ndo_set_vf_rate)(struct net_device *dev,
1200 int vf, int min_tx_rate,
1201 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001202 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1203 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001204 int (*ndo_set_vf_trust)(struct net_device *dev,
1205 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001206 int (*ndo_get_vf_config)(struct net_device *dev,
1207 int vf,
1208 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001209 int (*ndo_set_vf_link_state)(struct net_device *dev,
1210 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001211 int (*ndo_get_vf_stats)(struct net_device *dev,
1212 int vf,
1213 struct ifla_vf_stats
1214 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001215 int (*ndo_set_vf_port)(struct net_device *dev,
1216 int vf,
1217 struct nlattr *port[]);
1218 int (*ndo_get_vf_port)(struct net_device *dev,
1219 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001220 int (*ndo_set_vf_guid)(struct net_device *dev,
1221 int vf, u64 guid,
1222 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001223 int (*ndo_set_vf_rss_query_en)(
1224 struct net_device *dev,
1225 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001226 int (*ndo_setup_tc)(struct net_device *dev,
Jiri Pirkoa5fcf8a2017-06-06 17:00:16 +02001227 u32 handle, u32 chain_index,
John Fastabend16e5cc62016-02-16 21:16:43 -08001228 __be16 protocol,
1229 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001230#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001231 int (*ndo_fcoe_enable)(struct net_device *dev);
1232 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001233 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1234 u16 xid,
1235 struct scatterlist *sgl,
1236 unsigned int sgc);
1237 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1238 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001239 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1240 u16 xid,
1241 struct scatterlist *sgl,
1242 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001243 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1244 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001245#endif
1246
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001247#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001248#define NETDEV_FCOE_WWNN 0
1249#define NETDEV_FCOE_WWPN 1
1250 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1251 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001252#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001253
Ben Hutchingsc4454772011-01-19 11:03:53 +00001254#ifdef CONFIG_RFS_ACCEL
1255 int (*ndo_rx_flow_steer)(struct net_device *dev,
1256 const struct sk_buff *skb,
1257 u16 rxq_index,
1258 u32 flow_id);
1259#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001260 int (*ndo_add_slave)(struct net_device *dev,
1261 struct net_device *slave_dev);
1262 int (*ndo_del_slave)(struct net_device *dev,
1263 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001264 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1265 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001266 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001267 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001268 int (*ndo_neigh_construct)(struct net_device *dev,
1269 struct neighbour *n);
1270 void (*ndo_neigh_destroy)(struct net_device *dev,
1271 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001272
1273 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001274 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001275 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001276 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001277 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001278 u16 flags);
1279 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001280 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001281 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001282 const unsigned char *addr,
1283 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001284 int (*ndo_fdb_dump)(struct sk_buff *skb,
1285 struct netlink_callback *cb,
1286 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001287 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001288 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001289
1290 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001291 struct nlmsghdr *nlh,
1292 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001293 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1294 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001295 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001296 u32 filter_mask,
1297 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001298 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001299 struct nlmsghdr *nlh,
1300 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001301 int (*ndo_change_carrier)(struct net_device *dev,
1302 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001303 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001304 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001305 int (*ndo_get_phys_port_name)(struct net_device *dev,
1306 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001307 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1308 struct udp_tunnel_info *ti);
1309 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1310 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001311 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1312 struct net_device *dev);
1313 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1314 void *priv);
1315
1316 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1317 struct net_device *dev,
1318 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001319 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001320 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1321 int queue_index,
1322 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001323 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001324 int (*ndo_change_proto_down)(struct net_device *dev,
1325 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001326 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1327 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001328 void (*ndo_set_rx_headroom)(struct net_device *dev,
1329 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001330 int (*ndo_xdp)(struct net_device *dev,
1331 struct netdev_xdp *xdp);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001332};
1333
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001334/**
1335 * enum net_device_priv_flags - &struct net_device priv_flags
1336 *
1337 * These are the &struct net_device, they are only set internally
1338 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001339 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001340 * during any kernel release.
1341 *
1342 * You should have a pretty good reason to be extending these flags.
1343 *
1344 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1345 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001346 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001347 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001348 * @IFF_WAN_HDLC: WAN HDLC device
1349 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1350 * release skb->dst
1351 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1352 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1353 * @IFF_MACVLAN_PORT: device used as macvlan port
1354 * @IFF_BRIDGE_PORT: device used as bridge port
1355 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1356 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1357 * @IFF_UNICAST_FLT: Supports unicast filtering
1358 * @IFF_TEAM_PORT: device used as team port
1359 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1360 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1361 * change when it's running
1362 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001363 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1364 * underlying stacked devices
1365 * @IFF_IPVLAN_MASTER: IPvlan master device
1366 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001367 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001368 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001369 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001370 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001371 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001372 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001373 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1374 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001375 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001376 */
1377enum netdev_priv_flags {
1378 IFF_802_1Q_VLAN = 1<<0,
1379 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001380 IFF_BONDING = 1<<2,
1381 IFF_ISATAP = 1<<3,
1382 IFF_WAN_HDLC = 1<<4,
1383 IFF_XMIT_DST_RELEASE = 1<<5,
1384 IFF_DONT_BRIDGE = 1<<6,
1385 IFF_DISABLE_NETPOLL = 1<<7,
1386 IFF_MACVLAN_PORT = 1<<8,
1387 IFF_BRIDGE_PORT = 1<<9,
1388 IFF_OVS_DATAPATH = 1<<10,
1389 IFF_TX_SKB_SHARING = 1<<11,
1390 IFF_UNICAST_FLT = 1<<12,
1391 IFF_TEAM_PORT = 1<<13,
1392 IFF_SUPP_NOFCS = 1<<14,
1393 IFF_LIVE_ADDR_CHANGE = 1<<15,
1394 IFF_MACVLAN = 1<<16,
1395 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1396 IFF_IPVLAN_MASTER = 1<<18,
1397 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001398 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001399 IFF_NO_QUEUE = 1<<21,
1400 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001401 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001402 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001403 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001404 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001405 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001406};
1407
1408#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1409#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001410#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001411#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001412#define IFF_WAN_HDLC IFF_WAN_HDLC
1413#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1414#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1415#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1416#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1417#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1418#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1419#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1420#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1421#define IFF_TEAM_PORT IFF_TEAM_PORT
1422#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1423#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1424#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001425#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001426#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1427#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001428#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001429#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001430#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001431#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001432#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001433#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001434#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001435
Karoly Kemeny536721b2014-07-30 20:27:36 +02001436/**
1437 * struct net_device - The DEVICE structure.
1438 * Actually, this whole structure is a big mistake. It mixes I/O
1439 * data with strictly "high-level" data, and it has to know about
1440 * almost every data structure used in the INET module.
1441 *
1442 * @name: This is the first field of the "visible" part of this structure
1443 * (i.e. as seen by users in the "Space.c" file). It is the name
1444 * of the interface.
1445 *
1446 * @name_hlist: Device name hash chain, please keep it close to name[]
1447 * @ifalias: SNMP alias
1448 * @mem_end: Shared memory end
1449 * @mem_start: Shared memory start
1450 * @base_addr: Device I/O address
1451 * @irq: Device IRQ number
1452 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001453 * @carrier_changes: Stats to monitor carrier on<->off transitions
1454 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001455 * @state: Generic network queuing layer state, see netdev_state_t
1456 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001457 * @napi_list: List entry used for polling NAPI devices
1458 * @unreg_list: List entry when we are unregistering the
1459 * device; see the function unregister_netdev
1460 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001461 * @ptype_all: Device-specific packet handlers for all protocols
1462 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001463 *
1464 * @adj_list: Directly linked devices, like slaves for bonding
Karoly Kemeny536721b2014-07-30 20:27:36 +02001465 * @features: Currently active device features
1466 * @hw_features: User-changeable features
1467 *
1468 * @wanted_features: User-requested features
1469 * @vlan_features: Mask of features inheritable by VLAN devices
1470 *
1471 * @hw_enc_features: Mask of features inherited by encapsulating devices
1472 * This field indicates what encapsulation
1473 * offloads the hardware is capable of doing,
1474 * and drivers will need to set them appropriately.
1475 *
1476 * @mpls_features: Mask of features inheritable by MPLS
1477 *
1478 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001479 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001480 *
1481 * @stats: Statistics struct, which was left as a legacy, use
1482 * rtnl_link_stats64 instead
1483 *
1484 * @rx_dropped: Dropped packets by core network,
1485 * do not use this in drivers
1486 * @tx_dropped: Dropped packets by core network,
1487 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001488 * @rx_nohandler: nohandler dropped packets by core network on
1489 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001490 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001491 * @wireless_handlers: List of functions to handle Wireless Extensions,
1492 * instead of ioctl,
1493 * see <net/iw_handler.h> for details.
1494 * @wireless_data: Instance data managed by the core of wireless extensions
1495 *
1496 * @netdev_ops: Includes several pointers to callbacks,
1497 * if one wants to override the ndo_*() functions
1498 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001499 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1500 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001501 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001502 * of Layer 2 headers.
1503 *
1504 * @flags: Interface flags (a la BSD)
1505 * @priv_flags: Like 'flags' but invisible to userspace,
1506 * see if.h for the definitions
1507 * @gflags: Global flags ( kept as legacy )
1508 * @padded: How much padding added by alloc_netdev()
1509 * @operstate: RFC2863 operstate
1510 * @link_mode: Mapping policy to operstate
1511 * @if_port: Selectable AUI, TP, ...
1512 * @dma: DMA channel
1513 * @mtu: Interface MTU value
Jarod Wilson61e84622016-10-07 22:04:33 -04001514 * @min_mtu: Interface Minimum MTU value
1515 * @max_mtu: Interface Maximum MTU value
Karoly Kemeny536721b2014-07-30 20:27:36 +02001516 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001517 * @hard_header_len: Maximum hardware header length.
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05001518 * @min_header_len: Minimum hardware header length
Karoly Kemeny536721b2014-07-30 20:27:36 +02001519 *
1520 * @needed_headroom: Extra headroom the hardware may need, but not in all
1521 * cases can this be guaranteed
1522 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1523 * cases can this be guaranteed. Some cases also use
1524 * LL_MAX_HEADER instead to allocate the skb
1525 *
1526 * interface address info:
1527 *
1528 * @perm_addr: Permanent hw address
1529 * @addr_assign_type: Hw address assignment type
1530 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001531 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001532 * @dev_id: Used to differentiate devices that share
1533 * the same link layer address
1534 * @dev_port: Used to differentiate devices that share
1535 * the same function
1536 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001537 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001538 * has been enabled due to the need to listen to
1539 * additional unicast addresses in a device that
1540 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001541 * @uc: unicast mac addresses
1542 * @mc: multicast mac addresses
1543 * @dev_addrs: list of device hw addresses
1544 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001545 * @promiscuity: Number of times the NIC is told to work in
1546 * promiscuous mode; if it becomes 0 the NIC will
1547 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001548 * @allmulti: Counter, enables or disables allmulticast mode
1549 *
1550 * @vlan_info: VLAN info
1551 * @dsa_ptr: dsa specific data
1552 * @tipc_ptr: TIPC specific data
1553 * @atalk_ptr: AppleTalk link
1554 * @ip_ptr: IPv4 specific data
1555 * @dn_ptr: DECnet specific data
1556 * @ip6_ptr: IPv6 specific data
1557 * @ax25_ptr: AX.25 specific data
1558 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1559 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001560 * @dev_addr: Hw address (before bcast,
1561 * because most packets are unicast)
1562 *
1563 * @_rx: Array of RX queues
1564 * @num_rx_queues: Number of RX queues
1565 * allocated at register_netdev() time
1566 * @real_num_rx_queues: Number of RX queues currently active in device
1567 *
1568 * @rx_handler: handler for received packets
1569 * @rx_handler_data: XXX: need comments on this one
1570 * @ingress_queue: XXX: need comments on this one
1571 * @broadcast: hw bcast address
1572 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001573 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1574 * indexed by RX queue number. Assigned by driver.
1575 * This must only be set if the ndo_rx_flow_steer
1576 * operation is defined
1577 * @index_hlist: Device index hash chain
1578 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001579 * @_tx: Array of TX queues
1580 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1581 * @real_num_tx_queues: Number of TX queues currently active in device
1582 * @qdisc: Root qdisc from userspace point of view
1583 * @tx_queue_len: Max frames per queue allowed
1584 * @tx_global_lock: XXX: need comments on this one
1585 *
1586 * @xps_maps: XXX: need comments on this one
1587 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001588 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001589 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001590 * @watchdog_timer: List of timers
1591 *
1592 * @pcpu_refcnt: Number of references to this device
1593 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001594 * @link_watch_list: XXX: need comments on this one
1595 *
1596 * @reg_state: Register/unregister state machine
1597 * @dismantle: Device is going to be freed
1598 * @rtnl_link_state: This enum represents the phases of creating
1599 * a new link
1600 *
David S. Millercf124db2017-05-08 12:52:56 -04001601 * @needs_free_netdev: Should unregister perform free_netdev?
1602 * @priv_destructor: Called from unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001603 * @npinfo: XXX: need comments on this one
1604 * @nd_net: Network namespace this network device is inside
1605 *
1606 * @ml_priv: Mid-layer private
1607 * @lstats: Loopback statistics
1608 * @tstats: Tunnel statistics
1609 * @dstats: Dummy statistics
1610 * @vstats: Virtual ethernet statistics
1611 *
1612 * @garp_port: GARP
1613 * @mrp_port: MRP
1614 *
1615 * @dev: Class/net/name entry
1616 * @sysfs_groups: Space for optional device, statistics and wireless
1617 * sysfs groups
1618 *
1619 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1620 * @rtnl_link_ops: Rtnl_link_ops
1621 *
1622 * @gso_max_size: Maximum size of generic segmentation offload
1623 * @gso_max_segs: Maximum number of segments that can be passed to the
1624 * NIC for GSO
1625 *
1626 * @dcbnl_ops: Data Center Bridging netlink ops
1627 * @num_tc: Number of traffic classes in the net device
1628 * @tc_to_txq: XXX: need comments on this one
Randy Dunlap920c1cd2016-11-21 18:28:36 -08001629 * @prio_tc_map: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001630 *
1631 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1632 *
1633 * @priomap: XXX: need comments on this one
1634 * @phydev: Physical device may attach itself
1635 * for hardware timestamping
1636 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001637 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1638 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001639 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001640 * @proto_down: protocol port state information can be sent to the
1641 * switch driver and used to set the phys state of the
1642 * switch port.
1643 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 * FIXME: cleanup struct net_device such that network protocol info
1645 * moves out.
1646 */
1647
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001648struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001650 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001651 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 /*
1653 * I/O specific fields
1654 * FIXME: Merge these and struct ifmap into one
1655 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001656 unsigned long mem_end;
1657 unsigned long mem_start;
1658 unsigned long base_addr;
1659 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001661 atomic_t carrier_changes;
1662
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001664 * Some hardware also needs these fields (state,dev_list,
1665 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 * part of the usual set specified in Space.c.
1667 */
1668
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 unsigned long state;
1670
Pavel Emelianov7562f872007-05-03 15:13:45 -07001671 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001672 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001673 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001674 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001675 struct list_head ptype_all;
1676 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001677
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001678 struct {
1679 struct list_head upper;
1680 struct list_head lower;
1681 } adj_list;
1682
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001683 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001684 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001685 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001686 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001687 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001688 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001689 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001690
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001692 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693
Rusty Russellc45d2862007-03-28 14:29:08 -07001694 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001695
Eric Dumazet015f0682014-03-27 08:45:56 -07001696 atomic_long_t rx_dropped;
1697 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001698 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699
Johannes Bergb86e0282007-04-26 20:48:23 -07001700#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001701 const struct iw_handler_def *wireless_handlers;
1702 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001703#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001704 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001705 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001706#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001707 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001708#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001709#ifdef CONFIG_NET_L3_MASTER_DEV
1710 const struct l3mdev_ops *l3mdev_ops;
1711#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001712#if IS_ENABLED(CONFIG_IPV6)
1713 const struct ndisc_ops *ndisc_ops;
1714#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715
Steffen Klassertd77e38e2017-04-14 10:06:10 +02001716#ifdef CONFIG_XFRM
1717 const struct xfrmdev_ops *xfrmdev_ops;
1718#endif
1719
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001720 const struct header_ops *header_ops;
1721
Karoly Kemeny536721b2014-07-30 20:27:36 +02001722 unsigned int flags;
1723 unsigned int priv_flags;
1724
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001726 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727
Karoly Kemeny536721b2014-07-30 20:27:36 +02001728 unsigned char operstate;
1729 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001730
Karoly Kemeny536721b2014-07-30 20:27:36 +02001731 unsigned char if_port;
1732 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001733
Karoly Kemeny536721b2014-07-30 20:27:36 +02001734 unsigned int mtu;
Jarod Wilson61e84622016-10-07 22:04:33 -04001735 unsigned int min_mtu;
1736 unsigned int max_mtu;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001737 unsigned short type;
1738 unsigned short hard_header_len;
Alexey Dobriyand92be7a2017-04-10 11:25:26 +03001739 unsigned char min_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
Johannes Bergf5184d22008-05-12 20:48:31 -07001741 unsigned short needed_headroom;
1742 unsigned short needed_tailroom;
1743
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001745 unsigned char perm_addr[MAX_ADDR_LEN];
1746 unsigned char addr_assign_type;
1747 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001748 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001749 unsigned short dev_id;
1750 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001751 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001752 unsigned char name_assign_type;
1753 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001754 struct netdev_hw_addr_list uc;
1755 struct netdev_hw_addr_list mc;
1756 struct netdev_hw_addr_list dev_addrs;
1757
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001758#ifdef CONFIG_SYSFS
1759 struct kset *queues_kset;
1760#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001761 unsigned int promiscuity;
1762 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001765 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001766
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001767#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001768 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001769#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001770#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001771 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001772#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001773#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001774 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001775#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001776 void *atalk_ptr;
1777 struct in_device __rcu *ip_ptr;
1778 struct dn_dev __rcu *dn_ptr;
1779 struct inet6_dev __rcu *ip6_ptr;
1780 void *ax25_ptr;
1781 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001782 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001783#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1784 struct mpls_dev __rcu *mpls_ptr;
1785#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001787/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001788 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001789 */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001790 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001791 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001792
Michael Daltona953be52014-01-16 22:23:28 -08001793#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001794 struct netdev_rx_queue *_rx;
1795
Tom Herbert0a9627f2010-03-16 08:03:29 +00001796 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001797 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001798#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001799
Eric Dumazet7acedaf2017-04-25 11:36:52 -07001800 struct bpf_prog __rcu *xdp_prog;
Eric Dumazet3b47d302014-11-06 21:09:44 -08001801 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001802 rx_handler_func_t __rcu *rx_handler;
1803 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001804
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001805#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001806 struct tcf_proto __rcu *ingress_cl_list;
1807#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001808 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001809#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001810 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001811#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001812
Karoly Kemeny536721b2014-07-30 20:27:36 +02001813 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001814#ifdef CONFIG_RFS_ACCEL
1815 struct cpu_rmap *rx_cpu_rmap;
1816#endif
1817 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001818
1819/*
1820 * Cache lines mostly used on transmit path
1821 */
David S. Millere8a04642008-07-17 00:34:19 -07001822 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1823 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001824 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001825 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001826#ifdef CONFIG_NET_SCHED
1827 DECLARE_HASHTABLE (qdisc_hash, 4);
1828#endif
Alexey Dobriyan0cd29502017-05-17 13:30:44 +03001829 unsigned int tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001830 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001831 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001832
Tom Herbertbf264142010-11-26 08:36:09 +00001833#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001834 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001835#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001836#ifdef CONFIG_NET_CLS_ACT
1837 struct tcf_proto __rcu *egress_cl_list;
1838#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001839
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001840 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001841 struct timer_list watchdog_timer;
1842
Eric Dumazet29b44332010-10-11 10:22:12 +00001843 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845
Eric Dumazete014deb2009-11-17 05:59:21 +00001846 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001847
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001849 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 NETREG_UNREGISTERING, /* called unregister_netdevice */
1851 NETREG_UNREGISTERED, /* completed unregister todo */
1852 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001853 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001854 } reg_state:8;
1855
Karoly Kemeny536721b2014-07-30 20:27:36 +02001856 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001857
1858 enum {
1859 RTNL_LINK_INITIALIZED,
1860 RTNL_LINK_INITIALIZING,
1861 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862
David S. Millercf124db2017-05-08 12:52:56 -04001863 bool needs_free_netdev;
1864 void (*priv_destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001867 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868#endif
David S. Millereae792b2008-07-15 03:03:33 -07001869
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001870 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001871
David S. Miller49517042008-05-12 03:29:11 -07001872 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001873 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001874 void *ml_priv;
1875 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001876 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001877 struct pcpu_dstats __percpu *dstats;
1878 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001879 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001880
Tobias Klauserfb585b42017-02-10 16:43:50 +01001881#if IS_ENABLED(CONFIG_GARP)
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001882 struct garp_port __rcu *garp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001883#endif
1884#if IS_ENABLED(CONFIG_MRP)
David Wardfebf0182013-02-08 17:17:06 +00001885 struct mrp_port __rcu *mrp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001886#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001888 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001889 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001890 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001891
Patrick McHardy38f7b872007-06-13 12:03:51 -07001892 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001893
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001894 /* for setting kernel sock attribute on TCP connection setup */
1895#define GSO_MAX_SIZE 65536
1896 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001897#define GSO_MAX_SEGS 65535
1898 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001899
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001900#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001901 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001902#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001903 u8 num_tc;
1904 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1905 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001906
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001907#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001908 unsigned int fcoe_ddp_xid;
1909#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001910#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001911 struct netprio_map __rcu *priomap;
1912#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001913 struct phy_device *phydev;
1914 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001915 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001916 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001918#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919
David S. Millerb5cdae32017-04-18 15:36:58 -04001920static inline bool netif_elide_gro(const struct net_device *dev)
1921{
1922 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1923 return true;
1924 return false;
1925}
1926
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928
David S. Millere8a04642008-07-17 00:34:19 -07001929static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001930int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1931{
1932 return dev->prio_tc_map[prio & TC_BITMASK];
1933}
1934
1935static inline
1936int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1937{
1938 if (tc >= dev->num_tc)
1939 return -EINVAL;
1940
1941 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1942 return 0;
1943}
1944
Alexander Duyck8d059b02016-10-28 11:43:49 -04001945int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
Alexander Duyck9cf1f6a2016-10-28 11:43:20 -04001946void netdev_reset_tc(struct net_device *dev);
1947int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1948int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
John Fastabend4f57c082011-01-17 08:06:04 +00001949
1950static inline
1951int netdev_get_num_tc(struct net_device *dev)
1952{
1953 return dev->num_tc;
1954}
1955
1956static inline
David S. Millere8a04642008-07-17 00:34:19 -07001957struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1958 unsigned int index)
1959{
1960 return &dev->_tx[index];
1961}
1962
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001963static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1964 const struct sk_buff *skb)
1965{
1966 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1967}
1968
David S. Millere8a04642008-07-17 00:34:19 -07001969static inline void netdev_for_each_tx_queue(struct net_device *dev,
1970 void (*f)(struct net_device *,
1971 struct netdev_queue *,
1972 void *),
1973 void *arg)
1974{
1975 unsigned int i;
1976
1977 for (i = 0; i < dev->num_tx_queues; i++)
1978 f(dev, &dev->_tx[i], arg);
1979}
1980
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001981#define netdev_lockdep_set_classes(dev) \
1982{ \
1983 static struct lock_class_key qdisc_tx_busylock_key; \
1984 static struct lock_class_key qdisc_running_key; \
1985 static struct lock_class_key qdisc_xmit_lock_key; \
1986 static struct lock_class_key dev_addr_list_lock_key; \
1987 unsigned int i; \
1988 \
1989 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1990 (dev)->qdisc_running_key = &qdisc_running_key; \
1991 lockdep_set_class(&(dev)->addr_list_lock, \
1992 &dev_addr_list_lock_key); \
1993 for (i = 0; i < (dev)->num_tx_queues; i++) \
1994 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1995 &qdisc_xmit_lock_key); \
1996}
1997
Joe Perchesf629d202013-09-26 14:48:15 -07001998struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001999 struct sk_buff *skb,
2000 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00002001
Paolo Abeni871b6422016-02-26 10:45:37 +01002002/* returns the headroom that the master device needs to take in account
2003 * when forwarding to this dev
2004 */
2005static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2006{
2007 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2008}
2009
2010static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2011{
2012 if (dev->netdev_ops->ndo_set_rx_headroom)
2013 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2014}
2015
2016/* set the device rx headroom to the dev's default */
2017static inline void netdev_reset_rx_headroom(struct net_device *dev)
2018{
2019 netdev_set_rx_headroom(dev, -1);
2020}
2021
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002022/*
2023 * Net namespace inlines
2024 */
2025static inline
2026struct net *dev_net(const struct net_device *dev)
2027{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002028 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002029}
2030
2031static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002032void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002033{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002034 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002035}
2036
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002037/**
2038 * netdev_priv - access network device private data
2039 * @dev: network device
2040 *
2041 * Get network device private data
2042 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002043static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002045 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046}
2047
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048/* Set the sysfs physical device reference for the network logical device
2049 * if set prior to registration will cause a symlink during initialization.
2050 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002051#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052
Marcel Holtmann384912e2009-08-31 21:08:19 +00002053/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002054 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002055 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002056 */
2057#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2058
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002059/* Default NAPI poll() weight
2060 * Device drivers are strongly advised to not use bigger value
2061 */
2062#define NAPI_POLL_WEIGHT 64
2063
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002064/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002065 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002066 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002067 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002068 * @poll: polling function
2069 * @weight: default weight
2070 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002071 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2072 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002073 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002074void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2075 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002076
Alexander Duyckd8156532008-07-08 15:13:05 -07002077/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002078 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002079 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002080 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002081 * @poll: polling function
2082 * @weight: default weight
2083 *
2084 * This variant of netif_napi_add() should be used from drivers using NAPI
2085 * to exclusively poll a TX queue.
2086 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2087 */
2088static inline void netif_tx_napi_add(struct net_device *dev,
2089 struct napi_struct *napi,
2090 int (*poll)(struct napi_struct *, int),
2091 int weight)
2092{
2093 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2094 netif_napi_add(dev, napi, poll, weight);
2095}
2096
2097/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002098 * netif_napi_del - remove a NAPI context
2099 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002100 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002101 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002102 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002103void netif_napi_del(struct napi_struct *napi);
2104
2105struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002106 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002107 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002108
Herbert Xu74895942009-05-26 18:50:27 +00002109 /* Length of frag0. */
2110 unsigned int frag0_len;
2111
Herbert Xu86911732009-01-29 14:19:50 +00002112 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002113 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002114
Herbert Xud565b0a2008-12-15 23:38:52 -08002115 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002116 u16 flush;
2117
2118 /* Save the IP ID here and check when we get to the transport layer */
2119 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002120
2121 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002122 u16 count;
2123
Tom Herbert15e23962015-02-10 16:30:31 -08002124 /* Start offset for remote checksum offload */
2125 u16 gro_remcsum_start;
2126
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002127 /* jiffies when first packet was created/queued */
2128 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002129
Tom Herbertafe93322014-09-17 12:25:57 -07002130 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002131 u16 proto;
2132
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002133 /* This is non-zero if the packet may be of the same flow. */
2134 u8 same_flow:1;
2135
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002136 /* Used in tunnel GRO receive */
2137 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002138
2139 /* GRO checksum is valid */
2140 u8 csum_valid:1;
2141
Tom Herbert662880f2014-08-27 21:26:56 -07002142 /* Number of checksums via CHECKSUM_UNNECESSARY */
2143 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002144
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002145 /* Free the skb? */
2146 u8 free:2;
2147#define NAPI_GRO_FREE 1
2148#define NAPI_GRO_FREE_STOLEN_HEAD 2
2149
Tom Herbertefc98d02014-10-03 15:48:08 -07002150 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2151 u8 is_ipv6:1;
2152
Alexander Duycka0ca1532016-04-05 09:13:39 -07002153 /* Used in GRE, set in fou/gue_gro_receive */
2154 u8 is_fou:1;
2155
Alexander Duyck15305452016-04-10 21:44:57 -04002156 /* Used to determine if flush_id can be ignored */
2157 u8 is_atomic:1;
2158
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002159 /* Number of gro_receive callbacks this packet already went through */
2160 u8 recursion_counter:4;
2161
2162 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002163
Jerry Chubf5a7552014-01-07 10:23:19 -08002164 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2165 __wsum csum;
2166
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002167 /* used in skb_gro_receive() slow path */
2168 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002169};
2170
2171#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002172
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002173#define GRO_RECURSION_LIMIT 15
2174static inline int gro_recursion_inc_test(struct sk_buff *skb)
2175{
2176 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2177}
2178
2179typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2180static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2181 struct sk_buff **head,
2182 struct sk_buff *skb)
2183{
2184 if (unlikely(gro_recursion_inc_test(skb))) {
2185 NAPI_GRO_CB(skb)->flush |= 1;
2186 return NULL;
2187 }
2188
2189 return cb(head, skb);
2190}
2191
2192typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2193 struct sk_buff *);
2194static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2195 struct sock *sk,
2196 struct sk_buff **head,
2197 struct sk_buff *skb)
2198{
2199 if (unlikely(gro_recursion_inc_test(skb))) {
2200 NAPI_GRO_CB(skb)->flush |= 1;
2201 return NULL;
2202 }
2203
2204 return cb(sk, head, skb);
2205}
2206
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002208 __be16 type; /* This is really htons(ether_type). */
2209 struct net_device *dev; /* NULL is wildcarded here */
2210 int (*func) (struct sk_buff *,
2211 struct net_device *,
2212 struct packet_type *,
2213 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002214 bool (*id_match)(struct packet_type *ptype,
2215 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216 void *af_packet_priv;
2217 struct list_head list;
2218};
2219
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002220struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2222 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002224 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002225 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002226};
2227
2228struct packet_offload {
2229 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002230 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002231 struct offload_callbacks callbacks;
2232 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233};
2234
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002235/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002236struct pcpu_sw_netstats {
2237 u64 rx_packets;
2238 u64 rx_bytes;
2239 u64 tx_packets;
2240 u64 tx_bytes;
2241 struct u64_stats_sync syncp;
2242};
2243
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002244#define __netdev_alloc_pcpu_stats(type, gfp) \
2245({ \
2246 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2247 if (pcpu_stats) { \
2248 int __cpu; \
2249 for_each_possible_cpu(__cpu) { \
2250 typeof(type) *stat; \
2251 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2252 u64_stats_init(&stat->syncp); \
2253 } \
2254 } \
2255 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002256})
2257
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002258#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002259 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002260
Jiri Pirko764f5e52015-12-03 12:12:12 +01002261enum netdev_lag_tx_type {
2262 NETDEV_LAG_TX_TYPE_UNKNOWN,
2263 NETDEV_LAG_TX_TYPE_RANDOM,
2264 NETDEV_LAG_TX_TYPE_BROADCAST,
2265 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2266 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2267 NETDEV_LAG_TX_TYPE_HASH,
2268};
2269
2270struct netdev_lag_upper_info {
2271 enum netdev_lag_tx_type tx_type;
2272};
2273
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002274struct netdev_lag_lower_state_info {
2275 u8 link_up : 1,
2276 tx_enabled : 1;
2277};
2278
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279#include <linux/notifier.h>
2280
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002281/* netdevice notifier chain. Please remember to update the rtnetlink
2282 * notification exclusion list in rtnetlink_event() when adding new
2283 * types.
2284 */
2285#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2286#define NETDEV_DOWN 0x0002
2287#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2288 detected a hardware crash and restarted
2289 - we can use this eg to kick tcp sessions
2290 once done */
2291#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2292#define NETDEV_REGISTER 0x0005
2293#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002294#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002295#define NETDEV_CHANGEADDR 0x0008
2296#define NETDEV_GOING_DOWN 0x0009
2297#define NETDEV_CHANGENAME 0x000A
2298#define NETDEV_FEAT_CHANGE 0x000B
2299#define NETDEV_BONDING_FAILOVER 0x000C
2300#define NETDEV_PRE_UP 0x000D
2301#define NETDEV_PRE_TYPE_CHANGE 0x000E
2302#define NETDEV_POST_TYPE_CHANGE 0x000F
2303#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002304#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002305#define NETDEV_RELEASE 0x0012
2306#define NETDEV_NOTIFY_PEERS 0x0013
2307#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002308#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002309#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002310#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002311#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002312#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002313#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002314#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002315#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002316#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002317
Joe Perchesf629d202013-09-26 14:48:15 -07002318int register_netdevice_notifier(struct notifier_block *nb);
2319int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002320
2321struct netdev_notifier_info {
2322 struct net_device *dev;
2323};
2324
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002325struct netdev_notifier_change_info {
2326 struct netdev_notifier_info info; /* must be first */
2327 unsigned int flags_changed;
2328};
2329
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002330struct netdev_notifier_changeupper_info {
2331 struct netdev_notifier_info info; /* must be first */
2332 struct net_device *upper_dev; /* new upper dev */
2333 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002334 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002335 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002336};
2337
Jiri Pirko04d48262015-12-03 12:12:15 +01002338struct netdev_notifier_changelowerstate_info {
2339 struct netdev_notifier_info info; /* must be first */
2340 void *lower_state_info; /* is lower dev state */
2341};
2342
Cong Wang75538c22013-05-29 11:30:50 +08002343static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2344 struct net_device *dev)
2345{
2346 info->dev = dev;
2347}
2348
Jiri Pirko351638e2013-05-28 01:30:21 +00002349static inline struct net_device *
2350netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2351{
2352 return info->dev;
2353}
2354
Joe Perchesf629d202013-09-26 14:48:15 -07002355int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002356
2357
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358extern rwlock_t dev_base_lock; /* Device list lock */
2359
Eric W. Biederman881d9662007-09-17 11:56:21 -07002360#define for_each_netdev(net, d) \
2361 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002362#define for_each_netdev_reverse(net, d) \
2363 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002364#define for_each_netdev_rcu(net, d) \
2365 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002366#define for_each_netdev_safe(net, d, n) \
2367 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2368#define for_each_netdev_continue(net, d) \
2369 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002370#define for_each_netdev_continue_rcu(net, d) \
2371 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002372#define for_each_netdev_in_bond_rcu(bond, slave) \
2373 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002374 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002375#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2376
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002377static inline struct net_device *next_net_device(struct net_device *dev)
2378{
2379 struct list_head *lh;
2380 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002381
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002382 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002383 lh = dev->dev_list.next;
2384 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2385}
2386
Eric Dumazetce81b762009-11-11 17:34:30 +00002387static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2388{
2389 struct list_head *lh;
2390 struct net *net;
2391
2392 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002393 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002394 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2395}
2396
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002397static inline struct net_device *first_net_device(struct net *net)
2398{
2399 return list_empty(&net->dev_base_head) ? NULL :
2400 net_device_entry(net->dev_base_head.next);
2401}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002402
Eric Dumazetccf43432011-01-26 18:08:02 +00002403static inline struct net_device *first_net_device_rcu(struct net *net)
2404{
2405 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2406
2407 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2408}
2409
Joe Perchesf629d202013-09-26 14:48:15 -07002410int netdev_boot_setup_check(struct net_device *dev);
2411unsigned long netdev_boot_base(const char *prefix, int unit);
2412struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2413 const char *hwaddr);
2414struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2415struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2416void dev_add_pack(struct packet_type *pt);
2417void dev_remove_pack(struct packet_type *pt);
2418void __dev_remove_pack(struct packet_type *pt);
2419void dev_add_offload(struct packet_offload *po);
2420void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002422int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002423int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002424struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2425 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002426struct net_device *dev_get_by_name(struct net *net, const char *name);
2427struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2428struct net_device *__dev_get_by_name(struct net *net, const char *name);
2429int dev_alloc_name(struct net_device *dev, const char *name);
2430int dev_open(struct net_device *dev);
2431int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002432int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002433void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002434int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002435int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002436int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002437int register_netdevice(struct net_device *dev);
2438void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2439void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002440static inline void unregister_netdevice(struct net_device *dev)
2441{
2442 unregister_netdevice_queue(dev, NULL);
2443}
2444
Joe Perchesf629d202013-09-26 14:48:15 -07002445int netdev_refcnt_read(const struct net_device *dev);
2446void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002447void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002448void synchronize_net(void);
2449int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002450
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002451DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002452#define XMIT_RECURSION_LIMIT 10
2453
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002454static inline int dev_recursion_level(void)
2455{
2456 return this_cpu_read(xmit_recursion);
2457}
2458
Joe Perchesf629d202013-09-26 14:48:15 -07002459struct net_device *dev_get_by_index(struct net *net, int ifindex);
2460struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2461struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
Miroslav Lichvar90b602f2017-05-19 17:52:37 +02002462struct net_device *dev_get_by_napi_id(unsigned int napi_id);
Joe Perchesf629d202013-09-26 14:48:15 -07002463int netdev_get_name(struct net *net, char *name, int ifindex);
2464int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002465int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002466
2467static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2468{
2469 return NAPI_GRO_CB(skb)->data_offset;
2470}
2471
2472static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2473{
2474 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2475}
2476
2477static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2478{
2479 NAPI_GRO_CB(skb)->data_offset += len;
2480}
2481
Herbert Xua5b1cf22009-05-26 18:50:28 +00002482static inline void *skb_gro_header_fast(struct sk_buff *skb,
2483 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002484{
Herbert Xu78a478d2009-05-26 18:50:21 +00002485 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002486}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487
Herbert Xua5b1cf22009-05-26 18:50:28 +00002488static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2489{
2490 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2491}
2492
Herbert Xu57ea52a2017-01-10 12:24:15 -08002493static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2494{
2495 NAPI_GRO_CB(skb)->frag0 = NULL;
2496 NAPI_GRO_CB(skb)->frag0_len = 0;
2497}
2498
Herbert Xua5b1cf22009-05-26 18:50:28 +00002499static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2500 unsigned int offset)
2501{
Herbert Xu17dd7592011-07-27 06:16:28 -07002502 if (!pskb_may_pull(skb, hlen))
2503 return NULL;
2504
Herbert Xu57ea52a2017-01-10 12:24:15 -08002505 skb_gro_frag0_invalidate(skb);
Herbert Xu17dd7592011-07-27 06:16:28 -07002506 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002507}
2508
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002509static inline void *skb_gro_network_header(struct sk_buff *skb)
2510{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002511 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2512 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002513}
2514
Jerry Chubf5a7552014-01-07 10:23:19 -08002515static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2516 const void *start, unsigned int len)
2517{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002518 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002519 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2520 csum_partial(start, len, 0));
2521}
2522
Tom Herbert573e8fc2014-08-22 13:33:47 -07002523/* GRO checksum functions. These are logical equivalents of the normal
2524 * checksum functions (in skbuff.h) except that they operate on the GRO
2525 * offsets and fields in sk_buff.
2526 */
2527
2528__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2529
Tom Herbert15e23962015-02-10 16:30:31 -08002530static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2531{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002532 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002533}
2534
Tom Herbert573e8fc2014-08-22 13:33:47 -07002535static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2536 bool zero_okay,
2537 __sum16 check)
2538{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002539 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2540 skb_checksum_start_offset(skb) <
2541 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002542 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002543 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002544 (!zero_okay || check));
2545}
2546
2547static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2548 __wsum psum)
2549{
2550 if (NAPI_GRO_CB(skb)->csum_valid &&
2551 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2552 return 0;
2553
2554 NAPI_GRO_CB(skb)->csum = psum;
2555
2556 return __skb_gro_checksum_complete(skb);
2557}
2558
Tom Herbert573e8fc2014-08-22 13:33:47 -07002559static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2560{
Tom Herbert662880f2014-08-27 21:26:56 -07002561 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2562 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2563 NAPI_GRO_CB(skb)->csum_cnt--;
2564 } else {
2565 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2566 * verified a new top level checksum or an encapsulated one
2567 * during GRO. This saves work if we fallback to normal path.
2568 */
2569 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002570 }
2571}
2572
2573#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2574 compute_pseudo) \
2575({ \
2576 __sum16 __ret = 0; \
2577 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2578 __ret = __skb_gro_checksum_validate_complete(skb, \
2579 compute_pseudo(skb, proto)); \
Davide Caratti219f1d792017-05-18 15:44:39 +02002580 if (!__ret) \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002581 skb_gro_incr_csum_unnecessary(skb); \
2582 __ret; \
2583})
2584
2585#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2586 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2587
2588#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2589 compute_pseudo) \
2590 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2591
2592#define skb_gro_checksum_simple_validate(skb) \
2593 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2594
Tom Herbertd96535a2014-08-31 15:12:42 -07002595static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2596{
2597 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2598 !NAPI_GRO_CB(skb)->csum_valid);
2599}
2600
2601static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2602 __sum16 check, __wsum pseudo)
2603{
2604 NAPI_GRO_CB(skb)->csum = ~pseudo;
2605 NAPI_GRO_CB(skb)->csum_valid = 1;
2606}
2607
2608#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2609do { \
2610 if (__skb_gro_checksum_convert_check(skb)) \
2611 __skb_gro_checksum_convert(skb, check, \
2612 compute_pseudo(skb, proto)); \
2613} while (0)
2614
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002615struct gro_remcsum {
2616 int offset;
2617 __wsum delta;
2618};
2619
2620static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2621{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002622 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002623 grc->delta = 0;
2624}
2625
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002626static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2627 unsigned int off, size_t hdrlen,
2628 int start, int offset,
2629 struct gro_remcsum *grc,
2630 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002631{
2632 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002633 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002634
2635 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2636
Tom Herbert15e23962015-02-10 16:30:31 -08002637 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002638 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2639 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002640 }
2641
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002642 ptr = skb_gro_header_fast(skb, off);
2643 if (skb_gro_header_hard(skb, off + plen)) {
2644 ptr = skb_gro_header_slow(skb, off + plen, off);
2645 if (!ptr)
2646 return NULL;
2647 }
2648
2649 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2650 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002651
2652 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002653 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002654
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002655 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002656 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002657
2658 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002659}
2660
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002661static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2662 struct gro_remcsum *grc)
2663{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002664 void *ptr;
2665 size_t plen = grc->offset + sizeof(u16);
2666
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002667 if (!grc->delta)
2668 return;
2669
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002670 ptr = skb_gro_header_fast(skb, grc->offset);
2671 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2672 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2673 if (!ptr)
2674 return;
2675 }
2676
2677 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002678}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002679
Steffen Klassert25393d32017-02-15 09:39:44 +01002680#ifdef CONFIG_XFRM_OFFLOAD
2681static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2682{
2683 if (PTR_ERR(pp) != -EINPROGRESS)
2684 NAPI_GRO_CB(skb)->flush |= flush;
2685}
2686#else
Steffen Klassert5f114162017-02-15 09:39:39 +01002687static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2688{
2689 NAPI_GRO_CB(skb)->flush |= flush;
2690}
Steffen Klassert25393d32017-02-15 09:39:44 +01002691#endif
Steffen Klassert5f114162017-02-15 09:39:39 +01002692
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002693static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2694 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002695 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002696 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002697{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002698 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002699 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002700
2701 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002702}
2703
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002704static inline int dev_parse_header(const struct sk_buff *skb,
2705 unsigned char *haddr)
2706{
2707 const struct net_device *dev = skb->dev;
2708
Patrick McHardy1b833362007-10-18 05:09:28 -07002709 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002710 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002711 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002712}
2713
Willem de Bruijn2793a232016-03-09 21:58:32 -05002714/* ll_header must have at least hard_header_len allocated */
2715static inline bool dev_validate_header(const struct net_device *dev,
2716 char *ll_header, int len)
2717{
2718 if (likely(len >= dev->hard_header_len))
2719 return true;
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05002720 if (len < dev->min_header_len)
2721 return false;
Willem de Bruijn2793a232016-03-09 21:58:32 -05002722
2723 if (capable(CAP_SYS_RAWIO)) {
2724 memset(ll_header + len, 0, dev->hard_header_len - len);
2725 return true;
2726 }
2727
2728 if (dev->header_ops && dev->header_ops->validate)
2729 return dev->header_ops->validate(ll_header, len);
2730
2731 return false;
2732}
2733
Linus Torvalds1da177e2005-04-16 15:20:36 -07002734typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002735int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002736static inline int unregister_gifconf(unsigned int family)
2737{
2738 return register_gifconf(family, NULL);
2739}
2740
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002741#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002742#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002743struct sd_flow_limit {
2744 u64 count;
2745 unsigned int num_buckets;
2746 unsigned int history_head;
2747 u16 history[FLOW_LIMIT_HISTORY];
2748 u8 buckets[];
2749};
2750
2751extern int netdev_flow_limit_table_len;
2752#endif /* CONFIG_NET_FLOW_LIMIT */
2753
Linus Torvalds1da177e2005-04-16 15:20:36 -07002754/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002755 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002756 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002757struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002758 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002759 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002760
Changli Gaodee42872010-05-02 05:42:16 +00002761 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002762 unsigned int processed;
2763 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002764 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002765#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002766 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002767#endif
2768#ifdef CONFIG_NET_FLOW_LIMIT
2769 struct sd_flow_limit __rcu *flow_limit;
2770#endif
2771 struct Qdisc *output_queue;
2772 struct Qdisc **output_queue_tailp;
2773 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002774
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002775#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002776 /* input_queue_head should be written by cpu owning this struct,
2777 * and only read by other cpus. Worth using a cache line.
2778 */
2779 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2780
2781 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002782 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002783 struct softnet_data *rps_ipi_next;
2784 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002785 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002786#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002787 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002788 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002789 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002790
Linus Torvalds1da177e2005-04-16 15:20:36 -07002791};
2792
Tom Herbert76cc8b12010-05-20 18:37:59 +00002793static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002794{
2795#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002796 sd->input_queue_head++;
2797#endif
2798}
2799
2800static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2801 unsigned int *qtail)
2802{
2803#ifdef CONFIG_RPS
2804 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002805#endif
2806}
2807
Tom Herbert0a9627f2010-03-16 08:03:29 +00002808DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002809
Joe Perchesf629d202013-09-26 14:48:15 -07002810void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002811void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002812
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002813static inline void netif_tx_schedule_all(struct net_device *dev)
2814{
2815 unsigned int i;
2816
2817 for (i = 0; i < dev->num_tx_queues; i++)
2818 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2819}
2820
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002821static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002822{
Tom Herbert734664982011-11-28 16:32:44 +00002823 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002824}
2825
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002826/**
2827 * netif_start_queue - allow transmit
2828 * @dev: network device
2829 *
2830 * Allow upper layers to call the device hard_start_xmit routine.
2831 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002832static inline void netif_start_queue(struct net_device *dev)
2833{
David S. Millere8a04642008-07-17 00:34:19 -07002834 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002835}
2836
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002837static inline void netif_tx_start_all_queues(struct net_device *dev)
2838{
2839 unsigned int i;
2840
2841 for (i = 0; i < dev->num_tx_queues; i++) {
2842 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2843 netif_tx_start_queue(txq);
2844 }
2845}
2846
John Fastabend46e5da40a2014-09-12 20:04:52 -07002847void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002848
Dave Jonesd29f7492008-07-22 14:09:06 -07002849/**
2850 * netif_wake_queue - restart transmit
2851 * @dev: network device
2852 *
2853 * Allow upper layers to call the device hard_start_xmit routine.
2854 * Used for flow control when transmit resources are available.
2855 */
David S. Miller79d16382008-07-08 23:14:46 -07002856static inline void netif_wake_queue(struct net_device *dev)
2857{
David S. Millere8a04642008-07-17 00:34:19 -07002858 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859}
2860
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002861static inline void netif_tx_wake_all_queues(struct net_device *dev)
2862{
2863 unsigned int i;
2864
2865 for (i = 0; i < dev->num_tx_queues; i++) {
2866 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2867 netif_tx_wake_queue(txq);
2868 }
2869}
2870
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002871static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002872{
Tom Herbert734664982011-11-28 16:32:44 +00002873 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002874}
2875
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002876/**
2877 * netif_stop_queue - stop transmitted packets
2878 * @dev: network device
2879 *
2880 * Stop upper layers calling the device hard_start_xmit routine.
2881 * Used for flow control when transmit resources are unavailable.
2882 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002883static inline void netif_stop_queue(struct net_device *dev)
2884{
David S. Millere8a04642008-07-17 00:34:19 -07002885 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002886}
2887
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002888void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002889
David S. Miller4d295152012-03-07 21:02:35 -05002890static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002891{
Tom Herbert734664982011-11-28 16:32:44 +00002892 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002893}
2894
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002895/**
2896 * netif_queue_stopped - test if transmit queue is flowblocked
2897 * @dev: network device
2898 *
2899 * Test if transmit queue on device is currently unable to send.
2900 */
David S. Miller4d295152012-03-07 21:02:35 -05002901static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002902{
David S. Millere8a04642008-07-17 00:34:19 -07002903 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904}
2905
David S. Miller4d295152012-03-07 21:02:35 -05002906static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002907{
Tom Herbert734664982011-11-28 16:32:44 +00002908 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2909}
2910
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002911static inline bool
2912netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002913{
2914 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2915}
2916
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002917static inline bool
2918netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2919{
2920 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2921}
2922
Eric Dumazet53511452014-10-08 08:19:27 -07002923/**
2924 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2925 * @dev_queue: pointer to transmit queue
2926 *
2927 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002928 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002929 */
2930static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2931{
2932#ifdef CONFIG_BQL
2933 prefetchw(&dev_queue->dql.num_queued);
2934#endif
2935}
2936
2937/**
2938 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2939 * @dev_queue: pointer to transmit queue
2940 *
2941 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002942 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002943 */
2944static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2945{
2946#ifdef CONFIG_BQL
2947 prefetchw(&dev_queue->dql.limit);
2948#endif
2949}
2950
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002951static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2952 unsigned int bytes)
2953{
Tom Herbert114cf582011-11-28 16:33:09 +00002954#ifdef CONFIG_BQL
2955 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002956
2957 if (likely(dql_avail(&dev_queue->dql) >= 0))
2958 return;
2959
2960 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2961
2962 /*
2963 * The XOFF flag must be set before checking the dql_avail below,
2964 * because in netdev_tx_completed_queue we update the dql_completed
2965 * before checking the XOFF flag.
2966 */
2967 smp_mb();
2968
2969 /* check again in case another CPU has just made room avail */
2970 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2971 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002972#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002973}
2974
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002975/**
2976 * netdev_sent_queue - report the number of bytes queued to hardware
2977 * @dev: network device
2978 * @bytes: number of bytes queued to the hardware device queue
2979 *
2980 * Report the number of bytes queued for sending/completion to the network
2981 * device hardware queue. @bytes should be a good approximation and should
2982 * exactly match netdev_completed_queue() @bytes
2983 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002984static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2985{
2986 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2987}
2988
2989static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002990 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002991{
Tom Herbert114cf582011-11-28 16:33:09 +00002992#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002993 if (unlikely(!bytes))
2994 return;
2995
2996 dql_completed(&dev_queue->dql, bytes);
2997
2998 /*
2999 * Without the memory barrier there is a small possiblity that
3000 * netdev_tx_sent_queue will miss the update and cause the queue to
3001 * be stopped forever
3002 */
3003 smp_mb();
3004
3005 if (dql_avail(&dev_queue->dql) < 0)
3006 return;
3007
3008 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3009 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00003010#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003011}
3012
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003013/**
3014 * netdev_completed_queue - report bytes and packets completed by device
3015 * @dev: network device
3016 * @pkts: actual number of packets sent over the medium
3017 * @bytes: actual number of bytes sent over the medium
3018 *
3019 * Report the number of bytes and packets transmitted by the network device
3020 * hardware queue over the physical medium, @bytes must exactly match the
3021 * @bytes amount passed to netdev_sent_queue()
3022 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003023static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003024 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003025{
3026 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3027}
3028
3029static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3030{
Tom Herbert114cf582011-11-28 16:33:09 +00003031#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003032 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003033 dql_reset(&q->dql);
3034#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003035}
3036
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003037/**
3038 * netdev_reset_queue - reset the packets and bytes count of a network device
3039 * @dev_queue: network device
3040 *
3041 * Reset the bytes and packet count of a network device and clear the
3042 * software flow control OFF bit for this network device
3043 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003044static inline void netdev_reset_queue(struct net_device *dev_queue)
3045{
3046 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003047}
3048
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003049/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003050 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3051 * @dev: network device
3052 * @queue_index: given tx queue index
3053 *
3054 * Returns 0 if given tx queue index >= number of device tx queues,
3055 * otherwise returns the originally passed tx queue index.
3056 */
3057static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3058{
3059 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3060 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3061 dev->name, queue_index,
3062 dev->real_num_tx_queues);
3063 return 0;
3064 }
3065
3066 return queue_index;
3067}
3068
3069/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003070 * netif_running - test if up
3071 * @dev: network device
3072 *
3073 * Test if the device has been brought up.
3074 */
David S. Miller4d295152012-03-07 21:02:35 -05003075static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003076{
3077 return test_bit(__LINK_STATE_START, &dev->state);
3078}
3079
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003080/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003081 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003082 * stop, and a check if it's stopped. All other device management is
3083 * done at the overall netdevice level.
3084 * Also test the device if we're multiqueue.
3085 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003086
3087/**
3088 * netif_start_subqueue - allow sending packets on subqueue
3089 * @dev: network device
3090 * @queue_index: sub queue index
3091 *
3092 * Start individual transmit queue of a device with multiple transmit queues.
3093 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003094static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3095{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003096 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003097
3098 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003099}
3100
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003101/**
3102 * netif_stop_subqueue - stop sending packets on subqueue
3103 * @dev: network device
3104 * @queue_index: sub queue index
3105 *
3106 * Stop individual transmit queue of a device with multiple transmit queues.
3107 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003108static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3109{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003110 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003111 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003112}
3113
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003114/**
3115 * netif_subqueue_stopped - test status of subqueue
3116 * @dev: network device
3117 * @queue_index: sub queue index
3118 *
3119 * Check individual transmit queue of a device with multiple transmit queues.
3120 */
David S. Miller4d295152012-03-07 21:02:35 -05003121static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3122 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003123{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003124 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003125
3126 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003127}
3128
David S. Miller4d295152012-03-07 21:02:35 -05003129static inline bool netif_subqueue_stopped(const struct net_device *dev,
3130 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003131{
3132 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3133}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003134
Florian Fainelli738b35c2017-01-11 21:13:02 -08003135/**
3136 * netif_wake_subqueue - allow sending packets on subqueue
3137 * @dev: network device
3138 * @queue_index: sub queue index
3139 *
3140 * Resume individual transmit queue of a device with multiple transmit queues.
3141 */
3142static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3143{
3144 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3145
3146 netif_tx_wake_queue(txq);
3147}
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003148
Alexander Duyck537c00d2013-01-10 08:57:02 +00003149#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003150int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003151 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003152#else
3153static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003154 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003155 u16 index)
3156{
3157 return 0;
3158}
3159#endif
3160
Jiri Pirko5605c762015-05-12 14:56:12 +02003161u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3162 unsigned int num_tx_queues);
3163
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003164/*
3165 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3166 * as a distribution range limit for the returned value.
3167 */
3168static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003169 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003170{
3171 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3172}
3173
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003174/**
3175 * netif_is_multiqueue - test if device has multiple transmit queues
3176 * @dev: network device
3177 *
3178 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003179 */
David S. Miller4d295152012-03-07 21:02:35 -05003180static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003181{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003182 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003183}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003184
Joe Perchesf629d202013-09-26 14:48:15 -07003185int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003186
Michael Daltona953be52014-01-16 22:23:28 -08003187#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003188int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003189#else
3190static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3191 unsigned int rxq)
3192{
3193 return 0;
3194}
3195#endif
3196
Michael Daltona953be52014-01-16 22:23:28 -08003197#ifdef CONFIG_SYSFS
3198static inline unsigned int get_netdev_rx_queue_index(
3199 struct netdev_rx_queue *queue)
3200{
3201 struct net_device *dev = queue->dev;
3202 int index = queue - dev->_rx;
3203
3204 BUG_ON(index >= dev->num_rx_queues);
3205 return index;
3206}
3207#endif
3208
Yuval Mintz16917b82012-07-01 03:18:50 +00003209#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003210int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003211
Eric Dumazete6247022013-12-05 04:45:08 -08003212enum skb_free_reason {
3213 SKB_REASON_CONSUMED,
3214 SKB_REASON_DROPPED,
3215};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003216
Eric Dumazete6247022013-12-05 04:45:08 -08003217void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3218void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3219
3220/*
3221 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3222 * interrupt context or with hardware interrupts being disabled.
3223 * (in_irq() || irqs_disabled())
3224 *
3225 * We provide four helpers that can be used in following contexts :
3226 *
3227 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3228 * replacing kfree_skb(skb)
3229 *
3230 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3231 * Typically used in place of consume_skb(skb) in TX completion path
3232 *
3233 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3234 * replacing kfree_skb(skb)
3235 *
3236 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3237 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003238 */
Eric Dumazete6247022013-12-05 04:45:08 -08003239static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3240{
3241 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3242}
3243
3244static inline void dev_consume_skb_irq(struct sk_buff *skb)
3245{
3246 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3247}
3248
3249static inline void dev_kfree_skb_any(struct sk_buff *skb)
3250{
3251 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3252}
3253
3254static inline void dev_consume_skb_any(struct sk_buff *skb)
3255{
3256 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3257}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003258
Joe Perchesf629d202013-09-26 14:48:15 -07003259int netif_rx(struct sk_buff *skb);
3260int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003261int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003262gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3263void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3264struct sk_buff *napi_get_frags(struct napi_struct *napi);
3265gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003266struct packet_offload *gro_find_receive_by_type(__be16 type);
3267struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003268
3269static inline void napi_free_frags(struct napi_struct *napi)
3270{
3271 kfree_skb(napi->skb);
3272 napi->skb = NULL;
3273}
3274
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003275bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003276int netdev_rx_handler_register(struct net_device *dev,
3277 rx_handler_func_t *rx_handler,
3278 void *rx_handler_data);
3279void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003280
Joe Perchesf629d202013-09-26 14:48:15 -07003281bool dev_valid_name(const char *name);
3282int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3283int dev_ethtool(struct net *net, struct ifreq *);
3284unsigned int dev_get_flags(const struct net_device *);
3285int __dev_change_flags(struct net_device *, unsigned int flags);
3286int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003287void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3288 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003289int dev_change_name(struct net_device *, const char *);
3290int dev_set_alias(struct net_device *, const char *, size_t);
3291int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3292int dev_set_mtu(struct net_device *, int);
3293void dev_set_group(struct net_device *, int);
3294int dev_set_mac_address(struct net_device *, struct sockaddr *);
3295int dev_change_carrier(struct net_device *, bool new_carrier);
3296int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003297 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003298int dev_get_phys_port_name(struct net_device *dev,
3299 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003300int dev_change_proto_down(struct net_device *dev, bool proto_down);
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);
Daniel Borkmannd67b9cd2017-05-12 01:04:46 +02003304
3305typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3306int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3307 int fd, u32 flags);
Martin KaFai Lau58038692017-06-15 17:29:09 -07003308bool __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
Daniel Borkmannd67b9cd2017-05-12 01:04:46 +02003309
Herbert Xua0265d22014-04-17 13:45:03 +08003310int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003311int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003312bool is_skb_forwardable(const struct net_device *dev,
3313 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003314
Martin KaFai Lau4e3264d2016-11-09 15:36:33 -08003315static __always_inline int ____dev_forward_skb(struct net_device *dev,
3316 struct sk_buff *skb)
3317{
3318 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3319 unlikely(!is_skb_forwardable(dev, skb))) {
3320 atomic_long_inc(&dev->rx_dropped);
3321 kfree_skb(skb);
3322 return NET_RX_DROP;
3323 }
3324
3325 skb_scrub_packet(skb, true);
3326 skb->priority = 0;
3327 return 0;
3328}
3329
David Ahern74b20582016-05-10 11:19:50 -07003330void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3331
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003332extern int netdev_budget;
Matthew Whitehead7acf8a12017-04-19 12:37:10 -04003333extern unsigned int netdev_budget_usecs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003334
3335/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003336void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003337
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003338/**
3339 * dev_put - release reference to device
3340 * @dev: network device
3341 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003342 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003343 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003344static inline void dev_put(struct net_device *dev)
3345{
Christoph Lameter933393f2011-12-22 11:58:51 -06003346 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003347}
3348
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003349/**
3350 * dev_hold - get reference to device
3351 * @dev: network device
3352 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003353 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003354 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003355static inline void dev_hold(struct net_device *dev)
3356{
Christoph Lameter933393f2011-12-22 11:58:51 -06003357 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003358}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003359
3360/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3361 * and _off may be called from IRQ context, but it is caller
3362 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003363 *
3364 * The name carrier is inappropriate, these functions should really be
3365 * called netif_lowerlayer_*() because they represent the state of any
3366 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003367 */
3368
Joe Perchesf629d202013-09-26 14:48:15 -07003369void linkwatch_init_dev(struct net_device *dev);
3370void linkwatch_fire_event(struct net_device *dev);
3371void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003372
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003373/**
3374 * netif_carrier_ok - test if carrier present
3375 * @dev: network device
3376 *
3377 * Check if carrier is present on device
3378 */
David S. Miller4d295152012-03-07 21:02:35 -05003379static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003380{
3381 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3382}
3383
Joe Perchesf629d202013-09-26 14:48:15 -07003384unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003385
Joe Perchesf629d202013-09-26 14:48:15 -07003386void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003387
Joe Perchesf629d202013-09-26 14:48:15 -07003388void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003389
Joe Perchesf629d202013-09-26 14:48:15 -07003390void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003391
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003392/**
3393 * netif_dormant_on - mark device as dormant.
3394 * @dev: network device
3395 *
3396 * Mark device as dormant (as per RFC2863).
3397 *
3398 * The dormant state indicates that the relevant interface is not
3399 * actually in a condition to pass packets (i.e., it is not 'up') but is
3400 * in a "pending" state, waiting for some external event. For "on-
3401 * demand" interfaces, this new state identifies the situation where the
3402 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003403 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003404static inline void netif_dormant_on(struct net_device *dev)
3405{
3406 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3407 linkwatch_fire_event(dev);
3408}
3409
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003410/**
3411 * netif_dormant_off - set device as not dormant.
3412 * @dev: network device
3413 *
3414 * Device is not in dormant state.
3415 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003416static inline void netif_dormant_off(struct net_device *dev)
3417{
3418 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3419 linkwatch_fire_event(dev);
3420}
3421
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003422/**
Zhang Shengju8ecbc402017-04-26 11:05:12 +08003423 * netif_dormant - test if device is dormant
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003424 * @dev: network device
3425 *
Zhang Shengju8ecbc402017-04-26 11:05:12 +08003426 * Check if device is dormant.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003427 */
David S. Miller4d295152012-03-07 21:02:35 -05003428static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003429{
3430 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3431}
3432
3433
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003434/**
3435 * netif_oper_up - test if device is operational
3436 * @dev: network device
3437 *
3438 * Check if carrier is operational
3439 */
David S. Miller4d295152012-03-07 21:02:35 -05003440static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003441{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003442 return (dev->operstate == IF_OPER_UP ||
3443 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3444}
3445
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003446/**
3447 * netif_device_present - is device available or removed
3448 * @dev: network device
3449 *
3450 * Check if device has not been removed from system.
3451 */
David S. Miller4d295152012-03-07 21:02:35 -05003452static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003453{
3454 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3455}
3456
Joe Perchesf629d202013-09-26 14:48:15 -07003457void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003458
Joe Perchesf629d202013-09-26 14:48:15 -07003459void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003460
3461/*
3462 * Network interface message level settings
3463 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003464
3465enum {
3466 NETIF_MSG_DRV = 0x0001,
3467 NETIF_MSG_PROBE = 0x0002,
3468 NETIF_MSG_LINK = 0x0004,
3469 NETIF_MSG_TIMER = 0x0008,
3470 NETIF_MSG_IFDOWN = 0x0010,
3471 NETIF_MSG_IFUP = 0x0020,
3472 NETIF_MSG_RX_ERR = 0x0040,
3473 NETIF_MSG_TX_ERR = 0x0080,
3474 NETIF_MSG_TX_QUEUED = 0x0100,
3475 NETIF_MSG_INTR = 0x0200,
3476 NETIF_MSG_TX_DONE = 0x0400,
3477 NETIF_MSG_RX_STATUS = 0x0800,
3478 NETIF_MSG_PKTDATA = 0x1000,
3479 NETIF_MSG_HW = 0x2000,
3480 NETIF_MSG_WOL = 0x4000,
3481};
3482
3483#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3484#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3485#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3486#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3487#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3488#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3489#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3490#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3491#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3492#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3493#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3494#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3495#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3496#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3497#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3498
3499static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3500{
3501 /* use default */
3502 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3503 return default_msg_enable_bits;
3504 if (debug_value == 0) /* no output */
3505 return 0;
3506 /* set low N bits */
3507 return (1 << debug_value) - 1;
3508}
3509
David S. Millerc773e842008-07-08 23:13:53 -07003510static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003511{
David S. Millerc773e842008-07-08 23:13:53 -07003512 spin_lock(&txq->_xmit_lock);
3513 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003514}
3515
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003516static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3517{
3518 __acquire(&txq->_xmit_lock);
3519 return true;
3520}
3521
3522static inline void __netif_tx_release(struct netdev_queue *txq)
3523{
3524 __release(&txq->_xmit_lock);
3525}
3526
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003527static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3528{
3529 spin_lock_bh(&txq->_xmit_lock);
3530 txq->xmit_lock_owner = smp_processor_id();
3531}
3532
David S. Miller4d295152012-03-07 21:02:35 -05003533static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003534{
David S. Miller4d295152012-03-07 21:02:35 -05003535 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003536 if (likely(ok))
3537 txq->xmit_lock_owner = smp_processor_id();
3538 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003539}
3540
David S. Millerc773e842008-07-08 23:13:53 -07003541static inline void __netif_tx_unlock(struct netdev_queue *txq)
3542{
3543 txq->xmit_lock_owner = -1;
3544 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003545}
3546
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003547static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3548{
3549 txq->xmit_lock_owner = -1;
3550 spin_unlock_bh(&txq->_xmit_lock);
3551}
3552
Eric Dumazet08baf562009-05-25 22:58:01 -07003553static inline void txq_trans_update(struct netdev_queue *txq)
3554{
3555 if (txq->xmit_lock_owner != -1)
3556 txq->trans_start = jiffies;
3557}
3558
Florian Westphalba162f82016-05-03 16:31:00 +02003559/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3560static inline void netif_trans_update(struct net_device *dev)
3561{
Florian Westphal9b366272016-05-03 16:33:14 +02003562 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3563
3564 if (txq->trans_start != jiffies)
3565 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003566}
3567
David S. Millerc3f26a22008-07-31 16:58:50 -07003568/**
3569 * netif_tx_lock - grab network device transmit lock
3570 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003571 *
3572 * Get network device transmit lock
3573 */
3574static inline void netif_tx_lock(struct net_device *dev)
3575{
3576 unsigned int i;
3577 int cpu;
3578
3579 spin_lock(&dev->tx_global_lock);
3580 cpu = smp_processor_id();
3581 for (i = 0; i < dev->num_tx_queues; i++) {
3582 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3583
3584 /* We are the only thread of execution doing a
3585 * freeze, but we have to grab the _xmit_lock in
3586 * order to synchronize with threads which are in
3587 * the ->hard_start_xmit() handler and already
3588 * checked the frozen bit.
3589 */
3590 __netif_tx_lock(txq, cpu);
3591 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3592 __netif_tx_unlock(txq);
3593 }
3594}
3595
3596static inline void netif_tx_lock_bh(struct net_device *dev)
3597{
3598 local_bh_disable();
3599 netif_tx_lock(dev);
3600}
3601
Herbert Xu932ff272006-06-09 12:20:56 -07003602static inline void netif_tx_unlock(struct net_device *dev)
3603{
David S. Millere8a04642008-07-17 00:34:19 -07003604 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003605
David S. Millere8a04642008-07-17 00:34:19 -07003606 for (i = 0; i < dev->num_tx_queues; i++) {
3607 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003608
David S. Millerc3f26a22008-07-31 16:58:50 -07003609 /* No need to grab the _xmit_lock here. If the
3610 * queue is not stopped for another reason, we
3611 * force a schedule.
3612 */
3613 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003614 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003615 }
3616 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003617}
3618
3619static inline void netif_tx_unlock_bh(struct net_device *dev)
3620{
David S. Millere8a04642008-07-17 00:34:19 -07003621 netif_tx_unlock(dev);
3622 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003623}
3624
David S. Millerc773e842008-07-08 23:13:53 -07003625#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003626 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003627 __netif_tx_lock(txq, cpu); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003628 } else { \
3629 __netif_tx_acquire(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003630 } \
3631}
3632
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003633#define HARD_TX_TRYLOCK(dev, txq) \
3634 (((dev->features & NETIF_F_LLTX) == 0) ? \
3635 __netif_tx_trylock(txq) : \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003636 __netif_tx_acquire(txq))
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003637
David S. Millerc773e842008-07-08 23:13:53 -07003638#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003639 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003640 __netif_tx_unlock(txq); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003641 } else { \
3642 __netif_tx_release(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003643 } \
3644}
3645
Linus Torvalds1da177e2005-04-16 15:20:36 -07003646static inline void netif_tx_disable(struct net_device *dev)
3647{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003648 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003649 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003650
David S. Millerc3f26a22008-07-31 16:58:50 -07003651 local_bh_disable();
3652 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003653 for (i = 0; i < dev->num_tx_queues; i++) {
3654 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003655
3656 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003657 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003658 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003659 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003660 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003661}
3662
David S. Millere308a5d2008-07-15 00:13:44 -07003663static inline void netif_addr_lock(struct net_device *dev)
3664{
3665 spin_lock(&dev->addr_list_lock);
3666}
3667
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003668static inline void netif_addr_lock_nested(struct net_device *dev)
3669{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003670 int subclass = SINGLE_DEPTH_NESTING;
3671
3672 if (dev->netdev_ops->ndo_get_lock_subclass)
3673 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3674
3675 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003676}
3677
David S. Millere308a5d2008-07-15 00:13:44 -07003678static inline void netif_addr_lock_bh(struct net_device *dev)
3679{
3680 spin_lock_bh(&dev->addr_list_lock);
3681}
3682
3683static inline void netif_addr_unlock(struct net_device *dev)
3684{
3685 spin_unlock(&dev->addr_list_lock);
3686}
3687
3688static inline void netif_addr_unlock_bh(struct net_device *dev)
3689{
3690 spin_unlock_bh(&dev->addr_list_lock);
3691}
3692
Jiri Pirkof001fde2009-05-05 02:48:28 +00003693/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003694 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003695 * rcu_read_lock held.
3696 */
3697#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003698 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003699
Linus Torvalds1da177e2005-04-16 15:20:36 -07003700/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3701
Joe Perchesf629d202013-09-26 14:48:15 -07003702void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003703
3704/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003705struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003706 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003707 void (*setup)(struct net_device *),
3708 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003709#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3710 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003711
Tom Gundersenc835a672014-07-14 16:37:24 +02003712#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3713 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3714 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003715
Joe Perchesf629d202013-09-26 14:48:15 -07003716int register_netdev(struct net_device *dev);
3717void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003718
Jiri Pirko22bedad32010-04-01 21:22:57 +00003719/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003720int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3721 struct netdev_hw_addr_list *from_list, int addr_len);
3722void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3723 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003724int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3725 struct net_device *dev,
3726 int (*sync)(struct net_device *, const unsigned char *),
3727 int (*unsync)(struct net_device *,
3728 const unsigned char *));
3729void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3730 struct net_device *dev,
3731 int (*unsync)(struct net_device *,
3732 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003733void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003734
Jiri Pirkof001fde2009-05-05 02:48:28 +00003735/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003736int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3737 unsigned char addr_type);
3738int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3739 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003740void dev_addr_flush(struct net_device *dev);
3741int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003742
3743/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003744int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3745int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3746int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3747int dev_uc_sync(struct net_device *to, struct net_device *from);
3748int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3749void dev_uc_unsync(struct net_device *to, struct net_device *from);
3750void dev_uc_flush(struct net_device *dev);
3751void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003752
Alexander Duyck670e5b82014-05-28 18:44:46 -07003753/**
3754 * __dev_uc_sync - Synchonize device's unicast list
3755 * @dev: device to sync
3756 * @sync: function to call if address should be added
3757 * @unsync: function to call if address should be removed
3758 *
3759 * Add newly added addresses to the interface, and release
3760 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003761 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003762static inline int __dev_uc_sync(struct net_device *dev,
3763 int (*sync)(struct net_device *,
3764 const unsigned char *),
3765 int (*unsync)(struct net_device *,
3766 const unsigned char *))
3767{
3768 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3769}
3770
3771/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003772 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003773 * @dev: device to sync
3774 * @unsync: function to call if address should be removed
3775 *
3776 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003777 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003778static inline void __dev_uc_unsync(struct net_device *dev,
3779 int (*unsync)(struct net_device *,
3780 const unsigned char *))
3781{
3782 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3783}
3784
Jiri Pirko22bedad32010-04-01 21:22:57 +00003785/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003786int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3787int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3788int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3789int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3790int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3791int dev_mc_sync(struct net_device *to, struct net_device *from);
3792int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3793void dev_mc_unsync(struct net_device *to, struct net_device *from);
3794void dev_mc_flush(struct net_device *dev);
3795void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003796
Alexander Duyck670e5b82014-05-28 18:44:46 -07003797/**
3798 * __dev_mc_sync - Synchonize device's multicast list
3799 * @dev: device to sync
3800 * @sync: function to call if address should be added
3801 * @unsync: function to call if address should be removed
3802 *
3803 * Add newly added addresses to the interface, and release
3804 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003805 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003806static inline int __dev_mc_sync(struct net_device *dev,
3807 int (*sync)(struct net_device *,
3808 const unsigned char *),
3809 int (*unsync)(struct net_device *,
3810 const unsigned char *))
3811{
3812 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3813}
3814
3815/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003816 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003817 * @dev: device to sync
3818 * @unsync: function to call if address should be removed
3819 *
3820 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003821 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003822static inline void __dev_mc_unsync(struct net_device *dev,
3823 int (*unsync)(struct net_device *,
3824 const unsigned char *))
3825{
3826 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3827}
3828
Linus Torvalds1da177e2005-04-16 15:20:36 -07003829/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003830void dev_set_rx_mode(struct net_device *dev);
3831void __dev_set_rx_mode(struct net_device *dev);
3832int dev_set_promiscuity(struct net_device *dev, int inc);
3833int dev_set_allmulti(struct net_device *dev, int inc);
3834void netdev_state_change(struct net_device *dev);
3835void netdev_notify_peers(struct net_device *dev);
3836void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003837/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003838void dev_load(struct net *net, const char *name);
3839struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3840 struct rtnl_link_stats64 *storage);
3841void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3842 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003843
3844extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003845extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003846extern int weight_p;
Matthias Tafelmeier3d48b532016-12-29 21:37:21 +01003847extern int dev_weight_rx_bias;
3848extern int dev_weight_tx_bias;
3849extern int dev_rx_weight;
3850extern int dev_tx_weight;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003851
Joe Perchesf629d202013-09-26 14:48:15 -07003852bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003853struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3854 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003855struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3856 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003857
3858/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003859#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3860 for (iter = &(dev)->adj_list.upper, \
3861 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3862 updev; \
3863 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3864
David Ahern1a3f0602016-10-17 19:15:44 -07003865int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3866 int (*fn)(struct net_device *upper_dev,
3867 void *data),
3868 void *data);
3869
3870bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3871 struct net_device *upper_dev);
3872
Joe Perchesf629d202013-09-26 14:48:15 -07003873void *netdev_lower_get_next_private(struct net_device *dev,
3874 struct list_head **iter);
3875void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3876 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003877
3878#define netdev_for_each_lower_private(dev, priv, iter) \
3879 for (iter = (dev)->adj_list.lower.next, \
3880 priv = netdev_lower_get_next_private(dev, &(iter)); \
3881 priv; \
3882 priv = netdev_lower_get_next_private(dev, &(iter)))
3883
3884#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3885 for (iter = &(dev)->adj_list.lower, \
3886 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3887 priv; \
3888 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3889
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003890void *netdev_lower_get_next(struct net_device *dev,
3891 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003892
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003893#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003894 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003895 ldev = netdev_lower_get_next(dev, &(iter)); \
3896 ldev; \
3897 ldev = netdev_lower_get_next(dev, &(iter)))
3898
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003899struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3900 struct list_head **iter);
3901struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3902 struct list_head **iter);
3903
David Ahern1a3f0602016-10-17 19:15:44 -07003904int netdev_walk_all_lower_dev(struct net_device *dev,
3905 int (*fn)(struct net_device *lower_dev,
3906 void *data),
3907 void *data);
3908int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3909 int (*fn)(struct net_device *lower_dev,
3910 void *data),
3911 void *data);
3912
Joe Perchesf629d202013-09-26 14:48:15 -07003913void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003914void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003915struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3916struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3917int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3918int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003919 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003920 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003921void netdev_upper_dev_unlink(struct net_device *dev,
3922 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003923void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003924void *netdev_lower_dev_get_private(struct net_device *dev,
3925 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003926void netdev_lower_state_changed(struct net_device *lower_dev,
3927 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003928
3929/* RSS keys are 40 or 52 bytes long */
3930#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003931extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003932void netdev_rss_key_fill(void *buffer, size_t len);
3933
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003934int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003935int skb_checksum_help(struct sk_buff *skb);
Davide Carattib72b5bf2017-05-18 15:44:38 +02003936int skb_crc32c_csum_help(struct sk_buff *skb);
Davide Caratti43c26a12017-05-18 15:44:41 +02003937int skb_csum_hwoffload_help(struct sk_buff *skb,
3938 const netdev_features_t features);
3939
Joe Perchesf629d202013-09-26 14:48:15 -07003940struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3941 netdev_features_t features, bool tx_path);
3942struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3943 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003944
Moni Shoua61bd3852015-02-03 16:48:29 +02003945struct netdev_bonding_info {
3946 ifslave slave;
3947 ifbond master;
3948};
3949
3950struct netdev_notifier_bonding_info {
3951 struct netdev_notifier_info info; /* must be first */
3952 struct netdev_bonding_info bonding_info;
3953};
3954
3955void netdev_bonding_info_change(struct net_device *dev,
3956 struct netdev_bonding_info *bonding_info);
3957
Cong Wang12b00042013-02-05 16:36:38 +00003958static inline
3959struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3960{
3961 return __skb_gso_segment(skb, features, true);
3962}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003963__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003964
3965static inline bool can_checksum_protocol(netdev_features_t features,
3966 __be16 protocol)
3967{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003968 if (protocol == htons(ETH_P_FCOE))
3969 return !!(features & NETIF_F_FCOE_CRC);
3970
3971 /* Assume this is an IP checksum (not SCTP CRC) */
3972
3973 if (features & NETIF_F_HW_CSUM) {
3974 /* Can checksum everything */
3975 return true;
3976 }
3977
3978 switch (protocol) {
3979 case htons(ETH_P_IP):
3980 return !!(features & NETIF_F_IP_CSUM);
3981 case htons(ETH_P_IPV6):
3982 return !!(features & NETIF_F_IPV6_CSUM);
3983 default:
3984 return false;
3985 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003986}
Cong Wang12b00042013-02-05 16:36:38 +00003987
Herbert Xufb286bb2005-11-10 13:01:24 -08003988#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003989void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003990#else
3991static inline void netdev_rx_csum_fault(struct net_device *dev)
3992{
3993}
3994#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003995/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003996void net_enable_timestamp(void);
3997void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003998
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003999#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07004000int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00004001#else
4002#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004003#endif
4004
David S. Miller47982482014-08-22 16:21:53 -07004005static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004006 struct sk_buff *skb, struct net_device *dev,
4007 bool more)
David S. Miller47982482014-08-22 16:21:53 -07004008{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004009 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07004010 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07004011}
4012
David S. Miller10b3ad82014-08-29 21:07:24 -07004013static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004014 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07004015{
4016 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07004017 int rc;
David S. Miller47982482014-08-22 16:21:53 -07004018
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004019 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07004020 if (rc == NETDEV_TX_OK)
4021 txq_trans_update(txq);
4022
4023 return rc;
David S. Miller47982482014-08-22 16:21:53 -07004024}
4025
Linus Torvalds42a2d922013-11-13 17:40:34 +09004026int netdev_class_create_file_ns(struct class_attribute *class_attr,
4027 const void *ns);
4028void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4029 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04004030
4031static inline int netdev_class_create_file(struct class_attribute *class_attr)
4032{
4033 return netdev_class_create_file_ns(class_attr, NULL);
4034}
4035
4036static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4037{
4038 netdev_class_remove_file_ns(class_attr, NULL);
4039}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004040
Johannes Berg04600792010-08-05 17:45:15 +02004041extern struct kobj_ns_type_operations net_ns_type_operations;
4042
Joe Perchesf629d202013-09-26 14:48:15 -07004043const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004044
Joe Perchesf629d202013-09-26 14:48:15 -07004045void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004046
Michal Kubečekda081432014-05-20 08:29:25 +02004047static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4048 netdev_features_t f2)
4049{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004050 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4051 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004052 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004053 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004054 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004055 }
Michal Kubečekda081432014-05-20 08:29:25 +02004056
Tom Herbertc8cd0982015-12-14 11:19:44 -08004057 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004058}
4059
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004060static inline netdev_features_t netdev_get_wanted_features(
4061 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004062{
4063 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4064}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004065netdev_features_t netdev_increment_features(netdev_features_t all,
4066 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004067
4068/* Allow TSO being used on stacked device :
4069 * Performing the GSO segmentation before last device
4070 * is a performance improvement.
4071 */
4072static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4073 netdev_features_t mask)
4074{
4075 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4076}
4077
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004078int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004079void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004080void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004081
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004082void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4083 struct net_device *dev);
4084
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004085netdev_features_t passthru_features_check(struct sk_buff *skb,
4086 struct net_device *dev,
4087 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004088netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004089
David S. Miller4d295152012-03-07 21:02:35 -05004090static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004091{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004092 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004093
4094 /* check flags correspondence */
4095 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4096 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4097 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4098 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004099 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004100 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4101 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004102 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4103 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004104 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4105 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004106 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4107 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 -04004108 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004109 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004110 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Steffen Klassertc7ef8f02017-04-14 10:05:36 +02004111 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004112
Herbert Xubcd76112006-06-30 13:36:35 -07004113 return (features & feature) == feature;
4114}
4115
David S. Miller4d295152012-03-07 21:02:35 -05004116static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004117{
Herbert Xu278b2512009-06-03 21:20:51 -07004118 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004119 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004120}
4121
Johannes Berg8b86a612015-04-17 15:45:04 +02004122static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004123 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004124{
Jesse Grossfc741212011-01-09 06:23:32 +00004125 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004126 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4127 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004128}
4129
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004130static inline void netif_set_gso_max_size(struct net_device *dev,
4131 unsigned int size)
4132{
4133 dev->gso_max_size = size;
4134}
4135
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004136static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4137 int pulled_hlen, u16 mac_offset,
4138 int mac_len)
4139{
4140 skb->protocol = protocol;
4141 skb->encapsulation = 1;
4142 skb_push(skb, pulled_hlen);
4143 skb_reset_transport_header(skb);
4144 skb->mac_header = mac_offset;
4145 skb->network_header = skb->mac_header + mac_len;
4146 skb->mac_len = mac_len;
4147}
4148
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004149static inline bool netif_is_macsec(const struct net_device *dev)
4150{
4151 return dev->priv_flags & IFF_MACSEC;
4152}
4153
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004154static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004155{
4156 return dev->priv_flags & IFF_MACVLAN;
4157}
4158
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004159static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004160{
4161 return dev->priv_flags & IFF_MACVLAN_PORT;
4162}
4163
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004164static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004165{
4166 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4167}
4168
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004169static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004170{
4171 return dev->priv_flags & IFF_IPVLAN_MASTER;
4172}
4173
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004174static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004175{
4176 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4177}
4178
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004179static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004180{
4181 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4182}
4183
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004184static inline bool netif_supports_nofcs(struct net_device *dev)
4185{
4186 return dev->priv_flags & IFF_SUPP_NOFCS;
4187}
4188
David Ahern007979e2015-09-29 20:07:10 -07004189static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004190{
David Ahern007979e2015-09-29 20:07:10 -07004191 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004192}
4193
David Ahernfee6d4c2015-10-05 08:51:24 -07004194static inline bool netif_is_l3_slave(const struct net_device *dev)
4195{
4196 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4197}
4198
Jiri Pirko0894ae32015-08-27 09:31:19 +02004199static inline bool netif_is_bridge_master(const struct net_device *dev)
4200{
4201 return dev->priv_flags & IFF_EBRIDGE;
4202}
4203
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004204static inline bool netif_is_bridge_port(const struct net_device *dev)
4205{
4206 return dev->priv_flags & IFF_BRIDGE_PORT;
4207}
4208
Jiri Pirko35d4e172015-08-27 09:31:20 +02004209static inline bool netif_is_ovs_master(const struct net_device *dev)
4210{
4211 return dev->priv_flags & IFF_OPENVSWITCH;
4212}
4213
Jiri Pirko5be66142017-04-18 16:55:36 +02004214static inline bool netif_is_ovs_port(const struct net_device *dev)
4215{
4216 return dev->priv_flags & IFF_OVS_DATAPATH;
4217}
4218
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004219static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004220{
4221 return dev->priv_flags & IFF_TEAM;
4222}
4223
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004224static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004225{
4226 return dev->priv_flags & IFF_TEAM_PORT;
4227}
4228
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004229static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004230{
4231 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4232}
4233
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004234static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004235{
4236 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4237}
4238
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004239static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4240{
4241 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4242}
4243
Eric Dumazet02875872014-10-05 18:38:35 -07004244/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4245static inline void netif_keep_dst(struct net_device *dev)
4246{
4247 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4248}
4249
Paolo Abeni18d3df32016-07-14 18:00:10 +02004250/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4251static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4252{
4253 /* TODO: reserve and use an additional IFF bit, if we get more users */
4254 return dev->priv_flags & IFF_MACSEC;
4255}
4256
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004257extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004258
Joe Perches571ba422010-02-09 11:49:47 +00004259/* Logging, debugging and troubleshooting/diagnostic helpers. */
4260
4261/* netdev_printk helpers, similar to dev_printk */
4262
4263static inline const char *netdev_name(const struct net_device *dev)
4264{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004265 if (!dev->name[0] || strchr(dev->name, '%'))
4266 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004267 return dev->name;
4268}
4269
David Ahern8397ed32017-06-07 12:26:23 -06004270static inline bool netdev_unregistering(const struct net_device *dev)
4271{
4272 return dev->reg_state == NETREG_UNREGISTERING;
4273}
4274
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004275static inline const char *netdev_reg_state(const struct net_device *dev)
4276{
4277 switch (dev->reg_state) {
4278 case NETREG_UNINITIALIZED: return " (uninitialized)";
4279 case NETREG_REGISTERED: return "";
4280 case NETREG_UNREGISTERING: return " (unregistering)";
4281 case NETREG_UNREGISTERED: return " (unregistered)";
4282 case NETREG_RELEASED: return " (released)";
4283 case NETREG_DUMMY: return " (dummy)";
4284 }
4285
4286 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4287 return " (unknown)";
4288}
4289
Joe Perchesf629d202013-09-26 14:48:15 -07004290__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004291void netdev_printk(const char *level, const struct net_device *dev,
4292 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004293__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004294void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004295__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004296void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004297__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004298void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004299__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004300void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004301__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004302void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004303__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004304void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004305__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004306void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004307
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004308#define MODULE_ALIAS_NETDEV(device) \
4309 MODULE_ALIAS("netdev-" device)
4310
Jim Cromieb558c962011-12-19 17:11:18 -05004311#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004312#define netdev_dbg(__dev, format, args...) \
4313do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004314 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004315} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004316#elif defined(DEBUG)
4317#define netdev_dbg(__dev, format, args...) \
4318 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004319#else
4320#define netdev_dbg(__dev, format, args...) \
4321({ \
4322 if (0) \
4323 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004324})
4325#endif
4326
4327#if defined(VERBOSE_DEBUG)
4328#define netdev_vdbg netdev_dbg
4329#else
4330
4331#define netdev_vdbg(dev, format, args...) \
4332({ \
4333 if (0) \
4334 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4335 0; \
4336})
4337#endif
4338
4339/*
4340 * netdev_WARN() acts like dev_printk(), but with the key difference
4341 * of using a WARN/WARN_ON to get the message out, including the
4342 * file/line information and a backtrace.
4343 */
4344#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004345 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4346 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004347
Joe Perchesb3d95c52010-02-09 11:49:49 +00004348/* netif printk helpers, similar to netdev_printk */
4349
4350#define netif_printk(priv, type, level, dev, fmt, args...) \
4351do { \
4352 if (netif_msg_##type(priv)) \
4353 netdev_printk(level, (dev), fmt, ##args); \
4354} while (0)
4355
Joe Perchesf45f4322010-06-27 01:02:36 +00004356#define netif_level(level, priv, type, dev, fmt, args...) \
4357do { \
4358 if (netif_msg_##type(priv)) \
4359 netdev_##level(dev, fmt, ##args); \
4360} while (0)
4361
Joe Perchesb3d95c52010-02-09 11:49:49 +00004362#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004363 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004364#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004365 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004366#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004367 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004368#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004369 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004370#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004371 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004372#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004373 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004374#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004375 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004376
Joe Perches0053ea92012-05-30 07:43:34 +00004377#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004378#define netif_dbg(priv, type, netdev, format, args...) \
4379do { \
4380 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004381 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004382} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004383#elif defined(DEBUG)
4384#define netif_dbg(priv, type, dev, format, args...) \
4385 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004386#else
4387#define netif_dbg(priv, type, dev, format, args...) \
4388({ \
4389 if (0) \
4390 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4391 0; \
4392})
4393#endif
4394
Edward Creef617f272017-01-27 15:02:26 +00004395/* if @cond then downgrade to debug, else print at @level */
4396#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4397 do { \
4398 if (cond) \
4399 netif_dbg(priv, type, netdev, fmt, ##args); \
4400 else \
4401 netif_ ## level(priv, type, netdev, fmt, ##args); \
4402 } while (0)
4403
Joe Perchesb3d95c52010-02-09 11:49:49 +00004404#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004405#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004406#else
4407#define netif_vdbg(priv, type, dev, format, args...) \
4408({ \
4409 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004410 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004411 0; \
4412})
4413#endif
Joe Perches571ba422010-02-09 11:49:47 +00004414
Cong Wang900ff8c2013-02-18 19:20:33 +00004415/*
4416 * The list of packet types we will receive (as opposed to discard)
4417 * and the routines to invoke.
4418 *
4419 * Why 16. Because with 16 the only overlap we get on a hash of the
4420 * low nibble of the protocol value is RARP/SNAP/X.25.
4421 *
4422 * NOTE: That is no longer true with the addition of VLAN tags. Not
4423 * sure which should go first, but I bet it won't make much
4424 * difference if we are running VLANs. The good news is that
4425 * this protocol won't be in the list unless compiled in, so
4426 * the average user (w/out VLANs) will not be adversely affected.
4427 * --BLG
4428 *
4429 * 0800 IP
4430 * 8100 802.1Q VLAN
4431 * 0001 802.3
4432 * 0002 AX.25
4433 * 0004 802.2
4434 * 8035 RARP
4435 * 0005 SNAP
4436 * 0805 X.25
4437 * 0806 ARP
4438 * 8137 IPX
4439 * 0009 Localtalk
4440 * 86DD IPv6
4441 */
4442#define PTYPE_HASH_SIZE (16)
4443#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4444
Jiri Pirko385a1542009-05-27 15:48:07 -07004445#endif /* _LINUX_NETDEVICE_H */