blob: 03b1049082359b1eba3ee87bde49bcfc799bc356 [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;
John Fastabend814abfa2017-07-17 09:27:07 -070069struct xdp_buff;
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
Joe Perchesf629d202013-09-26 14:48:15 -070071void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000073
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000074/* Backlog congestion levels */
75#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76#define NET_RX_DROP 1 /* packet dropped */
77
Patrick McHardy572a9d72009-11-10 06:14:14 +000078/*
79 * Transmit return codes: transmit return codes originate from three different
80 * namespaces:
81 *
82 * - qdisc return codes
83 * - driver transmit return codes
84 * - errno values
85 *
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -050090 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
Patrick McHardy572a9d72009-11-10 06:14:14 +000092 * others are propagated to higher layers.
93 */
94
95/* qdisc ->enqueue() return codes. */
96#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000097#define NET_XMIT_DROP 0x01 /* skb dropped */
98#define NET_XMIT_CN 0x02 /* congestion notification */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000099#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -0200101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +0000104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000107/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000108#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000109
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000110enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000114};
115typedef enum netdev_tx netdev_tx_t;
116
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000117/*
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 */
121static inline bool dev_xmit_complete(int rc)
122{
123 /*
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 */
129 if (likely(rc < NET_XMIT_MASK))
130 return true;
131
132 return false;
133}
134
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500136 * Compute the worst-case header length according to the protocols
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 * used.
138 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800139
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800140#if defined(CONFIG_HYPERV_NET)
141# define LL_MAX_HEADER 128
142#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700143# if defined(CONFIG_MAC80211_MESH)
144# define LL_MAX_HEADER 128
145# else
146# define LL_MAX_HEADER 96
147# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700149# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150#endif
151
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000152#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154#define MAX_HEADER LL_MAX_HEADER
155#else
156#define MAX_HEADER (LL_MAX_HEADER + 48)
157#endif
158
159/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800163
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800164struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
188};
189
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190
191#include <linux/cache.h>
192#include <linux/skbuff.h>
193
Eric Dumazetadc93002011-11-17 03:13:26 +0000194#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100195#include <linux/static_key.h>
196extern struct static_key rps_needed;
Eric Dumazet13bfff22016-12-07 08:29:10 -0800197extern struct static_key rfs_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000198#endif
199
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200struct neighbour;
201struct neigh_parms;
202struct sk_buff;
203
Jiri Pirkof001fde2009-05-05 02:48:28 +0000204struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000208#define NETDEV_HW_ADDR_T_LAN 1
209#define NETDEV_HW_ADDR_T_SAN 2
210#define NETDEV_HW_ADDR_T_SLAVE 3
211#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000212#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000213 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000214 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700215 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400216 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000217 struct rcu_head rcu_head;
218};
219
Jiri Pirko31278e72009-06-17 01:12:19 +0000220struct netdev_hw_addr_list {
221 struct list_head list;
222 int count;
223};
224
Jiri Pirko22bedad32010-04-01 21:22:57 +0000225#define netdev_hw_addr_list_count(l) ((l)->count)
226#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227#define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
229
230#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800232#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800234
Jiri Pirko22bedad32010-04-01 21:22:57 +0000235#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700237#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece32010-02-04 10:22:25 -0800239
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800240struct hh_cache {
Alexey Dobriyan5b3dc2f2017-04-10 11:11:17 +0300241 unsigned int hh_len;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800242 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
244 /* cached hardware header; allow for machine alignment needs. */
245#define HH_DATA_MOD 16
246#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248#define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251};
252
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500253/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700266struct header_ops {
267 int (*create) (struct sk_buff *skb, struct net_device *dev,
268 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000269 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500275 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700276};
277
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278/* These flag bits are private to the generic network queueing
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500279 * layer; they may not be explicitly referenced by any other
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 * code.
281 */
282
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800283enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 __LINK_STATE_START,
285 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800287 __LINK_STATE_LINKWATCH_PENDING,
288 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289};
290
291
292/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500293 * This structure holds boot-time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800294 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 */
296struct netdev_boot_setup {
297 char name[IFNAMSIZ];
298 struct ifmap map;
299};
300#define NETDEV_BOOT_SETUP_MAX 8
301
Joe Perchesf629d202013-09-26 14:48:15 -0700302int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
304/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700305 * Structure for NAPI scheduling similar to tasklet but with weighting
306 */
307struct napi_struct {
308 /* The poll_list must only be managed by the entity which
309 * changes the state of the NAPI_STATE_SCHED bit. This means
310 * whoever atomically sets that bit can add this napi_struct
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500311 * to the per-CPU poll_list, and whoever clears that bit
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700312 * can remove from the list right before clearing the bit.
313 */
314 struct list_head poll_list;
315
316 unsigned long state;
317 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000318 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700319 int (*poll)(struct napi_struct *, int);
320#ifdef CONFIG_NETPOLL
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700321 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700322#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800323 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800324 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800325 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800326 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000327 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300328 struct hlist_node napi_hash_node;
329 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700330};
331
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800332enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700333 NAPI_STATE_SCHED, /* Poll is scheduled */
Eric Dumazet39e6c822017-02-28 10:34:50 -0800334 NAPI_STATE_MISSED, /* reschedule a napi */
David S. Millera0a46192008-01-07 20:35:07 -0800335 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Eric Dumazet217f6972016-11-15 10:15:11 -0800339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
340};
341
342enum {
Eric Dumazet39e6c822017-02-28 10:34:50 -0800343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700350};
351
Ben Hutchings5b252f02009-10-29 07:17:09 +0000352enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700353 GRO_MERGED,
354 GRO_MERGED_FREE,
355 GRO_HELD,
356 GRO_NORMAL,
357 GRO_DROP,
Steffen Klassert25393d32017-02-15 09:39:44 +0100358 GRO_CONSUMED,
Herbert Xud1c76af2009-03-16 10:50:02 -0700359};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000360typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700361
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000362/*
363 * enum rx_handler_result - Possible return values for rx_handlers.
364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * further.
366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
367 * case skb->dev was changed by rx_handler.
368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000370 *
371 * rx_handlers are functions called from inside __netif_receive_skb(), to do
372 * special processing of the skb, prior to delivery to protocol handlers.
373 *
374 * Currently, a net_device can only have a single rx_handler registered. Trying
375 * to register a second rx_handler will return -EBUSY.
376 *
377 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
378 * To unregister a rx_handler on a net_device, use
379 * netdev_rx_handler_unregister().
380 *
381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
382 * do with the skb.
383 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500384 * If the rx_handler consumed the skb in some way, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500386 * the skb to be delivered in some other way.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000387 *
388 * If the rx_handler changed skb->dev, to divert the skb to another
389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
390 * new device will be called if it exists.
391 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500392 * If the rx_handler decides the skb should be ignored, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400394 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000395 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000397 * delivered, it should return RX_HANDLER_PASS.
398 *
399 * A device without a registered rx_handler will behave as if rx_handler
400 * returned RX_HANDLER_PASS.
401 */
402
403enum rx_handler_result {
404 RX_HANDLER_CONSUMED,
405 RX_HANDLER_ANOTHER,
406 RX_HANDLER_EXACT,
407 RX_HANDLER_PASS,
408};
409typedef enum rx_handler_result rx_handler_result_t;
410typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000411
Joe Perchesf629d202013-09-26 14:48:15 -0700412void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700413void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700414
David S. Miller4d295152012-03-07 21:02:35 -0500415static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800416{
417 return test_bit(NAPI_STATE_DISABLE, &n->state);
418}
419
Eric Dumazet39e6c822017-02-28 10:34:50 -0800420bool napi_schedule_prep(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700421
422/**
423 * napi_schedule - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500424 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700425 *
426 * Schedule NAPI poll routine to be called if it is not already
427 * running.
428 */
429static inline void napi_schedule(struct napi_struct *n)
430{
431 if (napi_schedule_prep(n))
432 __napi_schedule(n);
433}
434
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700435/**
436 * napi_schedule_irqoff - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500437 * @n: NAPI context
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700438 *
439 * Variant of napi_schedule(), assuming hard irqs are masked.
440 */
441static inline void napi_schedule_irqoff(struct napi_struct *n)
442{
443 if (napi_schedule_prep(n))
444 __napi_schedule_irqoff(n);
445}
446
Roland Dreierbfe13f52007-10-09 15:47:37 -0700447/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500448static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700449{
450 if (napi_schedule_prep(napi)) {
451 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500452 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700453 }
David S. Miller4d295152012-03-07 21:02:35 -0500454 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700455}
456
Eric Dumazet364b6052016-11-15 10:15:13 -0800457bool napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700458/**
459 * napi_complete - NAPI processing complete
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500460 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700461 *
462 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800463 * Consider using napi_complete_done() instead.
Eric Dumazet364b6052016-11-15 10:15:13 -0800464 * Return false if device should avoid rearming interrupts.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700465 */
Eric Dumazet364b6052016-11-15 10:15:13 -0800466static inline bool napi_complete(struct napi_struct *n)
Eric Dumazet3b47d302014-11-06 21:09:44 -0800467{
468 return napi_complete_done(n, 0);
469}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700470
471/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300472 * napi_hash_del - remove a NAPI from global table
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500473 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300474 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500475 * Warning: caller must observe RCU grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800476 * before freeing memory containing @napi, if
477 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800478 * Note: core networking stack automatically calls it
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500479 * from netif_napi_del().
Eric Dumazet93d05d42015-11-18 06:31:03 -0800480 * Drivers might want to call this helper to combine all
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500481 * the needed RCU grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300482 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800483bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300484
485/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700486 * napi_disable - prevent NAPI from scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500487 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700488 *
489 * Stop NAPI from being scheduled on this context.
490 * Waits till any outstanding processing completes.
491 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800492void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700493
494/**
495 * napi_enable - enable NAPI scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500496 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700497 *
498 * Resume NAPI from being scheduled on this context.
499 * Must be paired with napi_disable.
500 */
501static inline void napi_enable(struct napi_struct *n)
502{
503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100504 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700505 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400506 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700507}
508
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700509/**
510 * napi_synchronize - wait until NAPI is not running
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500511 * @n: NAPI context
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700512 *
513 * Wait until NAPI is done being scheduled on this context.
514 * Waits till any outstanding processing completes but
515 * does not disable future activations.
516 */
517static inline void napi_synchronize(const struct napi_struct *n)
518{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100519 if (IS_ENABLED(CONFIG_SMP))
520 while (test_bit(NAPI_STATE_SCHED, &n->state))
521 msleep(1);
522 else
523 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700524}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700525
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800526enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000527 __QUEUE_STATE_DRV_XOFF,
528 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700529 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700530};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200531
532#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
533#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
534#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535
536#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
537#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 QUEUE_STATE_FROZEN)
539#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
540 QUEUE_STATE_FROZEN)
541
Tom Herbert734664982011-11-28 16:32:44 +0000542/*
543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
544 * netif_tx_* functions below are used to manipulate this flag. The
545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
546 * queue independently. The netif_xmit_*stopped functions below are called
547 * to check if the queue has been stopped by the driver or stack (either
548 * of the XOFF bits are set in the state). Drivers should not need to call
549 * netif_xmit*stopped functions, they should only be using netif_tx_*.
550 */
David S. Miller79d16382008-07-08 23:14:46 -0700551
David S. Millerbb949fb2008-07-08 16:55:56 -0700552struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700553/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500554 * read-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700555 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700556 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700557 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700558 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000559#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000560 struct kobject kobj;
561#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000562#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
563 int numa_node;
564#endif
Florian Westphalc0ef0792016-05-03 03:29:09 +0200565 unsigned long tx_maxrate;
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700571/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500572 * write-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700573 */
574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
575 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700576 /*
Florian Westphal9b366272016-05-03 16:33:14 +0200577 * Time (in jiffies) of last Tx
Eric Dumazet9d214932009-05-17 20:55:16 -0700578 */
579 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000580
Tom Herbert114cf582011-11-28 16:33:09 +0000581 unsigned long state;
582
583#ifdef CONFIG_BQL
584 struct dql dql;
585#endif
David S. Millere8a04642008-07-17 00:34:19 -0700586} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700587
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000588static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
589{
590#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 return q->numa_node;
592#else
Changli Gaob236da62010-12-14 03:09:15 +0000593 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000594#endif
595}
596
597static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 q->numa_node = node;
601#endif
602}
603
Eric Dumazetdf334542010-03-24 19:13:54 +0000604#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000605/*
606 * This structure holds an RPS map which can be of variable length. The
607 * map is an array of CPUs.
608 */
609struct rps_map {
610 unsigned int len;
611 struct rcu_head rcu;
612 u16 cpus[0];
613};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000614#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000615
Tom Herbertfec5e652010-04-16 16:01:27 -0700616/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
618 * tail pointer for that CPU's input queue at the time of last enqueue, and
619 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700620 */
621struct rps_dev_flow {
622 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000623 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700624 unsigned int last_qtail;
625};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000626#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700627
628/*
629 * The rps_dev_flow_table structure contains a table of flow mappings.
630 */
631struct rps_dev_flow_table {
632 unsigned int mask;
633 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700634 struct rps_dev_flow flows[0];
635};
636#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000637 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700638
639/*
640 * The rps_sock_flow_table contains mappings of flows to the last CPU
641 * on which they were processed by the application (set in recvmsg).
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500642 * Each entry is a 32bit value. Upper part is the high-order bits
643 * of flow hash, lower part is CPU number.
Eric Dumazet567e4b72015-02-06 12:59:01 -0800644 * rps_cpu_mask is used to partition the space, depending on number of
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
Eric Dumazet567e4b72015-02-06 12:59:01 -0800647 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700648 */
649struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800650 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800651
652 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700653};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800654#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700655
656#define RPS_NO_CPU 0xffff
657
Eric Dumazet567e4b72015-02-06 12:59:01 -0800658extern u32 rps_cpu_mask;
659extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
660
Tom Herbertfec5e652010-04-16 16:01:27 -0700661static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
662 u32 hash)
663{
664 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800665 unsigned int index = hash & table->mask;
666 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700667
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500668 /* We only give a hint, preemption can change CPU under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800669 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700670
Eric Dumazet567e4b72015-02-06 12:59:01 -0800671 if (table->ents[index] != val)
672 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700673 }
674}
675
Ben Hutchingsc4454772011-01-19 11:03:53 +0000676#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700677bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
678 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000679#endif
Michael Daltona953be52014-01-16 22:23:28 -0800680#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000681
Tom Herbert0a9627f2010-03-16 08:03:29 +0000682/* This structure contains an instance of an RX queue. */
683struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800684#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000685 struct rps_map __rcu *rps_map;
686 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800687#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000688 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000689 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000690} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800691
692/*
693 * RX queue sysfs structures and functions.
694 */
695struct rx_queue_attribute {
696 struct attribute attr;
697 ssize_t (*show)(struct netdev_rx_queue *queue,
698 struct rx_queue_attribute *attr, char *buf);
699 ssize_t (*store)(struct netdev_rx_queue *queue,
700 struct rx_queue_attribute *attr, const char *buf, size_t len);
701};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800702
Tom Herbertbf264142010-11-26 08:36:09 +0000703#ifdef CONFIG_XPS
704/*
705 * This structure holds an XPS map which can be of variable length. The
706 * map is an array of queues.
707 */
708struct xps_map {
709 unsigned int len;
710 unsigned int alloc_len;
711 struct rcu_head rcu;
712 u16 queues[0];
713};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000714#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100715#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
716 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000717
718/*
719 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 */
721struct xps_dev_maps {
722 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000723 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000724};
Alexander Duyck184c4492016-10-28 11:50:13 -0400725#define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
726 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
Tom Herbertbf264142010-11-26 08:36:09 +0000727#endif /* CONFIG_XPS */
728
John Fastabend4f57c082011-01-17 08:06:04 +0000729#define TC_MAX_QUEUE 16
730#define TC_BITMASK 15
731/* HW offloaded queuing disciplines txq count and offset maps */
732struct netdev_tc_txq {
733 u16 count;
734 u16 offset;
735};
736
Neerav Parikh68bad942012-01-04 20:23:39 +0000737#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738/*
739 * This structure is to hold information about the device
740 * configured to run FCoE protocol stack.
741 */
742struct netdev_fcoe_hbainfo {
743 char manufacturer[64];
744 char serial_number[64];
745 char hardware_version[64];
746 char driver_version[64];
747 char optionrom_version[64];
748 char firmware_version[64];
749 char model[256];
750 char model_description[256];
751};
752#endif
753
Jiri Pirko02637fc2014-11-28 14:34:16 +0100754#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200755
Jiri Pirko02637fc2014-11-28 14:34:16 +0100756/* This structure holds a unique identifier to identify some
757 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200758 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100759struct netdev_phys_item_id {
760 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200761 unsigned char id_len;
762};
763
Scott Feldmand754f982015-07-18 18:24:49 -0700764static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
765 struct netdev_phys_item_id *b)
766{
767 return a->id_len == b->id_len &&
768 memcmp(a->id, b->id, a->id_len) == 0;
769}
770
Daniel Borkmann99932d42014-02-16 15:55:20 +0100771typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
772 struct sk_buff *skb);
773
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800774/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800775 * that are being passed to the netdevice through the setup_tc op.
776 */
777enum {
778 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800779 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200780 TC_SETUP_CLSFLOWER,
Yotam Gigib87f7932016-07-21 12:03:12 +0200781 TC_SETUP_MATCHALL,
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100782 TC_SETUP_CLSBPF,
John Fastabend16e5cc62016-02-16 21:16:43 -0800783};
784
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800785struct tc_cls_u32_offload;
786
John Fastabend16e5cc62016-02-16 21:16:43 -0800787struct tc_to_netdev {
788 unsigned int type;
789 union {
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800790 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200791 struct tc_cls_flower_offload *cls_flower;
Yotam Gigib87f7932016-07-21 12:03:12 +0200792 struct tc_cls_matchall_offload *cls_mall;
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100793 struct tc_cls_bpf_offload *cls_bpf;
Amritha Nambiar56f36ac2017-03-15 10:39:25 -0700794 struct tc_mqprio_qopt *mqprio;
John Fastabend16e5cc62016-02-16 21:16:43 -0800795 };
Hadar Hen Zion7091d8c2016-12-01 14:06:37 +0200796 bool egress_dev;
John Fastabend16e5cc62016-02-16 21:16:43 -0800797};
798
Brenden Blancoa7862b42016-07-19 12:16:48 -0700799/* These structures hold the attributes of xdp state that are being passed
800 * to the netdevice through the xdp op.
801 */
802enum xdp_netdev_command {
803 /* Set or clear a bpf program used in the earliest stages of packet
804 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
805 * is responsible for calling bpf_prog_put on any old progs that are
806 * stored. In case of error, the callee need not release the new prog
807 * reference, but on success it takes ownership and must bpf_prog_put
808 * when it is no longer used.
809 */
810 XDP_SETUP_PROG,
Jakub Kicinskiee5d0322017-06-21 18:25:04 -0700811 XDP_SETUP_PROG_HW,
Brenden Blancoa7862b42016-07-19 12:16:48 -0700812 /* Check if a bpf program is set on the device. The callee should
Jakub Kicinskice158e52017-06-21 18:25:09 -0700813 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
814 * is equivalent to XDP_ATTACHED_DRV.
Brenden Blancoa7862b42016-07-19 12:16:48 -0700815 */
816 XDP_QUERY_PROG,
817};
818
Jakub Kicinskiddf9f972017-04-30 21:46:46 -0700819struct netlink_ext_ack;
820
Brenden Blancoa7862b42016-07-19 12:16:48 -0700821struct netdev_xdp {
822 enum xdp_netdev_command command;
823 union {
824 /* XDP_SETUP_PROG */
Jakub Kicinskiddf9f972017-04-30 21:46:46 -0700825 struct {
Jakub Kicinski32d60272017-06-21 18:25:03 -0700826 u32 flags;
Jakub Kicinskiddf9f972017-04-30 21:46:46 -0700827 struct bpf_prog *prog;
828 struct netlink_ext_ack *extack;
829 };
Brenden Blancoa7862b42016-07-19 12:16:48 -0700830 /* XDP_QUERY_PROG */
Martin KaFai Lau58038692017-06-15 17:29:09 -0700831 struct {
Jakub Kicinskice158e52017-06-21 18:25:09 -0700832 u8 prog_attached;
Martin KaFai Lau58038692017-06-15 17:29:09 -0700833 u32 prog_id;
834 };
Brenden Blancoa7862b42016-07-19 12:16:48 -0700835 };
836};
John Fastabend16e5cc62016-02-16 21:16:43 -0800837
Steffen Klassertd77e38e2017-04-14 10:06:10 +0200838#ifdef CONFIG_XFRM_OFFLOAD
839struct xfrmdev_ops {
840 int (*xdo_dev_state_add) (struct xfrm_state *x);
841 void (*xdo_dev_state_delete) (struct xfrm_state *x);
842 void (*xdo_dev_state_free) (struct xfrm_state *x);
843 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
844 struct xfrm_state *x);
845};
846#endif
847
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800848/*
849 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800850 * The following hooks can be defined; unless noted otherwise, they are
851 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800852 *
853 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500854 * This function is called once when a network device is registered.
855 * The network device can use this for any late stage initialization
856 * or semantic validation. It can fail with an error code which will
857 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800858 *
859 * void (*ndo_uninit)(struct net_device *dev);
860 * This function is called when device is unregistered or when registration
861 * fails. It is not called if init fails.
862 *
863 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500864 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800865 * state.
866 *
867 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500868 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800869 * state.
870 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000871 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
872 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800873 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030874 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
875 * the queue before that can happen; it's for obsolete devices and weird
876 * corner cases, but the stack really does a non-trivial amount
877 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500878 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800879 *
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -0800880 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
881 * struct net_device *dev
882 * netdev_features_t features);
883 * Called by core transmit path to determine if device is capable of
884 * performing offload operations on a given packet. This is to give
885 * the device an opportunity to implement any restrictions that cannot
886 * be otherwise expressed by feature flags. The check is called with
887 * the set of features that the stack has calculated and it returns
888 * those the driver believes to be appropriate.
Eric Dumazetcdba7562016-01-06 06:53:50 -0800889 *
Jason Wangf663dd92014-01-10 16:18:26 +0800890 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100891 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500892 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800893 * transmit queues.
894 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800895 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
896 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500897 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800898 *
899 * void (*ndo_set_rx_mode)(struct net_device *dev);
900 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000901 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500902 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800903 *
904 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
905 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700906 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500907 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800908 *
909 * int (*ndo_validate_addr)(struct net_device *dev);
910 * Test if Media Access Control address is valid for the device.
911 *
912 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500913 * Called when a user requests an ioctl which can't be handled by
914 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800915 * not supported error code.
916 *
917 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
918 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500919 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800920 * interface (PCI) for low level management.
921 *
922 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
923 * Called when a user wants to change the Maximum Transfer Unit
Magnus Dammdb46a0e2017-06-14 16:15:24 +0900924 * of a device.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800925 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800926 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500927 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800928 * for dev->watchdog ticks.
929 *
stephen hemmingerbc1f4472017-01-06 19:12:52 -0800930 * void (*ndo_get_stats64)(struct net_device *dev,
931 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700932 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800933 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000934 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000935 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
936 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700937 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000938 * (which should normally be dev->stats) and return a pointer to
939 * it. The structure may be changed asynchronously only if each
940 * field is written atomically.
941 * 3. Update dev->stats asynchronously and atomically, and define
942 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800943 *
Or Gerlitz3df5b3c2016-11-22 23:09:54 +0200944 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200945 * Return true if this device supports offload stats of this attr_id.
946 *
947 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
948 * void *attr_data)
949 * Get statistics for offload operations by attr_id. Write it into the
950 * attr_data pointer.
951 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530952 * int (*ndo_vlan_rx_add_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 registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800955 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530956 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500957 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000958 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800959 *
960 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000961 *
962 * SR-IOV management functions.
963 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300964 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
965 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400966 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
967 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000968 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000969 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000970 * int (*ndo_get_vf_config)(struct net_device *dev,
971 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300972 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700973 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
974 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300975 *
976 * Enable or disable the VF ability to query its RSS Redirection Table and
977 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500978 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300979 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700980 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
Jiri Pirkoa5fcf8a2017-06-06 17:00:16 +0200981 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle, u32 chain_index,
Florian Fainelli6a4bc2b2017-01-26 14:44:17 -0800982 * __be16 protocol, struct tc_to_netdev *tc);
983 * Called to setup any 'tc' scheduler, classifier or action on @dev.
984 * This is always called from the stack with the rtnl lock held and netif
985 * tx queues stopped. This allows the netdevice to perform queue
986 * management safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000987 *
Yi Zoue9bce842011-03-09 08:48:03 +0000988 * Fiber Channel over Ethernet (FCoE) offload functions.
989 * int (*ndo_fcoe_enable)(struct net_device *dev);
990 * Called when the FCoE protocol stack wants to start using LLD for FCoE
991 * so the underlying device can perform whatever needed configuration or
992 * initialization to support acceleration of FCoE traffic.
993 *
994 * int (*ndo_fcoe_disable)(struct net_device *dev);
995 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
996 * so the underlying device can perform whatever needed clean-ups to
997 * stop supporting acceleration of FCoE traffic.
998 *
999 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1000 * struct scatterlist *sgl, unsigned int sgc);
1001 * Called when the FCoE Initiator wants to initialize an I/O that
1002 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1003 * perform necessary setup and returns 1 to indicate the device is set up
1004 * successfully to perform DDP on this I/O, otherwise this returns 0.
1005 *
1006 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1007 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1008 * indicated by the FC exchange id 'xid', so the underlying device can
1009 * clean up and reuse resources for later DDP requests.
1010 *
1011 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1012 * struct scatterlist *sgl, unsigned int sgc);
1013 * Called when the FCoE Target wants to initialize an I/O that
1014 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1015 * perform necessary setup and returns 1 to indicate the device is set up
1016 * successfully to perform DDP on this I/O, otherwise this returns 0.
1017 *
Neerav Parikh68bad942012-01-04 20:23:39 +00001018 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1019 * struct netdev_fcoe_hbainfo *hbainfo);
1020 * Called when the FCoE Protocol stack wants information on the underlying
1021 * device. This information is utilized by the FCoE protocol stack to
1022 * register attributes with Fiber Channel management service as per the
1023 * FC-GS Fabric Device Management Information(FDMI) specification.
1024 *
Yi Zoue9bce842011-03-09 08:48:03 +00001025 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1026 * Called when the underlying device wants to override default World Wide
1027 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1028 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1029 * protocol stack to use.
1030 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001031 * RFS acceleration.
1032 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1033 * u16 rxq_index, u32 flow_id);
1034 * Set hardware filter for RFS. rxq_index is the target queue index;
1035 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1036 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001037 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001038 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001039 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1040 * Called to make another netdev an underling.
1041 *
1042 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1043 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001044 *
1045 * Feature/offload setting functions.
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -08001046 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1047 * netdev_features_t features);
1048 * Adjusts the requested feature flags according to device-specific
1049 * constraints, and returns the resulting flags. Must not modify
1050 * the device state.
1051 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001052 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001053 * Called to update device configuration to new features. Passed
1054 * feature set might be less than what was returned by ndo_fix_features()).
1055 * Must return >0 or -errno if it changed dev->features itself.
1056 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001057 * int (*ndo_fdb_add)(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, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001060 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001061 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1062 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001063 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001064 * Deletes the FDB entry from dev coresponding to addr.
1065 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001066 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001067 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001068 * Used to add FDB entries to dump requests. Implementers should add
1069 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001070 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001071 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1072 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001073 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001074 * struct net_device *dev, u32 filter_mask,
1075 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001076 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1077 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001078 *
1079 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1080 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1081 * which do not represent real hardware may define this to allow their
1082 * userspace components to manage their virtual carrier state. Devices
1083 * that determine carrier state from physical hardware properties (eg
1084 * network cables) or protocol-dependent mechanisms (eg
1085 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001086 *
1087 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001088 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001089 * Called to get ID of physical port of this device. If driver does
1090 * not implement this, it is assumed that the hw is not able to have
1091 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001092 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001093 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1094 * struct udp_tunnel_info *ti);
1095 * Called by UDP tunnel to notify a driver about the UDP port and socket
1096 * address family that a UDP tunnel is listnening to. It is called only
1097 * when a new port starts listening. The operation is protected by the
1098 * RTNL.
1099 *
1100 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1101 * struct udp_tunnel_info *ti);
1102 * Called by UDP tunnel to notify the driver about a UDP port and socket
1103 * address family that the UDP tunnel is not listening to anymore. The
1104 * operation is protected by the RTNL.
1105 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001106 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1107 * struct net_device *dev)
1108 * Called by upper layer devices to accelerate switching or other
1109 * station functionality into hardware. 'pdev is the lowerdev
1110 * to use for the offload and 'dev' is the net device that will
1111 * back the offload. Returns a pointer to the private structure
1112 * the upper layer will maintain.
1113 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1114 * Called by upper layer device to delete the station created
1115 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1116 * the station and priv is the structure returned by the add
1117 * operation.
John Fastabend822b3b22015-03-18 14:57:33 +02001118 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1119 * int queue_index, u32 maxrate);
1120 * Called when a user wants to set a max-rate limitation of specific
1121 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001122 * int (*ndo_get_iflink)(const struct net_device *dev);
1123 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001124 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001125 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001126 * This function is used to pass protocol port error state information
1127 * to the switch driver. The switch driver can react to the proto_down
1128 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001129 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1130 * This function is used to get egress tunnel information for given skb.
1131 * This is useful for retrieving outer tunnel header parameters while
1132 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001133 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1134 * This function is used to specify the headroom that the skb must
1135 * consider when allocation skb during packet reception. Setting
1136 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001137 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001138 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001139 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1140 * This function is used to set or query state related to XDP on the
1141 * netdevice. See definition of enum xdp_netdev_command for details.
John Fastabend814abfa2017-07-17 09:27:07 -07001142 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1143 * This function is used to submit a XDP packet for transmit on a
1144 * netdevice.
John Fastabend11393cc2017-07-17 09:29:40 -07001145 * void (*ndo_xdp_flush)(struct net_device *dev);
1146 * This function is used to inform the driver to flush a paticular
1147 * xpd tx queue. Must be called on same CPU as xdp_xmit.
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001148 */
1149struct net_device_ops {
1150 int (*ndo_init)(struct net_device *dev);
1151 void (*ndo_uninit)(struct net_device *dev);
1152 int (*ndo_open)(struct net_device *dev);
1153 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001154 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1155 struct net_device *dev);
1156 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1157 struct net_device *dev,
1158 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001159 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001160 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001161 void *accel_priv,
1162 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001163 void (*ndo_change_rx_flags)(struct net_device *dev,
1164 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001165 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001166 int (*ndo_set_mac_address)(struct net_device *dev,
1167 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001168 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001169 int (*ndo_do_ioctl)(struct net_device *dev,
1170 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001171 int (*ndo_set_config)(struct net_device *dev,
1172 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001173 int (*ndo_change_mtu)(struct net_device *dev,
1174 int new_mtu);
1175 int (*ndo_neigh_setup)(struct net_device *dev,
1176 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001177 void (*ndo_tx_timeout) (struct net_device *dev);
1178
stephen hemmingerbc1f4472017-01-06 19:12:52 -08001179 void (*ndo_get_stats64)(struct net_device *dev,
1180 struct rtnl_link_stats64 *storage);
Or Gerlitz3df5b3c2016-11-22 23:09:54 +02001181 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001182 int (*ndo_get_offload_stats)(int attr_id,
1183 const struct net_device *dev,
1184 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001185 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1186
Jiri Pirko8e586132011-12-08 19:52:37 -05001187 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001188 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001189 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001190 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001191#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001192 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001193 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001194 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001195 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001196#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001197 int (*ndo_set_vf_mac)(struct net_device *dev,
1198 int queue, u8 *mac);
1199 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001200 int queue, u16 vlan,
1201 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001202 int (*ndo_set_vf_rate)(struct net_device *dev,
1203 int vf, int min_tx_rate,
1204 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001205 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1206 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001207 int (*ndo_set_vf_trust)(struct net_device *dev,
1208 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001209 int (*ndo_get_vf_config)(struct net_device *dev,
1210 int vf,
1211 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001212 int (*ndo_set_vf_link_state)(struct net_device *dev,
1213 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001214 int (*ndo_get_vf_stats)(struct net_device *dev,
1215 int vf,
1216 struct ifla_vf_stats
1217 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001218 int (*ndo_set_vf_port)(struct net_device *dev,
1219 int vf,
1220 struct nlattr *port[]);
1221 int (*ndo_get_vf_port)(struct net_device *dev,
1222 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001223 int (*ndo_set_vf_guid)(struct net_device *dev,
1224 int vf, u64 guid,
1225 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001226 int (*ndo_set_vf_rss_query_en)(
1227 struct net_device *dev,
1228 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001229 int (*ndo_setup_tc)(struct net_device *dev,
Jiri Pirkoa5fcf8a2017-06-06 17:00:16 +02001230 u32 handle, u32 chain_index,
John Fastabend16e5cc62016-02-16 21:16:43 -08001231 __be16 protocol,
1232 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001233#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001234 int (*ndo_fcoe_enable)(struct net_device *dev);
1235 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001236 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1237 u16 xid,
1238 struct scatterlist *sgl,
1239 unsigned int sgc);
1240 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1241 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001242 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1243 u16 xid,
1244 struct scatterlist *sgl,
1245 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001246 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1247 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001248#endif
1249
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001250#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001251#define NETDEV_FCOE_WWNN 0
1252#define NETDEV_FCOE_WWPN 1
1253 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1254 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001255#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001256
Ben Hutchingsc4454772011-01-19 11:03:53 +00001257#ifdef CONFIG_RFS_ACCEL
1258 int (*ndo_rx_flow_steer)(struct net_device *dev,
1259 const struct sk_buff *skb,
1260 u16 rxq_index,
1261 u32 flow_id);
1262#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001263 int (*ndo_add_slave)(struct net_device *dev,
1264 struct net_device *slave_dev);
1265 int (*ndo_del_slave)(struct net_device *dev,
1266 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001267 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1268 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001269 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001270 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001271 int (*ndo_neigh_construct)(struct net_device *dev,
1272 struct neighbour *n);
1273 void (*ndo_neigh_destroy)(struct net_device *dev,
1274 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001275
1276 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001277 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001278 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001279 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001280 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001281 u16 flags);
1282 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001283 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001284 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001285 const unsigned char *addr,
1286 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001287 int (*ndo_fdb_dump)(struct sk_buff *skb,
1288 struct netlink_callback *cb,
1289 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001290 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001291 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001292
1293 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001294 struct nlmsghdr *nlh,
1295 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001296 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1297 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001298 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001299 u32 filter_mask,
1300 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001301 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001302 struct nlmsghdr *nlh,
1303 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001304 int (*ndo_change_carrier)(struct net_device *dev,
1305 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001306 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001307 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001308 int (*ndo_get_phys_port_name)(struct net_device *dev,
1309 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001310 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1311 struct udp_tunnel_info *ti);
1312 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1313 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001314 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1315 struct net_device *dev);
1316 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1317 void *priv);
1318
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);
John Fastabend814abfa2017-07-17 09:27:07 -07001332 int (*ndo_xdp_xmit)(struct net_device *dev,
1333 struct xdp_buff *xdp);
John Fastabend11393cc2017-07-17 09:29:40 -07001334 void (*ndo_xdp_flush)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001335};
1336
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001337/**
1338 * enum net_device_priv_flags - &struct net_device priv_flags
1339 *
1340 * These are the &struct net_device, they are only set internally
1341 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001342 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001343 * during any kernel release.
1344 *
1345 * You should have a pretty good reason to be extending these flags.
1346 *
1347 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1348 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001349 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001350 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001351 * @IFF_WAN_HDLC: WAN HDLC device
1352 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1353 * release skb->dst
1354 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1355 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1356 * @IFF_MACVLAN_PORT: device used as macvlan port
1357 * @IFF_BRIDGE_PORT: device used as bridge port
1358 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1359 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1360 * @IFF_UNICAST_FLT: Supports unicast filtering
1361 * @IFF_TEAM_PORT: device used as team port
1362 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1363 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1364 * change when it's running
1365 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001366 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1367 * underlying stacked devices
1368 * @IFF_IPVLAN_MASTER: IPvlan master device
1369 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001370 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001371 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001372 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001373 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001374 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001375 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001376 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1377 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001378 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001379 */
1380enum netdev_priv_flags {
1381 IFF_802_1Q_VLAN = 1<<0,
1382 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001383 IFF_BONDING = 1<<2,
1384 IFF_ISATAP = 1<<3,
1385 IFF_WAN_HDLC = 1<<4,
1386 IFF_XMIT_DST_RELEASE = 1<<5,
1387 IFF_DONT_BRIDGE = 1<<6,
1388 IFF_DISABLE_NETPOLL = 1<<7,
1389 IFF_MACVLAN_PORT = 1<<8,
1390 IFF_BRIDGE_PORT = 1<<9,
1391 IFF_OVS_DATAPATH = 1<<10,
1392 IFF_TX_SKB_SHARING = 1<<11,
1393 IFF_UNICAST_FLT = 1<<12,
1394 IFF_TEAM_PORT = 1<<13,
1395 IFF_SUPP_NOFCS = 1<<14,
1396 IFF_LIVE_ADDR_CHANGE = 1<<15,
1397 IFF_MACVLAN = 1<<16,
1398 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1399 IFF_IPVLAN_MASTER = 1<<18,
1400 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001401 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001402 IFF_NO_QUEUE = 1<<21,
1403 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001404 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001405 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001406 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001407 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001408 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001409};
1410
1411#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1412#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001413#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001414#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001415#define IFF_WAN_HDLC IFF_WAN_HDLC
1416#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1417#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1418#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1419#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1420#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1421#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1422#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1423#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1424#define IFF_TEAM_PORT IFF_TEAM_PORT
1425#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1426#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1427#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001428#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001429#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1430#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001431#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001432#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001433#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001434#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001435#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001436#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001437#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001438
Karoly Kemeny536721b2014-07-30 20:27:36 +02001439/**
1440 * struct net_device - The DEVICE structure.
Mauro Carvalho Chehabd6519832017-05-12 09:35:46 -03001441 *
1442 * Actually, this whole structure is a big mistake. It mixes I/O
1443 * data with strictly "high-level" data, and it has to know about
1444 * almost every data structure used in the INET module.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001445 *
1446 * @name: This is the first field of the "visible" part of this structure
1447 * (i.e. as seen by users in the "Space.c" file). It is the name
Mauro Carvalho Chehabd6519832017-05-12 09:35:46 -03001448 * of the interface.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001449 *
1450 * @name_hlist: Device name hash chain, please keep it close to name[]
1451 * @ifalias: SNMP alias
1452 * @mem_end: Shared memory end
1453 * @mem_start: Shared memory start
1454 * @base_addr: Device I/O address
1455 * @irq: Device IRQ number
1456 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001457 * @carrier_changes: Stats to monitor carrier on<->off transitions
1458 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001459 * @state: Generic network queuing layer state, see netdev_state_t
1460 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001461 * @napi_list: List entry used for polling NAPI devices
1462 * @unreg_list: List entry when we are unregistering the
1463 * device; see the function unregister_netdev
1464 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001465 * @ptype_all: Device-specific packet handlers for all protocols
1466 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001467 *
1468 * @adj_list: Directly linked devices, like slaves for bonding
Karoly Kemeny536721b2014-07-30 20:27:36 +02001469 * @features: Currently active device features
1470 * @hw_features: User-changeable features
1471 *
1472 * @wanted_features: User-requested features
1473 * @vlan_features: Mask of features inheritable by VLAN devices
1474 *
1475 * @hw_enc_features: Mask of features inherited by encapsulating devices
1476 * This field indicates what encapsulation
1477 * offloads the hardware is capable of doing,
1478 * and drivers will need to set them appropriately.
1479 *
1480 * @mpls_features: Mask of features inheritable by MPLS
1481 *
1482 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001483 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001484 *
1485 * @stats: Statistics struct, which was left as a legacy, use
1486 * rtnl_link_stats64 instead
1487 *
1488 * @rx_dropped: Dropped packets by core network,
1489 * do not use this in drivers
1490 * @tx_dropped: Dropped packets by core network,
1491 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001492 * @rx_nohandler: nohandler dropped packets by core network on
1493 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001494 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001495 * @wireless_handlers: List of functions to handle Wireless Extensions,
1496 * instead of ioctl,
1497 * see <net/iw_handler.h> for details.
1498 * @wireless_data: Instance data managed by the core of wireless extensions
1499 *
1500 * @netdev_ops: Includes several pointers to callbacks,
1501 * if one wants to override the ndo_*() functions
1502 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001503 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1504 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001505 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001506 * of Layer 2 headers.
1507 *
1508 * @flags: Interface flags (a la BSD)
1509 * @priv_flags: Like 'flags' but invisible to userspace,
1510 * see if.h for the definitions
1511 * @gflags: Global flags ( kept as legacy )
1512 * @padded: How much padding added by alloc_netdev()
1513 * @operstate: RFC2863 operstate
1514 * @link_mode: Mapping policy to operstate
1515 * @if_port: Selectable AUI, TP, ...
1516 * @dma: DMA channel
1517 * @mtu: Interface MTU value
Jarod Wilson61e84622016-10-07 22:04:33 -04001518 * @min_mtu: Interface Minimum MTU value
1519 * @max_mtu: Interface Maximum MTU value
Karoly Kemeny536721b2014-07-30 20:27:36 +02001520 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001521 * @hard_header_len: Maximum hardware header length.
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05001522 * @min_header_len: Minimum hardware header length
Karoly Kemeny536721b2014-07-30 20:27:36 +02001523 *
1524 * @needed_headroom: Extra headroom the hardware may need, but not in all
1525 * cases can this be guaranteed
1526 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1527 * cases can this be guaranteed. Some cases also use
1528 * LL_MAX_HEADER instead to allocate the skb
1529 *
1530 * interface address info:
1531 *
1532 * @perm_addr: Permanent hw address
1533 * @addr_assign_type: Hw address assignment type
1534 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001535 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001536 * @dev_id: Used to differentiate devices that share
1537 * the same link layer address
1538 * @dev_port: Used to differentiate devices that share
1539 * the same function
1540 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001541 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001542 * has been enabled due to the need to listen to
1543 * additional unicast addresses in a device that
1544 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001545 * @uc: unicast mac addresses
1546 * @mc: multicast mac addresses
1547 * @dev_addrs: list of device hw addresses
1548 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001549 * @promiscuity: Number of times the NIC is told to work in
1550 * promiscuous mode; if it becomes 0 the NIC will
1551 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001552 * @allmulti: Counter, enables or disables allmulticast mode
1553 *
1554 * @vlan_info: VLAN info
1555 * @dsa_ptr: dsa specific data
1556 * @tipc_ptr: TIPC specific data
1557 * @atalk_ptr: AppleTalk link
1558 * @ip_ptr: IPv4 specific data
1559 * @dn_ptr: DECnet specific data
1560 * @ip6_ptr: IPv6 specific data
1561 * @ax25_ptr: AX.25 specific data
1562 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1563 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001564 * @dev_addr: Hw address (before bcast,
1565 * because most packets are unicast)
1566 *
1567 * @_rx: Array of RX queues
1568 * @num_rx_queues: Number of RX queues
1569 * allocated at register_netdev() time
1570 * @real_num_rx_queues: Number of RX queues currently active in device
1571 *
1572 * @rx_handler: handler for received packets
1573 * @rx_handler_data: XXX: need comments on this one
1574 * @ingress_queue: XXX: need comments on this one
1575 * @broadcast: hw bcast address
1576 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001577 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1578 * indexed by RX queue number. Assigned by driver.
1579 * This must only be set if the ndo_rx_flow_steer
1580 * operation is defined
1581 * @index_hlist: Device index hash chain
1582 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001583 * @_tx: Array of TX queues
1584 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1585 * @real_num_tx_queues: Number of TX queues currently active in device
1586 * @qdisc: Root qdisc from userspace point of view
1587 * @tx_queue_len: Max frames per queue allowed
1588 * @tx_global_lock: XXX: need comments on this one
1589 *
1590 * @xps_maps: XXX: need comments on this one
1591 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001592 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001593 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001594 * @watchdog_timer: List of timers
1595 *
1596 * @pcpu_refcnt: Number of references to this device
1597 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001598 * @link_watch_list: XXX: need comments on this one
1599 *
1600 * @reg_state: Register/unregister state machine
1601 * @dismantle: Device is going to be freed
1602 * @rtnl_link_state: This enum represents the phases of creating
1603 * a new link
1604 *
David S. Millercf124db2017-05-08 12:52:56 -04001605 * @needs_free_netdev: Should unregister perform free_netdev?
1606 * @priv_destructor: Called from unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001607 * @npinfo: XXX: need comments on this one
1608 * @nd_net: Network namespace this network device is inside
1609 *
1610 * @ml_priv: Mid-layer private
1611 * @lstats: Loopback statistics
1612 * @tstats: Tunnel statistics
1613 * @dstats: Dummy statistics
1614 * @vstats: Virtual ethernet statistics
1615 *
1616 * @garp_port: GARP
1617 * @mrp_port: MRP
1618 *
1619 * @dev: Class/net/name entry
1620 * @sysfs_groups: Space for optional device, statistics and wireless
1621 * sysfs groups
1622 *
1623 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1624 * @rtnl_link_ops: Rtnl_link_ops
1625 *
1626 * @gso_max_size: Maximum size of generic segmentation offload
1627 * @gso_max_segs: Maximum number of segments that can be passed to the
1628 * NIC for GSO
1629 *
1630 * @dcbnl_ops: Data Center Bridging netlink ops
1631 * @num_tc: Number of traffic classes in the net device
1632 * @tc_to_txq: XXX: need comments on this one
Randy Dunlap920c1cd2016-11-21 18:28:36 -08001633 * @prio_tc_map: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001634 *
1635 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1636 *
1637 * @priomap: XXX: need comments on this one
1638 * @phydev: Physical device may attach itself
1639 * for hardware timestamping
1640 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001641 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1642 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001643 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001644 * @proto_down: protocol port state information can be sent to the
1645 * switch driver and used to set the phys state of the
1646 * switch port.
1647 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 * FIXME: cleanup struct net_device such that network protocol info
1649 * moves out.
1650 */
1651
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001652struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001654 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001655 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 /*
1657 * I/O specific fields
1658 * FIXME: Merge these and struct ifmap into one
1659 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001660 unsigned long mem_end;
1661 unsigned long mem_start;
1662 unsigned long base_addr;
1663 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001665 atomic_t carrier_changes;
1666
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001668 * Some hardware also needs these fields (state,dev_list,
1669 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 * part of the usual set specified in Space.c.
1671 */
1672
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 unsigned long state;
1674
Pavel Emelianov7562f872007-05-03 15:13:45 -07001675 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001676 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001677 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001678 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001679 struct list_head ptype_all;
1680 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001681
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001682 struct {
1683 struct list_head upper;
1684 struct list_head lower;
1685 } adj_list;
1686
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001687 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001688 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001689 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001690 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001691 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001692 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001693 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001694
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001696 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
Rusty Russellc45d2862007-03-28 14:29:08 -07001698 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001699
Eric Dumazet015f0682014-03-27 08:45:56 -07001700 atomic_long_t rx_dropped;
1701 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001702 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703
Johannes Bergb86e0282007-04-26 20:48:23 -07001704#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001705 const struct iw_handler_def *wireless_handlers;
1706 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001707#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001708 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001709 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001710#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001711 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001712#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001713#ifdef CONFIG_NET_L3_MASTER_DEV
1714 const struct l3mdev_ops *l3mdev_ops;
1715#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001716#if IS_ENABLED(CONFIG_IPV6)
1717 const struct ndisc_ops *ndisc_ops;
1718#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719
Steffen Klassertd77e38e2017-04-14 10:06:10 +02001720#ifdef CONFIG_XFRM
1721 const struct xfrmdev_ops *xfrmdev_ops;
1722#endif
1723
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001724 const struct header_ops *header_ops;
1725
Karoly Kemeny536721b2014-07-30 20:27:36 +02001726 unsigned int flags;
1727 unsigned int priv_flags;
1728
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001730 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
Karoly Kemeny536721b2014-07-30 20:27:36 +02001732 unsigned char operstate;
1733 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001734
Karoly Kemeny536721b2014-07-30 20:27:36 +02001735 unsigned char if_port;
1736 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001737
Karoly Kemeny536721b2014-07-30 20:27:36 +02001738 unsigned int mtu;
Jarod Wilson61e84622016-10-07 22:04:33 -04001739 unsigned int min_mtu;
1740 unsigned int max_mtu;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001741 unsigned short type;
1742 unsigned short hard_header_len;
Alexey Dobriyand92be7a2017-04-10 11:25:26 +03001743 unsigned char min_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744
Johannes Bergf5184d22008-05-12 20:48:31 -07001745 unsigned short needed_headroom;
1746 unsigned short needed_tailroom;
1747
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001749 unsigned char perm_addr[MAX_ADDR_LEN];
1750 unsigned char addr_assign_type;
1751 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001752 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001753 unsigned short dev_id;
1754 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001755 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001756 unsigned char name_assign_type;
1757 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001758 struct netdev_hw_addr_list uc;
1759 struct netdev_hw_addr_list mc;
1760 struct netdev_hw_addr_list dev_addrs;
1761
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001762#ifdef CONFIG_SYSFS
1763 struct kset *queues_kset;
1764#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001765 unsigned int promiscuity;
1766 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001769 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001770
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001771#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001772 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001773#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001774#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001775 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001776#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001777#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001778 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001779#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001780 void *atalk_ptr;
1781 struct in_device __rcu *ip_ptr;
1782 struct dn_dev __rcu *dn_ptr;
1783 struct inet6_dev __rcu *ip6_ptr;
1784 void *ax25_ptr;
1785 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001786 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001787#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1788 struct mpls_dev __rcu *mpls_ptr;
1789#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001791/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001792 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001793 */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001794 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001795 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001796
Michael Daltona953be52014-01-16 22:23:28 -08001797#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001798 struct netdev_rx_queue *_rx;
1799
Tom Herbert0a9627f2010-03-16 08:03:29 +00001800 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001801 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001802#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001803
Eric Dumazet7acedaf2017-04-25 11:36:52 -07001804 struct bpf_prog __rcu *xdp_prog;
Eric Dumazet3b47d302014-11-06 21:09:44 -08001805 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001806 rx_handler_func_t __rcu *rx_handler;
1807 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001808
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001809#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001810 struct tcf_proto __rcu *ingress_cl_list;
1811#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001812 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001813#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001814 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001815#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001816
Karoly Kemeny536721b2014-07-30 20:27:36 +02001817 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001818#ifdef CONFIG_RFS_ACCEL
1819 struct cpu_rmap *rx_cpu_rmap;
1820#endif
1821 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001822
1823/*
1824 * Cache lines mostly used on transmit path
1825 */
David S. Millere8a04642008-07-17 00:34:19 -07001826 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1827 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001828 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001829 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001830#ifdef CONFIG_NET_SCHED
1831 DECLARE_HASHTABLE (qdisc_hash, 4);
1832#endif
Alexey Dobriyan0cd29502017-05-17 13:30:44 +03001833 unsigned int tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001834 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001835 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001836
Tom Herbertbf264142010-11-26 08:36:09 +00001837#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001838 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001839#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001840#ifdef CONFIG_NET_CLS_ACT
1841 struct tcf_proto __rcu *egress_cl_list;
1842#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001843
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001844 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001845 struct timer_list watchdog_timer;
1846
Eric Dumazet29b44332010-10-11 10:22:12 +00001847 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849
Eric Dumazete014deb2009-11-17 05:59:21 +00001850 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001851
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001853 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854 NETREG_UNREGISTERING, /* called unregister_netdevice */
1855 NETREG_UNREGISTERED, /* completed unregister todo */
1856 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001857 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001858 } reg_state:8;
1859
Karoly Kemeny536721b2014-07-30 20:27:36 +02001860 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001861
1862 enum {
1863 RTNL_LINK_INITIALIZED,
1864 RTNL_LINK_INITIALIZING,
1865 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866
David S. Millercf124db2017-05-08 12:52:56 -04001867 bool needs_free_netdev;
1868 void (*priv_destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001871 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872#endif
David S. Millereae792b2008-07-15 03:03:33 -07001873
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001874 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001875
David S. Miller49517042008-05-12 03:29:11 -07001876 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001877 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001878 void *ml_priv;
1879 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001880 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001881 struct pcpu_dstats __percpu *dstats;
1882 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001883 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001884
Tobias Klauserfb585b42017-02-10 16:43:50 +01001885#if IS_ENABLED(CONFIG_GARP)
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001886 struct garp_port __rcu *garp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001887#endif
1888#if IS_ENABLED(CONFIG_MRP)
David Wardfebf0182013-02-08 17:17:06 +00001889 struct mrp_port __rcu *mrp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001890#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001892 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001893 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001894 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001895
Patrick McHardy38f7b872007-06-13 12:03:51 -07001896 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001897
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001898 /* for setting kernel sock attribute on TCP connection setup */
1899#define GSO_MAX_SIZE 65536
1900 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001901#define GSO_MAX_SEGS 65535
1902 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001903
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001904#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001905 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001906#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001907 u8 num_tc;
1908 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1909 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001910
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001911#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001912 unsigned int fcoe_ddp_xid;
1913#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001914#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001915 struct netprio_map __rcu *priomap;
1916#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001917 struct phy_device *phydev;
1918 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001919 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001920 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001922#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923
David S. Millerb5cdae32017-04-18 15:36:58 -04001924static inline bool netif_elide_gro(const struct net_device *dev)
1925{
1926 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1927 return true;
1928 return false;
1929}
1930
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932
David S. Millere8a04642008-07-17 00:34:19 -07001933static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001934int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1935{
1936 return dev->prio_tc_map[prio & TC_BITMASK];
1937}
1938
1939static inline
1940int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1941{
1942 if (tc >= dev->num_tc)
1943 return -EINVAL;
1944
1945 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1946 return 0;
1947}
1948
Alexander Duyck8d059b02016-10-28 11:43:49 -04001949int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
Alexander Duyck9cf1f6a2016-10-28 11:43:20 -04001950void netdev_reset_tc(struct net_device *dev);
1951int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1952int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
John Fastabend4f57c082011-01-17 08:06:04 +00001953
1954static inline
1955int netdev_get_num_tc(struct net_device *dev)
1956{
1957 return dev->num_tc;
1958}
1959
1960static inline
David S. Millere8a04642008-07-17 00:34:19 -07001961struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1962 unsigned int index)
1963{
1964 return &dev->_tx[index];
1965}
1966
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001967static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1968 const struct sk_buff *skb)
1969{
1970 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1971}
1972
David S. Millere8a04642008-07-17 00:34:19 -07001973static inline void netdev_for_each_tx_queue(struct net_device *dev,
1974 void (*f)(struct net_device *,
1975 struct netdev_queue *,
1976 void *),
1977 void *arg)
1978{
1979 unsigned int i;
1980
1981 for (i = 0; i < dev->num_tx_queues; i++)
1982 f(dev, &dev->_tx[i], arg);
1983}
1984
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001985#define netdev_lockdep_set_classes(dev) \
1986{ \
1987 static struct lock_class_key qdisc_tx_busylock_key; \
1988 static struct lock_class_key qdisc_running_key; \
1989 static struct lock_class_key qdisc_xmit_lock_key; \
1990 static struct lock_class_key dev_addr_list_lock_key; \
1991 unsigned int i; \
1992 \
1993 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1994 (dev)->qdisc_running_key = &qdisc_running_key; \
1995 lockdep_set_class(&(dev)->addr_list_lock, \
1996 &dev_addr_list_lock_key); \
1997 for (i = 0; i < (dev)->num_tx_queues; i++) \
1998 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1999 &qdisc_xmit_lock_key); \
2000}
2001
Joe Perchesf629d202013-09-26 14:48:15 -07002002struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08002003 struct sk_buff *skb,
2004 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00002005
Paolo Abeni871b6422016-02-26 10:45:37 +01002006/* returns the headroom that the master device needs to take in account
2007 * when forwarding to this dev
2008 */
2009static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2010{
2011 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2012}
2013
2014static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2015{
2016 if (dev->netdev_ops->ndo_set_rx_headroom)
2017 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2018}
2019
2020/* set the device rx headroom to the dev's default */
2021static inline void netdev_reset_rx_headroom(struct net_device *dev)
2022{
2023 netdev_set_rx_headroom(dev, -1);
2024}
2025
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002026/*
2027 * Net namespace inlines
2028 */
2029static inline
2030struct net *dev_net(const struct net_device *dev)
2031{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002032 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002033}
2034
2035static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002036void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002037{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002038 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002039}
2040
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002041/**
2042 * netdev_priv - access network device private data
2043 * @dev: network device
2044 *
2045 * Get network device private data
2046 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002047static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002049 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050}
2051
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052/* Set the sysfs physical device reference for the network logical device
2053 * if set prior to registration will cause a symlink during initialization.
2054 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002055#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056
Marcel Holtmann384912e2009-08-31 21:08:19 +00002057/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002058 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002059 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002060 */
2061#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2062
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002063/* Default NAPI poll() weight
2064 * Device drivers are strongly advised to not use bigger value
2065 */
2066#define NAPI_POLL_WEIGHT 64
2067
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002068/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002069 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002070 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002071 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002072 * @poll: polling function
2073 * @weight: default weight
2074 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002075 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2076 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002077 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002078void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2079 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002080
Alexander Duyckd8156532008-07-08 15:13:05 -07002081/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002082 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002083 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002084 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002085 * @poll: polling function
2086 * @weight: default weight
2087 *
2088 * This variant of netif_napi_add() should be used from drivers using NAPI
2089 * to exclusively poll a TX queue.
2090 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2091 */
2092static inline void netif_tx_napi_add(struct net_device *dev,
2093 struct napi_struct *napi,
2094 int (*poll)(struct napi_struct *, int),
2095 int weight)
2096{
2097 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2098 netif_napi_add(dev, napi, poll, weight);
2099}
2100
2101/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002102 * netif_napi_del - remove a NAPI context
2103 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002104 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002105 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002106 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002107void netif_napi_del(struct napi_struct *napi);
2108
2109struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002110 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002111 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002112
Herbert Xu74895942009-05-26 18:50:27 +00002113 /* Length of frag0. */
2114 unsigned int frag0_len;
2115
Herbert Xu86911732009-01-29 14:19:50 +00002116 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002117 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002118
Herbert Xud565b0a2008-12-15 23:38:52 -08002119 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002120 u16 flush;
2121
2122 /* Save the IP ID here and check when we get to the transport layer */
2123 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002124
2125 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002126 u16 count;
2127
Tom Herbert15e23962015-02-10 16:30:31 -08002128 /* Start offset for remote checksum offload */
2129 u16 gro_remcsum_start;
2130
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002131 /* jiffies when first packet was created/queued */
2132 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002133
Tom Herbertafe93322014-09-17 12:25:57 -07002134 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002135 u16 proto;
2136
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002137 /* This is non-zero if the packet may be of the same flow. */
2138 u8 same_flow:1;
2139
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002140 /* Used in tunnel GRO receive */
2141 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002142
2143 /* GRO checksum is valid */
2144 u8 csum_valid:1;
2145
Tom Herbert662880f2014-08-27 21:26:56 -07002146 /* Number of checksums via CHECKSUM_UNNECESSARY */
2147 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002148
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002149 /* Free the skb? */
2150 u8 free:2;
2151#define NAPI_GRO_FREE 1
2152#define NAPI_GRO_FREE_STOLEN_HEAD 2
2153
Tom Herbertefc98d02014-10-03 15:48:08 -07002154 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2155 u8 is_ipv6:1;
2156
Alexander Duycka0ca1532016-04-05 09:13:39 -07002157 /* Used in GRE, set in fou/gue_gro_receive */
2158 u8 is_fou:1;
2159
Alexander Duyck15305452016-04-10 21:44:57 -04002160 /* Used to determine if flush_id can be ignored */
2161 u8 is_atomic:1;
2162
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002163 /* Number of gro_receive callbacks this packet already went through */
2164 u8 recursion_counter:4;
2165
2166 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002167
Jerry Chubf5a7552014-01-07 10:23:19 -08002168 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2169 __wsum csum;
2170
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002171 /* used in skb_gro_receive() slow path */
2172 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002173};
2174
2175#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002176
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002177#define GRO_RECURSION_LIMIT 15
2178static inline int gro_recursion_inc_test(struct sk_buff *skb)
2179{
2180 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2181}
2182
2183typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2184static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2185 struct sk_buff **head,
2186 struct sk_buff *skb)
2187{
2188 if (unlikely(gro_recursion_inc_test(skb))) {
2189 NAPI_GRO_CB(skb)->flush |= 1;
2190 return NULL;
2191 }
2192
2193 return cb(head, skb);
2194}
2195
2196typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2197 struct sk_buff *);
2198static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2199 struct sock *sk,
2200 struct sk_buff **head,
2201 struct sk_buff *skb)
2202{
2203 if (unlikely(gro_recursion_inc_test(skb))) {
2204 NAPI_GRO_CB(skb)->flush |= 1;
2205 return NULL;
2206 }
2207
2208 return cb(sk, head, skb);
2209}
2210
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002212 __be16 type; /* This is really htons(ether_type). */
2213 struct net_device *dev; /* NULL is wildcarded here */
2214 int (*func) (struct sk_buff *,
2215 struct net_device *,
2216 struct packet_type *,
2217 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002218 bool (*id_match)(struct packet_type *ptype,
2219 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220 void *af_packet_priv;
2221 struct list_head list;
2222};
2223
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002224struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2226 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002228 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002229 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002230};
2231
2232struct packet_offload {
2233 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002234 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002235 struct offload_callbacks callbacks;
2236 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237};
2238
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002239/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002240struct pcpu_sw_netstats {
2241 u64 rx_packets;
2242 u64 rx_bytes;
2243 u64 tx_packets;
2244 u64 tx_bytes;
2245 struct u64_stats_sync syncp;
2246};
2247
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002248#define __netdev_alloc_pcpu_stats(type, gfp) \
2249({ \
2250 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2251 if (pcpu_stats) { \
2252 int __cpu; \
2253 for_each_possible_cpu(__cpu) { \
2254 typeof(type) *stat; \
2255 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2256 u64_stats_init(&stat->syncp); \
2257 } \
2258 } \
2259 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002260})
2261
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002262#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002263 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002264
Jiri Pirko764f5e52015-12-03 12:12:12 +01002265enum netdev_lag_tx_type {
2266 NETDEV_LAG_TX_TYPE_UNKNOWN,
2267 NETDEV_LAG_TX_TYPE_RANDOM,
2268 NETDEV_LAG_TX_TYPE_BROADCAST,
2269 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2270 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2271 NETDEV_LAG_TX_TYPE_HASH,
2272};
2273
2274struct netdev_lag_upper_info {
2275 enum netdev_lag_tx_type tx_type;
2276};
2277
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002278struct netdev_lag_lower_state_info {
2279 u8 link_up : 1,
2280 tx_enabled : 1;
2281};
2282
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283#include <linux/notifier.h>
2284
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002285/* netdevice notifier chain. Please remember to update the rtnetlink
2286 * notification exclusion list in rtnetlink_event() when adding new
2287 * types.
2288 */
2289#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2290#define NETDEV_DOWN 0x0002
2291#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2292 detected a hardware crash and restarted
2293 - we can use this eg to kick tcp sessions
2294 once done */
2295#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2296#define NETDEV_REGISTER 0x0005
2297#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002298#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002299#define NETDEV_CHANGEADDR 0x0008
2300#define NETDEV_GOING_DOWN 0x0009
2301#define NETDEV_CHANGENAME 0x000A
2302#define NETDEV_FEAT_CHANGE 0x000B
2303#define NETDEV_BONDING_FAILOVER 0x000C
2304#define NETDEV_PRE_UP 0x000D
2305#define NETDEV_PRE_TYPE_CHANGE 0x000E
2306#define NETDEV_POST_TYPE_CHANGE 0x000F
2307#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002308#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002309#define NETDEV_RELEASE 0x0012
2310#define NETDEV_NOTIFY_PEERS 0x0013
2311#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002312#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002313#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002314#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002315#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002316#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002317#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002318#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002319#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002320#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002321
Joe Perchesf629d202013-09-26 14:48:15 -07002322int register_netdevice_notifier(struct notifier_block *nb);
2323int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002324
2325struct netdev_notifier_info {
2326 struct net_device *dev;
2327};
2328
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002329struct netdev_notifier_change_info {
2330 struct netdev_notifier_info info; /* must be first */
2331 unsigned int flags_changed;
2332};
2333
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002334struct netdev_notifier_changeupper_info {
2335 struct netdev_notifier_info info; /* must be first */
2336 struct net_device *upper_dev; /* new upper dev */
2337 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002338 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002339 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002340};
2341
Jiri Pirko04d48262015-12-03 12:12:15 +01002342struct netdev_notifier_changelowerstate_info {
2343 struct netdev_notifier_info info; /* must be first */
2344 void *lower_state_info; /* is lower dev state */
2345};
2346
Cong Wang75538c22013-05-29 11:30:50 +08002347static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2348 struct net_device *dev)
2349{
2350 info->dev = dev;
2351}
2352
Jiri Pirko351638e2013-05-28 01:30:21 +00002353static inline struct net_device *
2354netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2355{
2356 return info->dev;
2357}
2358
Joe Perchesf629d202013-09-26 14:48:15 -07002359int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002360
2361
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362extern rwlock_t dev_base_lock; /* Device list lock */
2363
Eric W. Biederman881d9662007-09-17 11:56:21 -07002364#define for_each_netdev(net, d) \
2365 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002366#define for_each_netdev_reverse(net, d) \
2367 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002368#define for_each_netdev_rcu(net, d) \
2369 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002370#define for_each_netdev_safe(net, d, n) \
2371 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2372#define for_each_netdev_continue(net, d) \
2373 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002374#define for_each_netdev_continue_rcu(net, d) \
2375 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002376#define for_each_netdev_in_bond_rcu(bond, slave) \
2377 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002378 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002379#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2380
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002381static inline struct net_device *next_net_device(struct net_device *dev)
2382{
2383 struct list_head *lh;
2384 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002385
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002386 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002387 lh = dev->dev_list.next;
2388 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2389}
2390
Eric Dumazetce81b762009-11-11 17:34:30 +00002391static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2392{
2393 struct list_head *lh;
2394 struct net *net;
2395
2396 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002397 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002398 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2399}
2400
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002401static inline struct net_device *first_net_device(struct net *net)
2402{
2403 return list_empty(&net->dev_base_head) ? NULL :
2404 net_device_entry(net->dev_base_head.next);
2405}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002406
Eric Dumazetccf43432011-01-26 18:08:02 +00002407static inline struct net_device *first_net_device_rcu(struct net *net)
2408{
2409 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2410
2411 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2412}
2413
Joe Perchesf629d202013-09-26 14:48:15 -07002414int netdev_boot_setup_check(struct net_device *dev);
2415unsigned long netdev_boot_base(const char *prefix, int unit);
2416struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2417 const char *hwaddr);
2418struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2419struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2420void dev_add_pack(struct packet_type *pt);
2421void dev_remove_pack(struct packet_type *pt);
2422void __dev_remove_pack(struct packet_type *pt);
2423void dev_add_offload(struct packet_offload *po);
2424void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002426int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002427int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002428struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2429 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002430struct net_device *dev_get_by_name(struct net *net, const char *name);
2431struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2432struct net_device *__dev_get_by_name(struct net *net, const char *name);
2433int dev_alloc_name(struct net_device *dev, const char *name);
2434int dev_open(struct net_device *dev);
2435int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002436int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002437void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002438int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002439int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002440int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002441int register_netdevice(struct net_device *dev);
2442void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2443void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002444static inline void unregister_netdevice(struct net_device *dev)
2445{
2446 unregister_netdevice_queue(dev, NULL);
2447}
2448
Joe Perchesf629d202013-09-26 14:48:15 -07002449int netdev_refcnt_read(const struct net_device *dev);
2450void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002451void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002452void synchronize_net(void);
2453int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002454
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002455DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002456#define XMIT_RECURSION_LIMIT 10
2457
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002458static inline int dev_recursion_level(void)
2459{
2460 return this_cpu_read(xmit_recursion);
2461}
2462
Joe Perchesf629d202013-09-26 14:48:15 -07002463struct net_device *dev_get_by_index(struct net *net, int ifindex);
2464struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2465struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
Miroslav Lichvar90b602f2017-05-19 17:52:37 +02002466struct net_device *dev_get_by_napi_id(unsigned int napi_id);
Joe Perchesf629d202013-09-26 14:48:15 -07002467int netdev_get_name(struct net *net, char *name, int ifindex);
2468int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002469int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002470
2471static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2472{
2473 return NAPI_GRO_CB(skb)->data_offset;
2474}
2475
2476static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2477{
2478 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2479}
2480
2481static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2482{
2483 NAPI_GRO_CB(skb)->data_offset += len;
2484}
2485
Herbert Xua5b1cf22009-05-26 18:50:28 +00002486static inline void *skb_gro_header_fast(struct sk_buff *skb,
2487 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002488{
Herbert Xu78a478d2009-05-26 18:50:21 +00002489 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002490}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002491
Herbert Xua5b1cf22009-05-26 18:50:28 +00002492static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2493{
2494 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2495}
2496
Herbert Xu57ea52a2017-01-10 12:24:15 -08002497static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2498{
2499 NAPI_GRO_CB(skb)->frag0 = NULL;
2500 NAPI_GRO_CB(skb)->frag0_len = 0;
2501}
2502
Herbert Xua5b1cf22009-05-26 18:50:28 +00002503static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2504 unsigned int offset)
2505{
Herbert Xu17dd7592011-07-27 06:16:28 -07002506 if (!pskb_may_pull(skb, hlen))
2507 return NULL;
2508
Herbert Xu57ea52a2017-01-10 12:24:15 -08002509 skb_gro_frag0_invalidate(skb);
Herbert Xu17dd7592011-07-27 06:16:28 -07002510 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002511}
2512
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002513static inline void *skb_gro_network_header(struct sk_buff *skb)
2514{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002515 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2516 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002517}
2518
Jerry Chubf5a7552014-01-07 10:23:19 -08002519static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2520 const void *start, unsigned int len)
2521{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002522 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002523 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2524 csum_partial(start, len, 0));
2525}
2526
Tom Herbert573e8fc2014-08-22 13:33:47 -07002527/* GRO checksum functions. These are logical equivalents of the normal
2528 * checksum functions (in skbuff.h) except that they operate on the GRO
2529 * offsets and fields in sk_buff.
2530 */
2531
2532__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2533
Tom Herbert15e23962015-02-10 16:30:31 -08002534static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2535{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002536 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002537}
2538
Tom Herbert573e8fc2014-08-22 13:33:47 -07002539static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2540 bool zero_okay,
2541 __sum16 check)
2542{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002543 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2544 skb_checksum_start_offset(skb) <
2545 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002546 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002547 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002548 (!zero_okay || check));
2549}
2550
2551static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2552 __wsum psum)
2553{
2554 if (NAPI_GRO_CB(skb)->csum_valid &&
2555 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2556 return 0;
2557
2558 NAPI_GRO_CB(skb)->csum = psum;
2559
2560 return __skb_gro_checksum_complete(skb);
2561}
2562
Tom Herbert573e8fc2014-08-22 13:33:47 -07002563static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2564{
Tom Herbert662880f2014-08-27 21:26:56 -07002565 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2566 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2567 NAPI_GRO_CB(skb)->csum_cnt--;
2568 } else {
2569 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2570 * verified a new top level checksum or an encapsulated one
2571 * during GRO. This saves work if we fallback to normal path.
2572 */
2573 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002574 }
2575}
2576
2577#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2578 compute_pseudo) \
2579({ \
2580 __sum16 __ret = 0; \
2581 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2582 __ret = __skb_gro_checksum_validate_complete(skb, \
2583 compute_pseudo(skb, proto)); \
Davide Caratti219f1d792017-05-18 15:44:39 +02002584 if (!__ret) \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002585 skb_gro_incr_csum_unnecessary(skb); \
2586 __ret; \
2587})
2588
2589#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2590 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2591
2592#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2593 compute_pseudo) \
2594 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2595
2596#define skb_gro_checksum_simple_validate(skb) \
2597 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2598
Tom Herbertd96535a2014-08-31 15:12:42 -07002599static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2600{
2601 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2602 !NAPI_GRO_CB(skb)->csum_valid);
2603}
2604
2605static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2606 __sum16 check, __wsum pseudo)
2607{
2608 NAPI_GRO_CB(skb)->csum = ~pseudo;
2609 NAPI_GRO_CB(skb)->csum_valid = 1;
2610}
2611
2612#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2613do { \
2614 if (__skb_gro_checksum_convert_check(skb)) \
2615 __skb_gro_checksum_convert(skb, check, \
2616 compute_pseudo(skb, proto)); \
2617} while (0)
2618
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002619struct gro_remcsum {
2620 int offset;
2621 __wsum delta;
2622};
2623
2624static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2625{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002626 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002627 grc->delta = 0;
2628}
2629
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002630static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2631 unsigned int off, size_t hdrlen,
2632 int start, int offset,
2633 struct gro_remcsum *grc,
2634 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002635{
2636 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002637 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002638
2639 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2640
Tom Herbert15e23962015-02-10 16:30:31 -08002641 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002642 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2643 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002644 }
2645
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002646 ptr = skb_gro_header_fast(skb, off);
2647 if (skb_gro_header_hard(skb, off + plen)) {
2648 ptr = skb_gro_header_slow(skb, off + plen, off);
2649 if (!ptr)
2650 return NULL;
2651 }
2652
2653 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2654 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002655
2656 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002657 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002658
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002659 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002660 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002661
2662 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002663}
2664
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002665static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2666 struct gro_remcsum *grc)
2667{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002668 void *ptr;
2669 size_t plen = grc->offset + sizeof(u16);
2670
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002671 if (!grc->delta)
2672 return;
2673
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002674 ptr = skb_gro_header_fast(skb, grc->offset);
2675 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2676 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2677 if (!ptr)
2678 return;
2679 }
2680
2681 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002682}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002683
Steffen Klassert25393d32017-02-15 09:39:44 +01002684#ifdef CONFIG_XFRM_OFFLOAD
2685static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2686{
2687 if (PTR_ERR(pp) != -EINPROGRESS)
2688 NAPI_GRO_CB(skb)->flush |= flush;
2689}
2690#else
Steffen Klassert5f114162017-02-15 09:39:39 +01002691static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2692{
2693 NAPI_GRO_CB(skb)->flush |= flush;
2694}
Steffen Klassert25393d32017-02-15 09:39:44 +01002695#endif
Steffen Klassert5f114162017-02-15 09:39:39 +01002696
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002697static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2698 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002699 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002700 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002701{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002702 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002703 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002704
2705 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002706}
2707
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002708static inline int dev_parse_header(const struct sk_buff *skb,
2709 unsigned char *haddr)
2710{
2711 const struct net_device *dev = skb->dev;
2712
Patrick McHardy1b833362007-10-18 05:09:28 -07002713 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002714 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002715 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002716}
2717
Willem de Bruijn2793a232016-03-09 21:58:32 -05002718/* ll_header must have at least hard_header_len allocated */
2719static inline bool dev_validate_header(const struct net_device *dev,
2720 char *ll_header, int len)
2721{
2722 if (likely(len >= dev->hard_header_len))
2723 return true;
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05002724 if (len < dev->min_header_len)
2725 return false;
Willem de Bruijn2793a232016-03-09 21:58:32 -05002726
2727 if (capable(CAP_SYS_RAWIO)) {
2728 memset(ll_header + len, 0, dev->hard_header_len - len);
2729 return true;
2730 }
2731
2732 if (dev->header_ops && dev->header_ops->validate)
2733 return dev->header_ops->validate(ll_header, len);
2734
2735 return false;
2736}
2737
Linus Torvalds1da177e2005-04-16 15:20:36 -07002738typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002739int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002740static inline int unregister_gifconf(unsigned int family)
2741{
2742 return register_gifconf(family, NULL);
2743}
2744
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002745#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002746#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002747struct sd_flow_limit {
2748 u64 count;
2749 unsigned int num_buckets;
2750 unsigned int history_head;
2751 u16 history[FLOW_LIMIT_HISTORY];
2752 u8 buckets[];
2753};
2754
2755extern int netdev_flow_limit_table_len;
2756#endif /* CONFIG_NET_FLOW_LIMIT */
2757
Linus Torvalds1da177e2005-04-16 15:20:36 -07002758/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002759 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002760 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002761struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002762 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002763 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764
Changli Gaodee42872010-05-02 05:42:16 +00002765 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002766 unsigned int processed;
2767 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002768 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002769#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002770 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002771#endif
2772#ifdef CONFIG_NET_FLOW_LIMIT
2773 struct sd_flow_limit __rcu *flow_limit;
2774#endif
2775 struct Qdisc *output_queue;
2776 struct Qdisc **output_queue_tailp;
2777 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002778
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002779#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002780 /* input_queue_head should be written by cpu owning this struct,
2781 * and only read by other cpus. Worth using a cache line.
2782 */
2783 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2784
2785 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002786 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002787 struct softnet_data *rps_ipi_next;
2788 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002789 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002790#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002791 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002792 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002793 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002794
Linus Torvalds1da177e2005-04-16 15:20:36 -07002795};
2796
Tom Herbert76cc8b12010-05-20 18:37:59 +00002797static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002798{
2799#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002800 sd->input_queue_head++;
2801#endif
2802}
2803
2804static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2805 unsigned int *qtail)
2806{
2807#ifdef CONFIG_RPS
2808 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002809#endif
2810}
2811
Tom Herbert0a9627f2010-03-16 08:03:29 +00002812DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002813
Joe Perchesf629d202013-09-26 14:48:15 -07002814void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002815void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002816
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002817static inline void netif_tx_schedule_all(struct net_device *dev)
2818{
2819 unsigned int i;
2820
2821 for (i = 0; i < dev->num_tx_queues; i++)
2822 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2823}
2824
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002825static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002826{
Tom Herbert734664982011-11-28 16:32:44 +00002827 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002828}
2829
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002830/**
2831 * netif_start_queue - allow transmit
2832 * @dev: network device
2833 *
2834 * Allow upper layers to call the device hard_start_xmit routine.
2835 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002836static inline void netif_start_queue(struct net_device *dev)
2837{
David S. Millere8a04642008-07-17 00:34:19 -07002838 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002839}
2840
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002841static inline void netif_tx_start_all_queues(struct net_device *dev)
2842{
2843 unsigned int i;
2844
2845 for (i = 0; i < dev->num_tx_queues; i++) {
2846 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2847 netif_tx_start_queue(txq);
2848 }
2849}
2850
John Fastabend46e5da40a2014-09-12 20:04:52 -07002851void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002852
Dave Jonesd29f7492008-07-22 14:09:06 -07002853/**
2854 * netif_wake_queue - restart transmit
2855 * @dev: network device
2856 *
2857 * Allow upper layers to call the device hard_start_xmit routine.
2858 * Used for flow control when transmit resources are available.
2859 */
David S. Miller79d16382008-07-08 23:14:46 -07002860static inline void netif_wake_queue(struct net_device *dev)
2861{
David S. Millere8a04642008-07-17 00:34:19 -07002862 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002863}
2864
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002865static inline void netif_tx_wake_all_queues(struct net_device *dev)
2866{
2867 unsigned int i;
2868
2869 for (i = 0; i < dev->num_tx_queues; i++) {
2870 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2871 netif_tx_wake_queue(txq);
2872 }
2873}
2874
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002875static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002876{
Tom Herbert734664982011-11-28 16:32:44 +00002877 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002878}
2879
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002880/**
2881 * netif_stop_queue - stop transmitted packets
2882 * @dev: network device
2883 *
2884 * Stop upper layers calling the device hard_start_xmit routine.
2885 * Used for flow control when transmit resources are unavailable.
2886 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002887static inline void netif_stop_queue(struct net_device *dev)
2888{
David S. Millere8a04642008-07-17 00:34:19 -07002889 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002890}
2891
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002892void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002893
David S. Miller4d295152012-03-07 21:02:35 -05002894static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002895{
Tom Herbert734664982011-11-28 16:32:44 +00002896 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002897}
2898
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002899/**
2900 * netif_queue_stopped - test if transmit queue is flowblocked
2901 * @dev: network device
2902 *
2903 * Test if transmit queue on device is currently unable to send.
2904 */
David S. Miller4d295152012-03-07 21:02:35 -05002905static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002906{
David S. Millere8a04642008-07-17 00:34:19 -07002907 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002908}
2909
David S. Miller4d295152012-03-07 21:02:35 -05002910static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002911{
Tom Herbert734664982011-11-28 16:32:44 +00002912 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2913}
2914
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002915static inline bool
2916netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002917{
2918 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2919}
2920
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002921static inline bool
2922netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2923{
2924 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2925}
2926
Eric Dumazet53511452014-10-08 08:19:27 -07002927/**
2928 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2929 * @dev_queue: pointer to transmit queue
2930 *
2931 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002932 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002933 */
2934static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2935{
2936#ifdef CONFIG_BQL
2937 prefetchw(&dev_queue->dql.num_queued);
2938#endif
2939}
2940
2941/**
2942 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2943 * @dev_queue: pointer to transmit queue
2944 *
2945 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002946 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002947 */
2948static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2949{
2950#ifdef CONFIG_BQL
2951 prefetchw(&dev_queue->dql.limit);
2952#endif
2953}
2954
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002955static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2956 unsigned int bytes)
2957{
Tom Herbert114cf582011-11-28 16:33:09 +00002958#ifdef CONFIG_BQL
2959 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002960
2961 if (likely(dql_avail(&dev_queue->dql) >= 0))
2962 return;
2963
2964 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2965
2966 /*
2967 * The XOFF flag must be set before checking the dql_avail below,
2968 * because in netdev_tx_completed_queue we update the dql_completed
2969 * before checking the XOFF flag.
2970 */
2971 smp_mb();
2972
2973 /* check again in case another CPU has just made room avail */
2974 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2975 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002976#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002977}
2978
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002979/**
2980 * netdev_sent_queue - report the number of bytes queued to hardware
2981 * @dev: network device
2982 * @bytes: number of bytes queued to the hardware device queue
2983 *
2984 * Report the number of bytes queued for sending/completion to the network
2985 * device hardware queue. @bytes should be a good approximation and should
2986 * exactly match netdev_completed_queue() @bytes
2987 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002988static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2989{
2990 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2991}
2992
2993static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002994 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002995{
Tom Herbert114cf582011-11-28 16:33:09 +00002996#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002997 if (unlikely(!bytes))
2998 return;
2999
3000 dql_completed(&dev_queue->dql, bytes);
3001
3002 /*
3003 * Without the memory barrier there is a small possiblity that
3004 * netdev_tx_sent_queue will miss the update and cause the queue to
3005 * be stopped forever
3006 */
3007 smp_mb();
3008
3009 if (dql_avail(&dev_queue->dql) < 0)
3010 return;
3011
3012 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3013 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00003014#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003015}
3016
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003017/**
3018 * netdev_completed_queue - report bytes and packets completed by device
3019 * @dev: network device
3020 * @pkts: actual number of packets sent over the medium
3021 * @bytes: actual number of bytes sent over the medium
3022 *
3023 * Report the number of bytes and packets transmitted by the network device
3024 * hardware queue over the physical medium, @bytes must exactly match the
3025 * @bytes amount passed to netdev_sent_queue()
3026 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003027static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003028 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003029{
3030 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3031}
3032
3033static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3034{
Tom Herbert114cf582011-11-28 16:33:09 +00003035#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003036 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003037 dql_reset(&q->dql);
3038#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003039}
3040
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003041/**
3042 * netdev_reset_queue - reset the packets and bytes count of a network device
3043 * @dev_queue: network device
3044 *
3045 * Reset the bytes and packet count of a network device and clear the
3046 * software flow control OFF bit for this network device
3047 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003048static inline void netdev_reset_queue(struct net_device *dev_queue)
3049{
3050 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003051}
3052
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003053/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003054 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3055 * @dev: network device
3056 * @queue_index: given tx queue index
3057 *
3058 * Returns 0 if given tx queue index >= number of device tx queues,
3059 * otherwise returns the originally passed tx queue index.
3060 */
3061static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3062{
3063 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3064 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3065 dev->name, queue_index,
3066 dev->real_num_tx_queues);
3067 return 0;
3068 }
3069
3070 return queue_index;
3071}
3072
3073/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003074 * netif_running - test if up
3075 * @dev: network device
3076 *
3077 * Test if the device has been brought up.
3078 */
David S. Miller4d295152012-03-07 21:02:35 -05003079static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003080{
3081 return test_bit(__LINK_STATE_START, &dev->state);
3082}
3083
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003084/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003085 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003086 * stop, and a check if it's stopped. All other device management is
3087 * done at the overall netdevice level.
3088 * Also test the device if we're multiqueue.
3089 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003090
3091/**
3092 * netif_start_subqueue - allow sending packets on subqueue
3093 * @dev: network device
3094 * @queue_index: sub queue index
3095 *
3096 * Start individual transmit queue of a device with multiple transmit queues.
3097 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003098static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3099{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003100 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003101
3102 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003103}
3104
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003105/**
3106 * netif_stop_subqueue - stop sending packets on subqueue
3107 * @dev: network device
3108 * @queue_index: sub queue index
3109 *
3110 * Stop individual transmit queue of a device with multiple transmit queues.
3111 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003112static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3113{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003114 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003115 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003116}
3117
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003118/**
3119 * netif_subqueue_stopped - test status of subqueue
3120 * @dev: network device
3121 * @queue_index: sub queue index
3122 *
3123 * Check individual transmit queue of a device with multiple transmit queues.
3124 */
David S. Miller4d295152012-03-07 21:02:35 -05003125static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3126 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003127{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003128 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003129
3130 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003131}
3132
David S. Miller4d295152012-03-07 21:02:35 -05003133static inline bool netif_subqueue_stopped(const struct net_device *dev,
3134 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003135{
3136 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3137}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003138
Florian Fainelli738b35c2017-01-11 21:13:02 -08003139/**
3140 * netif_wake_subqueue - allow sending packets on subqueue
3141 * @dev: network device
3142 * @queue_index: sub queue index
3143 *
3144 * Resume individual transmit queue of a device with multiple transmit queues.
3145 */
3146static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3147{
3148 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3149
3150 netif_tx_wake_queue(txq);
3151}
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003152
Alexander Duyck537c00d2013-01-10 08:57:02 +00003153#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003154int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003155 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003156#else
3157static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003158 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003159 u16 index)
3160{
3161 return 0;
3162}
3163#endif
3164
Jiri Pirko5605c762015-05-12 14:56:12 +02003165u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3166 unsigned int num_tx_queues);
3167
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003168/*
3169 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3170 * as a distribution range limit for the returned value.
3171 */
3172static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003173 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003174{
3175 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3176}
3177
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003178/**
3179 * netif_is_multiqueue - test if device has multiple transmit queues
3180 * @dev: network device
3181 *
3182 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003183 */
David S. Miller4d295152012-03-07 21:02:35 -05003184static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003185{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003186 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003187}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003188
Joe Perchesf629d202013-09-26 14:48:15 -07003189int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003190
Michael Daltona953be52014-01-16 22:23:28 -08003191#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003192int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003193#else
3194static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3195 unsigned int rxq)
3196{
3197 return 0;
3198}
3199#endif
3200
Michael Daltona953be52014-01-16 22:23:28 -08003201#ifdef CONFIG_SYSFS
3202static inline unsigned int get_netdev_rx_queue_index(
3203 struct netdev_rx_queue *queue)
3204{
3205 struct net_device *dev = queue->dev;
3206 int index = queue - dev->_rx;
3207
3208 BUG_ON(index >= dev->num_rx_queues);
3209 return index;
3210}
3211#endif
3212
Yuval Mintz16917b82012-07-01 03:18:50 +00003213#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003214int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003215
Eric Dumazete6247022013-12-05 04:45:08 -08003216enum skb_free_reason {
3217 SKB_REASON_CONSUMED,
3218 SKB_REASON_DROPPED,
3219};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003220
Eric Dumazete6247022013-12-05 04:45:08 -08003221void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3222void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3223
3224/*
3225 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3226 * interrupt context or with hardware interrupts being disabled.
3227 * (in_irq() || irqs_disabled())
3228 *
3229 * We provide four helpers that can be used in following contexts :
3230 *
3231 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3232 * replacing kfree_skb(skb)
3233 *
3234 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3235 * Typically used in place of consume_skb(skb) in TX completion path
3236 *
3237 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3238 * replacing kfree_skb(skb)
3239 *
3240 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3241 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003242 */
Eric Dumazete6247022013-12-05 04:45:08 -08003243static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3244{
3245 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3246}
3247
3248static inline void dev_consume_skb_irq(struct sk_buff *skb)
3249{
3250 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3251}
3252
3253static inline void dev_kfree_skb_any(struct sk_buff *skb)
3254{
3255 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3256}
3257
3258static inline void dev_consume_skb_any(struct sk_buff *skb)
3259{
3260 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3261}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003262
Joe Perchesf629d202013-09-26 14:48:15 -07003263int netif_rx(struct sk_buff *skb);
3264int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003265int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003266gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3267void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3268struct sk_buff *napi_get_frags(struct napi_struct *napi);
3269gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003270struct packet_offload *gro_find_receive_by_type(__be16 type);
3271struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003272
3273static inline void napi_free_frags(struct napi_struct *napi)
3274{
3275 kfree_skb(napi->skb);
3276 napi->skb = NULL;
3277}
3278
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003279bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003280int netdev_rx_handler_register(struct net_device *dev,
3281 rx_handler_func_t *rx_handler,
3282 void *rx_handler_data);
3283void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003284
Joe Perchesf629d202013-09-26 14:48:15 -07003285bool dev_valid_name(const char *name);
3286int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3287int dev_ethtool(struct net *net, struct ifreq *);
3288unsigned int dev_get_flags(const struct net_device *);
3289int __dev_change_flags(struct net_device *, unsigned int flags);
3290int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003291void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3292 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003293int dev_change_name(struct net_device *, const char *);
3294int dev_set_alias(struct net_device *, const char *, size_t);
3295int dev_change_net_namespace(struct net_device *, struct net *, const char *);
WANG Congf51048c2017-07-06 15:01:57 -07003296int __dev_set_mtu(struct net_device *, int);
Joe Perchesf629d202013-09-26 14:48:15 -07003297int dev_set_mtu(struct net_device *, int);
3298void dev_set_group(struct net_device *, int);
3299int dev_set_mac_address(struct net_device *, struct sockaddr *);
3300int dev_change_carrier(struct net_device *, bool new_carrier);
3301int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003302 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003303int dev_get_phys_port_name(struct net_device *dev,
3304 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003305int dev_change_proto_down(struct net_device *dev, bool proto_down);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003306struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003307struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3308 struct netdev_queue *txq, int *ret);
Daniel Borkmannd67b9cd2017-05-12 01:04:46 +02003309
3310typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3311int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3312 int fd, u32 flags);
Jakub Kicinskice158e52017-06-21 18:25:09 -07003313u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
Daniel Borkmannd67b9cd2017-05-12 01:04:46 +02003314
Herbert Xua0265d22014-04-17 13:45:03 +08003315int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003316int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003317bool is_skb_forwardable(const struct net_device *dev,
3318 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003319
Martin KaFai Lau4e3264d2016-11-09 15:36:33 -08003320static __always_inline int ____dev_forward_skb(struct net_device *dev,
3321 struct sk_buff *skb)
3322{
3323 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3324 unlikely(!is_skb_forwardable(dev, skb))) {
3325 atomic_long_inc(&dev->rx_dropped);
3326 kfree_skb(skb);
3327 return NET_RX_DROP;
3328 }
3329
3330 skb_scrub_packet(skb, true);
3331 skb->priority = 0;
3332 return 0;
3333}
3334
David Ahern74b20582016-05-10 11:19:50 -07003335void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3336
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003337extern int netdev_budget;
Matthew Whitehead7acf8a12017-04-19 12:37:10 -04003338extern unsigned int netdev_budget_usecs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003339
3340/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003341void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003342
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003343/**
3344 * dev_put - release reference to device
3345 * @dev: network device
3346 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003347 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003348 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003349static inline void dev_put(struct net_device *dev)
3350{
Christoph Lameter933393f2011-12-22 11:58:51 -06003351 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003352}
3353
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003354/**
3355 * dev_hold - get reference to device
3356 * @dev: network device
3357 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003358 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003359 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003360static inline void dev_hold(struct net_device *dev)
3361{
Christoph Lameter933393f2011-12-22 11:58:51 -06003362 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003363}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003364
3365/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3366 * and _off may be called from IRQ context, but it is caller
3367 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003368 *
3369 * The name carrier is inappropriate, these functions should really be
3370 * called netif_lowerlayer_*() because they represent the state of any
3371 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003372 */
3373
Joe Perchesf629d202013-09-26 14:48:15 -07003374void linkwatch_init_dev(struct net_device *dev);
3375void linkwatch_fire_event(struct net_device *dev);
3376void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003377
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003378/**
3379 * netif_carrier_ok - test if carrier present
3380 * @dev: network device
3381 *
3382 * Check if carrier is present on device
3383 */
David S. Miller4d295152012-03-07 21:02:35 -05003384static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003385{
3386 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3387}
3388
Joe Perchesf629d202013-09-26 14:48:15 -07003389unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003390
Joe Perchesf629d202013-09-26 14:48:15 -07003391void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003392
Joe Perchesf629d202013-09-26 14:48:15 -07003393void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003394
Joe Perchesf629d202013-09-26 14:48:15 -07003395void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003396
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003397/**
3398 * netif_dormant_on - mark device as dormant.
3399 * @dev: network device
3400 *
3401 * Mark device as dormant (as per RFC2863).
3402 *
3403 * The dormant state indicates that the relevant interface is not
3404 * actually in a condition to pass packets (i.e., it is not 'up') but is
3405 * in a "pending" state, waiting for some external event. For "on-
3406 * demand" interfaces, this new state identifies the situation where the
3407 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003408 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003409static inline void netif_dormant_on(struct net_device *dev)
3410{
3411 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3412 linkwatch_fire_event(dev);
3413}
3414
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003415/**
3416 * netif_dormant_off - set device as not dormant.
3417 * @dev: network device
3418 *
3419 * Device is not in dormant state.
3420 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003421static inline void netif_dormant_off(struct net_device *dev)
3422{
3423 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3424 linkwatch_fire_event(dev);
3425}
3426
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003427/**
Zhang Shengju8ecbc402017-04-26 11:05:12 +08003428 * netif_dormant - test if device is dormant
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003429 * @dev: network device
3430 *
Zhang Shengju8ecbc402017-04-26 11:05:12 +08003431 * Check if device is dormant.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003432 */
David S. Miller4d295152012-03-07 21:02:35 -05003433static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003434{
3435 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3436}
3437
3438
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003439/**
3440 * netif_oper_up - test if device is operational
3441 * @dev: network device
3442 *
3443 * Check if carrier is operational
3444 */
David S. Miller4d295152012-03-07 21:02:35 -05003445static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003446{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003447 return (dev->operstate == IF_OPER_UP ||
3448 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3449}
3450
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003451/**
3452 * netif_device_present - is device available or removed
3453 * @dev: network device
3454 *
3455 * Check if device has not been removed from system.
3456 */
David S. Miller4d295152012-03-07 21:02:35 -05003457static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003458{
3459 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3460}
3461
Joe Perchesf629d202013-09-26 14:48:15 -07003462void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003463
Joe Perchesf629d202013-09-26 14:48:15 -07003464void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003465
3466/*
3467 * Network interface message level settings
3468 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003469
3470enum {
3471 NETIF_MSG_DRV = 0x0001,
3472 NETIF_MSG_PROBE = 0x0002,
3473 NETIF_MSG_LINK = 0x0004,
3474 NETIF_MSG_TIMER = 0x0008,
3475 NETIF_MSG_IFDOWN = 0x0010,
3476 NETIF_MSG_IFUP = 0x0020,
3477 NETIF_MSG_RX_ERR = 0x0040,
3478 NETIF_MSG_TX_ERR = 0x0080,
3479 NETIF_MSG_TX_QUEUED = 0x0100,
3480 NETIF_MSG_INTR = 0x0200,
3481 NETIF_MSG_TX_DONE = 0x0400,
3482 NETIF_MSG_RX_STATUS = 0x0800,
3483 NETIF_MSG_PKTDATA = 0x1000,
3484 NETIF_MSG_HW = 0x2000,
3485 NETIF_MSG_WOL = 0x4000,
3486};
3487
3488#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3489#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3490#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3491#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3492#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3493#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3494#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3495#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3496#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3497#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3498#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3499#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3500#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3501#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3502#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3503
3504static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3505{
3506 /* use default */
3507 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3508 return default_msg_enable_bits;
3509 if (debug_value == 0) /* no output */
3510 return 0;
3511 /* set low N bits */
3512 return (1 << debug_value) - 1;
3513}
3514
David S. Millerc773e842008-07-08 23:13:53 -07003515static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003516{
David S. Millerc773e842008-07-08 23:13:53 -07003517 spin_lock(&txq->_xmit_lock);
3518 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003519}
3520
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003521static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3522{
3523 __acquire(&txq->_xmit_lock);
3524 return true;
3525}
3526
3527static inline void __netif_tx_release(struct netdev_queue *txq)
3528{
3529 __release(&txq->_xmit_lock);
3530}
3531
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003532static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3533{
3534 spin_lock_bh(&txq->_xmit_lock);
3535 txq->xmit_lock_owner = smp_processor_id();
3536}
3537
David S. Miller4d295152012-03-07 21:02:35 -05003538static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003539{
David S. Miller4d295152012-03-07 21:02:35 -05003540 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003541 if (likely(ok))
3542 txq->xmit_lock_owner = smp_processor_id();
3543 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003544}
3545
David S. Millerc773e842008-07-08 23:13:53 -07003546static inline void __netif_tx_unlock(struct netdev_queue *txq)
3547{
3548 txq->xmit_lock_owner = -1;
3549 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003550}
3551
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003552static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3553{
3554 txq->xmit_lock_owner = -1;
3555 spin_unlock_bh(&txq->_xmit_lock);
3556}
3557
Eric Dumazet08baf562009-05-25 22:58:01 -07003558static inline void txq_trans_update(struct netdev_queue *txq)
3559{
3560 if (txq->xmit_lock_owner != -1)
3561 txq->trans_start = jiffies;
3562}
3563
Florian Westphalba162f82016-05-03 16:31:00 +02003564/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3565static inline void netif_trans_update(struct net_device *dev)
3566{
Florian Westphal9b366272016-05-03 16:33:14 +02003567 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3568
3569 if (txq->trans_start != jiffies)
3570 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003571}
3572
David S. Millerc3f26a22008-07-31 16:58:50 -07003573/**
3574 * netif_tx_lock - grab network device transmit lock
3575 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003576 *
3577 * Get network device transmit lock
3578 */
3579static inline void netif_tx_lock(struct net_device *dev)
3580{
3581 unsigned int i;
3582 int cpu;
3583
3584 spin_lock(&dev->tx_global_lock);
3585 cpu = smp_processor_id();
3586 for (i = 0; i < dev->num_tx_queues; i++) {
3587 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3588
3589 /* We are the only thread of execution doing a
3590 * freeze, but we have to grab the _xmit_lock in
3591 * order to synchronize with threads which are in
3592 * the ->hard_start_xmit() handler and already
3593 * checked the frozen bit.
3594 */
3595 __netif_tx_lock(txq, cpu);
3596 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3597 __netif_tx_unlock(txq);
3598 }
3599}
3600
3601static inline void netif_tx_lock_bh(struct net_device *dev)
3602{
3603 local_bh_disable();
3604 netif_tx_lock(dev);
3605}
3606
Herbert Xu932ff272006-06-09 12:20:56 -07003607static inline void netif_tx_unlock(struct net_device *dev)
3608{
David S. Millere8a04642008-07-17 00:34:19 -07003609 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003610
David S. Millere8a04642008-07-17 00:34:19 -07003611 for (i = 0; i < dev->num_tx_queues; i++) {
3612 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003613
David S. Millerc3f26a22008-07-31 16:58:50 -07003614 /* No need to grab the _xmit_lock here. If the
3615 * queue is not stopped for another reason, we
3616 * force a schedule.
3617 */
3618 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003619 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003620 }
3621 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003622}
3623
3624static inline void netif_tx_unlock_bh(struct net_device *dev)
3625{
David S. Millere8a04642008-07-17 00:34:19 -07003626 netif_tx_unlock(dev);
3627 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003628}
3629
David S. Millerc773e842008-07-08 23:13:53 -07003630#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003631 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003632 __netif_tx_lock(txq, cpu); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003633 } else { \
3634 __netif_tx_acquire(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003635 } \
3636}
3637
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003638#define HARD_TX_TRYLOCK(dev, txq) \
3639 (((dev->features & NETIF_F_LLTX) == 0) ? \
3640 __netif_tx_trylock(txq) : \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003641 __netif_tx_acquire(txq))
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003642
David S. Millerc773e842008-07-08 23:13:53 -07003643#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003644 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003645 __netif_tx_unlock(txq); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003646 } else { \
3647 __netif_tx_release(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003648 } \
3649}
3650
Linus Torvalds1da177e2005-04-16 15:20:36 -07003651static inline void netif_tx_disable(struct net_device *dev)
3652{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003653 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003654 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003655
David S. Millerc3f26a22008-07-31 16:58:50 -07003656 local_bh_disable();
3657 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003658 for (i = 0; i < dev->num_tx_queues; i++) {
3659 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003660
3661 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003662 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003663 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003664 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003665 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003666}
3667
David S. Millere308a5d2008-07-15 00:13:44 -07003668static inline void netif_addr_lock(struct net_device *dev)
3669{
3670 spin_lock(&dev->addr_list_lock);
3671}
3672
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003673static inline void netif_addr_lock_nested(struct net_device *dev)
3674{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003675 int subclass = SINGLE_DEPTH_NESTING;
3676
3677 if (dev->netdev_ops->ndo_get_lock_subclass)
3678 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3679
3680 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003681}
3682
David S. Millere308a5d2008-07-15 00:13:44 -07003683static inline void netif_addr_lock_bh(struct net_device *dev)
3684{
3685 spin_lock_bh(&dev->addr_list_lock);
3686}
3687
3688static inline void netif_addr_unlock(struct net_device *dev)
3689{
3690 spin_unlock(&dev->addr_list_lock);
3691}
3692
3693static inline void netif_addr_unlock_bh(struct net_device *dev)
3694{
3695 spin_unlock_bh(&dev->addr_list_lock);
3696}
3697
Jiri Pirkof001fde2009-05-05 02:48:28 +00003698/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003699 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003700 * rcu_read_lock held.
3701 */
3702#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003703 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003704
Linus Torvalds1da177e2005-04-16 15:20:36 -07003705/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3706
Joe Perchesf629d202013-09-26 14:48:15 -07003707void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003708
3709/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003710struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003711 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003712 void (*setup)(struct net_device *),
3713 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003714#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3715 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003716
Tom Gundersenc835a672014-07-14 16:37:24 +02003717#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3718 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3719 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003720
Joe Perchesf629d202013-09-26 14:48:15 -07003721int register_netdev(struct net_device *dev);
3722void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003723
Jiri Pirko22bedad32010-04-01 21:22:57 +00003724/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003725int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3726 struct netdev_hw_addr_list *from_list, int addr_len);
3727void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3728 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003729int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3730 struct net_device *dev,
3731 int (*sync)(struct net_device *, const unsigned char *),
3732 int (*unsync)(struct net_device *,
3733 const unsigned char *));
3734void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3735 struct net_device *dev,
3736 int (*unsync)(struct net_device *,
3737 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003738void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003739
Jiri Pirkof001fde2009-05-05 02:48:28 +00003740/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003741int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3742 unsigned char addr_type);
3743int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3744 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003745void dev_addr_flush(struct net_device *dev);
3746int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003747
3748/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003749int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3750int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3751int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3752int dev_uc_sync(struct net_device *to, struct net_device *from);
3753int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3754void dev_uc_unsync(struct net_device *to, struct net_device *from);
3755void dev_uc_flush(struct net_device *dev);
3756void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003757
Alexander Duyck670e5b82014-05-28 18:44:46 -07003758/**
3759 * __dev_uc_sync - Synchonize device's unicast list
3760 * @dev: device to sync
3761 * @sync: function to call if address should be added
3762 * @unsync: function to call if address should be removed
3763 *
3764 * Add newly added addresses to the interface, and release
3765 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003766 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003767static inline int __dev_uc_sync(struct net_device *dev,
3768 int (*sync)(struct net_device *,
3769 const unsigned char *),
3770 int (*unsync)(struct net_device *,
3771 const unsigned char *))
3772{
3773 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3774}
3775
3776/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003777 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003778 * @dev: device to sync
3779 * @unsync: function to call if address should be removed
3780 *
3781 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003782 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003783static inline void __dev_uc_unsync(struct net_device *dev,
3784 int (*unsync)(struct net_device *,
3785 const unsigned char *))
3786{
3787 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3788}
3789
Jiri Pirko22bedad32010-04-01 21:22:57 +00003790/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003791int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3792int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3793int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3794int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3795int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3796int dev_mc_sync(struct net_device *to, struct net_device *from);
3797int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3798void dev_mc_unsync(struct net_device *to, struct net_device *from);
3799void dev_mc_flush(struct net_device *dev);
3800void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003801
Alexander Duyck670e5b82014-05-28 18:44:46 -07003802/**
3803 * __dev_mc_sync - Synchonize device's multicast list
3804 * @dev: device to sync
3805 * @sync: function to call if address should be added
3806 * @unsync: function to call if address should be removed
3807 *
3808 * Add newly added addresses to the interface, and release
3809 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003810 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003811static inline int __dev_mc_sync(struct net_device *dev,
3812 int (*sync)(struct net_device *,
3813 const unsigned char *),
3814 int (*unsync)(struct net_device *,
3815 const unsigned char *))
3816{
3817 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3818}
3819
3820/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003821 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003822 * @dev: device to sync
3823 * @unsync: function to call if address should be removed
3824 *
3825 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003826 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003827static inline void __dev_mc_unsync(struct net_device *dev,
3828 int (*unsync)(struct net_device *,
3829 const unsigned char *))
3830{
3831 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3832}
3833
Linus Torvalds1da177e2005-04-16 15:20:36 -07003834/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003835void dev_set_rx_mode(struct net_device *dev);
3836void __dev_set_rx_mode(struct net_device *dev);
3837int dev_set_promiscuity(struct net_device *dev, int inc);
3838int dev_set_allmulti(struct net_device *dev, int inc);
3839void netdev_state_change(struct net_device *dev);
3840void netdev_notify_peers(struct net_device *dev);
3841void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003842/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003843void dev_load(struct net *net, const char *name);
3844struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3845 struct rtnl_link_stats64 *storage);
3846void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3847 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003848
3849extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003850extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003851extern int weight_p;
Matthias Tafelmeier3d48b532016-12-29 21:37:21 +01003852extern int dev_weight_rx_bias;
3853extern int dev_weight_tx_bias;
3854extern int dev_rx_weight;
3855extern int dev_tx_weight;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003856
Joe Perchesf629d202013-09-26 14:48:15 -07003857bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003858struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3859 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003860struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3861 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003862
3863/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003864#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3865 for (iter = &(dev)->adj_list.upper, \
3866 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3867 updev; \
3868 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3869
David Ahern1a3f0602016-10-17 19:15:44 -07003870int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3871 int (*fn)(struct net_device *upper_dev,
3872 void *data),
3873 void *data);
3874
3875bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3876 struct net_device *upper_dev);
3877
Joe Perchesf629d202013-09-26 14:48:15 -07003878void *netdev_lower_get_next_private(struct net_device *dev,
3879 struct list_head **iter);
3880void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3881 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003882
3883#define netdev_for_each_lower_private(dev, priv, iter) \
3884 for (iter = (dev)->adj_list.lower.next, \
3885 priv = netdev_lower_get_next_private(dev, &(iter)); \
3886 priv; \
3887 priv = netdev_lower_get_next_private(dev, &(iter)))
3888
3889#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3890 for (iter = &(dev)->adj_list.lower, \
3891 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3892 priv; \
3893 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3894
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003895void *netdev_lower_get_next(struct net_device *dev,
3896 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003897
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003898#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003899 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003900 ldev = netdev_lower_get_next(dev, &(iter)); \
3901 ldev; \
3902 ldev = netdev_lower_get_next(dev, &(iter)))
3903
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003904struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3905 struct list_head **iter);
3906struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3907 struct list_head **iter);
3908
David Ahern1a3f0602016-10-17 19:15:44 -07003909int netdev_walk_all_lower_dev(struct net_device *dev,
3910 int (*fn)(struct net_device *lower_dev,
3911 void *data),
3912 void *data);
3913int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3914 int (*fn)(struct net_device *lower_dev,
3915 void *data),
3916 void *data);
3917
Joe Perchesf629d202013-09-26 14:48:15 -07003918void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003919void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003920struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3921struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3922int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3923int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003924 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003925 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003926void netdev_upper_dev_unlink(struct net_device *dev,
3927 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003928void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003929void *netdev_lower_dev_get_private(struct net_device *dev,
3930 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003931void netdev_lower_state_changed(struct net_device *lower_dev,
3932 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003933
3934/* RSS keys are 40 or 52 bytes long */
3935#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003936extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003937void netdev_rss_key_fill(void *buffer, size_t len);
3938
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003939int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003940int skb_checksum_help(struct sk_buff *skb);
Davide Carattib72b5bf2017-05-18 15:44:38 +02003941int skb_crc32c_csum_help(struct sk_buff *skb);
Davide Caratti43c26a12017-05-18 15:44:41 +02003942int skb_csum_hwoffload_help(struct sk_buff *skb,
3943 const netdev_features_t features);
3944
Joe Perchesf629d202013-09-26 14:48:15 -07003945struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3946 netdev_features_t features, bool tx_path);
3947struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3948 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003949
Moni Shoua61bd3852015-02-03 16:48:29 +02003950struct netdev_bonding_info {
3951 ifslave slave;
3952 ifbond master;
3953};
3954
3955struct netdev_notifier_bonding_info {
3956 struct netdev_notifier_info info; /* must be first */
3957 struct netdev_bonding_info bonding_info;
3958};
3959
3960void netdev_bonding_info_change(struct net_device *dev,
3961 struct netdev_bonding_info *bonding_info);
3962
Cong Wang12b00042013-02-05 16:36:38 +00003963static inline
3964struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3965{
3966 return __skb_gso_segment(skb, features, true);
3967}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003968__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003969
3970static inline bool can_checksum_protocol(netdev_features_t features,
3971 __be16 protocol)
3972{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003973 if (protocol == htons(ETH_P_FCOE))
3974 return !!(features & NETIF_F_FCOE_CRC);
3975
3976 /* Assume this is an IP checksum (not SCTP CRC) */
3977
3978 if (features & NETIF_F_HW_CSUM) {
3979 /* Can checksum everything */
3980 return true;
3981 }
3982
3983 switch (protocol) {
3984 case htons(ETH_P_IP):
3985 return !!(features & NETIF_F_IP_CSUM);
3986 case htons(ETH_P_IPV6):
3987 return !!(features & NETIF_F_IPV6_CSUM);
3988 default:
3989 return false;
3990 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003991}
Cong Wang12b00042013-02-05 16:36:38 +00003992
Herbert Xufb286bb2005-11-10 13:01:24 -08003993#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003994void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003995#else
3996static inline void netdev_rx_csum_fault(struct net_device *dev)
3997{
3998}
3999#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07004000/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07004001void net_enable_timestamp(void);
4002void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004003
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004004#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07004005int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00004006#else
4007#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004008#endif
4009
David S. Miller47982482014-08-22 16:21:53 -07004010static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004011 struct sk_buff *skb, struct net_device *dev,
4012 bool more)
David S. Miller47982482014-08-22 16:21:53 -07004013{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004014 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07004015 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07004016}
4017
David S. Miller10b3ad82014-08-29 21:07:24 -07004018static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004019 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07004020{
4021 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07004022 int rc;
David S. Miller47982482014-08-22 16:21:53 -07004023
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004024 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07004025 if (rc == NETDEV_TX_OK)
4026 txq_trans_update(txq);
4027
4028 return rc;
David S. Miller47982482014-08-22 16:21:53 -07004029}
4030
Linus Torvalds42a2d922013-11-13 17:40:34 +09004031int netdev_class_create_file_ns(struct class_attribute *class_attr,
4032 const void *ns);
4033void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4034 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04004035
4036static inline int netdev_class_create_file(struct class_attribute *class_attr)
4037{
4038 return netdev_class_create_file_ns(class_attr, NULL);
4039}
4040
4041static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4042{
4043 netdev_class_remove_file_ns(class_attr, NULL);
4044}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004045
Johannes Berg04600792010-08-05 17:45:15 +02004046extern struct kobj_ns_type_operations net_ns_type_operations;
4047
Joe Perchesf629d202013-09-26 14:48:15 -07004048const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004049
Joe Perchesf629d202013-09-26 14:48:15 -07004050void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004051
Michal Kubečekda081432014-05-20 08:29:25 +02004052static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4053 netdev_features_t f2)
4054{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004055 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4056 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004057 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004058 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004059 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004060 }
Michal Kubečekda081432014-05-20 08:29:25 +02004061
Tom Herbertc8cd0982015-12-14 11:19:44 -08004062 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004063}
4064
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004065static inline netdev_features_t netdev_get_wanted_features(
4066 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004067{
4068 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4069}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004070netdev_features_t netdev_increment_features(netdev_features_t all,
4071 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004072
4073/* Allow TSO being used on stacked device :
4074 * Performing the GSO segmentation before last device
4075 * is a performance improvement.
4076 */
4077static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4078 netdev_features_t mask)
4079{
4080 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4081}
4082
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004083int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004084void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004085void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004086
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004087void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4088 struct net_device *dev);
4089
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004090netdev_features_t passthru_features_check(struct sk_buff *skb,
4091 struct net_device *dev,
4092 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004093netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004094
David S. Miller4d295152012-03-07 21:02:35 -05004095static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004096{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004097 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004098
4099 /* check flags correspondence */
4100 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4101 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4102 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4103 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004104 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004105 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4106 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004107 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4108 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004109 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4110 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004111 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4112 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 -04004113 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004114 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004115 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Steffen Klassertc7ef8f02017-04-14 10:05:36 +02004116 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004117
Herbert Xubcd76112006-06-30 13:36:35 -07004118 return (features & feature) == feature;
4119}
4120
David S. Miller4d295152012-03-07 21:02:35 -05004121static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004122{
Herbert Xu278b2512009-06-03 21:20:51 -07004123 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004124 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004125}
4126
Johannes Berg8b86a612015-04-17 15:45:04 +02004127static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004128 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004129{
Jesse Grossfc741212011-01-09 06:23:32 +00004130 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004131 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4132 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004133}
4134
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004135static inline void netif_set_gso_max_size(struct net_device *dev,
4136 unsigned int size)
4137{
4138 dev->gso_max_size = size;
4139}
4140
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004141static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4142 int pulled_hlen, u16 mac_offset,
4143 int mac_len)
4144{
4145 skb->protocol = protocol;
4146 skb->encapsulation = 1;
4147 skb_push(skb, pulled_hlen);
4148 skb_reset_transport_header(skb);
4149 skb->mac_header = mac_offset;
4150 skb->network_header = skb->mac_header + mac_len;
4151 skb->mac_len = mac_len;
4152}
4153
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004154static inline bool netif_is_macsec(const struct net_device *dev)
4155{
4156 return dev->priv_flags & IFF_MACSEC;
4157}
4158
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004159static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004160{
4161 return dev->priv_flags & IFF_MACVLAN;
4162}
4163
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004164static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004165{
4166 return dev->priv_flags & IFF_MACVLAN_PORT;
4167}
4168
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004169static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004170{
4171 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4172}
4173
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004174static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004175{
4176 return dev->priv_flags & IFF_IPVLAN_MASTER;
4177}
4178
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004179static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004180{
4181 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4182}
4183
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004184static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004185{
4186 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4187}
4188
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004189static inline bool netif_supports_nofcs(struct net_device *dev)
4190{
4191 return dev->priv_flags & IFF_SUPP_NOFCS;
4192}
4193
David Ahern007979e2015-09-29 20:07:10 -07004194static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004195{
David Ahern007979e2015-09-29 20:07:10 -07004196 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004197}
4198
David Ahernfee6d4c2015-10-05 08:51:24 -07004199static inline bool netif_is_l3_slave(const struct net_device *dev)
4200{
4201 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4202}
4203
Jiri Pirko0894ae32015-08-27 09:31:19 +02004204static inline bool netif_is_bridge_master(const struct net_device *dev)
4205{
4206 return dev->priv_flags & IFF_EBRIDGE;
4207}
4208
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004209static inline bool netif_is_bridge_port(const struct net_device *dev)
4210{
4211 return dev->priv_flags & IFF_BRIDGE_PORT;
4212}
4213
Jiri Pirko35d4e172015-08-27 09:31:20 +02004214static inline bool netif_is_ovs_master(const struct net_device *dev)
4215{
4216 return dev->priv_flags & IFF_OPENVSWITCH;
4217}
4218
Jiri Pirko5be66142017-04-18 16:55:36 +02004219static inline bool netif_is_ovs_port(const struct net_device *dev)
4220{
4221 return dev->priv_flags & IFF_OVS_DATAPATH;
4222}
4223
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004224static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004225{
4226 return dev->priv_flags & IFF_TEAM;
4227}
4228
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004229static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004230{
4231 return dev->priv_flags & IFF_TEAM_PORT;
4232}
4233
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004234static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004235{
4236 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4237}
4238
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004239static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004240{
4241 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4242}
4243
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004244static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4245{
4246 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4247}
4248
Eric Dumazet02875872014-10-05 18:38:35 -07004249/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4250static inline void netif_keep_dst(struct net_device *dev)
4251{
4252 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4253}
4254
Paolo Abeni18d3df32016-07-14 18:00:10 +02004255/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4256static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4257{
4258 /* TODO: reserve and use an additional IFF bit, if we get more users */
4259 return dev->priv_flags & IFF_MACSEC;
4260}
4261
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004262extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004263
Joe Perches571ba422010-02-09 11:49:47 +00004264/* Logging, debugging and troubleshooting/diagnostic helpers. */
4265
4266/* netdev_printk helpers, similar to dev_printk */
4267
4268static inline const char *netdev_name(const struct net_device *dev)
4269{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004270 if (!dev->name[0] || strchr(dev->name, '%'))
4271 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004272 return dev->name;
4273}
4274
David Ahern8397ed32017-06-07 12:26:23 -06004275static inline bool netdev_unregistering(const struct net_device *dev)
4276{
4277 return dev->reg_state == NETREG_UNREGISTERING;
4278}
4279
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004280static inline const char *netdev_reg_state(const struct net_device *dev)
4281{
4282 switch (dev->reg_state) {
4283 case NETREG_UNINITIALIZED: return " (uninitialized)";
4284 case NETREG_REGISTERED: return "";
4285 case NETREG_UNREGISTERING: return " (unregistering)";
4286 case NETREG_UNREGISTERED: return " (unregistered)";
4287 case NETREG_RELEASED: return " (released)";
4288 case NETREG_DUMMY: return " (dummy)";
4289 }
4290
4291 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4292 return " (unknown)";
4293}
4294
Joe Perchesf629d202013-09-26 14:48:15 -07004295__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004296void netdev_printk(const char *level, const struct net_device *dev,
4297 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004298__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004299void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004300__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004301void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004302__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004303void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004304__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004305void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004306__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004307void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004308__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004309void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004310__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004311void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004312
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004313#define MODULE_ALIAS_NETDEV(device) \
4314 MODULE_ALIAS("netdev-" device)
4315
Jim Cromieb558c962011-12-19 17:11:18 -05004316#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004317#define netdev_dbg(__dev, format, args...) \
4318do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004319 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004320} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004321#elif defined(DEBUG)
4322#define netdev_dbg(__dev, format, args...) \
4323 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004324#else
4325#define netdev_dbg(__dev, format, args...) \
4326({ \
4327 if (0) \
4328 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004329})
4330#endif
4331
4332#if defined(VERBOSE_DEBUG)
4333#define netdev_vdbg netdev_dbg
4334#else
4335
4336#define netdev_vdbg(dev, format, args...) \
4337({ \
4338 if (0) \
4339 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4340 0; \
4341})
4342#endif
4343
4344/*
4345 * netdev_WARN() acts like dev_printk(), but with the key difference
4346 * of using a WARN/WARN_ON to get the message out, including the
4347 * file/line information and a backtrace.
4348 */
4349#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004350 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4351 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004352
Joe Perchesb3d95c52010-02-09 11:49:49 +00004353/* netif printk helpers, similar to netdev_printk */
4354
4355#define netif_printk(priv, type, level, dev, fmt, args...) \
4356do { \
4357 if (netif_msg_##type(priv)) \
4358 netdev_printk(level, (dev), fmt, ##args); \
4359} while (0)
4360
Joe Perchesf45f4322010-06-27 01:02:36 +00004361#define netif_level(level, priv, type, dev, fmt, args...) \
4362do { \
4363 if (netif_msg_##type(priv)) \
4364 netdev_##level(dev, fmt, ##args); \
4365} while (0)
4366
Joe Perchesb3d95c52010-02-09 11:49:49 +00004367#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004368 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004369#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004370 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004371#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004372 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004373#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004374 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004375#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004376 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004377#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004378 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004379#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004380 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004381
Joe Perches0053ea92012-05-30 07:43:34 +00004382#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004383#define netif_dbg(priv, type, netdev, format, args...) \
4384do { \
4385 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004386 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004387} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004388#elif defined(DEBUG)
4389#define netif_dbg(priv, type, dev, format, args...) \
4390 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004391#else
4392#define netif_dbg(priv, type, dev, format, args...) \
4393({ \
4394 if (0) \
4395 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4396 0; \
4397})
4398#endif
4399
Edward Creef617f272017-01-27 15:02:26 +00004400/* if @cond then downgrade to debug, else print at @level */
4401#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4402 do { \
4403 if (cond) \
4404 netif_dbg(priv, type, netdev, fmt, ##args); \
4405 else \
4406 netif_ ## level(priv, type, netdev, fmt, ##args); \
4407 } while (0)
4408
Joe Perchesb3d95c52010-02-09 11:49:49 +00004409#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004410#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004411#else
4412#define netif_vdbg(priv, type, dev, format, args...) \
4413({ \
4414 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004415 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004416 0; \
4417})
4418#endif
Joe Perches571ba422010-02-09 11:49:47 +00004419
Cong Wang900ff8c2013-02-18 19:20:33 +00004420/*
4421 * The list of packet types we will receive (as opposed to discard)
4422 * and the routines to invoke.
4423 *
4424 * Why 16. Because with 16 the only overlap we get on a hash of the
4425 * low nibble of the protocol value is RARP/SNAP/X.25.
4426 *
4427 * NOTE: That is no longer true with the addition of VLAN tags. Not
4428 * sure which should go first, but I bet it won't make much
4429 * difference if we are running VLANs. The good news is that
4430 * this protocol won't be in the list unless compiled in, so
4431 * the average user (w/out VLANs) will not be adversely affected.
4432 * --BLG
4433 *
4434 * 0800 IP
4435 * 8100 802.1Q VLAN
4436 * 0001 802.3
4437 * 0002 AX.25
4438 * 0004 802.2
4439 * 8035 RARP
4440 * 0005 SNAP
4441 * 0805 X.25
4442 * 0806 ARP
4443 * 8137 IPX
4444 * 0009 Localtalk
4445 * 86DD IPv6
4446 */
4447#define PTYPE_HASH_SIZE (16)
4448#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4449
Jiri Pirko385a1542009-05-27 15:48:07 -07004450#endif /* _LINUX_NETDEVICE_H */