blob: 91af73c9dd5108a68b9fe8b4dd4ef2b035b81ace [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
Alan Cox113aa832008-10-13 19:01:08 -070014 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
Al Virod7fe0f22006-12-03 23:15:30 -050028#include <linux/timer.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050029#include <linux/bug.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070030#include <linux/delay.h>
Arun Sharma600634972011-07-26 16:09:06 -070031#include <linux/atomic.h>
Eric Dumazet53511452014-10-08 08:19:27 -070032#include <linux/prefetch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <asm/cache.h>
34#include <asm/byteorder.h>
35
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/percpu.h>
David S. Miller4d5b78c2009-05-06 16:52:51 -070037#include <linux/rculist.h>
Chris Leechdb217332006-06-17 21:24:58 -070038#include <linux/dmaengine.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070039#include <linux/workqueue.h>
Tom Herbert114cf582011-11-28 16:33:09 +000040#include <linux/dynamic_queue_limits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Patrick McHardyb1b67dd2009-04-20 04:49:28 +000042#include <linux/ethtool.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020043#include <net/net_namespace.h>
Lennert Buytenhekcf85d082008-10-07 13:45:02 +000044#include <net/dsa.h>
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -080045#ifdef CONFIG_DCB
Alexander Duyck2f90b862008-11-20 20:52:10 -080046#include <net/dcbnl.h>
47#endif
Neil Horman5bc14212011-11-22 05:10:51 +000048#include <net/netprio_cgroup.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020049
Michał Mirosława59e2ec2011-11-15 15:29:55 +000050#include <linux/netdev_features.h>
John Fastabend77162022012-04-15 06:43:56 +000051#include <linux/neighbour.h>
David Howells607ca462012-10-13 10:46:48 +010052#include <uapi/linux/netdevice.h>
Moni Shoua61bd3852015-02-03 16:48:29 +020053#include <uapi/linux/if_bonding.h>
John Fastabende4c67342016-02-16 21:16:15 -080054#include <uapi/linux/pkt_cls.h>
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;
Johannes Berg704232c2007-04-23 12:20:05 -070059/* 802.11 specific */
60struct wireless_dev;
Alexander Aring98a18b62014-11-02 06:44:54 +010061/* 802.15.4 specific */
62struct wpan_dev;
Robert Shearman03c57742015-04-22 11:14:37 +010063struct mpls_dev;
Alexander Duyck7c46a642016-06-16 12:21:00 -070064/* UDP Tunnel offloads */
65struct udp_tunnel_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
Joe Perchesf629d202013-09-26 14:48:15 -070067void netdev_set_default_ethtool_ops(struct net_device *dev,
68 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000069
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000070/* Backlog congestion levels */
71#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
72#define NET_RX_DROP 1 /* packet dropped */
73
Patrick McHardy572a9d72009-11-10 06:14:14 +000074/*
75 * Transmit return codes: transmit return codes originate from three different
76 * namespaces:
77 *
78 * - qdisc return codes
79 * - driver transmit return codes
80 * - errno values
81 *
82 * Drivers are allowed to return any one of those in their hard_start_xmit()
83 * function. Real network devices commonly used with qdiscs should only return
84 * the driver transmit return codes though - when qdiscs are used, the actual
85 * transmission happens asynchronously, so the value is not propagated to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -050086 * higher layers. Virtual network devices transmit synchronously; in this case
87 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
Patrick McHardy572a9d72009-11-10 06:14:14 +000088 * others are propagated to higher layers.
89 */
90
91/* qdisc ->enqueue() return codes. */
92#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000093#define NET_XMIT_DROP 0x01 /* skb dropped */
94#define NET_XMIT_CN 0x02 /* congestion notification */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000095#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -020097/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
98 * indicates that the device will soon be dropping packets, or already drops
99 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +0000100#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
102
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000103/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000104#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000105
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000106enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000107 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000108 NETDEV_TX_OK = 0x00, /* driver took care of packet */
109 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000110};
111typedef enum netdev_tx netdev_tx_t;
112
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000113/*
114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 */
117static inline bool dev_xmit_complete(int rc)
118{
119 /*
120 * Positive cases with an skb consumed by a driver:
121 * - successful transmission (rc == NETDEV_TX_OK)
122 * - error while transmitting (rc < 0)
123 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 */
125 if (likely(rc < NET_XMIT_MASK))
126 return true;
127
128 return false;
129}
130
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500132 * Compute the worst-case header length according to the protocols
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 * used.
134 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800135
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800136#if defined(CONFIG_HYPERV_NET)
137# define LL_MAX_HEADER 128
138#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700139# if defined(CONFIG_MAC80211_MESH)
140# define LL_MAX_HEADER 128
141# else
142# define LL_MAX_HEADER 96
143# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700145# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146#endif
147
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000148#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
149 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150#define MAX_HEADER LL_MAX_HEADER
151#else
152#define MAX_HEADER (LL_MAX_HEADER + 48)
153#endif
154
155/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000156 * Old network device statistics. Fields are native words
157 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800159
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800160struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000161 unsigned long rx_packets;
162 unsigned long tx_packets;
163 unsigned long rx_bytes;
164 unsigned long tx_bytes;
165 unsigned long rx_errors;
166 unsigned long tx_errors;
167 unsigned long rx_dropped;
168 unsigned long tx_dropped;
169 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000172 unsigned long rx_over_errors;
173 unsigned long rx_crc_errors;
174 unsigned long rx_frame_errors;
175 unsigned long rx_fifo_errors;
176 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 unsigned long tx_aborted_errors;
178 unsigned long tx_carrier_errors;
179 unsigned long tx_fifo_errors;
180 unsigned long tx_heartbeat_errors;
181 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 unsigned long rx_compressed;
183 unsigned long tx_compressed;
184};
185
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
187#include <linux/cache.h>
188#include <linux/skbuff.h>
189
Eric Dumazetadc93002011-11-17 03:13:26 +0000190#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100191#include <linux/static_key.h>
192extern struct static_key rps_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000193#endif
194
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195struct neighbour;
196struct neigh_parms;
197struct sk_buff;
198
Jiri Pirkof001fde2009-05-05 02:48:28 +0000199struct netdev_hw_addr {
200 struct list_head list;
201 unsigned char addr[MAX_ADDR_LEN];
202 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000203#define NETDEV_HW_ADDR_T_LAN 1
204#define NETDEV_HW_ADDR_T_SAN 2
205#define NETDEV_HW_ADDR_T_SLAVE 3
206#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000207#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000208 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000209 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700210 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400211 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000212 struct rcu_head rcu_head;
213};
214
Jiri Pirko31278e72009-06-17 01:12:19 +0000215struct netdev_hw_addr_list {
216 struct list_head list;
217 int count;
218};
219
Jiri Pirko22bedad32010-04-01 21:22:57 +0000220#define netdev_hw_addr_list_count(l) ((l)->count)
221#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
222#define netdev_hw_addr_list_for_each(ha, l) \
223 list_for_each_entry(ha, &(l)->list, list)
224
225#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
226#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800227#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000228 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800229
Jiri Pirko22bedad32010-04-01 21:22:57 +0000230#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
231#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700232#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000233 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece32010-02-04 10:22:25 -0800234
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800235struct hh_cache {
David S. Millerf6b72b622011-07-14 07:53:20 -0700236 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700237 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800238 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240 /* cached hardware header; allow for machine alignment needs. */
241#define HH_DATA_MOD 16
242#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700243 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244#define HH_DATA_ALIGN(__len) \
245 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
246 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
247};
248
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500249/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 * Alternative is:
251 * dev->hard_header_len ? (dev->hard_header_len +
252 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
253 *
254 * We could use other alignment values, but we must maintain the
255 * relationship HH alignment <= LL alignment.
256 */
257#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700258 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700260 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700262struct header_ops {
263 int (*create) (struct sk_buff *skb, struct net_device *dev,
264 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000265 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700266 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700267 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700268 void (*cache_update)(struct hh_cache *hh,
269 const struct net_device *dev,
270 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500271 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700272};
273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274/* These flag bits are private to the generic network queueing
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500275 * layer; they may not be explicitly referenced by any other
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 * code.
277 */
278
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800279enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 __LINK_STATE_START,
281 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800283 __LINK_STATE_LINKWATCH_PENDING,
284 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285};
286
287
288/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500289 * This structure holds boot-time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800290 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 */
292struct netdev_boot_setup {
293 char name[IFNAMSIZ];
294 struct ifmap map;
295};
296#define NETDEV_BOOT_SETUP_MAX 8
297
Joe Perchesf629d202013-09-26 14:48:15 -0700298int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
300/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700301 * Structure for NAPI scheduling similar to tasklet but with weighting
302 */
303struct napi_struct {
304 /* The poll_list must only be managed by the entity which
305 * changes the state of the NAPI_STATE_SCHED bit. This means
306 * whoever atomically sets that bit can add this napi_struct
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500307 * to the per-CPU poll_list, and whoever clears that bit
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700308 * can remove from the list right before clearing the bit.
309 */
310 struct list_head poll_list;
311
312 unsigned long state;
313 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000314 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700315 int (*poll)(struct napi_struct *, int);
316#ifdef CONFIG_NETPOLL
317 spinlock_t poll_lock;
318 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700319#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800320 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800321 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800322 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800323 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000324 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300325 struct hlist_node napi_hash_node;
326 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700327};
328
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800329enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700330 NAPI_STATE_SCHED, /* Poll is scheduled */
David S. Millera0a46192008-01-07 20:35:07 -0800331 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800332 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800333 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
334 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700335};
336
Ben Hutchings5b252f02009-10-29 07:17:09 +0000337enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700338 GRO_MERGED,
339 GRO_MERGED_FREE,
340 GRO_HELD,
341 GRO_NORMAL,
342 GRO_DROP,
343};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000344typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700345
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000346/*
347 * enum rx_handler_result - Possible return values for rx_handlers.
348 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
349 * further.
350 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
351 * case skb->dev was changed by rx_handler.
352 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500353 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000354 *
355 * rx_handlers are functions called from inside __netif_receive_skb(), to do
356 * special processing of the skb, prior to delivery to protocol handlers.
357 *
358 * Currently, a net_device can only have a single rx_handler registered. Trying
359 * to register a second rx_handler will return -EBUSY.
360 *
361 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
362 * To unregister a rx_handler on a net_device, use
363 * netdev_rx_handler_unregister().
364 *
365 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
366 * do with the skb.
367 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500368 * If the rx_handler consumed the skb in some way, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000369 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500370 * the skb to be delivered in some other way.
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000371 *
372 * If the rx_handler changed skb->dev, to divert the skb to another
373 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
374 * new device will be called if it exists.
375 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500376 * If the rx_handler decides the skb should be ignored, it should return
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000377 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400378 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000379 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500380 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000381 * delivered, it should return RX_HANDLER_PASS.
382 *
383 * A device without a registered rx_handler will behave as if rx_handler
384 * returned RX_HANDLER_PASS.
385 */
386
387enum rx_handler_result {
388 RX_HANDLER_CONSUMED,
389 RX_HANDLER_ANOTHER,
390 RX_HANDLER_EXACT,
391 RX_HANDLER_PASS,
392};
393typedef enum rx_handler_result rx_handler_result_t;
394typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000395
Joe Perchesf629d202013-09-26 14:48:15 -0700396void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700397void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700398
David S. Miller4d295152012-03-07 21:02:35 -0500399static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800400{
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
402}
403
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700404/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500405 * napi_schedule_prep - check if NAPI can be scheduled
406 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700407 *
408 * Test if NAPI routine is already running, and if not mark
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500409 * it as running. This is used as a condition variable to
David S. Millera0a46192008-01-07 20:35:07 -0800410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700412 */
David S. Miller4d295152012-03-07 21:02:35 -0500413static inline bool napi_schedule_prep(struct napi_struct *n)
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700414{
David S. Millera0a46192008-01-07 20:35:07 -0800415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700417}
418
419/**
420 * napi_schedule - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500421 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700422 *
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
425 */
426static inline void napi_schedule(struct napi_struct *n)
427{
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
430}
431
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700432/**
433 * napi_schedule_irqoff - schedule NAPI poll
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500434 * @n: NAPI context
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700435 *
436 * Variant of napi_schedule(), assuming hard irqs are masked.
437 */
438static inline void napi_schedule_irqoff(struct napi_struct *n)
439{
440 if (napi_schedule_prep(n))
441 __napi_schedule_irqoff(n);
442}
443
Roland Dreierbfe13f52007-10-09 15:47:37 -0700444/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500445static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700446{
447 if (napi_schedule_prep(napi)) {
448 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500449 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700450 }
David S. Miller4d295152012-03-07 21:02:35 -0500451 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700452}
453
Eric Dumazet3b47d302014-11-06 21:09:44 -0800454void __napi_complete(struct napi_struct *n);
455void napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700456/**
457 * napi_complete - NAPI processing complete
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500458 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700459 *
460 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800461 * Consider using napi_complete_done() instead.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700462 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800463static inline void napi_complete(struct napi_struct *n)
464{
465 return napi_complete_done(n, 0);
466}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700467
468/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300469 * napi_hash_add - add a NAPI to global hashtable
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500470 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300471 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500472 * Generate a new napi_id and store a @napi under it in napi_hash.
473 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
Eric Dumazet93d05d42015-11-18 06:31:03 -0800474 * Note: This is normally automatically done from netif_napi_add(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500475 * so might disappear in a future Linux version.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300476 */
Joe Perchesf629d202013-09-26 14:48:15 -0700477void napi_hash_add(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300478
479/**
480 * napi_hash_del - remove a NAPI from global table
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500481 * @napi: NAPI context
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300482 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500483 * Warning: caller must observe RCU grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800484 * before freeing memory containing @napi, if
485 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800486 * Note: core networking stack automatically calls it
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500487 * from netif_napi_del().
Eric Dumazet93d05d42015-11-18 06:31:03 -0800488 * Drivers might want to call this helper to combine all
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500489 * the needed RCU grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300490 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800491bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300492
493/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700494 * napi_disable - prevent NAPI from scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500495 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700496 *
497 * Stop NAPI from being scheduled on this context.
498 * Waits till any outstanding processing completes.
499 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800500void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700501
502/**
503 * napi_enable - enable NAPI scheduling
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500504 * @n: NAPI context
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700505 *
506 * Resume NAPI from being scheduled on this context.
507 * Must be paired with napi_disable.
508 */
509static inline void napi_enable(struct napi_struct *n)
510{
511 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100512 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700513 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400514 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700515}
516
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700517/**
518 * napi_synchronize - wait until NAPI is not running
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500519 * @n: NAPI context
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700520 *
521 * Wait until NAPI is done being scheduled on this context.
522 * Waits till any outstanding processing completes but
523 * does not disable future activations.
524 */
525static inline void napi_synchronize(const struct napi_struct *n)
526{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100527 if (IS_ENABLED(CONFIG_SMP))
528 while (test_bit(NAPI_STATE_SCHED, &n->state))
529 msleep(1);
530 else
531 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700532}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700533
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800534enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000535 __QUEUE_STATE_DRV_XOFF,
536 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700537 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700538};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200539
540#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
541#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
542#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
543
544#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
545#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
546 QUEUE_STATE_FROZEN)
547#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 QUEUE_STATE_FROZEN)
549
Tom Herbert734664982011-11-28 16:32:44 +0000550/*
551 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
552 * netif_tx_* functions below are used to manipulate this flag. The
553 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
554 * queue independently. The netif_xmit_*stopped functions below are called
555 * to check if the queue has been stopped by the driver or stack (either
556 * of the XOFF bits are set in the state). Drivers should not need to call
557 * netif_xmit*stopped functions, they should only be using netif_tx_*.
558 */
David S. Miller79d16382008-07-08 23:14:46 -0700559
David S. Millerbb949fb2008-07-08 16:55:56 -0700560struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700561/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500562 * read-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700563 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700564 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700565 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700566 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000567#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000568 struct kobject kobj;
569#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000570#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
571 int numa_node;
572#endif
Florian Westphalc0ef0792016-05-03 03:29:09 +0200573 unsigned long tx_maxrate;
574 /*
575 * Number of TX timeouts for this queue
576 * (/sys/class/net/DEV/Q/trans_timeout)
577 */
578 unsigned long trans_timeout;
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700579/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500580 * write-mostly part
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700581 */
582 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
583 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700584 /*
Florian Westphal9b366272016-05-03 16:33:14 +0200585 * Time (in jiffies) of last Tx
Eric Dumazet9d214932009-05-17 20:55:16 -0700586 */
587 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000588
Tom Herbert114cf582011-11-28 16:33:09 +0000589 unsigned long state;
590
591#ifdef CONFIG_BQL
592 struct dql dql;
593#endif
David S. Millere8a04642008-07-17 00:34:19 -0700594} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700595
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000596static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
597{
598#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 return q->numa_node;
600#else
Changli Gaob236da62010-12-14 03:09:15 +0000601 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000602#endif
603}
604
605static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
606{
607#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 q->numa_node = node;
609#endif
610}
611
Eric Dumazetdf334542010-03-24 19:13:54 +0000612#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000613/*
614 * This structure holds an RPS map which can be of variable length. The
615 * map is an array of CPUs.
616 */
617struct rps_map {
618 unsigned int len;
619 struct rcu_head rcu;
620 u16 cpus[0];
621};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000622#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000623
Tom Herbertfec5e652010-04-16 16:01:27 -0700624/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000625 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
626 * tail pointer for that CPU's input queue at the time of last enqueue, and
627 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700628 */
629struct rps_dev_flow {
630 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000631 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700632 unsigned int last_qtail;
633};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000634#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700635
636/*
637 * The rps_dev_flow_table structure contains a table of flow mappings.
638 */
639struct rps_dev_flow_table {
640 unsigned int mask;
641 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700642 struct rps_dev_flow flows[0];
643};
644#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000645 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700646
647/*
648 * The rps_sock_flow_table contains mappings of flows to the last CPU
649 * on which they were processed by the application (set in recvmsg).
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500650 * Each entry is a 32bit value. Upper part is the high-order bits
651 * of flow hash, lower part is CPU number.
Eric Dumazet567e4b72015-02-06 12:59:01 -0800652 * rps_cpu_mask is used to partition the space, depending on number of
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500653 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
654 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
Eric Dumazet567e4b72015-02-06 12:59:01 -0800655 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700656 */
657struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800658 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800659
660 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700661};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800662#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700663
664#define RPS_NO_CPU 0xffff
665
Eric Dumazet567e4b72015-02-06 12:59:01 -0800666extern u32 rps_cpu_mask;
667extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
668
Tom Herbertfec5e652010-04-16 16:01:27 -0700669static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
670 u32 hash)
671{
672 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800673 unsigned int index = hash & table->mask;
674 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700675
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500676 /* We only give a hint, preemption can change CPU under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800677 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700678
Eric Dumazet567e4b72015-02-06 12:59:01 -0800679 if (table->ents[index] != val)
680 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700681 }
682}
683
Ben Hutchingsc4454772011-01-19 11:03:53 +0000684#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700685bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000687#endif
Michael Daltona953be52014-01-16 22:23:28 -0800688#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000689
Tom Herbert0a9627f2010-03-16 08:03:29 +0000690/* This structure contains an instance of an RX queue. */
691struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800692#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000693 struct rps_map __rcu *rps_map;
694 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800695#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000696 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000697 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000698} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800699
700/*
701 * RX queue sysfs structures and functions.
702 */
703struct rx_queue_attribute {
704 struct attribute attr;
705 ssize_t (*show)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, char *buf);
707 ssize_t (*store)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, const char *buf, size_t len);
709};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800710
Tom Herbertbf264142010-11-26 08:36:09 +0000711#ifdef CONFIG_XPS
712/*
713 * This structure holds an XPS map which can be of variable length. The
714 * map is an array of queues.
715 */
716struct xps_map {
717 unsigned int len;
718 unsigned int alloc_len;
719 struct rcu_head rcu;
720 u16 queues[0];
721};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000722#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100723#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
724 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000725
726/*
727 * This structure holds all XPS maps for device. Maps are indexed by CPU.
728 */
729struct xps_dev_maps {
730 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000731 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000732};
733#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
734 (nr_cpu_ids * sizeof(struct xps_map *)))
735#endif /* CONFIG_XPS */
736
John Fastabend4f57c082011-01-17 08:06:04 +0000737#define TC_MAX_QUEUE 16
738#define TC_BITMASK 15
739/* HW offloaded queuing disciplines txq count and offset maps */
740struct netdev_tc_txq {
741 u16 count;
742 u16 offset;
743};
744
Neerav Parikh68bad942012-01-04 20:23:39 +0000745#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
746/*
747 * This structure is to hold information about the device
748 * configured to run FCoE protocol stack.
749 */
750struct netdev_fcoe_hbainfo {
751 char manufacturer[64];
752 char serial_number[64];
753 char hardware_version[64];
754 char driver_version[64];
755 char optionrom_version[64];
756 char firmware_version[64];
757 char model[256];
758 char model_description[256];
759};
760#endif
761
Jiri Pirko02637fc2014-11-28 14:34:16 +0100762#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200763
Jiri Pirko02637fc2014-11-28 14:34:16 +0100764/* This structure holds a unique identifier to identify some
765 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200766 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100767struct netdev_phys_item_id {
768 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200769 unsigned char id_len;
770};
771
Scott Feldmand754f982015-07-18 18:24:49 -0700772static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
773 struct netdev_phys_item_id *b)
774{
775 return a->id_len == b->id_len &&
776 memcmp(a->id, b->id, a->id_len) == 0;
777}
778
Daniel Borkmann99932d42014-02-16 15:55:20 +0100779typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
780 struct sk_buff *skb);
781
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800782/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800783 * that are being passed to the netdevice through the setup_tc op.
784 */
785enum {
786 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800787 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200788 TC_SETUP_CLSFLOWER,
John Fastabend16e5cc62016-02-16 21:16:43 -0800789};
790
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800791struct tc_cls_u32_offload;
792
John Fastabend16e5cc62016-02-16 21:16:43 -0800793struct tc_to_netdev {
794 unsigned int type;
795 union {
796 u8 tc;
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800797 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200798 struct tc_cls_flower_offload *cls_flower;
John Fastabend16e5cc62016-02-16 21:16:43 -0800799 };
800};
801
802
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800803/*
804 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800805 * The following hooks can be defined; unless noted otherwise, they are
806 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800807 *
808 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500809 * This function is called once when a network device is registered.
810 * The network device can use this for any late stage initialization
811 * or semantic validation. It can fail with an error code which will
812 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800813 *
814 * void (*ndo_uninit)(struct net_device *dev);
815 * This function is called when device is unregistered or when registration
816 * fails. It is not called if init fails.
817 *
818 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500819 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800820 * state.
821 *
822 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500823 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800824 * state.
825 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000826 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
827 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800828 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030829 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
830 * the queue before that can happen; it's for obsolete devices and weird
831 * corner cases, but the stack really does a non-trivial amount
832 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500833 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800834 *
Eric Dumazetcdba7562016-01-06 06:53:50 -0800835 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
836 * netdev_features_t features);
837 * Adjusts the requested feature flags according to device-specific
838 * constraints, and returns the resulting flags. Must not modify
839 * the device state.
840 *
Jason Wangf663dd92014-01-10 16:18:26 +0800841 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100842 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500843 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800844 * transmit queues.
845 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800846 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
847 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500848 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800849 *
850 * void (*ndo_set_rx_mode)(struct net_device *dev);
851 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000852 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500853 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800854 *
855 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
856 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700857 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500858 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800859 *
860 * int (*ndo_validate_addr)(struct net_device *dev);
861 * Test if Media Access Control address is valid for the device.
862 *
863 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500864 * Called when a user requests an ioctl which can't be handled by
865 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800866 * not supported error code.
867 *
868 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
869 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500870 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800871 * interface (PCI) for low level management.
872 *
873 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
874 * Called when a user wants to change the Maximum Transfer Unit
875 * of a device. If not defined, any request to change MTU will
876 * will return an error.
877 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800878 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500879 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800880 * for dev->watchdog ticks.
881 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000882 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700883 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700884 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800885 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000886 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000887 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
888 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700889 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000890 * (which should normally be dev->stats) and return a pointer to
891 * it. The structure may be changed asynchronously only if each
892 * field is written atomically.
893 * 3. Update dev->stats asynchronously and atomically, and define
894 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800895 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530896 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500897 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000898 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800899 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530900 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500901 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000902 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800903 *
904 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000905 *
906 * SR-IOV management functions.
907 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
908 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400909 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
910 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000911 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000912 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000913 * int (*ndo_get_vf_config)(struct net_device *dev,
914 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300915 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700916 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
917 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300918 *
919 * Enable or disable the VF ability to query its RSS Redirection Table and
920 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500921 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300922 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700923 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000924 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
925 * Called to setup 'tc' number of traffic classes in the net device. This
926 * is always called from the stack with the rtnl lock held and netif tx
927 * queues stopped. This allows the netdevice to perform queue management
928 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000929 *
Yi Zoue9bce842011-03-09 08:48:03 +0000930 * Fiber Channel over Ethernet (FCoE) offload functions.
931 * int (*ndo_fcoe_enable)(struct net_device *dev);
932 * Called when the FCoE protocol stack wants to start using LLD for FCoE
933 * so the underlying device can perform whatever needed configuration or
934 * initialization to support acceleration of FCoE traffic.
935 *
936 * int (*ndo_fcoe_disable)(struct net_device *dev);
937 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
938 * so the underlying device can perform whatever needed clean-ups to
939 * stop supporting acceleration of FCoE traffic.
940 *
941 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
942 * struct scatterlist *sgl, unsigned int sgc);
943 * Called when the FCoE Initiator wants to initialize an I/O that
944 * is a possible candidate for Direct Data Placement (DDP). The LLD can
945 * perform necessary setup and returns 1 to indicate the device is set up
946 * successfully to perform DDP on this I/O, otherwise this returns 0.
947 *
948 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
949 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
950 * indicated by the FC exchange id 'xid', so the underlying device can
951 * clean up and reuse resources for later DDP requests.
952 *
953 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
954 * struct scatterlist *sgl, unsigned int sgc);
955 * Called when the FCoE Target wants to initialize an I/O that
956 * is a possible candidate for Direct Data Placement (DDP). The LLD can
957 * perform necessary setup and returns 1 to indicate the device is set up
958 * successfully to perform DDP on this I/O, otherwise this returns 0.
959 *
Neerav Parikh68bad942012-01-04 20:23:39 +0000960 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
961 * struct netdev_fcoe_hbainfo *hbainfo);
962 * Called when the FCoE Protocol stack wants information on the underlying
963 * device. This information is utilized by the FCoE protocol stack to
964 * register attributes with Fiber Channel management service as per the
965 * FC-GS Fabric Device Management Information(FDMI) specification.
966 *
Yi Zoue9bce842011-03-09 08:48:03 +0000967 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
968 * Called when the underlying device wants to override default World Wide
969 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
970 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
971 * protocol stack to use.
972 *
Ben Hutchingsc4454772011-01-19 11:03:53 +0000973 * RFS acceleration.
974 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
975 * u16 rxq_index, u32 flow_id);
976 * Set hardware filter for RFS. rxq_index is the target queue index;
977 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
978 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000979 *
Jiri Pirko8b98a702013-01-03 22:49:02 +0000980 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000981 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
982 * Called to make another netdev an underling.
983 *
984 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
985 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +0000986 *
987 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000988 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +0000989 * Called to update device configuration to new features. Passed
990 * feature set might be less than what was returned by ndo_fix_features()).
991 * Must return >0 or -errno if it changed dev->features itself.
992 *
stephen hemmingeredc7d572012-10-01 12:32:33 +0000993 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
994 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100995 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +0000996 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +0000997 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
998 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100999 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001000 * Deletes the FDB entry from dev coresponding to addr.
1001 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001002 * struct net_device *dev, struct net_device *filter_dev,
1003 * int idx)
John Fastabend77162022012-04-15 06:43:56 +00001004 * Used to add FDB entries to dump requests. Implementers should add
1005 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001006 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001007 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1008 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001009 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001010 * struct net_device *dev, u32 filter_mask,
1011 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001012 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1013 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001014 *
1015 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1016 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1017 * which do not represent real hardware may define this to allow their
1018 * userspace components to manage their virtual carrier state. Devices
1019 * that determine carrier state from physical hardware properties (eg
1020 * network cables) or protocol-dependent mechanisms (eg
1021 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001022 *
1023 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001024 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001025 * Called to get ID of physical port of this device. If driver does
1026 * not implement this, it is assumed that the hw is not able to have
1027 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001028 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001029 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1030 * struct udp_tunnel_info *ti);
1031 * Called by UDP tunnel to notify a driver about the UDP port and socket
1032 * address family that a UDP tunnel is listnening to. It is called only
1033 * when a new port starts listening. The operation is protected by the
1034 * RTNL.
1035 *
1036 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1037 * struct udp_tunnel_info *ti);
1038 * Called by UDP tunnel to notify the driver about a UDP port and socket
1039 * address family that the UDP tunnel is not listening to anymore. The
1040 * operation is protected by the RTNL.
1041 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001042 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1043 * struct net_device *dev)
1044 * Called by upper layer devices to accelerate switching or other
1045 * station functionality into hardware. 'pdev is the lowerdev
1046 * to use for the offload and 'dev' is the net device that will
1047 * back the offload. Returns a pointer to the private structure
1048 * the upper layer will maintain.
1049 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1050 * Called by upper layer device to delete the station created
1051 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1052 * the station and priv is the structure returned by the add
1053 * operation.
1054 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1055 * struct net_device *dev,
1056 * void *priv);
1057 * Callback to use for xmit over the accelerated station. This
1058 * is used in place of ndo_start_xmit on accelerated net
1059 * devices.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001060 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1061 * struct net_device *dev
1062 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001063 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001064 * performing offload operations on a given packet. This is to give
1065 * the device an opportunity to implement any restrictions that cannot
1066 * be otherwise expressed by feature flags. The check is called with
1067 * the set of features that the stack has calculated and it returns
1068 * those the driver believes to be appropriate.
John Fastabend822b3b22015-03-18 14:57:33 +02001069 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1070 * int queue_index, u32 maxrate);
1071 * Called when a user wants to set a max-rate limitation of specific
1072 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001073 * int (*ndo_get_iflink)(const struct net_device *dev);
1074 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001075 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001076 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001077 * This function is used to pass protocol port error state information
1078 * to the switch driver. The switch driver can react to the proto_down
1079 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001080 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1081 * This function is used to get egress tunnel information for given skb.
1082 * This is useful for retrieving outer tunnel header parameters while
1083 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001084 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1085 * This function is used to specify the headroom that the skb must
1086 * consider when allocation skb during packet reception. Setting
1087 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001088 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001089 * default value.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001090 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001091 */
1092struct net_device_ops {
1093 int (*ndo_init)(struct net_device *dev);
1094 void (*ndo_uninit)(struct net_device *dev);
1095 int (*ndo_open)(struct net_device *dev);
1096 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001097 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1098 struct net_device *dev);
1099 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1100 struct net_device *dev,
1101 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001102 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001103 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001104 void *accel_priv,
1105 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001106 void (*ndo_change_rx_flags)(struct net_device *dev,
1107 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001108 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001109 int (*ndo_set_mac_address)(struct net_device *dev,
1110 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001111 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001112 int (*ndo_do_ioctl)(struct net_device *dev,
1113 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001114 int (*ndo_set_config)(struct net_device *dev,
1115 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001116 int (*ndo_change_mtu)(struct net_device *dev,
1117 int new_mtu);
1118 int (*ndo_neigh_setup)(struct net_device *dev,
1119 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001120 void (*ndo_tx_timeout) (struct net_device *dev);
1121
Eric Dumazet28172732010-07-07 14:58:56 -07001122 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1123 struct rtnl_link_stats64 *storage);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001124 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1125
Jiri Pirko8e586132011-12-08 19:52:37 -05001126 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001127 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001128 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001129 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001130#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001131 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001132 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001133 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001134 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001135#endif
Cong Wange0d10952013-08-01 11:10:25 +08001136#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001137 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001138#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001139 int (*ndo_set_vf_mac)(struct net_device *dev,
1140 int queue, u8 *mac);
1141 int (*ndo_set_vf_vlan)(struct net_device *dev,
1142 int queue, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001143 int (*ndo_set_vf_rate)(struct net_device *dev,
1144 int vf, int min_tx_rate,
1145 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001146 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1147 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001148 int (*ndo_set_vf_trust)(struct net_device *dev,
1149 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001150 int (*ndo_get_vf_config)(struct net_device *dev,
1151 int vf,
1152 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001153 int (*ndo_set_vf_link_state)(struct net_device *dev,
1154 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001155 int (*ndo_get_vf_stats)(struct net_device *dev,
1156 int vf,
1157 struct ifla_vf_stats
1158 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001159 int (*ndo_set_vf_port)(struct net_device *dev,
1160 int vf,
1161 struct nlattr *port[]);
1162 int (*ndo_get_vf_port)(struct net_device *dev,
1163 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001164 int (*ndo_set_vf_guid)(struct net_device *dev,
1165 int vf, u64 guid,
1166 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001167 int (*ndo_set_vf_rss_query_en)(
1168 struct net_device *dev,
1169 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001170 int (*ndo_setup_tc)(struct net_device *dev,
1171 u32 handle,
1172 __be16 protocol,
1173 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001174#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001175 int (*ndo_fcoe_enable)(struct net_device *dev);
1176 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001177 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1178 u16 xid,
1179 struct scatterlist *sgl,
1180 unsigned int sgc);
1181 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1182 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001183 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1184 u16 xid,
1185 struct scatterlist *sgl,
1186 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001187 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1188 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001189#endif
1190
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001191#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001192#define NETDEV_FCOE_WWNN 0
1193#define NETDEV_FCOE_WWPN 1
1194 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1195 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001196#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001197
Ben Hutchingsc4454772011-01-19 11:03:53 +00001198#ifdef CONFIG_RFS_ACCEL
1199 int (*ndo_rx_flow_steer)(struct net_device *dev,
1200 const struct sk_buff *skb,
1201 u16 rxq_index,
1202 u32 flow_id);
1203#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001204 int (*ndo_add_slave)(struct net_device *dev,
1205 struct net_device *slave_dev);
1206 int (*ndo_del_slave)(struct net_device *dev,
1207 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001208 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1209 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001210 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001211 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001212 int (*ndo_neigh_construct)(struct net_device *dev,
1213 struct neighbour *n);
1214 void (*ndo_neigh_destroy)(struct net_device *dev,
1215 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001216
1217 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001218 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001219 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001220 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001221 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001222 u16 flags);
1223 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001224 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001225 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001226 const unsigned char *addr,
1227 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001228 int (*ndo_fdb_dump)(struct sk_buff *skb,
1229 struct netlink_callback *cb,
1230 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001231 struct net_device *filter_dev,
John Fastabend77162022012-04-15 06:43:56 +00001232 int idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001233
1234 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001235 struct nlmsghdr *nlh,
1236 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001237 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1238 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001239 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001240 u32 filter_mask,
1241 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001242 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001243 struct nlmsghdr *nlh,
1244 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001245 int (*ndo_change_carrier)(struct net_device *dev,
1246 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001247 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001248 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001249 int (*ndo_get_phys_port_name)(struct net_device *dev,
1250 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001251 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1252 struct udp_tunnel_info *ti);
1253 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1254 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001255 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1256 struct net_device *dev);
1257 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1258 void *priv);
1259
1260 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1261 struct net_device *dev,
1262 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001263 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001264 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1265 int queue_index,
1266 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001267 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001268 int (*ndo_change_proto_down)(struct net_device *dev,
1269 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001270 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1271 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001272 void (*ndo_set_rx_headroom)(struct net_device *dev,
1273 int needed_headroom);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001274};
1275
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001276/**
1277 * enum net_device_priv_flags - &struct net_device priv_flags
1278 *
1279 * These are the &struct net_device, they are only set internally
1280 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001281 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001282 * during any kernel release.
1283 *
1284 * You should have a pretty good reason to be extending these flags.
1285 *
1286 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1287 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001288 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001289 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001290 * @IFF_WAN_HDLC: WAN HDLC device
1291 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1292 * release skb->dst
1293 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1294 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1295 * @IFF_MACVLAN_PORT: device used as macvlan port
1296 * @IFF_BRIDGE_PORT: device used as bridge port
1297 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1298 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1299 * @IFF_UNICAST_FLT: Supports unicast filtering
1300 * @IFF_TEAM_PORT: device used as team port
1301 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1302 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1303 * change when it's running
1304 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001305 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1306 * underlying stacked devices
1307 * @IFF_IPVLAN_MASTER: IPvlan master device
1308 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001309 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001310 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001311 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001312 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001313 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001314 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001315 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1316 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001317 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001318 */
1319enum netdev_priv_flags {
1320 IFF_802_1Q_VLAN = 1<<0,
1321 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001322 IFF_BONDING = 1<<2,
1323 IFF_ISATAP = 1<<3,
1324 IFF_WAN_HDLC = 1<<4,
1325 IFF_XMIT_DST_RELEASE = 1<<5,
1326 IFF_DONT_BRIDGE = 1<<6,
1327 IFF_DISABLE_NETPOLL = 1<<7,
1328 IFF_MACVLAN_PORT = 1<<8,
1329 IFF_BRIDGE_PORT = 1<<9,
1330 IFF_OVS_DATAPATH = 1<<10,
1331 IFF_TX_SKB_SHARING = 1<<11,
1332 IFF_UNICAST_FLT = 1<<12,
1333 IFF_TEAM_PORT = 1<<13,
1334 IFF_SUPP_NOFCS = 1<<14,
1335 IFF_LIVE_ADDR_CHANGE = 1<<15,
1336 IFF_MACVLAN = 1<<16,
1337 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1338 IFF_IPVLAN_MASTER = 1<<18,
1339 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001340 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001341 IFF_NO_QUEUE = 1<<21,
1342 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001343 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001344 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001345 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001346 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001347 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001348};
1349
1350#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1351#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001352#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001353#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001354#define IFF_WAN_HDLC IFF_WAN_HDLC
1355#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1356#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1357#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1358#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1359#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1360#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1361#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1362#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1363#define IFF_TEAM_PORT IFF_TEAM_PORT
1364#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1365#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1366#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001367#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001368#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1369#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001370#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001371#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001372#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001373#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001374#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001375#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001376#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001377
Karoly Kemeny536721b2014-07-30 20:27:36 +02001378/**
1379 * struct net_device - The DEVICE structure.
1380 * Actually, this whole structure is a big mistake. It mixes I/O
1381 * data with strictly "high-level" data, and it has to know about
1382 * almost every data structure used in the INET module.
1383 *
1384 * @name: This is the first field of the "visible" part of this structure
1385 * (i.e. as seen by users in the "Space.c" file). It is the name
1386 * of the interface.
1387 *
1388 * @name_hlist: Device name hash chain, please keep it close to name[]
1389 * @ifalias: SNMP alias
1390 * @mem_end: Shared memory end
1391 * @mem_start: Shared memory start
1392 * @base_addr: Device I/O address
1393 * @irq: Device IRQ number
1394 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001395 * @carrier_changes: Stats to monitor carrier on<->off transitions
1396 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001397 * @state: Generic network queuing layer state, see netdev_state_t
1398 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001399 * @napi_list: List entry used for polling NAPI devices
1400 * @unreg_list: List entry when we are unregistering the
1401 * device; see the function unregister_netdev
1402 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001403 * @ptype_all: Device-specific packet handlers for all protocols
1404 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001405 *
1406 * @adj_list: Directly linked devices, like slaves for bonding
1407 * @all_adj_list: All linked devices, *including* neighbours
1408 * @features: Currently active device features
1409 * @hw_features: User-changeable features
1410 *
1411 * @wanted_features: User-requested features
1412 * @vlan_features: Mask of features inheritable by VLAN devices
1413 *
1414 * @hw_enc_features: Mask of features inherited by encapsulating devices
1415 * This field indicates what encapsulation
1416 * offloads the hardware is capable of doing,
1417 * and drivers will need to set them appropriately.
1418 *
1419 * @mpls_features: Mask of features inheritable by MPLS
1420 *
1421 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001422 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001423 *
1424 * @stats: Statistics struct, which was left as a legacy, use
1425 * rtnl_link_stats64 instead
1426 *
1427 * @rx_dropped: Dropped packets by core network,
1428 * do not use this in drivers
1429 * @tx_dropped: Dropped packets by core network,
1430 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001431 * @rx_nohandler: nohandler dropped packets by core network on
1432 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001433 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001434 * @wireless_handlers: List of functions to handle Wireless Extensions,
1435 * instead of ioctl,
1436 * see <net/iw_handler.h> for details.
1437 * @wireless_data: Instance data managed by the core of wireless extensions
1438 *
1439 * @netdev_ops: Includes several pointers to callbacks,
1440 * if one wants to override the ndo_*() functions
1441 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001442 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1443 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001444 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001445 * of Layer 2 headers.
1446 *
1447 * @flags: Interface flags (a la BSD)
1448 * @priv_flags: Like 'flags' but invisible to userspace,
1449 * see if.h for the definitions
1450 * @gflags: Global flags ( kept as legacy )
1451 * @padded: How much padding added by alloc_netdev()
1452 * @operstate: RFC2863 operstate
1453 * @link_mode: Mapping policy to operstate
1454 * @if_port: Selectable AUI, TP, ...
1455 * @dma: DMA channel
1456 * @mtu: Interface MTU value
1457 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001458 * @hard_header_len: Maximum hardware header length.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001459 *
1460 * @needed_headroom: Extra headroom the hardware may need, but not in all
1461 * cases can this be guaranteed
1462 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1463 * cases can this be guaranteed. Some cases also use
1464 * LL_MAX_HEADER instead to allocate the skb
1465 *
1466 * interface address info:
1467 *
1468 * @perm_addr: Permanent hw address
1469 * @addr_assign_type: Hw address assignment type
1470 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001471 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001472 * @dev_id: Used to differentiate devices that share
1473 * the same link layer address
1474 * @dev_port: Used to differentiate devices that share
1475 * the same function
1476 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001477 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001478 * has been enabled due to the need to listen to
1479 * additional unicast addresses in a device that
1480 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001481 * @uc: unicast mac addresses
1482 * @mc: multicast mac addresses
1483 * @dev_addrs: list of device hw addresses
1484 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001485 * @promiscuity: Number of times the NIC is told to work in
1486 * promiscuous mode; if it becomes 0 the NIC will
1487 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001488 * @allmulti: Counter, enables or disables allmulticast mode
1489 *
1490 * @vlan_info: VLAN info
1491 * @dsa_ptr: dsa specific data
1492 * @tipc_ptr: TIPC specific data
1493 * @atalk_ptr: AppleTalk link
1494 * @ip_ptr: IPv4 specific data
1495 * @dn_ptr: DECnet specific data
1496 * @ip6_ptr: IPv6 specific data
1497 * @ax25_ptr: AX.25 specific data
1498 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1499 *
1500 * @last_rx: Time of last Rx
1501 * @dev_addr: Hw address (before bcast,
1502 * because most packets are unicast)
1503 *
1504 * @_rx: Array of RX queues
1505 * @num_rx_queues: Number of RX queues
1506 * allocated at register_netdev() time
1507 * @real_num_rx_queues: Number of RX queues currently active in device
1508 *
1509 * @rx_handler: handler for received packets
1510 * @rx_handler_data: XXX: need comments on this one
1511 * @ingress_queue: XXX: need comments on this one
1512 * @broadcast: hw bcast address
1513 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001514 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1515 * indexed by RX queue number. Assigned by driver.
1516 * This must only be set if the ndo_rx_flow_steer
1517 * operation is defined
1518 * @index_hlist: Device index hash chain
1519 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001520 * @_tx: Array of TX queues
1521 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1522 * @real_num_tx_queues: Number of TX queues currently active in device
1523 * @qdisc: Root qdisc from userspace point of view
1524 * @tx_queue_len: Max frames per queue allowed
1525 * @tx_global_lock: XXX: need comments on this one
1526 *
1527 * @xps_maps: XXX: need comments on this one
1528 *
Scott Feldman0c4f6912015-07-18 18:24:48 -07001529 * @offload_fwd_mark: Offload device fwding mark
1530 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001531 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001532 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001533 * @watchdog_timer: List of timers
1534 *
1535 * @pcpu_refcnt: Number of references to this device
1536 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001537 * @link_watch_list: XXX: need comments on this one
1538 *
1539 * @reg_state: Register/unregister state machine
1540 * @dismantle: Device is going to be freed
1541 * @rtnl_link_state: This enum represents the phases of creating
1542 * a new link
1543 *
1544 * @destructor: Called from unregister,
1545 * can be used to call free_netdev
1546 * @npinfo: XXX: need comments on this one
1547 * @nd_net: Network namespace this network device is inside
1548 *
1549 * @ml_priv: Mid-layer private
1550 * @lstats: Loopback statistics
1551 * @tstats: Tunnel statistics
1552 * @dstats: Dummy statistics
1553 * @vstats: Virtual ethernet statistics
1554 *
1555 * @garp_port: GARP
1556 * @mrp_port: MRP
1557 *
1558 * @dev: Class/net/name entry
1559 * @sysfs_groups: Space for optional device, statistics and wireless
1560 * sysfs groups
1561 *
1562 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1563 * @rtnl_link_ops: Rtnl_link_ops
1564 *
1565 * @gso_max_size: Maximum size of generic segmentation offload
1566 * @gso_max_segs: Maximum number of segments that can be passed to the
1567 * NIC for GSO
1568 *
1569 * @dcbnl_ops: Data Center Bridging netlink ops
1570 * @num_tc: Number of traffic classes in the net device
1571 * @tc_to_txq: XXX: need comments on this one
1572 * @prio_tc_map XXX: need comments on this one
1573 *
1574 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1575 *
1576 * @priomap: XXX: need comments on this one
1577 * @phydev: Physical device may attach itself
1578 * for hardware timestamping
1579 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001580 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1581 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001582 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001583 * @proto_down: protocol port state information can be sent to the
1584 * switch driver and used to set the phys state of the
1585 * switch port.
1586 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 * FIXME: cleanup struct net_device such that network protocol info
1588 * moves out.
1589 */
1590
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001591struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001593 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001594 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 /*
1596 * I/O specific fields
1597 * FIXME: Merge these and struct ifmap into one
1598 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001599 unsigned long mem_end;
1600 unsigned long mem_start;
1601 unsigned long base_addr;
1602 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001604 atomic_t carrier_changes;
1605
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001607 * Some hardware also needs these fields (state,dev_list,
1608 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 * part of the usual set specified in Space.c.
1610 */
1611
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612 unsigned long state;
1613
Pavel Emelianov7562f872007-05-03 15:13:45 -07001614 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001615 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001616 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001617 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001618 struct list_head ptype_all;
1619 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001620
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001621 struct {
1622 struct list_head upper;
1623 struct list_head lower;
1624 } adj_list;
1625
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001626 struct {
1627 struct list_head upper;
1628 struct list_head lower;
1629 } all_adj_list;
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001630
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001631 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001632 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001633 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001634 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001635 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001636 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001637 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001638
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001640 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641
Rusty Russellc45d2862007-03-28 14:29:08 -07001642 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001643
Eric Dumazet015f0682014-03-27 08:45:56 -07001644 atomic_long_t rx_dropped;
1645 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001646 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
Johannes Bergb86e0282007-04-26 20:48:23 -07001648#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001649 const struct iw_handler_def *wireless_handlers;
1650 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001651#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001652 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001653 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001654#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001655 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001656#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001657#ifdef CONFIG_NET_L3_MASTER_DEV
1658 const struct l3mdev_ops *l3mdev_ops;
1659#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001660#if IS_ENABLED(CONFIG_IPV6)
1661 const struct ndisc_ops *ndisc_ops;
1662#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001664 const struct header_ops *header_ops;
1665
Karoly Kemeny536721b2014-07-30 20:27:36 +02001666 unsigned int flags;
1667 unsigned int priv_flags;
1668
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001670 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671
Karoly Kemeny536721b2014-07-30 20:27:36 +02001672 unsigned char operstate;
1673 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001674
Karoly Kemeny536721b2014-07-30 20:27:36 +02001675 unsigned char if_port;
1676 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001677
Karoly Kemeny536721b2014-07-30 20:27:36 +02001678 unsigned int mtu;
1679 unsigned short type;
1680 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681
Johannes Bergf5184d22008-05-12 20:48:31 -07001682 unsigned short needed_headroom;
1683 unsigned short needed_tailroom;
1684
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001686 unsigned char perm_addr[MAX_ADDR_LEN];
1687 unsigned char addr_assign_type;
1688 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001689 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001690 unsigned short dev_id;
1691 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001692 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001693 unsigned char name_assign_type;
1694 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001695 struct netdev_hw_addr_list uc;
1696 struct netdev_hw_addr_list mc;
1697 struct netdev_hw_addr_list dev_addrs;
1698
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001699#ifdef CONFIG_SYSFS
1700 struct kset *queues_kset;
1701#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001702 unsigned int promiscuity;
1703 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001706 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001707
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001708#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001709 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001710#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001711#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001712 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001713#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001714#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001715 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001716#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001717 void *atalk_ptr;
1718 struct in_device __rcu *ip_ptr;
1719 struct dn_dev __rcu *dn_ptr;
1720 struct inet6_dev __rcu *ip6_ptr;
1721 void *ax25_ptr;
1722 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001723 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001724#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1725 struct mpls_dev __rcu *mpls_ptr;
1726#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001728/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001729 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001730 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001731 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001732
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001733 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001734 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001735
Michael Daltona953be52014-01-16 22:23:28 -08001736#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001737 struct netdev_rx_queue *_rx;
1738
Tom Herbert0a9627f2010-03-16 08:03:29 +00001739 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001740 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001741#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001742
Eric Dumazet3b47d302014-11-06 21:09:44 -08001743 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001744 rx_handler_func_t __rcu *rx_handler;
1745 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001746
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001747#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001748 struct tcf_proto __rcu *ingress_cl_list;
1749#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001750 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001751#ifdef CONFIG_NETFILTER_INGRESS
1752 struct list_head nf_hooks_ingress;
1753#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001754
Karoly Kemeny536721b2014-07-30 20:27:36 +02001755 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001756#ifdef CONFIG_RFS_ACCEL
1757 struct cpu_rmap *rx_cpu_rmap;
1758#endif
1759 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001760
1761/*
1762 * Cache lines mostly used on transmit path
1763 */
David S. Millere8a04642008-07-17 00:34:19 -07001764 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1765 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001766 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001767 struct Qdisc *qdisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001768 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001769 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001770 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001771
Tom Herbertbf264142010-11-26 08:36:09 +00001772#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001773 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001774#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001775#ifdef CONFIG_NET_CLS_ACT
1776 struct tcf_proto __rcu *egress_cl_list;
1777#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001778#ifdef CONFIG_NET_SWITCHDEV
1779 u32 offload_fwd_mark;
1780#endif
1781
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001782 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001783 struct timer_list watchdog_timer;
1784
Eric Dumazet29b44332010-10-11 10:22:12 +00001785 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787
Eric Dumazete014deb2009-11-17 05:59:21 +00001788 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001789
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001791 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 NETREG_UNREGISTERING, /* called unregister_netdevice */
1793 NETREG_UNREGISTERED, /* completed unregister todo */
1794 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001795 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001796 } reg_state:8;
1797
Karoly Kemeny536721b2014-07-30 20:27:36 +02001798 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001799
1800 enum {
1801 RTNL_LINK_INITIALIZED,
1802 RTNL_LINK_INITIALIZING,
1803 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001805 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001808 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809#endif
David S. Millereae792b2008-07-15 03:03:33 -07001810
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001811 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001812
David S. Miller49517042008-05-12 03:29:11 -07001813 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001814 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001815 void *ml_priv;
1816 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001817 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001818 struct pcpu_dstats __percpu *dstats;
1819 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001820 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001821
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001822 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001823 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001825 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001826 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001827 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001828
Patrick McHardy38f7b872007-06-13 12:03:51 -07001829 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001830
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001831 /* for setting kernel sock attribute on TCP connection setup */
1832#define GSO_MAX_SIZE 65536
1833 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001834#define GSO_MAX_SEGS 65535
1835 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001836
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001837#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001838 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001839#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001840 u8 num_tc;
1841 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1842 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001843
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001844#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001845 unsigned int fcoe_ddp_xid;
1846#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001847#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001848 struct netprio_map __rcu *priomap;
1849#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001850 struct phy_device *phydev;
1851 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001852 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001853 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001855#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856
1857#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858
David S. Millere8a04642008-07-17 00:34:19 -07001859static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001860int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1861{
1862 return dev->prio_tc_map[prio & TC_BITMASK];
1863}
1864
1865static inline
1866int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1867{
1868 if (tc >= dev->num_tc)
1869 return -EINVAL;
1870
1871 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1872 return 0;
1873}
1874
1875static inline
1876void netdev_reset_tc(struct net_device *dev)
1877{
1878 dev->num_tc = 0;
1879 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1880 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1881}
1882
1883static inline
1884int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1885{
1886 if (tc >= dev->num_tc)
1887 return -EINVAL;
1888
1889 dev->tc_to_txq[tc].count = count;
1890 dev->tc_to_txq[tc].offset = offset;
1891 return 0;
1892}
1893
1894static inline
1895int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1896{
1897 if (num_tc > TC_MAX_QUEUE)
1898 return -EINVAL;
1899
1900 dev->num_tc = num_tc;
1901 return 0;
1902}
1903
1904static inline
1905int netdev_get_num_tc(struct net_device *dev)
1906{
1907 return dev->num_tc;
1908}
1909
1910static inline
David S. Millere8a04642008-07-17 00:34:19 -07001911struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1912 unsigned int index)
1913{
1914 return &dev->_tx[index];
1915}
1916
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001917static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1918 const struct sk_buff *skb)
1919{
1920 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1921}
1922
David S. Millere8a04642008-07-17 00:34:19 -07001923static inline void netdev_for_each_tx_queue(struct net_device *dev,
1924 void (*f)(struct net_device *,
1925 struct netdev_queue *,
1926 void *),
1927 void *arg)
1928{
1929 unsigned int i;
1930
1931 for (i = 0; i < dev->num_tx_queues; i++)
1932 f(dev, &dev->_tx[i], arg);
1933}
1934
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001935#define netdev_lockdep_set_classes(dev) \
1936{ \
1937 static struct lock_class_key qdisc_tx_busylock_key; \
1938 static struct lock_class_key qdisc_running_key; \
1939 static struct lock_class_key qdisc_xmit_lock_key; \
1940 static struct lock_class_key dev_addr_list_lock_key; \
1941 unsigned int i; \
1942 \
1943 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1944 (dev)->qdisc_running_key = &qdisc_running_key; \
1945 lockdep_set_class(&(dev)->addr_list_lock, \
1946 &dev_addr_list_lock_key); \
1947 for (i = 0; i < (dev)->num_tx_queues; i++) \
1948 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1949 &qdisc_xmit_lock_key); \
1950}
1951
Joe Perchesf629d202013-09-26 14:48:15 -07001952struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001953 struct sk_buff *skb,
1954 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001955
Paolo Abeni871b6422016-02-26 10:45:37 +01001956/* returns the headroom that the master device needs to take in account
1957 * when forwarding to this dev
1958 */
1959static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1960{
1961 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1962}
1963
1964static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1965{
1966 if (dev->netdev_ops->ndo_set_rx_headroom)
1967 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1968}
1969
1970/* set the device rx headroom to the dev's default */
1971static inline void netdev_reset_rx_headroom(struct net_device *dev)
1972{
1973 netdev_set_rx_headroom(dev, -1);
1974}
1975
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001976/*
1977 * Net namespace inlines
1978 */
1979static inline
1980struct net *dev_net(const struct net_device *dev)
1981{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001982 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001983}
1984
1985static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001986void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001987{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001988 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001989}
1990
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07001991static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001992{
Alexander Duyck3fc88672014-09-12 23:12:46 -07001993#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001994 if (dev->dsa_ptr != NULL)
1995 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001996#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001997 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00001998}
1999
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002000/**
2001 * netdev_priv - access network device private data
2002 * @dev: network device
2003 *
2004 * Get network device private data
2005 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002006static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002008 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009}
2010
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011/* Set the sysfs physical device reference for the network logical device
2012 * if set prior to registration will cause a symlink during initialization.
2013 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002014#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015
Marcel Holtmann384912e2009-08-31 21:08:19 +00002016/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002017 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002018 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002019 */
2020#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2021
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002022/* Default NAPI poll() weight
2023 * Device drivers are strongly advised to not use bigger value
2024 */
2025#define NAPI_POLL_WEIGHT 64
2026
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002027/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002028 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002029 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002030 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002031 * @poll: polling function
2032 * @weight: default weight
2033 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002034 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2035 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002036 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002037void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2038 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002039
Alexander Duyckd8156532008-07-08 15:13:05 -07002040/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002041 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002042 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002043 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002044 * @poll: polling function
2045 * @weight: default weight
2046 *
2047 * This variant of netif_napi_add() should be used from drivers using NAPI
2048 * to exclusively poll a TX queue.
2049 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2050 */
2051static inline void netif_tx_napi_add(struct net_device *dev,
2052 struct napi_struct *napi,
2053 int (*poll)(struct napi_struct *, int),
2054 int weight)
2055{
2056 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2057 netif_napi_add(dev, napi, poll, weight);
2058}
2059
2060/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002061 * netif_napi_del - remove a NAPI context
2062 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002063 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002064 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002065 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002066void netif_napi_del(struct napi_struct *napi);
2067
2068struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002069 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002070 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002071
Herbert Xu74895942009-05-26 18:50:27 +00002072 /* Length of frag0. */
2073 unsigned int frag0_len;
2074
Herbert Xu86911732009-01-29 14:19:50 +00002075 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002076 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002077
Herbert Xud565b0a2008-12-15 23:38:52 -08002078 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002079 u16 flush;
2080
2081 /* Save the IP ID here and check when we get to the transport layer */
2082 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002083
2084 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002085 u16 count;
2086
Tom Herbert15e23962015-02-10 16:30:31 -08002087 /* Start offset for remote checksum offload */
2088 u16 gro_remcsum_start;
2089
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002090 /* jiffies when first packet was created/queued */
2091 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002092
Tom Herbertafe93322014-09-17 12:25:57 -07002093 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002094 u16 proto;
2095
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002096 /* This is non-zero if the packet may be of the same flow. */
2097 u8 same_flow:1;
2098
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002099 /* Used in tunnel GRO receive */
2100 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002101
2102 /* GRO checksum is valid */
2103 u8 csum_valid:1;
2104
Tom Herbert662880f2014-08-27 21:26:56 -07002105 /* Number of checksums via CHECKSUM_UNNECESSARY */
2106 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002107
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002108 /* Free the skb? */
2109 u8 free:2;
2110#define NAPI_GRO_FREE 1
2111#define NAPI_GRO_FREE_STOLEN_HEAD 2
2112
Tom Herbertefc98d02014-10-03 15:48:08 -07002113 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2114 u8 is_ipv6:1;
2115
Alexander Duycka0ca1532016-04-05 09:13:39 -07002116 /* Used in GRE, set in fou/gue_gro_receive */
2117 u8 is_fou:1;
2118
Alexander Duyck15305452016-04-10 21:44:57 -04002119 /* Used to determine if flush_id can be ignored */
2120 u8 is_atomic:1;
2121
2122 /* 5 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002123
Jerry Chubf5a7552014-01-07 10:23:19 -08002124 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2125 __wsum csum;
2126
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002127 /* used in skb_gro_receive() slow path */
2128 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002129};
2130
2131#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002132
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002134 __be16 type; /* This is really htons(ether_type). */
2135 struct net_device *dev; /* NULL is wildcarded here */
2136 int (*func) (struct sk_buff *,
2137 struct net_device *,
2138 struct packet_type *,
2139 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002140 bool (*id_match)(struct packet_type *ptype,
2141 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 void *af_packet_priv;
2143 struct list_head list;
2144};
2145
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002146struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2148 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002150 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002151 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002152};
2153
2154struct packet_offload {
2155 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002156 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002157 struct offload_callbacks callbacks;
2158 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159};
2160
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002161/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002162struct pcpu_sw_netstats {
2163 u64 rx_packets;
2164 u64 rx_bytes;
2165 u64 tx_packets;
2166 u64 tx_bytes;
2167 struct u64_stats_sync syncp;
2168};
2169
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002170#define __netdev_alloc_pcpu_stats(type, gfp) \
2171({ \
2172 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2173 if (pcpu_stats) { \
2174 int __cpu; \
2175 for_each_possible_cpu(__cpu) { \
2176 typeof(type) *stat; \
2177 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2178 u64_stats_init(&stat->syncp); \
2179 } \
2180 } \
2181 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002182})
2183
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002184#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002185 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002186
Jiri Pirko764f5e52015-12-03 12:12:12 +01002187enum netdev_lag_tx_type {
2188 NETDEV_LAG_TX_TYPE_UNKNOWN,
2189 NETDEV_LAG_TX_TYPE_RANDOM,
2190 NETDEV_LAG_TX_TYPE_BROADCAST,
2191 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2192 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2193 NETDEV_LAG_TX_TYPE_HASH,
2194};
2195
2196struct netdev_lag_upper_info {
2197 enum netdev_lag_tx_type tx_type;
2198};
2199
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002200struct netdev_lag_lower_state_info {
2201 u8 link_up : 1,
2202 tx_enabled : 1;
2203};
2204
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205#include <linux/notifier.h>
2206
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002207/* netdevice notifier chain. Please remember to update the rtnetlink
2208 * notification exclusion list in rtnetlink_event() when adding new
2209 * types.
2210 */
2211#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2212#define NETDEV_DOWN 0x0002
2213#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2214 detected a hardware crash and restarted
2215 - we can use this eg to kick tcp sessions
2216 once done */
2217#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2218#define NETDEV_REGISTER 0x0005
2219#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002220#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002221#define NETDEV_CHANGEADDR 0x0008
2222#define NETDEV_GOING_DOWN 0x0009
2223#define NETDEV_CHANGENAME 0x000A
2224#define NETDEV_FEAT_CHANGE 0x000B
2225#define NETDEV_BONDING_FAILOVER 0x000C
2226#define NETDEV_PRE_UP 0x000D
2227#define NETDEV_PRE_TYPE_CHANGE 0x000E
2228#define NETDEV_POST_TYPE_CHANGE 0x000F
2229#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002230#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002231#define NETDEV_RELEASE 0x0012
2232#define NETDEV_NOTIFY_PEERS 0x0013
2233#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002234#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002235#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002236#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002237#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002238#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002239#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002240#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002241#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Jason Wang08294a22016-06-30 14:45:35 +08002242#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002243
Joe Perchesf629d202013-09-26 14:48:15 -07002244int register_netdevice_notifier(struct notifier_block *nb);
2245int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002246
2247struct netdev_notifier_info {
2248 struct net_device *dev;
2249};
2250
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002251struct netdev_notifier_change_info {
2252 struct netdev_notifier_info info; /* must be first */
2253 unsigned int flags_changed;
2254};
2255
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002256struct netdev_notifier_changeupper_info {
2257 struct netdev_notifier_info info; /* must be first */
2258 struct net_device *upper_dev; /* new upper dev */
2259 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002260 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002261 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002262};
2263
Jiri Pirko04d48262015-12-03 12:12:15 +01002264struct netdev_notifier_changelowerstate_info {
2265 struct netdev_notifier_info info; /* must be first */
2266 void *lower_state_info; /* is lower dev state */
2267};
2268
Cong Wang75538c22013-05-29 11:30:50 +08002269static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2270 struct net_device *dev)
2271{
2272 info->dev = dev;
2273}
2274
Jiri Pirko351638e2013-05-28 01:30:21 +00002275static inline struct net_device *
2276netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2277{
2278 return info->dev;
2279}
2280
Joe Perchesf629d202013-09-26 14:48:15 -07002281int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002282
2283
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284extern rwlock_t dev_base_lock; /* Device list lock */
2285
Eric W. Biederman881d9662007-09-17 11:56:21 -07002286#define for_each_netdev(net, d) \
2287 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002288#define for_each_netdev_reverse(net, d) \
2289 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002290#define for_each_netdev_rcu(net, d) \
2291 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002292#define for_each_netdev_safe(net, d, n) \
2293 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2294#define for_each_netdev_continue(net, d) \
2295 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002296#define for_each_netdev_continue_rcu(net, d) \
2297 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002298#define for_each_netdev_in_bond_rcu(bond, slave) \
2299 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002300 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002301#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2302
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002303static inline struct net_device *next_net_device(struct net_device *dev)
2304{
2305 struct list_head *lh;
2306 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002307
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002308 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002309 lh = dev->dev_list.next;
2310 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2311}
2312
Eric Dumazetce81b762009-11-11 17:34:30 +00002313static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2314{
2315 struct list_head *lh;
2316 struct net *net;
2317
2318 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002319 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002320 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2321}
2322
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002323static inline struct net_device *first_net_device(struct net *net)
2324{
2325 return list_empty(&net->dev_base_head) ? NULL :
2326 net_device_entry(net->dev_base_head.next);
2327}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002328
Eric Dumazetccf43432011-01-26 18:08:02 +00002329static inline struct net_device *first_net_device_rcu(struct net *net)
2330{
2331 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2332
2333 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2334}
2335
Joe Perchesf629d202013-09-26 14:48:15 -07002336int netdev_boot_setup_check(struct net_device *dev);
2337unsigned long netdev_boot_base(const char *prefix, int unit);
2338struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2339 const char *hwaddr);
2340struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2341struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2342void dev_add_pack(struct packet_type *pt);
2343void dev_remove_pack(struct packet_type *pt);
2344void __dev_remove_pack(struct packet_type *pt);
2345void dev_add_offload(struct packet_offload *po);
2346void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002348int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002349int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002350struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2351 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002352struct net_device *dev_get_by_name(struct net *net, const char *name);
2353struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2354struct net_device *__dev_get_by_name(struct net *net, const char *name);
2355int dev_alloc_name(struct net_device *dev, const char *name);
2356int dev_open(struct net_device *dev);
2357int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002358int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002359void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002360int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002361int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002362int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002363int register_netdevice(struct net_device *dev);
2364void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2365void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002366static inline void unregister_netdevice(struct net_device *dev)
2367{
2368 unregister_netdevice_queue(dev, NULL);
2369}
2370
Joe Perchesf629d202013-09-26 14:48:15 -07002371int netdev_refcnt_read(const struct net_device *dev);
2372void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002373void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002374void synchronize_net(void);
2375int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002376
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002377DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002378#define XMIT_RECURSION_LIMIT 10
2379
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002380static inline int dev_recursion_level(void)
2381{
2382 return this_cpu_read(xmit_recursion);
2383}
2384
Joe Perchesf629d202013-09-26 14:48:15 -07002385struct net_device *dev_get_by_index(struct net *net, int ifindex);
2386struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2387struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2388int netdev_get_name(struct net *net, char *name, int ifindex);
2389int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002390int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002391
2392static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2393{
2394 return NAPI_GRO_CB(skb)->data_offset;
2395}
2396
2397static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2398{
2399 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2400}
2401
2402static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2403{
2404 NAPI_GRO_CB(skb)->data_offset += len;
2405}
2406
Herbert Xua5b1cf22009-05-26 18:50:28 +00002407static inline void *skb_gro_header_fast(struct sk_buff *skb,
2408 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002409{
Herbert Xu78a478d2009-05-26 18:50:21 +00002410 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002411}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412
Herbert Xua5b1cf22009-05-26 18:50:28 +00002413static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2414{
2415 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2416}
2417
2418static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2419 unsigned int offset)
2420{
Herbert Xu17dd7592011-07-27 06:16:28 -07002421 if (!pskb_may_pull(skb, hlen))
2422 return NULL;
2423
Herbert Xua5b1cf22009-05-26 18:50:28 +00002424 NAPI_GRO_CB(skb)->frag0 = NULL;
2425 NAPI_GRO_CB(skb)->frag0_len = 0;
Herbert Xu17dd7592011-07-27 06:16:28 -07002426 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002427}
2428
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002429static inline void *skb_gro_network_header(struct sk_buff *skb)
2430{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002431 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2432 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002433}
2434
Jerry Chubf5a7552014-01-07 10:23:19 -08002435static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2436 const void *start, unsigned int len)
2437{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002438 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002439 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2440 csum_partial(start, len, 0));
2441}
2442
Tom Herbert573e8fc2014-08-22 13:33:47 -07002443/* GRO checksum functions. These are logical equivalents of the normal
2444 * checksum functions (in skbuff.h) except that they operate on the GRO
2445 * offsets and fields in sk_buff.
2446 */
2447
2448__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2449
Tom Herbert15e23962015-02-10 16:30:31 -08002450static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2451{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002452 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002453}
2454
Tom Herbert573e8fc2014-08-22 13:33:47 -07002455static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2456 bool zero_okay,
2457 __sum16 check)
2458{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002459 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2460 skb_checksum_start_offset(skb) <
2461 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002462 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002463 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002464 (!zero_okay || check));
2465}
2466
2467static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2468 __wsum psum)
2469{
2470 if (NAPI_GRO_CB(skb)->csum_valid &&
2471 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2472 return 0;
2473
2474 NAPI_GRO_CB(skb)->csum = psum;
2475
2476 return __skb_gro_checksum_complete(skb);
2477}
2478
Tom Herbert573e8fc2014-08-22 13:33:47 -07002479static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2480{
Tom Herbert662880f2014-08-27 21:26:56 -07002481 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2482 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2483 NAPI_GRO_CB(skb)->csum_cnt--;
2484 } else {
2485 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2486 * verified a new top level checksum or an encapsulated one
2487 * during GRO. This saves work if we fallback to normal path.
2488 */
2489 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002490 }
2491}
2492
2493#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2494 compute_pseudo) \
2495({ \
2496 __sum16 __ret = 0; \
2497 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2498 __ret = __skb_gro_checksum_validate_complete(skb, \
2499 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002500 if (__ret) \
2501 __skb_mark_checksum_bad(skb); \
2502 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002503 skb_gro_incr_csum_unnecessary(skb); \
2504 __ret; \
2505})
2506
2507#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2508 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2509
2510#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2511 compute_pseudo) \
2512 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2513
2514#define skb_gro_checksum_simple_validate(skb) \
2515 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2516
Tom Herbertd96535a2014-08-31 15:12:42 -07002517static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2518{
2519 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2520 !NAPI_GRO_CB(skb)->csum_valid);
2521}
2522
2523static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2524 __sum16 check, __wsum pseudo)
2525{
2526 NAPI_GRO_CB(skb)->csum = ~pseudo;
2527 NAPI_GRO_CB(skb)->csum_valid = 1;
2528}
2529
2530#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2531do { \
2532 if (__skb_gro_checksum_convert_check(skb)) \
2533 __skb_gro_checksum_convert(skb, check, \
2534 compute_pseudo(skb, proto)); \
2535} while (0)
2536
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002537struct gro_remcsum {
2538 int offset;
2539 __wsum delta;
2540};
2541
2542static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2543{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002544 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002545 grc->delta = 0;
2546}
2547
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002548static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2549 unsigned int off, size_t hdrlen,
2550 int start, int offset,
2551 struct gro_remcsum *grc,
2552 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002553{
2554 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002555 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002556
2557 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2558
Tom Herbert15e23962015-02-10 16:30:31 -08002559 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002560 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2561 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002562 }
2563
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002564 ptr = skb_gro_header_fast(skb, off);
2565 if (skb_gro_header_hard(skb, off + plen)) {
2566 ptr = skb_gro_header_slow(skb, off + plen, off);
2567 if (!ptr)
2568 return NULL;
2569 }
2570
2571 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2572 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002573
2574 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002575 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002576
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002577 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002578 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002579
2580 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002581}
2582
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002583static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2584 struct gro_remcsum *grc)
2585{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002586 void *ptr;
2587 size_t plen = grc->offset + sizeof(u16);
2588
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002589 if (!grc->delta)
2590 return;
2591
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002592 ptr = skb_gro_header_fast(skb, grc->offset);
2593 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2594 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2595 if (!ptr)
2596 return;
2597 }
2598
2599 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002600}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002601
Tom Herbert6ae23ad2015-12-14 11:19:46 -08002602struct skb_csum_offl_spec {
2603 __u16 ipv4_okay:1,
2604 ipv6_okay:1,
2605 encap_okay:1,
2606 ip_options_okay:1,
2607 ext_hdrs_okay:1,
2608 tcp_okay:1,
2609 udp_okay:1,
2610 sctp_okay:1,
2611 vlan_okay:1,
2612 no_encapped_ipv6:1,
2613 no_not_encapped:1;
2614};
2615
2616bool __skb_csum_offload_chk(struct sk_buff *skb,
2617 const struct skb_csum_offl_spec *spec,
2618 bool *csum_encapped,
2619 bool csum_help);
2620
2621static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2622 const struct skb_csum_offl_spec *spec,
2623 bool *csum_encapped,
2624 bool csum_help)
2625{
2626 if (skb->ip_summed != CHECKSUM_PARTIAL)
2627 return false;
2628
2629 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2630}
2631
2632static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2633 const struct skb_csum_offl_spec *spec)
2634{
2635 bool csum_encapped;
2636
2637 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2638}
2639
2640static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2641{
2642 static const struct skb_csum_offl_spec csum_offl_spec = {
2643 .ipv4_okay = 1,
2644 .ip_options_okay = 1,
2645 .ipv6_okay = 1,
2646 .vlan_okay = 1,
2647 .tcp_okay = 1,
2648 .udp_okay = 1,
2649 };
2650
2651 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2652}
2653
2654static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2655{
2656 static const struct skb_csum_offl_spec csum_offl_spec = {
2657 .ipv4_okay = 1,
2658 .ip_options_okay = 1,
2659 .tcp_okay = 1,
2660 .udp_okay = 1,
2661 .vlan_okay = 1,
2662 };
2663
2664 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2665}
2666
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002667static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2668 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002669 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002670 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002671{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002672 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002673 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002674
2675 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002676}
2677
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002678static inline int dev_parse_header(const struct sk_buff *skb,
2679 unsigned char *haddr)
2680{
2681 const struct net_device *dev = skb->dev;
2682
Patrick McHardy1b833362007-10-18 05:09:28 -07002683 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002684 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002685 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002686}
2687
Willem de Bruijn2793a232016-03-09 21:58:32 -05002688/* ll_header must have at least hard_header_len allocated */
2689static inline bool dev_validate_header(const struct net_device *dev,
2690 char *ll_header, int len)
2691{
2692 if (likely(len >= dev->hard_header_len))
2693 return true;
2694
2695 if (capable(CAP_SYS_RAWIO)) {
2696 memset(ll_header + len, 0, dev->hard_header_len - len);
2697 return true;
2698 }
2699
2700 if (dev->header_ops && dev->header_ops->validate)
2701 return dev->header_ops->validate(ll_header, len);
2702
2703 return false;
2704}
2705
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002707int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002708static inline int unregister_gifconf(unsigned int family)
2709{
2710 return register_gifconf(family, NULL);
2711}
2712
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002713#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002714#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002715struct sd_flow_limit {
2716 u64 count;
2717 unsigned int num_buckets;
2718 unsigned int history_head;
2719 u16 history[FLOW_LIMIT_HISTORY];
2720 u8 buckets[];
2721};
2722
2723extern int netdev_flow_limit_table_len;
2724#endif /* CONFIG_NET_FLOW_LIMIT */
2725
Linus Torvalds1da177e2005-04-16 15:20:36 -07002726/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002727 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002728 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002729struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002730 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002731 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732
Changli Gaodee42872010-05-02 05:42:16 +00002733 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002734 unsigned int processed;
2735 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002736 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002737#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002738 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002739#endif
2740#ifdef CONFIG_NET_FLOW_LIMIT
2741 struct sd_flow_limit __rcu *flow_limit;
2742#endif
2743 struct Qdisc *output_queue;
2744 struct Qdisc **output_queue_tailp;
2745 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002746
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002747#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002748 /* input_queue_head should be written by cpu owning this struct,
2749 * and only read by other cpus. Worth using a cache line.
2750 */
2751 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2752
2753 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002754 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002755 struct softnet_data *rps_ipi_next;
2756 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002757 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002758#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002759 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002760 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002761 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002762
Linus Torvalds1da177e2005-04-16 15:20:36 -07002763};
2764
Tom Herbert76cc8b12010-05-20 18:37:59 +00002765static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002766{
2767#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002768 sd->input_queue_head++;
2769#endif
2770}
2771
2772static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2773 unsigned int *qtail)
2774{
2775#ifdef CONFIG_RPS
2776 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002777#endif
2778}
2779
Tom Herbert0a9627f2010-03-16 08:03:29 +00002780DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002781
Joe Perchesf629d202013-09-26 14:48:15 -07002782void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002783void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002784
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002785static inline void netif_tx_schedule_all(struct net_device *dev)
2786{
2787 unsigned int i;
2788
2789 for (i = 0; i < dev->num_tx_queues; i++)
2790 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2791}
2792
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002793static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002794{
Tom Herbert734664982011-11-28 16:32:44 +00002795 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002796}
2797
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002798/**
2799 * netif_start_queue - allow transmit
2800 * @dev: network device
2801 *
2802 * Allow upper layers to call the device hard_start_xmit routine.
2803 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804static inline void netif_start_queue(struct net_device *dev)
2805{
David S. Millere8a04642008-07-17 00:34:19 -07002806 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002807}
2808
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002809static inline void netif_tx_start_all_queues(struct net_device *dev)
2810{
2811 unsigned int i;
2812
2813 for (i = 0; i < dev->num_tx_queues; i++) {
2814 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2815 netif_tx_start_queue(txq);
2816 }
2817}
2818
John Fastabend46e5da40a2014-09-12 20:04:52 -07002819void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002820
Dave Jonesd29f7492008-07-22 14:09:06 -07002821/**
2822 * netif_wake_queue - restart transmit
2823 * @dev: network device
2824 *
2825 * Allow upper layers to call the device hard_start_xmit routine.
2826 * Used for flow control when transmit resources are available.
2827 */
David S. Miller79d16382008-07-08 23:14:46 -07002828static inline void netif_wake_queue(struct net_device *dev)
2829{
David S. Millere8a04642008-07-17 00:34:19 -07002830 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002831}
2832
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002833static inline void netif_tx_wake_all_queues(struct net_device *dev)
2834{
2835 unsigned int i;
2836
2837 for (i = 0; i < dev->num_tx_queues; i++) {
2838 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2839 netif_tx_wake_queue(txq);
2840 }
2841}
2842
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002843static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002844{
Tom Herbert734664982011-11-28 16:32:44 +00002845 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002846}
2847
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002848/**
2849 * netif_stop_queue - stop transmitted packets
2850 * @dev: network device
2851 *
2852 * Stop upper layers calling the device hard_start_xmit routine.
2853 * Used for flow control when transmit resources are unavailable.
2854 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002855static inline void netif_stop_queue(struct net_device *dev)
2856{
David S. Millere8a04642008-07-17 00:34:19 -07002857 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858}
2859
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002860void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002861
David S. Miller4d295152012-03-07 21:02:35 -05002862static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002863{
Tom Herbert734664982011-11-28 16:32:44 +00002864 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002865}
2866
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002867/**
2868 * netif_queue_stopped - test if transmit queue is flowblocked
2869 * @dev: network device
2870 *
2871 * Test if transmit queue on device is currently unable to send.
2872 */
David S. Miller4d295152012-03-07 21:02:35 -05002873static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002874{
David S. Millere8a04642008-07-17 00:34:19 -07002875 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876}
2877
David S. Miller4d295152012-03-07 21:02:35 -05002878static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002879{
Tom Herbert734664982011-11-28 16:32:44 +00002880 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2881}
2882
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002883static inline bool
2884netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002885{
2886 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2887}
2888
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002889static inline bool
2890netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2891{
2892 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2893}
2894
Eric Dumazet53511452014-10-08 08:19:27 -07002895/**
2896 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2897 * @dev_queue: pointer to transmit queue
2898 *
2899 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002900 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002901 */
2902static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2903{
2904#ifdef CONFIG_BQL
2905 prefetchw(&dev_queue->dql.num_queued);
2906#endif
2907}
2908
2909/**
2910 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2911 * @dev_queue: pointer to transmit queue
2912 *
2913 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002914 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002915 */
2916static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2917{
2918#ifdef CONFIG_BQL
2919 prefetchw(&dev_queue->dql.limit);
2920#endif
2921}
2922
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002923static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2924 unsigned int bytes)
2925{
Tom Herbert114cf582011-11-28 16:33:09 +00002926#ifdef CONFIG_BQL
2927 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002928
2929 if (likely(dql_avail(&dev_queue->dql) >= 0))
2930 return;
2931
2932 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2933
2934 /*
2935 * The XOFF flag must be set before checking the dql_avail below,
2936 * because in netdev_tx_completed_queue we update the dql_completed
2937 * before checking the XOFF flag.
2938 */
2939 smp_mb();
2940
2941 /* check again in case another CPU has just made room avail */
2942 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2943 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002944#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002945}
2946
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002947/**
2948 * netdev_sent_queue - report the number of bytes queued to hardware
2949 * @dev: network device
2950 * @bytes: number of bytes queued to the hardware device queue
2951 *
2952 * Report the number of bytes queued for sending/completion to the network
2953 * device hardware queue. @bytes should be a good approximation and should
2954 * exactly match netdev_completed_queue() @bytes
2955 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002956static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2957{
2958 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2959}
2960
2961static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002962 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002963{
Tom Herbert114cf582011-11-28 16:33:09 +00002964#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002965 if (unlikely(!bytes))
2966 return;
2967
2968 dql_completed(&dev_queue->dql, bytes);
2969
2970 /*
2971 * Without the memory barrier there is a small possiblity that
2972 * netdev_tx_sent_queue will miss the update and cause the queue to
2973 * be stopped forever
2974 */
2975 smp_mb();
2976
2977 if (dql_avail(&dev_queue->dql) < 0)
2978 return;
2979
2980 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2981 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002982#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002983}
2984
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002985/**
2986 * netdev_completed_queue - report bytes and packets completed by device
2987 * @dev: network device
2988 * @pkts: actual number of packets sent over the medium
2989 * @bytes: actual number of bytes sent over the medium
2990 *
2991 * Report the number of bytes and packets transmitted by the network device
2992 * hardware queue over the physical medium, @bytes must exactly match the
2993 * @bytes amount passed to netdev_sent_queue()
2994 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002995static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00002996 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002997{
2998 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2999}
3000
3001static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3002{
Tom Herbert114cf582011-11-28 16:33:09 +00003003#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003004 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003005 dql_reset(&q->dql);
3006#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003007}
3008
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003009/**
3010 * netdev_reset_queue - reset the packets and bytes count of a network device
3011 * @dev_queue: network device
3012 *
3013 * Reset the bytes and packet count of a network device and clear the
3014 * software flow control OFF bit for this network device
3015 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003016static inline void netdev_reset_queue(struct net_device *dev_queue)
3017{
3018 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003019}
3020
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003021/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003022 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3023 * @dev: network device
3024 * @queue_index: given tx queue index
3025 *
3026 * Returns 0 if given tx queue index >= number of device tx queues,
3027 * otherwise returns the originally passed tx queue index.
3028 */
3029static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3030{
3031 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3032 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3033 dev->name, queue_index,
3034 dev->real_num_tx_queues);
3035 return 0;
3036 }
3037
3038 return queue_index;
3039}
3040
3041/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003042 * netif_running - test if up
3043 * @dev: network device
3044 *
3045 * Test if the device has been brought up.
3046 */
David S. Miller4d295152012-03-07 21:02:35 -05003047static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003048{
3049 return test_bit(__LINK_STATE_START, &dev->state);
3050}
3051
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003052/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003053 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003054 * stop, and a check if it's stopped. All other device management is
3055 * done at the overall netdevice level.
3056 * Also test the device if we're multiqueue.
3057 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003058
3059/**
3060 * netif_start_subqueue - allow sending packets on subqueue
3061 * @dev: network device
3062 * @queue_index: sub queue index
3063 *
3064 * Start individual transmit queue of a device with multiple transmit queues.
3065 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003066static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3067{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003068 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003069
3070 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003071}
3072
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003073/**
3074 * netif_stop_subqueue - stop sending packets on subqueue
3075 * @dev: network device
3076 * @queue_index: sub queue index
3077 *
3078 * Stop individual transmit queue of a device with multiple transmit queues.
3079 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003080static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3081{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003082 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003083 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003084}
3085
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003086/**
3087 * netif_subqueue_stopped - test status of subqueue
3088 * @dev: network device
3089 * @queue_index: sub queue index
3090 *
3091 * Check individual transmit queue of a device with multiple transmit queues.
3092 */
David S. Miller4d295152012-03-07 21:02:35 -05003093static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3094 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003095{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003096 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003097
3098 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003099}
3100
David S. Miller4d295152012-03-07 21:02:35 -05003101static inline bool netif_subqueue_stopped(const struct net_device *dev,
3102 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003103{
3104 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3105}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003106
John Fastabend46e5da40a2014-09-12 20:04:52 -07003107void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003108
Alexander Duyck537c00d2013-01-10 08:57:02 +00003109#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003110int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003111 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003112#else
3113static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003114 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003115 u16 index)
3116{
3117 return 0;
3118}
3119#endif
3120
Jiri Pirko5605c762015-05-12 14:56:12 +02003121u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3122 unsigned int num_tx_queues);
3123
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003124/*
3125 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3126 * as a distribution range limit for the returned value.
3127 */
3128static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003129 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003130{
3131 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3132}
3133
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003134/**
3135 * netif_is_multiqueue - test if device has multiple transmit queues
3136 * @dev: network device
3137 *
3138 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003139 */
David S. Miller4d295152012-03-07 21:02:35 -05003140static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003141{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003142 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003143}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003144
Joe Perchesf629d202013-09-26 14:48:15 -07003145int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003146
Michael Daltona953be52014-01-16 22:23:28 -08003147#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003148int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003149#else
3150static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3151 unsigned int rxq)
3152{
3153 return 0;
3154}
3155#endif
3156
Michael Daltona953be52014-01-16 22:23:28 -08003157#ifdef CONFIG_SYSFS
3158static inline unsigned int get_netdev_rx_queue_index(
3159 struct netdev_rx_queue *queue)
3160{
3161 struct net_device *dev = queue->dev;
3162 int index = queue - dev->_rx;
3163
3164 BUG_ON(index >= dev->num_rx_queues);
3165 return index;
3166}
3167#endif
3168
Yuval Mintz16917b82012-07-01 03:18:50 +00003169#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003170int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003171
Eric Dumazete6247022013-12-05 04:45:08 -08003172enum skb_free_reason {
3173 SKB_REASON_CONSUMED,
3174 SKB_REASON_DROPPED,
3175};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003176
Eric Dumazete6247022013-12-05 04:45:08 -08003177void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3178void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3179
3180/*
3181 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3182 * interrupt context or with hardware interrupts being disabled.
3183 * (in_irq() || irqs_disabled())
3184 *
3185 * We provide four helpers that can be used in following contexts :
3186 *
3187 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3188 * replacing kfree_skb(skb)
3189 *
3190 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3191 * Typically used in place of consume_skb(skb) in TX completion path
3192 *
3193 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3194 * replacing kfree_skb(skb)
3195 *
3196 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3197 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003198 */
Eric Dumazete6247022013-12-05 04:45:08 -08003199static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3200{
3201 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3202}
3203
3204static inline void dev_consume_skb_irq(struct sk_buff *skb)
3205{
3206 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3207}
3208
3209static inline void dev_kfree_skb_any(struct sk_buff *skb)
3210{
3211 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3212}
3213
3214static inline void dev_consume_skb_any(struct sk_buff *skb)
3215{
3216 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3217}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003218
Joe Perchesf629d202013-09-26 14:48:15 -07003219int netif_rx(struct sk_buff *skb);
3220int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003221int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003222gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3223void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3224struct sk_buff *napi_get_frags(struct napi_struct *napi);
3225gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003226struct packet_offload *gro_find_receive_by_type(__be16 type);
3227struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003228
3229static inline void napi_free_frags(struct napi_struct *napi)
3230{
3231 kfree_skb(napi->skb);
3232 napi->skb = NULL;
3233}
3234
Joe Perchesf629d202013-09-26 14:48:15 -07003235int netdev_rx_handler_register(struct net_device *dev,
3236 rx_handler_func_t *rx_handler,
3237 void *rx_handler_data);
3238void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003239
Joe Perchesf629d202013-09-26 14:48:15 -07003240bool dev_valid_name(const char *name);
3241int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3242int dev_ethtool(struct net *net, struct ifreq *);
3243unsigned int dev_get_flags(const struct net_device *);
3244int __dev_change_flags(struct net_device *, unsigned int flags);
3245int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003246void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3247 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003248int dev_change_name(struct net_device *, const char *);
3249int dev_set_alias(struct net_device *, const char *, size_t);
3250int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3251int dev_set_mtu(struct net_device *, int);
3252void dev_set_group(struct net_device *, int);
3253int dev_set_mac_address(struct net_device *, struct sockaddr *);
3254int dev_change_carrier(struct net_device *, bool new_carrier);
3255int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003256 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003257int dev_get_phys_port_name(struct net_device *dev,
3258 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003259int dev_change_proto_down(struct net_device *dev, bool proto_down);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003260struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003261struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3262 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003263int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003264int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003265bool is_skb_forwardable(const struct net_device *dev,
3266 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003267
David Ahern74b20582016-05-10 11:19:50 -07003268void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3269
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003270extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003271
3272/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003273void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003274
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003275/**
3276 * dev_put - release reference to device
3277 * @dev: network device
3278 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003279 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003280 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003281static inline void dev_put(struct net_device *dev)
3282{
Christoph Lameter933393f2011-12-22 11:58:51 -06003283 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003284}
3285
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003286/**
3287 * dev_hold - get reference to device
3288 * @dev: network device
3289 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003290 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003291 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003292static inline void dev_hold(struct net_device *dev)
3293{
Christoph Lameter933393f2011-12-22 11:58:51 -06003294 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003295}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003296
3297/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3298 * and _off may be called from IRQ context, but it is caller
3299 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003300 *
3301 * The name carrier is inappropriate, these functions should really be
3302 * called netif_lowerlayer_*() because they represent the state of any
3303 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003304 */
3305
Joe Perchesf629d202013-09-26 14:48:15 -07003306void linkwatch_init_dev(struct net_device *dev);
3307void linkwatch_fire_event(struct net_device *dev);
3308void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003309
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003310/**
3311 * netif_carrier_ok - test if carrier present
3312 * @dev: network device
3313 *
3314 * Check if carrier is present on device
3315 */
David S. Miller4d295152012-03-07 21:02:35 -05003316static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003317{
3318 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3319}
3320
Joe Perchesf629d202013-09-26 14:48:15 -07003321unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003322
Joe Perchesf629d202013-09-26 14:48:15 -07003323void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003324
Joe Perchesf629d202013-09-26 14:48:15 -07003325void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003326
Joe Perchesf629d202013-09-26 14:48:15 -07003327void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003328
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003329/**
3330 * netif_dormant_on - mark device as dormant.
3331 * @dev: network device
3332 *
3333 * Mark device as dormant (as per RFC2863).
3334 *
3335 * The dormant state indicates that the relevant interface is not
3336 * actually in a condition to pass packets (i.e., it is not 'up') but is
3337 * in a "pending" state, waiting for some external event. For "on-
3338 * demand" interfaces, this new state identifies the situation where the
3339 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003340 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003341static inline void netif_dormant_on(struct net_device *dev)
3342{
3343 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3344 linkwatch_fire_event(dev);
3345}
3346
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003347/**
3348 * netif_dormant_off - set device as not dormant.
3349 * @dev: network device
3350 *
3351 * Device is not in dormant state.
3352 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003353static inline void netif_dormant_off(struct net_device *dev)
3354{
3355 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3356 linkwatch_fire_event(dev);
3357}
3358
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003359/**
3360 * netif_dormant - test if carrier present
3361 * @dev: network device
3362 *
3363 * Check if carrier is present on device
3364 */
David S. Miller4d295152012-03-07 21:02:35 -05003365static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003366{
3367 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3368}
3369
3370
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003371/**
3372 * netif_oper_up - test if device is operational
3373 * @dev: network device
3374 *
3375 * Check if carrier is operational
3376 */
David S. Miller4d295152012-03-07 21:02:35 -05003377static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003378{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003379 return (dev->operstate == IF_OPER_UP ||
3380 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3381}
3382
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003383/**
3384 * netif_device_present - is device available or removed
3385 * @dev: network device
3386 *
3387 * Check if device has not been removed from system.
3388 */
David S. Miller4d295152012-03-07 21:02:35 -05003389static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003390{
3391 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3392}
3393
Joe Perchesf629d202013-09-26 14:48:15 -07003394void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003395
Joe Perchesf629d202013-09-26 14:48:15 -07003396void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003397
3398/*
3399 * Network interface message level settings
3400 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003401
3402enum {
3403 NETIF_MSG_DRV = 0x0001,
3404 NETIF_MSG_PROBE = 0x0002,
3405 NETIF_MSG_LINK = 0x0004,
3406 NETIF_MSG_TIMER = 0x0008,
3407 NETIF_MSG_IFDOWN = 0x0010,
3408 NETIF_MSG_IFUP = 0x0020,
3409 NETIF_MSG_RX_ERR = 0x0040,
3410 NETIF_MSG_TX_ERR = 0x0080,
3411 NETIF_MSG_TX_QUEUED = 0x0100,
3412 NETIF_MSG_INTR = 0x0200,
3413 NETIF_MSG_TX_DONE = 0x0400,
3414 NETIF_MSG_RX_STATUS = 0x0800,
3415 NETIF_MSG_PKTDATA = 0x1000,
3416 NETIF_MSG_HW = 0x2000,
3417 NETIF_MSG_WOL = 0x4000,
3418};
3419
3420#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3421#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3422#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3423#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3424#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3425#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3426#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3427#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3428#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3429#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3430#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3431#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3432#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3433#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3434#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3435
3436static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3437{
3438 /* use default */
3439 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3440 return default_msg_enable_bits;
3441 if (debug_value == 0) /* no output */
3442 return 0;
3443 /* set low N bits */
3444 return (1 << debug_value) - 1;
3445}
3446
David S. Millerc773e842008-07-08 23:13:53 -07003447static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003448{
David S. Millerc773e842008-07-08 23:13:53 -07003449 spin_lock(&txq->_xmit_lock);
3450 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003451}
3452
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003453static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3454{
3455 spin_lock_bh(&txq->_xmit_lock);
3456 txq->xmit_lock_owner = smp_processor_id();
3457}
3458
David S. Miller4d295152012-03-07 21:02:35 -05003459static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003460{
David S. Miller4d295152012-03-07 21:02:35 -05003461 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003462 if (likely(ok))
3463 txq->xmit_lock_owner = smp_processor_id();
3464 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003465}
3466
David S. Millerc773e842008-07-08 23:13:53 -07003467static inline void __netif_tx_unlock(struct netdev_queue *txq)
3468{
3469 txq->xmit_lock_owner = -1;
3470 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003471}
3472
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003473static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3474{
3475 txq->xmit_lock_owner = -1;
3476 spin_unlock_bh(&txq->_xmit_lock);
3477}
3478
Eric Dumazet08baf562009-05-25 22:58:01 -07003479static inline void txq_trans_update(struct netdev_queue *txq)
3480{
3481 if (txq->xmit_lock_owner != -1)
3482 txq->trans_start = jiffies;
3483}
3484
Florian Westphalba162f82016-05-03 16:31:00 +02003485/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3486static inline void netif_trans_update(struct net_device *dev)
3487{
Florian Westphal9b366272016-05-03 16:33:14 +02003488 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3489
3490 if (txq->trans_start != jiffies)
3491 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003492}
3493
David S. Millerc3f26a22008-07-31 16:58:50 -07003494/**
3495 * netif_tx_lock - grab network device transmit lock
3496 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003497 *
3498 * Get network device transmit lock
3499 */
3500static inline void netif_tx_lock(struct net_device *dev)
3501{
3502 unsigned int i;
3503 int cpu;
3504
3505 spin_lock(&dev->tx_global_lock);
3506 cpu = smp_processor_id();
3507 for (i = 0; i < dev->num_tx_queues; i++) {
3508 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3509
3510 /* We are the only thread of execution doing a
3511 * freeze, but we have to grab the _xmit_lock in
3512 * order to synchronize with threads which are in
3513 * the ->hard_start_xmit() handler and already
3514 * checked the frozen bit.
3515 */
3516 __netif_tx_lock(txq, cpu);
3517 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3518 __netif_tx_unlock(txq);
3519 }
3520}
3521
3522static inline void netif_tx_lock_bh(struct net_device *dev)
3523{
3524 local_bh_disable();
3525 netif_tx_lock(dev);
3526}
3527
Herbert Xu932ff272006-06-09 12:20:56 -07003528static inline void netif_tx_unlock(struct net_device *dev)
3529{
David S. Millere8a04642008-07-17 00:34:19 -07003530 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003531
David S. Millere8a04642008-07-17 00:34:19 -07003532 for (i = 0; i < dev->num_tx_queues; i++) {
3533 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003534
David S. Millerc3f26a22008-07-31 16:58:50 -07003535 /* No need to grab the _xmit_lock here. If the
3536 * queue is not stopped for another reason, we
3537 * force a schedule.
3538 */
3539 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003540 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003541 }
3542 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003543}
3544
3545static inline void netif_tx_unlock_bh(struct net_device *dev)
3546{
David S. Millere8a04642008-07-17 00:34:19 -07003547 netif_tx_unlock(dev);
3548 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003549}
3550
David S. Millerc773e842008-07-08 23:13:53 -07003551#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003552 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003553 __netif_tx_lock(txq, cpu); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003554 } \
3555}
3556
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003557#define HARD_TX_TRYLOCK(dev, txq) \
3558 (((dev->features & NETIF_F_LLTX) == 0) ? \
3559 __netif_tx_trylock(txq) : \
3560 true )
3561
David S. Millerc773e842008-07-08 23:13:53 -07003562#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003563 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003564 __netif_tx_unlock(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003565 } \
3566}
3567
Linus Torvalds1da177e2005-04-16 15:20:36 -07003568static inline void netif_tx_disable(struct net_device *dev)
3569{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003570 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003571 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003572
David S. Millerc3f26a22008-07-31 16:58:50 -07003573 local_bh_disable();
3574 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003575 for (i = 0; i < dev->num_tx_queues; i++) {
3576 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003577
3578 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003579 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003580 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003581 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003582 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003583}
3584
David S. Millere308a5d2008-07-15 00:13:44 -07003585static inline void netif_addr_lock(struct net_device *dev)
3586{
3587 spin_lock(&dev->addr_list_lock);
3588}
3589
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003590static inline void netif_addr_lock_nested(struct net_device *dev)
3591{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003592 int subclass = SINGLE_DEPTH_NESTING;
3593
3594 if (dev->netdev_ops->ndo_get_lock_subclass)
3595 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3596
3597 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003598}
3599
David S. Millere308a5d2008-07-15 00:13:44 -07003600static inline void netif_addr_lock_bh(struct net_device *dev)
3601{
3602 spin_lock_bh(&dev->addr_list_lock);
3603}
3604
3605static inline void netif_addr_unlock(struct net_device *dev)
3606{
3607 spin_unlock(&dev->addr_list_lock);
3608}
3609
3610static inline void netif_addr_unlock_bh(struct net_device *dev)
3611{
3612 spin_unlock_bh(&dev->addr_list_lock);
3613}
3614
Jiri Pirkof001fde2009-05-05 02:48:28 +00003615/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003616 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003617 * rcu_read_lock held.
3618 */
3619#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003620 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003621
Linus Torvalds1da177e2005-04-16 15:20:36 -07003622/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3623
Joe Perchesf629d202013-09-26 14:48:15 -07003624void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003625
3626/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003627struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003628 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003629 void (*setup)(struct net_device *),
3630 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003631#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3632 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003633
Tom Gundersenc835a672014-07-14 16:37:24 +02003634#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3635 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3636 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003637
Joe Perchesf629d202013-09-26 14:48:15 -07003638int register_netdev(struct net_device *dev);
3639void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003640
Jiri Pirko22bedad32010-04-01 21:22:57 +00003641/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003642int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3643 struct netdev_hw_addr_list *from_list, int addr_len);
3644void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3645 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003646int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3647 struct net_device *dev,
3648 int (*sync)(struct net_device *, const unsigned char *),
3649 int (*unsync)(struct net_device *,
3650 const unsigned char *));
3651void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3652 struct net_device *dev,
3653 int (*unsync)(struct net_device *,
3654 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003655void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003656
Jiri Pirkof001fde2009-05-05 02:48:28 +00003657/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003658int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3659 unsigned char addr_type);
3660int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3661 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003662void dev_addr_flush(struct net_device *dev);
3663int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003664
3665/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003666int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3667int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3668int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3669int dev_uc_sync(struct net_device *to, struct net_device *from);
3670int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3671void dev_uc_unsync(struct net_device *to, struct net_device *from);
3672void dev_uc_flush(struct net_device *dev);
3673void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003674
Alexander Duyck670e5b82014-05-28 18:44:46 -07003675/**
3676 * __dev_uc_sync - Synchonize device's unicast list
3677 * @dev: device to sync
3678 * @sync: function to call if address should be added
3679 * @unsync: function to call if address should be removed
3680 *
3681 * Add newly added addresses to the interface, and release
3682 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003683 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003684static inline int __dev_uc_sync(struct net_device *dev,
3685 int (*sync)(struct net_device *,
3686 const unsigned char *),
3687 int (*unsync)(struct net_device *,
3688 const unsigned char *))
3689{
3690 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3691}
3692
3693/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003694 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003695 * @dev: device to sync
3696 * @unsync: function to call if address should be removed
3697 *
3698 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003699 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003700static inline void __dev_uc_unsync(struct net_device *dev,
3701 int (*unsync)(struct net_device *,
3702 const unsigned char *))
3703{
3704 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3705}
3706
Jiri Pirko22bedad32010-04-01 21:22:57 +00003707/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003708int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3709int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3710int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3711int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3712int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3713int dev_mc_sync(struct net_device *to, struct net_device *from);
3714int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3715void dev_mc_unsync(struct net_device *to, struct net_device *from);
3716void dev_mc_flush(struct net_device *dev);
3717void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003718
Alexander Duyck670e5b82014-05-28 18:44:46 -07003719/**
3720 * __dev_mc_sync - Synchonize device's multicast list
3721 * @dev: device to sync
3722 * @sync: function to call if address should be added
3723 * @unsync: function to call if address should be removed
3724 *
3725 * Add newly added addresses to the interface, and release
3726 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003727 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003728static inline int __dev_mc_sync(struct net_device *dev,
3729 int (*sync)(struct net_device *,
3730 const unsigned char *),
3731 int (*unsync)(struct net_device *,
3732 const unsigned char *))
3733{
3734 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3735}
3736
3737/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003738 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003739 * @dev: device to sync
3740 * @unsync: function to call if address should be removed
3741 *
3742 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003743 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003744static inline void __dev_mc_unsync(struct net_device *dev,
3745 int (*unsync)(struct net_device *,
3746 const unsigned char *))
3747{
3748 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3749}
3750
Linus Torvalds1da177e2005-04-16 15:20:36 -07003751/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003752void dev_set_rx_mode(struct net_device *dev);
3753void __dev_set_rx_mode(struct net_device *dev);
3754int dev_set_promiscuity(struct net_device *dev, int inc);
3755int dev_set_allmulti(struct net_device *dev, int inc);
3756void netdev_state_change(struct net_device *dev);
3757void netdev_notify_peers(struct net_device *dev);
3758void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003759/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003760void dev_load(struct net *net, const char *name);
3761struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3762 struct rtnl_link_stats64 *storage);
3763void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3764 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003765
3766extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003767extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003768extern int weight_p;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003769
Joe Perchesf629d202013-09-26 14:48:15 -07003770bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003771struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3772 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003773struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3774 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003775
3776/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003777#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3778 for (iter = &(dev)->adj_list.upper, \
3779 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3780 updev; \
3781 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3782
3783/* iterate through upper list, must be called under RCU read lock */
Veaceslav Falico2f268f12013-09-25 09:20:07 +02003784#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3785 for (iter = &(dev)->all_adj_list.upper, \
3786 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3787 updev; \
3788 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003789
Joe Perchesf629d202013-09-26 14:48:15 -07003790void *netdev_lower_get_next_private(struct net_device *dev,
3791 struct list_head **iter);
3792void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3793 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003794
3795#define netdev_for_each_lower_private(dev, priv, iter) \
3796 for (iter = (dev)->adj_list.lower.next, \
3797 priv = netdev_lower_get_next_private(dev, &(iter)); \
3798 priv; \
3799 priv = netdev_lower_get_next_private(dev, &(iter)))
3800
3801#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3802 for (iter = &(dev)->adj_list.lower, \
3803 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3804 priv; \
3805 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3806
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003807void *netdev_lower_get_next(struct net_device *dev,
3808 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003809
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003810#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003811 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003812 ldev = netdev_lower_get_next(dev, &(iter)); \
3813 ldev; \
3814 ldev = netdev_lower_get_next(dev, &(iter)))
3815
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003816struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3817 struct list_head **iter);
3818struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3819 struct list_head **iter);
3820
3821#define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3822 for (iter = (dev)->all_adj_list.lower.next, \
3823 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3824 ldev; \
3825 ldev = netdev_all_lower_get_next(dev, &(iter)))
3826
3827#define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3828 for (iter = (dev)->all_adj_list.lower.next, \
3829 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3830 ldev; \
3831 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3832
Joe Perchesf629d202013-09-26 14:48:15 -07003833void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003834void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003835struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3836struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3837int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3838int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003839 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003840 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003841void netdev_upper_dev_unlink(struct net_device *dev,
3842 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003843void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003844void *netdev_lower_dev_get_private(struct net_device *dev,
3845 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003846void netdev_lower_state_changed(struct net_device *lower_dev,
3847 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003848
3849/* RSS keys are 40 or 52 bytes long */
3850#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003851extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003852void netdev_rss_key_fill(void *buffer, size_t len);
3853
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003854int dev_get_nest_level(struct net_device *dev,
Jiri Pirkob618aaa2015-12-04 15:01:31 +01003855 bool (*type_check)(const struct net_device *dev));
Joe Perchesf629d202013-09-26 14:48:15 -07003856int skb_checksum_help(struct sk_buff *skb);
3857struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3858 netdev_features_t features, bool tx_path);
3859struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3860 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003861
Moni Shoua61bd3852015-02-03 16:48:29 +02003862struct netdev_bonding_info {
3863 ifslave slave;
3864 ifbond master;
3865};
3866
3867struct netdev_notifier_bonding_info {
3868 struct netdev_notifier_info info; /* must be first */
3869 struct netdev_bonding_info bonding_info;
3870};
3871
3872void netdev_bonding_info_change(struct net_device *dev,
3873 struct netdev_bonding_info *bonding_info);
3874
Cong Wang12b00042013-02-05 16:36:38 +00003875static inline
3876struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3877{
3878 return __skb_gso_segment(skb, features, true);
3879}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003880__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003881
3882static inline bool can_checksum_protocol(netdev_features_t features,
3883 __be16 protocol)
3884{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003885 if (protocol == htons(ETH_P_FCOE))
3886 return !!(features & NETIF_F_FCOE_CRC);
3887
3888 /* Assume this is an IP checksum (not SCTP CRC) */
3889
3890 if (features & NETIF_F_HW_CSUM) {
3891 /* Can checksum everything */
3892 return true;
3893 }
3894
3895 switch (protocol) {
3896 case htons(ETH_P_IP):
3897 return !!(features & NETIF_F_IP_CSUM);
3898 case htons(ETH_P_IPV6):
3899 return !!(features & NETIF_F_IPV6_CSUM);
3900 default:
3901 return false;
3902 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003903}
Cong Wang12b00042013-02-05 16:36:38 +00003904
Tom Herbert6ae23ad2015-12-14 11:19:46 -08003905/* Map an ethertype into IP protocol if possible */
3906static inline int eproto_to_ipproto(int eproto)
3907{
3908 switch (eproto) {
3909 case htons(ETH_P_IP):
3910 return IPPROTO_IP;
3911 case htons(ETH_P_IPV6):
3912 return IPPROTO_IPV6;
3913 default:
3914 return -1;
3915 }
3916}
3917
Herbert Xufb286bb2005-11-10 13:01:24 -08003918#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003919void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003920#else
3921static inline void netdev_rx_csum_fault(struct net_device *dev)
3922{
3923}
3924#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003925/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003926void net_enable_timestamp(void);
3927void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003928
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003929#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003930int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003931#else
3932#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003933#endif
3934
David S. Miller47982482014-08-22 16:21:53 -07003935static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003936 struct sk_buff *skb, struct net_device *dev,
3937 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003938{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003939 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003940 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003941}
3942
David S. Miller10b3ad82014-08-29 21:07:24 -07003943static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003944 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003945{
3946 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003947 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003948
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003949 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003950 if (rc == NETDEV_TX_OK)
3951 txq_trans_update(txq);
3952
3953 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003954}
3955
Linus Torvalds42a2d922013-11-13 17:40:34 +09003956int netdev_class_create_file_ns(struct class_attribute *class_attr,
3957 const void *ns);
3958void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3959 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003960
3961static inline int netdev_class_create_file(struct class_attribute *class_attr)
3962{
3963 return netdev_class_create_file_ns(class_attr, NULL);
3964}
3965
3966static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3967{
3968 netdev_class_remove_file_ns(class_attr, NULL);
3969}
Jay Vosburghb8a97872008-06-13 18:12:04 -07003970
Johannes Berg04600792010-08-05 17:45:15 +02003971extern struct kobj_ns_type_operations net_ns_type_operations;
3972
Joe Perchesf629d202013-09-26 14:48:15 -07003973const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07003974
Joe Perchesf629d202013-09-26 14:48:15 -07003975void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003976
Michal Kubečekda081432014-05-20 08:29:25 +02003977static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3978 netdev_features_t f2)
3979{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003980 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3981 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08003982 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08003983 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08003984 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08003985 }
Michal Kubečekda081432014-05-20 08:29:25 +02003986
Tom Herbertc8cd0982015-12-14 11:19:44 -08003987 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02003988}
3989
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003990static inline netdev_features_t netdev_get_wanted_features(
3991 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00003992{
3993 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3994}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003995netdev_features_t netdev_increment_features(netdev_features_t all,
3996 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00003997
3998/* Allow TSO being used on stacked device :
3999 * Performing the GSO segmentation before last device
4000 * is a performance improvement.
4001 */
4002static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4003 netdev_features_t mask)
4004{
4005 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4006}
4007
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004008int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004009void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004010void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004011
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004012void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4013 struct net_device *dev);
4014
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004015netdev_features_t passthru_features_check(struct sk_buff *skb,
4016 struct net_device *dev,
4017 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004018netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004019
David S. Miller4d295152012-03-07 21:02:35 -05004020static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004021{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004022 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004023
4024 /* check flags correspondence */
4025 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4026 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4027 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4028 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004029 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004030 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4031 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004032 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4033 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004034 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4035 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004036 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4037 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 -04004038 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004039 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004040 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004041
Herbert Xubcd76112006-06-30 13:36:35 -07004042 return (features & feature) == feature;
4043}
4044
David S. Miller4d295152012-03-07 21:02:35 -05004045static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004046{
Herbert Xu278b2512009-06-03 21:20:51 -07004047 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004048 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004049}
4050
Johannes Berg8b86a612015-04-17 15:45:04 +02004051static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004052 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004053{
Jesse Grossfc741212011-01-09 06:23:32 +00004054 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004055 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4056 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004057}
4058
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004059static inline void netif_set_gso_max_size(struct net_device *dev,
4060 unsigned int size)
4061{
4062 dev->gso_max_size = size;
4063}
4064
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004065static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4066 int pulled_hlen, u16 mac_offset,
4067 int mac_len)
4068{
4069 skb->protocol = protocol;
4070 skb->encapsulation = 1;
4071 skb_push(skb, pulled_hlen);
4072 skb_reset_transport_header(skb);
4073 skb->mac_header = mac_offset;
4074 skb->network_header = skb->mac_header + mac_len;
4075 skb->mac_len = mac_len;
4076}
4077
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004078static inline bool netif_is_macsec(const struct net_device *dev)
4079{
4080 return dev->priv_flags & IFF_MACSEC;
4081}
4082
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004083static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004084{
4085 return dev->priv_flags & IFF_MACVLAN;
4086}
4087
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004088static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004089{
4090 return dev->priv_flags & IFF_MACVLAN_PORT;
4091}
4092
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004093static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004094{
4095 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4096}
4097
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004098static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004099{
4100 return dev->priv_flags & IFF_IPVLAN_MASTER;
4101}
4102
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004103static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004104{
4105 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4106}
4107
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004108static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004109{
4110 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4111}
4112
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004113static inline bool netif_supports_nofcs(struct net_device *dev)
4114{
4115 return dev->priv_flags & IFF_SUPP_NOFCS;
4116}
4117
David Ahern007979e2015-09-29 20:07:10 -07004118static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004119{
David Ahern007979e2015-09-29 20:07:10 -07004120 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004121}
4122
David Ahernfee6d4c2015-10-05 08:51:24 -07004123static inline bool netif_is_l3_slave(const struct net_device *dev)
4124{
4125 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4126}
4127
Jiri Pirko0894ae32015-08-27 09:31:19 +02004128static inline bool netif_is_bridge_master(const struct net_device *dev)
4129{
4130 return dev->priv_flags & IFF_EBRIDGE;
4131}
4132
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004133static inline bool netif_is_bridge_port(const struct net_device *dev)
4134{
4135 return dev->priv_flags & IFF_BRIDGE_PORT;
4136}
4137
Jiri Pirko35d4e172015-08-27 09:31:20 +02004138static inline bool netif_is_ovs_master(const struct net_device *dev)
4139{
4140 return dev->priv_flags & IFF_OPENVSWITCH;
4141}
4142
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004143static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004144{
4145 return dev->priv_flags & IFF_TEAM;
4146}
4147
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004148static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004149{
4150 return dev->priv_flags & IFF_TEAM_PORT;
4151}
4152
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004153static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004154{
4155 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4156}
4157
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004158static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004159{
4160 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4161}
4162
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004163static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4164{
4165 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4166}
4167
Eric Dumazet02875872014-10-05 18:38:35 -07004168/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4169static inline void netif_keep_dst(struct net_device *dev)
4170{
4171 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4172}
4173
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004174extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004175
Joe Perches571ba422010-02-09 11:49:47 +00004176/* Logging, debugging and troubleshooting/diagnostic helpers. */
4177
4178/* netdev_printk helpers, similar to dev_printk */
4179
4180static inline const char *netdev_name(const struct net_device *dev)
4181{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004182 if (!dev->name[0] || strchr(dev->name, '%'))
4183 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004184 return dev->name;
4185}
4186
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004187static inline const char *netdev_reg_state(const struct net_device *dev)
4188{
4189 switch (dev->reg_state) {
4190 case NETREG_UNINITIALIZED: return " (uninitialized)";
4191 case NETREG_REGISTERED: return "";
4192 case NETREG_UNREGISTERING: return " (unregistering)";
4193 case NETREG_UNREGISTERED: return " (unregistered)";
4194 case NETREG_RELEASED: return " (released)";
4195 case NETREG_DUMMY: return " (dummy)";
4196 }
4197
4198 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4199 return " (unknown)";
4200}
4201
Joe Perchesf629d202013-09-26 14:48:15 -07004202__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004203void netdev_printk(const char *level, const struct net_device *dev,
4204 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004205__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004206void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004207__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004208void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004209__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004210void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004211__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004212void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004213__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004214void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004215__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004216void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004217__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004218void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004219
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004220#define MODULE_ALIAS_NETDEV(device) \
4221 MODULE_ALIAS("netdev-" device)
4222
Jim Cromieb558c962011-12-19 17:11:18 -05004223#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004224#define netdev_dbg(__dev, format, args...) \
4225do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004226 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004227} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004228#elif defined(DEBUG)
4229#define netdev_dbg(__dev, format, args...) \
4230 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004231#else
4232#define netdev_dbg(__dev, format, args...) \
4233({ \
4234 if (0) \
4235 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004236})
4237#endif
4238
4239#if defined(VERBOSE_DEBUG)
4240#define netdev_vdbg netdev_dbg
4241#else
4242
4243#define netdev_vdbg(dev, format, args...) \
4244({ \
4245 if (0) \
4246 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4247 0; \
4248})
4249#endif
4250
4251/*
4252 * netdev_WARN() acts like dev_printk(), but with the key difference
4253 * of using a WARN/WARN_ON to get the message out, including the
4254 * file/line information and a backtrace.
4255 */
4256#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004257 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4258 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004259
Joe Perchesb3d95c52010-02-09 11:49:49 +00004260/* netif printk helpers, similar to netdev_printk */
4261
4262#define netif_printk(priv, type, level, dev, fmt, args...) \
4263do { \
4264 if (netif_msg_##type(priv)) \
4265 netdev_printk(level, (dev), fmt, ##args); \
4266} while (0)
4267
Joe Perchesf45f4322010-06-27 01:02:36 +00004268#define netif_level(level, priv, type, dev, fmt, args...) \
4269do { \
4270 if (netif_msg_##type(priv)) \
4271 netdev_##level(dev, fmt, ##args); \
4272} while (0)
4273
Joe Perchesb3d95c52010-02-09 11:49:49 +00004274#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004275 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004276#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004277 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004278#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004279 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004280#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004281 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004282#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004283 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004284#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004285 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004286#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004287 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004288
Joe Perches0053ea92012-05-30 07:43:34 +00004289#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004290#define netif_dbg(priv, type, netdev, format, args...) \
4291do { \
4292 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004293 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004294} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004295#elif defined(DEBUG)
4296#define netif_dbg(priv, type, dev, format, args...) \
4297 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004298#else
4299#define netif_dbg(priv, type, dev, format, args...) \
4300({ \
4301 if (0) \
4302 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4303 0; \
4304})
4305#endif
4306
4307#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004308#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004309#else
4310#define netif_vdbg(priv, type, dev, format, args...) \
4311({ \
4312 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004313 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004314 0; \
4315})
4316#endif
Joe Perches571ba422010-02-09 11:49:47 +00004317
Cong Wang900ff8c2013-02-18 19:20:33 +00004318/*
4319 * The list of packet types we will receive (as opposed to discard)
4320 * and the routines to invoke.
4321 *
4322 * Why 16. Because with 16 the only overlap we get on a hash of the
4323 * low nibble of the protocol value is RARP/SNAP/X.25.
4324 *
4325 * NOTE: That is no longer true with the addition of VLAN tags. Not
4326 * sure which should go first, but I bet it won't make much
4327 * difference if we are running VLANs. The good news is that
4328 * this protocol won't be in the list unless compiled in, so
4329 * the average user (w/out VLANs) will not be adversely affected.
4330 * --BLG
4331 *
4332 * 0800 IP
4333 * 8100 802.1Q VLAN
4334 * 0001 802.3
4335 * 0002 AX.25
4336 * 0004 802.2
4337 * 8035 RARP
4338 * 0005 SNAP
4339 * 0805 X.25
4340 * 0806 ARP
4341 * 8137 IPX
4342 * 0009 Localtalk
4343 * 86DD IPv6
4344 */
4345#define PTYPE_HASH_SIZE (16)
4346#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4347
Jiri Pirko385a1542009-05-27 15:48:07 -07004348#endif /* _LINUX_NETDEVICE_H */