Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NET3 Protocol independent device support routines. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public License | |
6 | * as published by the Free Software Foundation; either version | |
7 | * 2 of the License, or (at your option) any later version. | |
8 | * | |
9 | * Derived from the non IP parts of dev.c 1.0.19 | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
12 | * Mark Evans, <evansmp@uhura.aston.ac.uk> | |
13 | * | |
14 | * Additional Authors: | |
15 | * Florian la Roche <rzsfl@rz.uni-sb.de> | |
16 | * Alan Cox <gw4pts@gw4pts.ampr.org> | |
17 | * David Hinds <dahinds@users.sourceforge.net> | |
18 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> | |
19 | * Adam Sulmicki <adam@cfar.umd.edu> | |
20 | * Pekka Riikonen <priikone@poesidon.pspt.fi> | |
21 | * | |
22 | * Changes: | |
23 | * D.J. Barrow : Fixed bug where dev->refcnt gets set | |
24 | * to 2 if register_netdev gets called | |
25 | * before net_dev_init & also removed a | |
26 | * few lines of code in the process. | |
27 | * Alan Cox : device private ioctl copies fields back. | |
28 | * Alan Cox : Transmit queue code does relevant | |
29 | * stunts to keep the queue safe. | |
30 | * Alan Cox : Fixed double lock. | |
31 | * Alan Cox : Fixed promisc NULL pointer trap | |
32 | * ???????? : Support the full private ioctl range | |
33 | * Alan Cox : Moved ioctl permission check into | |
34 | * drivers | |
35 | * Tim Kordas : SIOCADDMULTI/SIOCDELMULTI | |
36 | * Alan Cox : 100 backlog just doesn't cut it when | |
37 | * you start doing multicast video 8) | |
38 | * Alan Cox : Rewrote net_bh and list manager. | |
39 | * Alan Cox : Fix ETH_P_ALL echoback lengths. | |
40 | * Alan Cox : Took out transmit every packet pass | |
41 | * Saved a few bytes in the ioctl handler | |
42 | * Alan Cox : Network driver sets packet type before | |
43 | * calling netif_rx. Saves a function | |
44 | * call a packet. | |
45 | * Alan Cox : Hashed net_bh() | |
46 | * Richard Kooijman: Timestamp fixes. | |
47 | * Alan Cox : Wrong field in SIOCGIFDSTADDR | |
48 | * Alan Cox : Device lock protection. | |
49 | * Alan Cox : Fixed nasty side effect of device close | |
50 | * changes. | |
51 | * Rudi Cilibrasi : Pass the right thing to | |
52 | * set_mac_address() | |
53 | * Dave Miller : 32bit quantity for the device lock to | |
54 | * make it work out on a Sparc. | |
55 | * Bjorn Ekwall : Added KERNELD hack. | |
56 | * Alan Cox : Cleaned up the backlog initialise. | |
57 | * Craig Metz : SIOCGIFCONF fix if space for under | |
58 | * 1 device. | |
59 | * Thomas Bogendoerfer : Return ENODEV for dev_open, if there | |
60 | * is no device open function. | |
61 | * Andi Kleen : Fix error reporting for SIOCGIFCONF | |
62 | * Michael Chastain : Fix signed/unsigned for SIOCGIFCONF | |
63 | * Cyrus Durgin : Cleaned for KMOD | |
64 | * Adam Sulmicki : Bug Fix : Network Device Unload | |
65 | * A network device unload needs to purge | |
66 | * the backlog queue. | |
67 | * Paul Rusty Russell : SIOCSIFNAME | |
68 | * Pekka Riikonen : Netdev boot-time settings code | |
69 | * Andrew Morton : Make unregister_netdevice wait | |
70 | * indefinitely on dev->refcnt | |
71 | * J Hadi Salim : - Backlog queue sampling | |
72 | * - netif_rx() feedback | |
73 | */ | |
74 | ||
75 | #include <asm/uaccess.h> | |
76 | #include <asm/system.h> | |
77 | #include <linux/bitops.h> | |
4fc268d2 | 78 | #include <linux/capability.h> |
1da177e4 LT |
79 | #include <linux/cpu.h> |
80 | #include <linux/types.h> | |
81 | #include <linux/kernel.h> | |
82 | #include <linux/sched.h> | |
4a3e2f71 | 83 | #include <linux/mutex.h> |
1da177e4 LT |
84 | #include <linux/string.h> |
85 | #include <linux/mm.h> | |
86 | #include <linux/socket.h> | |
87 | #include <linux/sockios.h> | |
88 | #include <linux/errno.h> | |
89 | #include <linux/interrupt.h> | |
90 | #include <linux/if_ether.h> | |
91 | #include <linux/netdevice.h> | |
92 | #include <linux/etherdevice.h> | |
0187bdfb | 93 | #include <linux/ethtool.h> |
1da177e4 LT |
94 | #include <linux/notifier.h> |
95 | #include <linux/skbuff.h> | |
457c4cbc | 96 | #include <net/net_namespace.h> |
1da177e4 LT |
97 | #include <net/sock.h> |
98 | #include <linux/rtnetlink.h> | |
99 | #include <linux/proc_fs.h> | |
100 | #include <linux/seq_file.h> | |
101 | #include <linux/stat.h> | |
102 | #include <linux/if_bridge.h> | |
b863ceb7 | 103 | #include <linux/if_macvlan.h> |
1da177e4 LT |
104 | #include <net/dst.h> |
105 | #include <net/pkt_sched.h> | |
106 | #include <net/checksum.h> | |
107 | #include <linux/highmem.h> | |
108 | #include <linux/init.h> | |
109 | #include <linux/kmod.h> | |
110 | #include <linux/module.h> | |
1da177e4 LT |
111 | #include <linux/netpoll.h> |
112 | #include <linux/rcupdate.h> | |
113 | #include <linux/delay.h> | |
295f4a1f | 114 | #include <net/wext.h> |
1da177e4 | 115 | #include <net/iw_handler.h> |
1da177e4 | 116 | #include <asm/current.h> |
5bdb9886 | 117 | #include <linux/audit.h> |
db217334 | 118 | #include <linux/dmaengine.h> |
f6a78bfc | 119 | #include <linux/err.h> |
c7fa9d18 | 120 | #include <linux/ctype.h> |
723e98b7 | 121 | #include <linux/if_arp.h> |
6de329e2 | 122 | #include <linux/if_vlan.h> |
8f0f2223 | 123 | #include <linux/ip.h> |
ad55dcaf | 124 | #include <net/ip.h> |
8f0f2223 DM |
125 | #include <linux/ipv6.h> |
126 | #include <linux/in.h> | |
b6b2fed1 DM |
127 | #include <linux/jhash.h> |
128 | #include <linux/random.h> | |
9cbc1cb8 | 129 | #include <trace/events/napi.h> |
1da177e4 | 130 | |
342709ef PE |
131 | #include "net-sysfs.h" |
132 | ||
d565b0a1 HX |
133 | /* Instead of increasing this, you should create a hash table. */ |
134 | #define MAX_GRO_SKBS 8 | |
135 | ||
5d38a079 HX |
136 | /* This should be increased if a protocol with a bigger head is added. */ |
137 | #define GRO_MAX_HEAD (MAX_HEADER + 128) | |
138 | ||
1da177e4 LT |
139 | /* |
140 | * The list of packet types we will receive (as opposed to discard) | |
141 | * and the routines to invoke. | |
142 | * | |
143 | * Why 16. Because with 16 the only overlap we get on a hash of the | |
144 | * low nibble of the protocol value is RARP/SNAP/X.25. | |
145 | * | |
146 | * NOTE: That is no longer true with the addition of VLAN tags. Not | |
147 | * sure which should go first, but I bet it won't make much | |
148 | * difference if we are running VLANs. The good news is that | |
149 | * this protocol won't be in the list unless compiled in, so | |
3041a069 | 150 | * the average user (w/out VLANs) will not be adversely affected. |
1da177e4 LT |
151 | * --BLG |
152 | * | |
153 | * 0800 IP | |
154 | * 8100 802.1Q VLAN | |
155 | * 0001 802.3 | |
156 | * 0002 AX.25 | |
157 | * 0004 802.2 | |
158 | * 8035 RARP | |
159 | * 0005 SNAP | |
160 | * 0805 X.25 | |
161 | * 0806 ARP | |
162 | * 8137 IPX | |
163 | * 0009 Localtalk | |
164 | * 86DD IPv6 | |
165 | */ | |
166 | ||
82d8a867 PE |
167 | #define PTYPE_HASH_SIZE (16) |
168 | #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) | |
169 | ||
1da177e4 | 170 | static DEFINE_SPINLOCK(ptype_lock); |
82d8a867 | 171 | static struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; |
6b2bedc3 | 172 | static struct list_head ptype_all __read_mostly; /* Taps */ |
1da177e4 | 173 | |
1da177e4 | 174 | /* |
7562f876 | 175 | * The @dev_base_head list is protected by @dev_base_lock and the rtnl |
1da177e4 LT |
176 | * semaphore. |
177 | * | |
178 | * Pure readers hold dev_base_lock for reading. | |
179 | * | |
180 | * Writers must hold the rtnl semaphore while they loop through the | |
7562f876 | 181 | * dev_base_head list, and hold dev_base_lock for writing when they do the |
1da177e4 LT |
182 | * actual updates. This allows pure readers to access the list even |
183 | * while a writer is preparing to update it. | |
184 | * | |
185 | * To put it another way, dev_base_lock is held for writing only to | |
186 | * protect against pure readers; the rtnl semaphore provides the | |
187 | * protection against other writers. | |
188 | * | |
189 | * See, for example usages, register_netdevice() and | |
190 | * unregister_netdevice(), which must be called with the rtnl | |
191 | * semaphore held. | |
192 | */ | |
1da177e4 | 193 | DEFINE_RWLOCK(dev_base_lock); |
1da177e4 LT |
194 | EXPORT_SYMBOL(dev_base_lock); |
195 | ||
196 | #define NETDEV_HASHBITS 8 | |
881d966b | 197 | #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS) |
1da177e4 | 198 | |
881d966b | 199 | static inline struct hlist_head *dev_name_hash(struct net *net, const char *name) |
1da177e4 LT |
200 | { |
201 | unsigned hash = full_name_hash(name, strnlen(name, IFNAMSIZ)); | |
881d966b | 202 | return &net->dev_name_head[hash & ((1 << NETDEV_HASHBITS) - 1)]; |
1da177e4 LT |
203 | } |
204 | ||
881d966b | 205 | static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex) |
1da177e4 | 206 | { |
881d966b | 207 | return &net->dev_index_head[ifindex & ((1 << NETDEV_HASHBITS) - 1)]; |
1da177e4 LT |
208 | } |
209 | ||
ce286d32 EB |
210 | /* Device list insertion */ |
211 | static int list_netdevice(struct net_device *dev) | |
212 | { | |
c346dca1 | 213 | struct net *net = dev_net(dev); |
ce286d32 EB |
214 | |
215 | ASSERT_RTNL(); | |
216 | ||
217 | write_lock_bh(&dev_base_lock); | |
218 | list_add_tail(&dev->dev_list, &net->dev_base_head); | |
219 | hlist_add_head(&dev->name_hlist, dev_name_hash(net, dev->name)); | |
220 | hlist_add_head(&dev->index_hlist, dev_index_hash(net, dev->ifindex)); | |
221 | write_unlock_bh(&dev_base_lock); | |
222 | return 0; | |
223 | } | |
224 | ||
225 | /* Device list removal */ | |
226 | static void unlist_netdevice(struct net_device *dev) | |
227 | { | |
228 | ASSERT_RTNL(); | |
229 | ||
230 | /* Unlink dev from the device chain */ | |
231 | write_lock_bh(&dev_base_lock); | |
232 | list_del(&dev->dev_list); | |
233 | hlist_del(&dev->name_hlist); | |
234 | hlist_del(&dev->index_hlist); | |
235 | write_unlock_bh(&dev_base_lock); | |
236 | } | |
237 | ||
1da177e4 LT |
238 | /* |
239 | * Our notifier list | |
240 | */ | |
241 | ||
f07d5b94 | 242 | static RAW_NOTIFIER_HEAD(netdev_chain); |
1da177e4 LT |
243 | |
244 | /* | |
245 | * Device drivers call our routines to queue packets here. We empty the | |
246 | * queue in the local softnet handler. | |
247 | */ | |
bea3348e SH |
248 | |
249 | DEFINE_PER_CPU(struct softnet_data, softnet_data); | |
d1b19dff | 250 | EXPORT_PER_CPU_SYMBOL(softnet_data); |
1da177e4 | 251 | |
cf508b12 | 252 | #ifdef CONFIG_LOCKDEP |
723e98b7 | 253 | /* |
c773e847 | 254 | * register_netdevice() inits txq->_xmit_lock and sets lockdep class |
723e98b7 JP |
255 | * according to dev->type |
256 | */ | |
257 | static const unsigned short netdev_lock_type[] = | |
258 | {ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25, | |
259 | ARPHRD_PRONET, ARPHRD_CHAOS, ARPHRD_IEEE802, ARPHRD_ARCNET, | |
260 | ARPHRD_APPLETLK, ARPHRD_DLCI, ARPHRD_ATM, ARPHRD_METRICOM, | |
261 | ARPHRD_IEEE1394, ARPHRD_EUI64, ARPHRD_INFINIBAND, ARPHRD_SLIP, | |
262 | ARPHRD_CSLIP, ARPHRD_SLIP6, ARPHRD_CSLIP6, ARPHRD_RSRVD, | |
263 | ARPHRD_ADAPT, ARPHRD_ROSE, ARPHRD_X25, ARPHRD_HWX25, | |
264 | ARPHRD_PPP, ARPHRD_CISCO, ARPHRD_LAPB, ARPHRD_DDCMP, | |
265 | ARPHRD_RAWHDLC, ARPHRD_TUNNEL, ARPHRD_TUNNEL6, ARPHRD_FRAD, | |
266 | ARPHRD_SKIP, ARPHRD_LOOPBACK, ARPHRD_LOCALTLK, ARPHRD_FDDI, | |
267 | ARPHRD_BIF, ARPHRD_SIT, ARPHRD_IPDDP, ARPHRD_IPGRE, | |
268 | ARPHRD_PIMREG, ARPHRD_HIPPI, ARPHRD_ASH, ARPHRD_ECONET, | |
269 | ARPHRD_IRDA, ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL, | |
270 | ARPHRD_FCFABRIC, ARPHRD_IEEE802_TR, ARPHRD_IEEE80211, | |
2d91d78b | 271 | ARPHRD_IEEE80211_PRISM, ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, |
929122cd | 272 | ARPHRD_PHONET_PIPE, ARPHRD_IEEE802154, |
fcb94e42 | 273 | ARPHRD_VOID, ARPHRD_NONE}; |
723e98b7 | 274 | |
36cbd3dc | 275 | static const char *const netdev_lock_name[] = |
723e98b7 JP |
276 | {"_xmit_NETROM", "_xmit_ETHER", "_xmit_EETHER", "_xmit_AX25", |
277 | "_xmit_PRONET", "_xmit_CHAOS", "_xmit_IEEE802", "_xmit_ARCNET", | |
278 | "_xmit_APPLETLK", "_xmit_DLCI", "_xmit_ATM", "_xmit_METRICOM", | |
279 | "_xmit_IEEE1394", "_xmit_EUI64", "_xmit_INFINIBAND", "_xmit_SLIP", | |
280 | "_xmit_CSLIP", "_xmit_SLIP6", "_xmit_CSLIP6", "_xmit_RSRVD", | |
281 | "_xmit_ADAPT", "_xmit_ROSE", "_xmit_X25", "_xmit_HWX25", | |
282 | "_xmit_PPP", "_xmit_CISCO", "_xmit_LAPB", "_xmit_DDCMP", | |
283 | "_xmit_RAWHDLC", "_xmit_TUNNEL", "_xmit_TUNNEL6", "_xmit_FRAD", | |
284 | "_xmit_SKIP", "_xmit_LOOPBACK", "_xmit_LOCALTLK", "_xmit_FDDI", | |
285 | "_xmit_BIF", "_xmit_SIT", "_xmit_IPDDP", "_xmit_IPGRE", | |
286 | "_xmit_PIMREG", "_xmit_HIPPI", "_xmit_ASH", "_xmit_ECONET", | |
287 | "_xmit_IRDA", "_xmit_FCPP", "_xmit_FCAL", "_xmit_FCPL", | |
288 | "_xmit_FCFABRIC", "_xmit_IEEE802_TR", "_xmit_IEEE80211", | |
2d91d78b | 289 | "_xmit_IEEE80211_PRISM", "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", |
929122cd | 290 | "_xmit_PHONET_PIPE", "_xmit_IEEE802154", |
fcb94e42 | 291 | "_xmit_VOID", "_xmit_NONE"}; |
723e98b7 JP |
292 | |
293 | static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)]; | |
cf508b12 | 294 | static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)]; |
723e98b7 JP |
295 | |
296 | static inline unsigned short netdev_lock_pos(unsigned short dev_type) | |
297 | { | |
298 | int i; | |
299 | ||
300 | for (i = 0; i < ARRAY_SIZE(netdev_lock_type); i++) | |
301 | if (netdev_lock_type[i] == dev_type) | |
302 | return i; | |
303 | /* the last key is used by default */ | |
304 | return ARRAY_SIZE(netdev_lock_type) - 1; | |
305 | } | |
306 | ||
cf508b12 DM |
307 | static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock, |
308 | unsigned short dev_type) | |
723e98b7 JP |
309 | { |
310 | int i; | |
311 | ||
312 | i = netdev_lock_pos(dev_type); | |
313 | lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i], | |
314 | netdev_lock_name[i]); | |
315 | } | |
cf508b12 DM |
316 | |
317 | static inline void netdev_set_addr_lockdep_class(struct net_device *dev) | |
318 | { | |
319 | int i; | |
320 | ||
321 | i = netdev_lock_pos(dev->type); | |
322 | lockdep_set_class_and_name(&dev->addr_list_lock, | |
323 | &netdev_addr_lock_key[i], | |
324 | netdev_lock_name[i]); | |
325 | } | |
723e98b7 | 326 | #else |
cf508b12 DM |
327 | static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock, |
328 | unsigned short dev_type) | |
329 | { | |
330 | } | |
331 | static inline void netdev_set_addr_lockdep_class(struct net_device *dev) | |
723e98b7 JP |
332 | { |
333 | } | |
334 | #endif | |
1da177e4 LT |
335 | |
336 | /******************************************************************************* | |
337 | ||
338 | Protocol management and registration routines | |
339 | ||
340 | *******************************************************************************/ | |
341 | ||
1da177e4 LT |
342 | /* |
343 | * Add a protocol ID to the list. Now that the input handler is | |
344 | * smarter we can dispense with all the messy stuff that used to be | |
345 | * here. | |
346 | * | |
347 | * BEWARE!!! Protocol handlers, mangling input packets, | |
348 | * MUST BE last in hash buckets and checking protocol handlers | |
349 | * MUST start from promiscuous ptype_all chain in net_bh. | |
350 | * It is true now, do not change it. | |
351 | * Explanation follows: if protocol handler, mangling packet, will | |
352 | * be the first on list, it is not able to sense, that packet | |
353 | * is cloned and should be copied-on-write, so that it will | |
354 | * change it and subsequent readers will get broken packet. | |
355 | * --ANK (980803) | |
356 | */ | |
357 | ||
358 | /** | |
359 | * dev_add_pack - add packet handler | |
360 | * @pt: packet type declaration | |
361 | * | |
362 | * Add a protocol handler to the networking stack. The passed &packet_type | |
363 | * is linked into kernel lists and may not be freed until it has been | |
364 | * removed from the kernel lists. | |
365 | * | |
4ec93edb | 366 | * This call does not sleep therefore it can not |
1da177e4 LT |
367 | * guarantee all CPU's that are in middle of receiving packets |
368 | * will see the new packet type (until the next received packet). | |
369 | */ | |
370 | ||
371 | void dev_add_pack(struct packet_type *pt) | |
372 | { | |
373 | int hash; | |
374 | ||
375 | spin_lock_bh(&ptype_lock); | |
9be9a6b9 | 376 | if (pt->type == htons(ETH_P_ALL)) |
1da177e4 | 377 | list_add_rcu(&pt->list, &ptype_all); |
9be9a6b9 | 378 | else { |
82d8a867 | 379 | hash = ntohs(pt->type) & PTYPE_HASH_MASK; |
1da177e4 LT |
380 | list_add_rcu(&pt->list, &ptype_base[hash]); |
381 | } | |
382 | spin_unlock_bh(&ptype_lock); | |
383 | } | |
d1b19dff | 384 | EXPORT_SYMBOL(dev_add_pack); |
1da177e4 | 385 | |
1da177e4 LT |
386 | /** |
387 | * __dev_remove_pack - remove packet handler | |
388 | * @pt: packet type declaration | |
389 | * | |
390 | * Remove a protocol handler that was previously added to the kernel | |
391 | * protocol handlers by dev_add_pack(). The passed &packet_type is removed | |
392 | * from the kernel lists and can be freed or reused once this function | |
4ec93edb | 393 | * returns. |
1da177e4 LT |
394 | * |
395 | * The packet type might still be in use by receivers | |
396 | * and must not be freed until after all the CPU's have gone | |
397 | * through a quiescent state. | |
398 | */ | |
399 | void __dev_remove_pack(struct packet_type *pt) | |
400 | { | |
401 | struct list_head *head; | |
402 | struct packet_type *pt1; | |
403 | ||
404 | spin_lock_bh(&ptype_lock); | |
405 | ||
9be9a6b9 | 406 | if (pt->type == htons(ETH_P_ALL)) |
1da177e4 | 407 | head = &ptype_all; |
9be9a6b9 | 408 | else |
82d8a867 | 409 | head = &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK]; |
1da177e4 LT |
410 | |
411 | list_for_each_entry(pt1, head, list) { | |
412 | if (pt == pt1) { | |
413 | list_del_rcu(&pt->list); | |
414 | goto out; | |
415 | } | |
416 | } | |
417 | ||
418 | printk(KERN_WARNING "dev_remove_pack: %p not found.\n", pt); | |
419 | out: | |
420 | spin_unlock_bh(&ptype_lock); | |
421 | } | |
d1b19dff ED |
422 | EXPORT_SYMBOL(__dev_remove_pack); |
423 | ||
1da177e4 LT |
424 | /** |
425 | * dev_remove_pack - remove packet handler | |
426 | * @pt: packet type declaration | |
427 | * | |
428 | * Remove a protocol handler that was previously added to the kernel | |
429 | * protocol handlers by dev_add_pack(). The passed &packet_type is removed | |
430 | * from the kernel lists and can be freed or reused once this function | |
431 | * returns. | |
432 | * | |
433 | * This call sleeps to guarantee that no CPU is looking at the packet | |
434 | * type after return. | |
435 | */ | |
436 | void dev_remove_pack(struct packet_type *pt) | |
437 | { | |
438 | __dev_remove_pack(pt); | |
4ec93edb | 439 | |
1da177e4 LT |
440 | synchronize_net(); |
441 | } | |
d1b19dff | 442 | EXPORT_SYMBOL(dev_remove_pack); |
1da177e4 LT |
443 | |
444 | /****************************************************************************** | |
445 | ||
446 | Device Boot-time Settings Routines | |
447 | ||
448 | *******************************************************************************/ | |
449 | ||
450 | /* Boot time configuration table */ | |
451 | static struct netdev_boot_setup dev_boot_setup[NETDEV_BOOT_SETUP_MAX]; | |
452 | ||
453 | /** | |
454 | * netdev_boot_setup_add - add new setup entry | |
455 | * @name: name of the device | |
456 | * @map: configured settings for the device | |
457 | * | |
458 | * Adds new setup entry to the dev_boot_setup list. The function | |
459 | * returns 0 on error and 1 on success. This is a generic routine to | |
460 | * all netdevices. | |
461 | */ | |
462 | static int netdev_boot_setup_add(char *name, struct ifmap *map) | |
463 | { | |
464 | struct netdev_boot_setup *s; | |
465 | int i; | |
466 | ||
467 | s = dev_boot_setup; | |
468 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { | |
469 | if (s[i].name[0] == '\0' || s[i].name[0] == ' ') { | |
470 | memset(s[i].name, 0, sizeof(s[i].name)); | |
93b3cff9 | 471 | strlcpy(s[i].name, name, IFNAMSIZ); |
1da177e4 LT |
472 | memcpy(&s[i].map, map, sizeof(s[i].map)); |
473 | break; | |
474 | } | |
475 | } | |
476 | ||
477 | return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1; | |
478 | } | |
479 | ||
480 | /** | |
481 | * netdev_boot_setup_check - check boot time settings | |
482 | * @dev: the netdevice | |
483 | * | |
484 | * Check boot time settings for the device. | |
485 | * The found settings are set for the device to be used | |
486 | * later in the device probing. | |
487 | * Returns 0 if no settings found, 1 if they are. | |
488 | */ | |
489 | int netdev_boot_setup_check(struct net_device *dev) | |
490 | { | |
491 | struct netdev_boot_setup *s = dev_boot_setup; | |
492 | int i; | |
493 | ||
494 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { | |
495 | if (s[i].name[0] != '\0' && s[i].name[0] != ' ' && | |
93b3cff9 | 496 | !strcmp(dev->name, s[i].name)) { |
1da177e4 LT |
497 | dev->irq = s[i].map.irq; |
498 | dev->base_addr = s[i].map.base_addr; | |
499 | dev->mem_start = s[i].map.mem_start; | |
500 | dev->mem_end = s[i].map.mem_end; | |
501 | return 1; | |
502 | } | |
503 | } | |
504 | return 0; | |
505 | } | |
d1b19dff | 506 | EXPORT_SYMBOL(netdev_boot_setup_check); |
1da177e4 LT |
507 | |
508 | ||
509 | /** | |
510 | * netdev_boot_base - get address from boot time settings | |
511 | * @prefix: prefix for network device | |
512 | * @unit: id for network device | |
513 | * | |
514 | * Check boot time settings for the base address of device. | |
515 | * The found settings are set for the device to be used | |
516 | * later in the device probing. | |
517 | * Returns 0 if no settings found. | |
518 | */ | |
519 | unsigned long netdev_boot_base(const char *prefix, int unit) | |
520 | { | |
521 | const struct netdev_boot_setup *s = dev_boot_setup; | |
522 | char name[IFNAMSIZ]; | |
523 | int i; | |
524 | ||
525 | sprintf(name, "%s%d", prefix, unit); | |
526 | ||
527 | /* | |
528 | * If device already registered then return base of 1 | |
529 | * to indicate not to probe for this interface | |
530 | */ | |
881d966b | 531 | if (__dev_get_by_name(&init_net, name)) |
1da177e4 LT |
532 | return 1; |
533 | ||
534 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) | |
535 | if (!strcmp(name, s[i].name)) | |
536 | return s[i].map.base_addr; | |
537 | return 0; | |
538 | } | |
539 | ||
540 | /* | |
541 | * Saves at boot time configured settings for any netdevice. | |
542 | */ | |
543 | int __init netdev_boot_setup(char *str) | |
544 | { | |
545 | int ints[5]; | |
546 | struct ifmap map; | |
547 | ||
548 | str = get_options(str, ARRAY_SIZE(ints), ints); | |
549 | if (!str || !*str) | |
550 | return 0; | |
551 | ||
552 | /* Save settings */ | |
553 | memset(&map, 0, sizeof(map)); | |
554 | if (ints[0] > 0) | |
555 | map.irq = ints[1]; | |
556 | if (ints[0] > 1) | |
557 | map.base_addr = ints[2]; | |
558 | if (ints[0] > 2) | |
559 | map.mem_start = ints[3]; | |
560 | if (ints[0] > 3) | |
561 | map.mem_end = ints[4]; | |
562 | ||
563 | /* Add new entry to the list */ | |
564 | return netdev_boot_setup_add(str, &map); | |
565 | } | |
566 | ||
567 | __setup("netdev=", netdev_boot_setup); | |
568 | ||
569 | /******************************************************************************* | |
570 | ||
571 | Device Interface Subroutines | |
572 | ||
573 | *******************************************************************************/ | |
574 | ||
575 | /** | |
576 | * __dev_get_by_name - find a device by its name | |
c4ea43c5 | 577 | * @net: the applicable net namespace |
1da177e4 LT |
578 | * @name: name to find |
579 | * | |
580 | * Find an interface by name. Must be called under RTNL semaphore | |
581 | * or @dev_base_lock. If the name is found a pointer to the device | |
582 | * is returned. If the name is not found then %NULL is returned. The | |
583 | * reference counters are not incremented so the caller must be | |
584 | * careful with locks. | |
585 | */ | |
586 | ||
881d966b | 587 | struct net_device *__dev_get_by_name(struct net *net, const char *name) |
1da177e4 LT |
588 | { |
589 | struct hlist_node *p; | |
590 | ||
881d966b | 591 | hlist_for_each(p, dev_name_hash(net, name)) { |
1da177e4 LT |
592 | struct net_device *dev |
593 | = hlist_entry(p, struct net_device, name_hlist); | |
594 | if (!strncmp(dev->name, name, IFNAMSIZ)) | |
595 | return dev; | |
596 | } | |
597 | return NULL; | |
598 | } | |
d1b19dff | 599 | EXPORT_SYMBOL(__dev_get_by_name); |
1da177e4 LT |
600 | |
601 | /** | |
602 | * dev_get_by_name - find a device by its name | |
c4ea43c5 | 603 | * @net: the applicable net namespace |
1da177e4 LT |
604 | * @name: name to find |
605 | * | |
606 | * Find an interface by name. This can be called from any | |
607 | * context and does its own locking. The returned handle has | |
608 | * the usage count incremented and the caller must use dev_put() to | |
609 | * release it when it is no longer needed. %NULL is returned if no | |
610 | * matching device is found. | |
611 | */ | |
612 | ||
881d966b | 613 | struct net_device *dev_get_by_name(struct net *net, const char *name) |
1da177e4 LT |
614 | { |
615 | struct net_device *dev; | |
616 | ||
617 | read_lock(&dev_base_lock); | |
881d966b | 618 | dev = __dev_get_by_name(net, name); |
1da177e4 LT |
619 | if (dev) |
620 | dev_hold(dev); | |
621 | read_unlock(&dev_base_lock); | |
622 | return dev; | |
623 | } | |
d1b19dff | 624 | EXPORT_SYMBOL(dev_get_by_name); |
1da177e4 LT |
625 | |
626 | /** | |
627 | * __dev_get_by_index - find a device by its ifindex | |
c4ea43c5 | 628 | * @net: the applicable net namespace |
1da177e4 LT |
629 | * @ifindex: index of device |
630 | * | |
631 | * Search for an interface by index. Returns %NULL if the device | |
632 | * is not found or a pointer to the device. The device has not | |
633 | * had its reference counter increased so the caller must be careful | |
634 | * about locking. The caller must hold either the RTNL semaphore | |
635 | * or @dev_base_lock. | |
636 | */ | |
637 | ||
881d966b | 638 | struct net_device *__dev_get_by_index(struct net *net, int ifindex) |
1da177e4 LT |
639 | { |
640 | struct hlist_node *p; | |
641 | ||
881d966b | 642 | hlist_for_each(p, dev_index_hash(net, ifindex)) { |
1da177e4 LT |
643 | struct net_device *dev |
644 | = hlist_entry(p, struct net_device, index_hlist); | |
645 | if (dev->ifindex == ifindex) | |
646 | return dev; | |
647 | } | |
648 | return NULL; | |
649 | } | |
d1b19dff | 650 | EXPORT_SYMBOL(__dev_get_by_index); |
1da177e4 LT |
651 | |
652 | ||
653 | /** | |
654 | * dev_get_by_index - find a device by its ifindex | |
c4ea43c5 | 655 | * @net: the applicable net namespace |
1da177e4 LT |
656 | * @ifindex: index of device |
657 | * | |
658 | * Search for an interface by index. Returns NULL if the device | |
659 | * is not found or a pointer to the device. The device returned has | |
660 | * had a reference added and the pointer is safe until the user calls | |
661 | * dev_put to indicate they have finished with it. | |
662 | */ | |
663 | ||
881d966b | 664 | struct net_device *dev_get_by_index(struct net *net, int ifindex) |
1da177e4 LT |
665 | { |
666 | struct net_device *dev; | |
667 | ||
668 | read_lock(&dev_base_lock); | |
881d966b | 669 | dev = __dev_get_by_index(net, ifindex); |
1da177e4 LT |
670 | if (dev) |
671 | dev_hold(dev); | |
672 | read_unlock(&dev_base_lock); | |
673 | return dev; | |
674 | } | |
d1b19dff | 675 | EXPORT_SYMBOL(dev_get_by_index); |
1da177e4 LT |
676 | |
677 | /** | |
678 | * dev_getbyhwaddr - find a device by its hardware address | |
c4ea43c5 | 679 | * @net: the applicable net namespace |
1da177e4 LT |
680 | * @type: media type of device |
681 | * @ha: hardware address | |
682 | * | |
683 | * Search for an interface by MAC address. Returns NULL if the device | |
684 | * is not found or a pointer to the device. The caller must hold the | |
685 | * rtnl semaphore. The returned device has not had its ref count increased | |
686 | * and the caller must therefore be careful about locking | |
687 | * | |
688 | * BUGS: | |
689 | * If the API was consistent this would be __dev_get_by_hwaddr | |
690 | */ | |
691 | ||
881d966b | 692 | struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, char *ha) |
1da177e4 LT |
693 | { |
694 | struct net_device *dev; | |
695 | ||
696 | ASSERT_RTNL(); | |
697 | ||
81103a52 | 698 | for_each_netdev(net, dev) |
1da177e4 LT |
699 | if (dev->type == type && |
700 | !memcmp(dev->dev_addr, ha, dev->addr_len)) | |
7562f876 PE |
701 | return dev; |
702 | ||
703 | return NULL; | |
1da177e4 | 704 | } |
cf309e3f JF |
705 | EXPORT_SYMBOL(dev_getbyhwaddr); |
706 | ||
881d966b | 707 | struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type) |
1da177e4 LT |
708 | { |
709 | struct net_device *dev; | |
710 | ||
4e9cac2b | 711 | ASSERT_RTNL(); |
881d966b | 712 | for_each_netdev(net, dev) |
4e9cac2b | 713 | if (dev->type == type) |
7562f876 PE |
714 | return dev; |
715 | ||
716 | return NULL; | |
4e9cac2b | 717 | } |
4e9cac2b PM |
718 | EXPORT_SYMBOL(__dev_getfirstbyhwtype); |
719 | ||
881d966b | 720 | struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type) |
4e9cac2b PM |
721 | { |
722 | struct net_device *dev; | |
723 | ||
724 | rtnl_lock(); | |
881d966b | 725 | dev = __dev_getfirstbyhwtype(net, type); |
4e9cac2b PM |
726 | if (dev) |
727 | dev_hold(dev); | |
1da177e4 LT |
728 | rtnl_unlock(); |
729 | return dev; | |
730 | } | |
1da177e4 LT |
731 | EXPORT_SYMBOL(dev_getfirstbyhwtype); |
732 | ||
733 | /** | |
734 | * dev_get_by_flags - find any device with given flags | |
c4ea43c5 | 735 | * @net: the applicable net namespace |
1da177e4 LT |
736 | * @if_flags: IFF_* values |
737 | * @mask: bitmask of bits in if_flags to check | |
738 | * | |
739 | * Search for any interface with the given flags. Returns NULL if a device | |
4ec93edb | 740 | * is not found or a pointer to the device. The device returned has |
1da177e4 LT |
741 | * had a reference added and the pointer is safe until the user calls |
742 | * dev_put to indicate they have finished with it. | |
743 | */ | |
744 | ||
d1b19dff ED |
745 | struct net_device *dev_get_by_flags(struct net *net, unsigned short if_flags, |
746 | unsigned short mask) | |
1da177e4 | 747 | { |
7562f876 | 748 | struct net_device *dev, *ret; |
1da177e4 | 749 | |
7562f876 | 750 | ret = NULL; |
1da177e4 | 751 | read_lock(&dev_base_lock); |
881d966b | 752 | for_each_netdev(net, dev) { |
1da177e4 LT |
753 | if (((dev->flags ^ if_flags) & mask) == 0) { |
754 | dev_hold(dev); | |
7562f876 | 755 | ret = dev; |
1da177e4 LT |
756 | break; |
757 | } | |
758 | } | |
759 | read_unlock(&dev_base_lock); | |
7562f876 | 760 | return ret; |
1da177e4 | 761 | } |
d1b19dff | 762 | EXPORT_SYMBOL(dev_get_by_flags); |
1da177e4 LT |
763 | |
764 | /** | |
765 | * dev_valid_name - check if name is okay for network device | |
766 | * @name: name string | |
767 | * | |
768 | * Network device names need to be valid file names to | |
c7fa9d18 DM |
769 | * to allow sysfs to work. We also disallow any kind of |
770 | * whitespace. | |
1da177e4 | 771 | */ |
c2373ee9 | 772 | int dev_valid_name(const char *name) |
1da177e4 | 773 | { |
c7fa9d18 DM |
774 | if (*name == '\0') |
775 | return 0; | |
b6fe17d6 SH |
776 | if (strlen(name) >= IFNAMSIZ) |
777 | return 0; | |
c7fa9d18 DM |
778 | if (!strcmp(name, ".") || !strcmp(name, "..")) |
779 | return 0; | |
780 | ||
781 | while (*name) { | |
782 | if (*name == '/' || isspace(*name)) | |
783 | return 0; | |
784 | name++; | |
785 | } | |
786 | return 1; | |
1da177e4 | 787 | } |
d1b19dff | 788 | EXPORT_SYMBOL(dev_valid_name); |
1da177e4 LT |
789 | |
790 | /** | |
b267b179 EB |
791 | * __dev_alloc_name - allocate a name for a device |
792 | * @net: network namespace to allocate the device name in | |
1da177e4 | 793 | * @name: name format string |
b267b179 | 794 | * @buf: scratch buffer and result name string |
1da177e4 LT |
795 | * |
796 | * Passed a format string - eg "lt%d" it will try and find a suitable | |
3041a069 SH |
797 | * id. It scans list of devices to build up a free map, then chooses |
798 | * the first empty slot. The caller must hold the dev_base or rtnl lock | |
799 | * while allocating the name and adding the device in order to avoid | |
800 | * duplicates. | |
801 | * Limited to bits_per_byte * page size devices (ie 32K on most platforms). | |
802 | * Returns the number of the unit assigned or a negative errno code. | |
1da177e4 LT |
803 | */ |
804 | ||
b267b179 | 805 | static int __dev_alloc_name(struct net *net, const char *name, char *buf) |
1da177e4 LT |
806 | { |
807 | int i = 0; | |
1da177e4 LT |
808 | const char *p; |
809 | const int max_netdevices = 8*PAGE_SIZE; | |
cfcabdcc | 810 | unsigned long *inuse; |
1da177e4 LT |
811 | struct net_device *d; |
812 | ||
813 | p = strnchr(name, IFNAMSIZ-1, '%'); | |
814 | if (p) { | |
815 | /* | |
816 | * Verify the string as this thing may have come from | |
817 | * the user. There must be either one "%d" and no other "%" | |
818 | * characters. | |
819 | */ | |
820 | if (p[1] != 'd' || strchr(p + 2, '%')) | |
821 | return -EINVAL; | |
822 | ||
823 | /* Use one page as a bit array of possible slots */ | |
cfcabdcc | 824 | inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC); |
1da177e4 LT |
825 | if (!inuse) |
826 | return -ENOMEM; | |
827 | ||
881d966b | 828 | for_each_netdev(net, d) { |
1da177e4 LT |
829 | if (!sscanf(d->name, name, &i)) |
830 | continue; | |
831 | if (i < 0 || i >= max_netdevices) | |
832 | continue; | |
833 | ||
834 | /* avoid cases where sscanf is not exact inverse of printf */ | |
b267b179 | 835 | snprintf(buf, IFNAMSIZ, name, i); |
1da177e4 LT |
836 | if (!strncmp(buf, d->name, IFNAMSIZ)) |
837 | set_bit(i, inuse); | |
838 | } | |
839 | ||
840 | i = find_first_zero_bit(inuse, max_netdevices); | |
841 | free_page((unsigned long) inuse); | |
842 | } | |
843 | ||
b267b179 EB |
844 | snprintf(buf, IFNAMSIZ, name, i); |
845 | if (!__dev_get_by_name(net, buf)) | |
1da177e4 | 846 | return i; |
1da177e4 LT |
847 | |
848 | /* It is possible to run out of possible slots | |
849 | * when the name is long and there isn't enough space left | |
850 | * for the digits, or if all bits are used. | |
851 | */ | |
852 | return -ENFILE; | |
853 | } | |
854 | ||
b267b179 EB |
855 | /** |
856 | * dev_alloc_name - allocate a name for a device | |
857 | * @dev: device | |
858 | * @name: name format string | |
859 | * | |
860 | * Passed a format string - eg "lt%d" it will try and find a suitable | |
861 | * id. It scans list of devices to build up a free map, then chooses | |
862 | * the first empty slot. The caller must hold the dev_base or rtnl lock | |
863 | * while allocating the name and adding the device in order to avoid | |
864 | * duplicates. | |
865 | * Limited to bits_per_byte * page size devices (ie 32K on most platforms). | |
866 | * Returns the number of the unit assigned or a negative errno code. | |
867 | */ | |
868 | ||
869 | int dev_alloc_name(struct net_device *dev, const char *name) | |
870 | { | |
871 | char buf[IFNAMSIZ]; | |
872 | struct net *net; | |
873 | int ret; | |
874 | ||
c346dca1 YH |
875 | BUG_ON(!dev_net(dev)); |
876 | net = dev_net(dev); | |
b267b179 EB |
877 | ret = __dev_alloc_name(net, name, buf); |
878 | if (ret >= 0) | |
879 | strlcpy(dev->name, buf, IFNAMSIZ); | |
880 | return ret; | |
881 | } | |
d1b19dff | 882 | EXPORT_SYMBOL(dev_alloc_name); |
b267b179 | 883 | |
1da177e4 LT |
884 | |
885 | /** | |
886 | * dev_change_name - change name of a device | |
887 | * @dev: device | |
888 | * @newname: name (or format string) must be at least IFNAMSIZ | |
889 | * | |
890 | * Change name of a device, can pass format strings "eth%d". | |
891 | * for wildcarding. | |
892 | */ | |
cf04a4c7 | 893 | int dev_change_name(struct net_device *dev, const char *newname) |
1da177e4 | 894 | { |
fcc5a03a | 895 | char oldname[IFNAMSIZ]; |
1da177e4 | 896 | int err = 0; |
fcc5a03a | 897 | int ret; |
881d966b | 898 | struct net *net; |
1da177e4 LT |
899 | |
900 | ASSERT_RTNL(); | |
c346dca1 | 901 | BUG_ON(!dev_net(dev)); |
1da177e4 | 902 | |
c346dca1 | 903 | net = dev_net(dev); |
1da177e4 LT |
904 | if (dev->flags & IFF_UP) |
905 | return -EBUSY; | |
906 | ||
907 | if (!dev_valid_name(newname)) | |
908 | return -EINVAL; | |
909 | ||
c8d90dca SH |
910 | if (strncmp(newname, dev->name, IFNAMSIZ) == 0) |
911 | return 0; | |
912 | ||
fcc5a03a HX |
913 | memcpy(oldname, dev->name, IFNAMSIZ); |
914 | ||
1da177e4 LT |
915 | if (strchr(newname, '%')) { |
916 | err = dev_alloc_name(dev, newname); | |
917 | if (err < 0) | |
918 | return err; | |
d1b19dff | 919 | } else if (__dev_get_by_name(net, newname)) |
1da177e4 LT |
920 | return -EEXIST; |
921 | else | |
922 | strlcpy(dev->name, newname, IFNAMSIZ); | |
923 | ||
fcc5a03a | 924 | rollback: |
3891845e EB |
925 | /* For now only devices in the initial network namespace |
926 | * are in sysfs. | |
927 | */ | |
928 | if (net == &init_net) { | |
929 | ret = device_rename(&dev->dev, dev->name); | |
930 | if (ret) { | |
931 | memcpy(dev->name, oldname, IFNAMSIZ); | |
932 | return ret; | |
933 | } | |
dcc99773 | 934 | } |
7f988eab HX |
935 | |
936 | write_lock_bh(&dev_base_lock); | |
92749821 | 937 | hlist_del(&dev->name_hlist); |
881d966b | 938 | hlist_add_head(&dev->name_hlist, dev_name_hash(net, dev->name)); |
7f988eab HX |
939 | write_unlock_bh(&dev_base_lock); |
940 | ||
056925ab | 941 | ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev); |
fcc5a03a HX |
942 | ret = notifier_to_errno(ret); |
943 | ||
944 | if (ret) { | |
945 | if (err) { | |
946 | printk(KERN_ERR | |
947 | "%s: name change rollback failed: %d.\n", | |
948 | dev->name, ret); | |
949 | } else { | |
950 | err = ret; | |
951 | memcpy(dev->name, oldname, IFNAMSIZ); | |
952 | goto rollback; | |
953 | } | |
954 | } | |
1da177e4 LT |
955 | |
956 | return err; | |
957 | } | |
958 | ||
0b815a1a SH |
959 | /** |
960 | * dev_set_alias - change ifalias of a device | |
961 | * @dev: device | |
962 | * @alias: name up to IFALIASZ | |
f0db275a | 963 | * @len: limit of bytes to copy from info |
0b815a1a SH |
964 | * |
965 | * Set ifalias for a device, | |
966 | */ | |
967 | int dev_set_alias(struct net_device *dev, const char *alias, size_t len) | |
968 | { | |
969 | ASSERT_RTNL(); | |
970 | ||
971 | if (len >= IFALIASZ) | |
972 | return -EINVAL; | |
973 | ||
96ca4a2c OH |
974 | if (!len) { |
975 | if (dev->ifalias) { | |
976 | kfree(dev->ifalias); | |
977 | dev->ifalias = NULL; | |
978 | } | |
979 | return 0; | |
980 | } | |
981 | ||
d1b19dff | 982 | dev->ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL); |
0b815a1a SH |
983 | if (!dev->ifalias) |
984 | return -ENOMEM; | |
985 | ||
986 | strlcpy(dev->ifalias, alias, len+1); | |
987 | return len; | |
988 | } | |
989 | ||
990 | ||
d8a33ac4 | 991 | /** |
3041a069 | 992 | * netdev_features_change - device changes features |
d8a33ac4 SH |
993 | * @dev: device to cause notification |
994 | * | |
995 | * Called to indicate a device has changed features. | |
996 | */ | |
997 | void netdev_features_change(struct net_device *dev) | |
998 | { | |
056925ab | 999 | call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev); |
d8a33ac4 SH |
1000 | } |
1001 | EXPORT_SYMBOL(netdev_features_change); | |
1002 | ||
1da177e4 LT |
1003 | /** |
1004 | * netdev_state_change - device changes state | |
1005 | * @dev: device to cause notification | |
1006 | * | |
1007 | * Called to indicate a device has changed state. This function calls | |
1008 | * the notifier chains for netdev_chain and sends a NEWLINK message | |
1009 | * to the routing socket. | |
1010 | */ | |
1011 | void netdev_state_change(struct net_device *dev) | |
1012 | { | |
1013 | if (dev->flags & IFF_UP) { | |
056925ab | 1014 | call_netdevice_notifiers(NETDEV_CHANGE, dev); |
1da177e4 LT |
1015 | rtmsg_ifinfo(RTM_NEWLINK, dev, 0); |
1016 | } | |
1017 | } | |
d1b19dff | 1018 | EXPORT_SYMBOL(netdev_state_change); |
1da177e4 | 1019 | |
c1da4ac7 OG |
1020 | void netdev_bonding_change(struct net_device *dev) |
1021 | { | |
1022 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, dev); | |
1023 | } | |
1024 | EXPORT_SYMBOL(netdev_bonding_change); | |
1025 | ||
1da177e4 LT |
1026 | /** |
1027 | * dev_load - load a network module | |
c4ea43c5 | 1028 | * @net: the applicable net namespace |
1da177e4 LT |
1029 | * @name: name of interface |
1030 | * | |
1031 | * If a network interface is not present and the process has suitable | |
1032 | * privileges this function loads the module. If module loading is not | |
1033 | * available in this kernel then it becomes a nop. | |
1034 | */ | |
1035 | ||
881d966b | 1036 | void dev_load(struct net *net, const char *name) |
1da177e4 | 1037 | { |
4ec93edb | 1038 | struct net_device *dev; |
1da177e4 LT |
1039 | |
1040 | read_lock(&dev_base_lock); | |
881d966b | 1041 | dev = __dev_get_by_name(net, name); |
1da177e4 LT |
1042 | read_unlock(&dev_base_lock); |
1043 | ||
a8f80e8f | 1044 | if (!dev && capable(CAP_NET_ADMIN)) |
1da177e4 LT |
1045 | request_module("%s", name); |
1046 | } | |
d1b19dff | 1047 | EXPORT_SYMBOL(dev_load); |
1da177e4 | 1048 | |
1da177e4 LT |
1049 | /** |
1050 | * dev_open - prepare an interface for use. | |
1051 | * @dev: device to open | |
1052 | * | |
1053 | * Takes a device from down to up state. The device's private open | |
1054 | * function is invoked and then the multicast lists are loaded. Finally | |
1055 | * the device is moved into the up state and a %NETDEV_UP message is | |
1056 | * sent to the netdev notifier chain. | |
1057 | * | |
1058 | * Calling this function on an active interface is a nop. On a failure | |
1059 | * a negative errno code is returned. | |
1060 | */ | |
1061 | int dev_open(struct net_device *dev) | |
1062 | { | |
d314774c | 1063 | const struct net_device_ops *ops = dev->netdev_ops; |
3b8bcfd5 | 1064 | int ret; |
1da177e4 | 1065 | |
e46b66bc BH |
1066 | ASSERT_RTNL(); |
1067 | ||
1da177e4 LT |
1068 | /* |
1069 | * Is it already up? | |
1070 | */ | |
1071 | ||
1072 | if (dev->flags & IFF_UP) | |
1073 | return 0; | |
1074 | ||
1075 | /* | |
1076 | * Is it even present? | |
1077 | */ | |
1078 | if (!netif_device_present(dev)) | |
1079 | return -ENODEV; | |
1080 | ||
3b8bcfd5 JB |
1081 | ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev); |
1082 | ret = notifier_to_errno(ret); | |
1083 | if (ret) | |
1084 | return ret; | |
1085 | ||
1da177e4 LT |
1086 | /* |
1087 | * Call device private open method | |
1088 | */ | |
1089 | set_bit(__LINK_STATE_START, &dev->state); | |
bada339b | 1090 | |
d314774c SH |
1091 | if (ops->ndo_validate_addr) |
1092 | ret = ops->ndo_validate_addr(dev); | |
bada339b | 1093 | |
d314774c SH |
1094 | if (!ret && ops->ndo_open) |
1095 | ret = ops->ndo_open(dev); | |
1da177e4 | 1096 | |
4ec93edb | 1097 | /* |
1da177e4 LT |
1098 | * If it went open OK then: |
1099 | */ | |
1100 | ||
bada339b JG |
1101 | if (ret) |
1102 | clear_bit(__LINK_STATE_START, &dev->state); | |
1103 | else { | |
1da177e4 LT |
1104 | /* |
1105 | * Set the flags. | |
1106 | */ | |
1107 | dev->flags |= IFF_UP; | |
1108 | ||
649274d9 DW |
1109 | /* |
1110 | * Enable NET_DMA | |
1111 | */ | |
b4bd07c2 | 1112 | net_dmaengine_get(); |
649274d9 | 1113 | |
1da177e4 LT |
1114 | /* |
1115 | * Initialize multicasting status | |
1116 | */ | |
4417da66 | 1117 | dev_set_rx_mode(dev); |
1da177e4 LT |
1118 | |
1119 | /* | |
1120 | * Wakeup transmit queue engine | |
1121 | */ | |
1122 | dev_activate(dev); | |
1123 | ||
1124 | /* | |
1125 | * ... and announce new interface. | |
1126 | */ | |
056925ab | 1127 | call_netdevice_notifiers(NETDEV_UP, dev); |
1da177e4 | 1128 | } |
bada339b | 1129 | |
1da177e4 LT |
1130 | return ret; |
1131 | } | |
d1b19dff | 1132 | EXPORT_SYMBOL(dev_open); |
1da177e4 LT |
1133 | |
1134 | /** | |
1135 | * dev_close - shutdown an interface. | |
1136 | * @dev: device to shutdown | |
1137 | * | |
1138 | * This function moves an active device into down state. A | |
1139 | * %NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device | |
1140 | * is then deactivated and finally a %NETDEV_DOWN is sent to the notifier | |
1141 | * chain. | |
1142 | */ | |
1143 | int dev_close(struct net_device *dev) | |
1144 | { | |
d314774c | 1145 | const struct net_device_ops *ops = dev->netdev_ops; |
e46b66bc BH |
1146 | ASSERT_RTNL(); |
1147 | ||
9d5010db DM |
1148 | might_sleep(); |
1149 | ||
1da177e4 LT |
1150 | if (!(dev->flags & IFF_UP)) |
1151 | return 0; | |
1152 | ||
1153 | /* | |
1154 | * Tell people we are going down, so that they can | |
1155 | * prepare to death, when device is still operating. | |
1156 | */ | |
056925ab | 1157 | call_netdevice_notifiers(NETDEV_GOING_DOWN, dev); |
1da177e4 | 1158 | |
1da177e4 LT |
1159 | clear_bit(__LINK_STATE_START, &dev->state); |
1160 | ||
1161 | /* Synchronize to scheduled poll. We cannot touch poll list, | |
bea3348e SH |
1162 | * it can be even on different cpu. So just clear netif_running(). |
1163 | * | |
1164 | * dev->stop() will invoke napi_disable() on all of it's | |
1165 | * napi_struct instances on this device. | |
1166 | */ | |
1da177e4 | 1167 | smp_mb__after_clear_bit(); /* Commit netif_running(). */ |
1da177e4 | 1168 | |
d8b2a4d2 ML |
1169 | dev_deactivate(dev); |
1170 | ||
1da177e4 LT |
1171 | /* |
1172 | * Call the device specific close. This cannot fail. | |
1173 | * Only if device is UP | |
1174 | * | |
1175 | * We allow it to be called even after a DETACH hot-plug | |
1176 | * event. | |
1177 | */ | |
d314774c SH |
1178 | if (ops->ndo_stop) |
1179 | ops->ndo_stop(dev); | |
1da177e4 LT |
1180 | |
1181 | /* | |
1182 | * Device is now down. | |
1183 | */ | |
1184 | ||
1185 | dev->flags &= ~IFF_UP; | |
1186 | ||
1187 | /* | |
1188 | * Tell people we are down | |
1189 | */ | |
056925ab | 1190 | call_netdevice_notifiers(NETDEV_DOWN, dev); |
1da177e4 | 1191 | |
649274d9 DW |
1192 | /* |
1193 | * Shutdown NET_DMA | |
1194 | */ | |
b4bd07c2 | 1195 | net_dmaengine_put(); |
649274d9 | 1196 | |
1da177e4 LT |
1197 | return 0; |
1198 | } | |
d1b19dff | 1199 | EXPORT_SYMBOL(dev_close); |
1da177e4 LT |
1200 | |
1201 | ||
0187bdfb BH |
1202 | /** |
1203 | * dev_disable_lro - disable Large Receive Offload on a device | |
1204 | * @dev: device | |
1205 | * | |
1206 | * Disable Large Receive Offload (LRO) on a net device. Must be | |
1207 | * called under RTNL. This is needed if received packets may be | |
1208 | * forwarded to another interface. | |
1209 | */ | |
1210 | void dev_disable_lro(struct net_device *dev) | |
1211 | { | |
1212 | if (dev->ethtool_ops && dev->ethtool_ops->get_flags && | |
1213 | dev->ethtool_ops->set_flags) { | |
1214 | u32 flags = dev->ethtool_ops->get_flags(dev); | |
1215 | if (flags & ETH_FLAG_LRO) { | |
1216 | flags &= ~ETH_FLAG_LRO; | |
1217 | dev->ethtool_ops->set_flags(dev, flags); | |
1218 | } | |
1219 | } | |
1220 | WARN_ON(dev->features & NETIF_F_LRO); | |
1221 | } | |
1222 | EXPORT_SYMBOL(dev_disable_lro); | |
1223 | ||
1224 | ||
881d966b EB |
1225 | static int dev_boot_phase = 1; |
1226 | ||
1da177e4 LT |
1227 | /* |
1228 | * Device change register/unregister. These are not inline or static | |
1229 | * as we export them to the world. | |
1230 | */ | |
1231 | ||
1232 | /** | |
1233 | * register_netdevice_notifier - register a network notifier block | |
1234 | * @nb: notifier | |
1235 | * | |
1236 | * Register a notifier to be called when network device events occur. | |
1237 | * The notifier passed is linked into the kernel structures and must | |
1238 | * not be reused until it has been unregistered. A negative errno code | |
1239 | * is returned on a failure. | |
1240 | * | |
1241 | * When registered all registration and up events are replayed | |
4ec93edb | 1242 | * to the new notifier to allow device to have a race free |
1da177e4 LT |
1243 | * view of the network device list. |
1244 | */ | |
1245 | ||
1246 | int register_netdevice_notifier(struct notifier_block *nb) | |
1247 | { | |
1248 | struct net_device *dev; | |
fcc5a03a | 1249 | struct net_device *last; |
881d966b | 1250 | struct net *net; |
1da177e4 LT |
1251 | int err; |
1252 | ||
1253 | rtnl_lock(); | |
f07d5b94 | 1254 | err = raw_notifier_chain_register(&netdev_chain, nb); |
fcc5a03a HX |
1255 | if (err) |
1256 | goto unlock; | |
881d966b EB |
1257 | if (dev_boot_phase) |
1258 | goto unlock; | |
1259 | for_each_net(net) { | |
1260 | for_each_netdev(net, dev) { | |
1261 | err = nb->notifier_call(nb, NETDEV_REGISTER, dev); | |
1262 | err = notifier_to_errno(err); | |
1263 | if (err) | |
1264 | goto rollback; | |
1265 | ||
1266 | if (!(dev->flags & IFF_UP)) | |
1267 | continue; | |
1da177e4 | 1268 | |
881d966b EB |
1269 | nb->notifier_call(nb, NETDEV_UP, dev); |
1270 | } | |
1da177e4 | 1271 | } |
fcc5a03a HX |
1272 | |
1273 | unlock: | |
1da177e4 LT |
1274 | rtnl_unlock(); |
1275 | return err; | |
fcc5a03a HX |
1276 | |
1277 | rollback: | |
1278 | last = dev; | |
881d966b EB |
1279 | for_each_net(net) { |
1280 | for_each_netdev(net, dev) { | |
1281 | if (dev == last) | |
1282 | break; | |
fcc5a03a | 1283 | |
881d966b EB |
1284 | if (dev->flags & IFF_UP) { |
1285 | nb->notifier_call(nb, NETDEV_GOING_DOWN, dev); | |
1286 | nb->notifier_call(nb, NETDEV_DOWN, dev); | |
1287 | } | |
1288 | nb->notifier_call(nb, NETDEV_UNREGISTER, dev); | |
fcc5a03a | 1289 | } |
fcc5a03a | 1290 | } |
c67625a1 PE |
1291 | |
1292 | raw_notifier_chain_unregister(&netdev_chain, nb); | |
fcc5a03a | 1293 | goto unlock; |
1da177e4 | 1294 | } |
d1b19dff | 1295 | EXPORT_SYMBOL(register_netdevice_notifier); |
1da177e4 LT |
1296 | |
1297 | /** | |
1298 | * unregister_netdevice_notifier - unregister a network notifier block | |
1299 | * @nb: notifier | |
1300 | * | |
1301 | * Unregister a notifier previously registered by | |
1302 | * register_netdevice_notifier(). The notifier is unlinked into the | |
1303 | * kernel structures and may then be reused. A negative errno code | |
1304 | * is returned on a failure. | |
1305 | */ | |
1306 | ||
1307 | int unregister_netdevice_notifier(struct notifier_block *nb) | |
1308 | { | |
9f514950 HX |
1309 | int err; |
1310 | ||
1311 | rtnl_lock(); | |
f07d5b94 | 1312 | err = raw_notifier_chain_unregister(&netdev_chain, nb); |
9f514950 HX |
1313 | rtnl_unlock(); |
1314 | return err; | |
1da177e4 | 1315 | } |
d1b19dff | 1316 | EXPORT_SYMBOL(unregister_netdevice_notifier); |
1da177e4 LT |
1317 | |
1318 | /** | |
1319 | * call_netdevice_notifiers - call all network notifier blocks | |
1320 | * @val: value passed unmodified to notifier function | |
c4ea43c5 | 1321 | * @dev: net_device pointer passed unmodified to notifier function |
1da177e4 LT |
1322 | * |
1323 | * Call all network notifier blocks. Parameters and return value | |
f07d5b94 | 1324 | * are as for raw_notifier_call_chain(). |
1da177e4 LT |
1325 | */ |
1326 | ||
ad7379d4 | 1327 | int call_netdevice_notifiers(unsigned long val, struct net_device *dev) |
1da177e4 | 1328 | { |
ad7379d4 | 1329 | return raw_notifier_call_chain(&netdev_chain, val, dev); |
1da177e4 LT |
1330 | } |
1331 | ||
1332 | /* When > 0 there are consumers of rx skb time stamps */ | |
1333 | static atomic_t netstamp_needed = ATOMIC_INIT(0); | |
1334 | ||
1335 | void net_enable_timestamp(void) | |
1336 | { | |
1337 | atomic_inc(&netstamp_needed); | |
1338 | } | |
d1b19dff | 1339 | EXPORT_SYMBOL(net_enable_timestamp); |
1da177e4 LT |
1340 | |
1341 | void net_disable_timestamp(void) | |
1342 | { | |
1343 | atomic_dec(&netstamp_needed); | |
1344 | } | |
d1b19dff | 1345 | EXPORT_SYMBOL(net_disable_timestamp); |
1da177e4 | 1346 | |
a61bbcf2 | 1347 | static inline void net_timestamp(struct sk_buff *skb) |
1da177e4 LT |
1348 | { |
1349 | if (atomic_read(&netstamp_needed)) | |
a61bbcf2 | 1350 | __net_timestamp(skb); |
b7aa0bf7 ED |
1351 | else |
1352 | skb->tstamp.tv64 = 0; | |
1da177e4 LT |
1353 | } |
1354 | ||
1355 | /* | |
1356 | * Support routine. Sends outgoing frames to any network | |
1357 | * taps currently in use. | |
1358 | */ | |
1359 | ||
f6a78bfc | 1360 | static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
1361 | { |
1362 | struct packet_type *ptype; | |
a61bbcf2 | 1363 | |
8caf1539 JP |
1364 | #ifdef CONFIG_NET_CLS_ACT |
1365 | if (!(skb->tstamp.tv64 && (G_TC_FROM(skb->tc_verd) & AT_INGRESS))) | |
1366 | net_timestamp(skb); | |
1367 | #else | |
a61bbcf2 | 1368 | net_timestamp(skb); |
8caf1539 | 1369 | #endif |
1da177e4 LT |
1370 | |
1371 | rcu_read_lock(); | |
1372 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
1373 | /* Never send packets back to the socket | |
1374 | * they originated from - MvS (miquels@drinkel.ow.org) | |
1375 | */ | |
1376 | if ((ptype->dev == dev || !ptype->dev) && | |
1377 | (ptype->af_packet_priv == NULL || | |
1378 | (struct sock *)ptype->af_packet_priv != skb->sk)) { | |
d1b19dff | 1379 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); |
1da177e4 LT |
1380 | if (!skb2) |
1381 | break; | |
1382 | ||
1383 | /* skb->nh should be correctly | |
1384 | set by sender, so that the second statement is | |
1385 | just protection against buggy protocols. | |
1386 | */ | |
459a98ed | 1387 | skb_reset_mac_header(skb2); |
1da177e4 | 1388 | |
d56f90a7 | 1389 | if (skb_network_header(skb2) < skb2->data || |
27a884dc | 1390 | skb2->network_header > skb2->tail) { |
1da177e4 LT |
1391 | if (net_ratelimit()) |
1392 | printk(KERN_CRIT "protocol %04x is " | |
1393 | "buggy, dev %s\n", | |
1394 | skb2->protocol, dev->name); | |
c1d2bbe1 | 1395 | skb_reset_network_header(skb2); |
1da177e4 LT |
1396 | } |
1397 | ||
b0e380b1 | 1398 | skb2->transport_header = skb2->network_header; |
1da177e4 | 1399 | skb2->pkt_type = PACKET_OUTGOING; |
f2ccd8fa | 1400 | ptype->func(skb2, skb->dev, ptype, skb->dev); |
1da177e4 LT |
1401 | } |
1402 | } | |
1403 | rcu_read_unlock(); | |
1404 | } | |
1405 | ||
56079431 | 1406 | |
def82a1d | 1407 | static inline void __netif_reschedule(struct Qdisc *q) |
56079431 | 1408 | { |
def82a1d JP |
1409 | struct softnet_data *sd; |
1410 | unsigned long flags; | |
56079431 | 1411 | |
def82a1d JP |
1412 | local_irq_save(flags); |
1413 | sd = &__get_cpu_var(softnet_data); | |
1414 | q->next_sched = sd->output_queue; | |
1415 | sd->output_queue = q; | |
1416 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
1417 | local_irq_restore(flags); | |
1418 | } | |
1419 | ||
1420 | void __netif_schedule(struct Qdisc *q) | |
1421 | { | |
1422 | if (!test_and_set_bit(__QDISC_STATE_SCHED, &q->state)) | |
1423 | __netif_reschedule(q); | |
56079431 DV |
1424 | } |
1425 | EXPORT_SYMBOL(__netif_schedule); | |
1426 | ||
bea3348e | 1427 | void dev_kfree_skb_irq(struct sk_buff *skb) |
56079431 | 1428 | { |
bea3348e SH |
1429 | if (atomic_dec_and_test(&skb->users)) { |
1430 | struct softnet_data *sd; | |
1431 | unsigned long flags; | |
56079431 | 1432 | |
bea3348e SH |
1433 | local_irq_save(flags); |
1434 | sd = &__get_cpu_var(softnet_data); | |
1435 | skb->next = sd->completion_queue; | |
1436 | sd->completion_queue = skb; | |
1437 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
1438 | local_irq_restore(flags); | |
1439 | } | |
56079431 | 1440 | } |
bea3348e | 1441 | EXPORT_SYMBOL(dev_kfree_skb_irq); |
56079431 DV |
1442 | |
1443 | void dev_kfree_skb_any(struct sk_buff *skb) | |
1444 | { | |
1445 | if (in_irq() || irqs_disabled()) | |
1446 | dev_kfree_skb_irq(skb); | |
1447 | else | |
1448 | dev_kfree_skb(skb); | |
1449 | } | |
1450 | EXPORT_SYMBOL(dev_kfree_skb_any); | |
1451 | ||
1452 | ||
bea3348e SH |
1453 | /** |
1454 | * netif_device_detach - mark device as removed | |
1455 | * @dev: network device | |
1456 | * | |
1457 | * Mark device as removed from system and therefore no longer available. | |
1458 | */ | |
56079431 DV |
1459 | void netif_device_detach(struct net_device *dev) |
1460 | { | |
1461 | if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) && | |
1462 | netif_running(dev)) { | |
d543103a | 1463 | netif_tx_stop_all_queues(dev); |
56079431 DV |
1464 | } |
1465 | } | |
1466 | EXPORT_SYMBOL(netif_device_detach); | |
1467 | ||
bea3348e SH |
1468 | /** |
1469 | * netif_device_attach - mark device as attached | |
1470 | * @dev: network device | |
1471 | * | |
1472 | * Mark device as attached from system and restart if needed. | |
1473 | */ | |
56079431 DV |
1474 | void netif_device_attach(struct net_device *dev) |
1475 | { | |
1476 | if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) && | |
1477 | netif_running(dev)) { | |
d543103a | 1478 | netif_tx_wake_all_queues(dev); |
4ec93edb | 1479 | __netdev_watchdog_up(dev); |
56079431 DV |
1480 | } |
1481 | } | |
1482 | EXPORT_SYMBOL(netif_device_attach); | |
1483 | ||
6de329e2 BH |
1484 | static bool can_checksum_protocol(unsigned long features, __be16 protocol) |
1485 | { | |
1486 | return ((features & NETIF_F_GEN_CSUM) || | |
1487 | ((features & NETIF_F_IP_CSUM) && | |
1488 | protocol == htons(ETH_P_IP)) || | |
1489 | ((features & NETIF_F_IPV6_CSUM) && | |
1c8dbcf6 YZ |
1490 | protocol == htons(ETH_P_IPV6)) || |
1491 | ((features & NETIF_F_FCOE_CRC) && | |
1492 | protocol == htons(ETH_P_FCOE))); | |
6de329e2 BH |
1493 | } |
1494 | ||
1495 | static bool dev_can_checksum(struct net_device *dev, struct sk_buff *skb) | |
1496 | { | |
1497 | if (can_checksum_protocol(dev->features, skb->protocol)) | |
1498 | return true; | |
1499 | ||
1500 | if (skb->protocol == htons(ETH_P_8021Q)) { | |
1501 | struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data; | |
1502 | if (can_checksum_protocol(dev->features & dev->vlan_features, | |
1503 | veh->h_vlan_encapsulated_proto)) | |
1504 | return true; | |
1505 | } | |
1506 | ||
1507 | return false; | |
1508 | } | |
56079431 | 1509 | |
1da177e4 LT |
1510 | /* |
1511 | * Invalidate hardware checksum when packet is to be mangled, and | |
1512 | * complete checksum manually on outgoing path. | |
1513 | */ | |
84fa7933 | 1514 | int skb_checksum_help(struct sk_buff *skb) |
1da177e4 | 1515 | { |
d3bc23e7 | 1516 | __wsum csum; |
663ead3b | 1517 | int ret = 0, offset; |
1da177e4 | 1518 | |
84fa7933 | 1519 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
a430a43d HX |
1520 | goto out_set_summed; |
1521 | ||
1522 | if (unlikely(skb_shinfo(skb)->gso_size)) { | |
a430a43d HX |
1523 | /* Let GSO fix up the checksum. */ |
1524 | goto out_set_summed; | |
1da177e4 LT |
1525 | } |
1526 | ||
a030847e HX |
1527 | offset = skb->csum_start - skb_headroom(skb); |
1528 | BUG_ON(offset >= skb_headlen(skb)); | |
1529 | csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
1530 | ||
1531 | offset += skb->csum_offset; | |
1532 | BUG_ON(offset + sizeof(__sum16) > skb_headlen(skb)); | |
1533 | ||
1534 | if (skb_cloned(skb) && | |
1535 | !skb_clone_writable(skb, offset + sizeof(__sum16))) { | |
1da177e4 LT |
1536 | ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC); |
1537 | if (ret) | |
1538 | goto out; | |
1539 | } | |
1540 | ||
a030847e | 1541 | *(__sum16 *)(skb->data + offset) = csum_fold(csum); |
a430a43d | 1542 | out_set_summed: |
1da177e4 | 1543 | skb->ip_summed = CHECKSUM_NONE; |
4ec93edb | 1544 | out: |
1da177e4 LT |
1545 | return ret; |
1546 | } | |
d1b19dff | 1547 | EXPORT_SYMBOL(skb_checksum_help); |
1da177e4 | 1548 | |
f6a78bfc HX |
1549 | /** |
1550 | * skb_gso_segment - Perform segmentation on skb. | |
1551 | * @skb: buffer to segment | |
576a30eb | 1552 | * @features: features for the output path (see dev->features) |
f6a78bfc HX |
1553 | * |
1554 | * This function segments the given skb and returns a list of segments. | |
576a30eb HX |
1555 | * |
1556 | * It may return NULL if the skb requires no segmentation. This is | |
1557 | * only possible when GSO is used for verifying header integrity. | |
f6a78bfc | 1558 | */ |
576a30eb | 1559 | struct sk_buff *skb_gso_segment(struct sk_buff *skb, int features) |
f6a78bfc HX |
1560 | { |
1561 | struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT); | |
1562 | struct packet_type *ptype; | |
252e3346 | 1563 | __be16 type = skb->protocol; |
a430a43d | 1564 | int err; |
f6a78bfc | 1565 | |
459a98ed | 1566 | skb_reset_mac_header(skb); |
b0e380b1 | 1567 | skb->mac_len = skb->network_header - skb->mac_header; |
f6a78bfc HX |
1568 | __skb_pull(skb, skb->mac_len); |
1569 | ||
67fd1a73 HX |
1570 | if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { |
1571 | struct net_device *dev = skb->dev; | |
1572 | struct ethtool_drvinfo info = {}; | |
1573 | ||
1574 | if (dev && dev->ethtool_ops && dev->ethtool_ops->get_drvinfo) | |
1575 | dev->ethtool_ops->get_drvinfo(dev, &info); | |
1576 | ||
1577 | WARN(1, "%s: caps=(0x%lx, 0x%lx) len=%d data_len=%d " | |
1578 | "ip_summed=%d", | |
1579 | info.driver, dev ? dev->features : 0L, | |
1580 | skb->sk ? skb->sk->sk_route_caps : 0L, | |
1581 | skb->len, skb->data_len, skb->ip_summed); | |
1582 | ||
a430a43d HX |
1583 | if (skb_header_cloned(skb) && |
1584 | (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) | |
1585 | return ERR_PTR(err); | |
1586 | } | |
1587 | ||
f6a78bfc | 1588 | rcu_read_lock(); |
82d8a867 PE |
1589 | list_for_each_entry_rcu(ptype, |
1590 | &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { | |
f6a78bfc | 1591 | if (ptype->type == type && !ptype->dev && ptype->gso_segment) { |
84fa7933 | 1592 | if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { |
a430a43d HX |
1593 | err = ptype->gso_send_check(skb); |
1594 | segs = ERR_PTR(err); | |
1595 | if (err || skb_gso_ok(skb, features)) | |
1596 | break; | |
d56f90a7 ACM |
1597 | __skb_push(skb, (skb->data - |
1598 | skb_network_header(skb))); | |
a430a43d | 1599 | } |
576a30eb | 1600 | segs = ptype->gso_segment(skb, features); |
f6a78bfc HX |
1601 | break; |
1602 | } | |
1603 | } | |
1604 | rcu_read_unlock(); | |
1605 | ||
98e399f8 | 1606 | __skb_push(skb, skb->data - skb_mac_header(skb)); |
576a30eb | 1607 | |
f6a78bfc HX |
1608 | return segs; |
1609 | } | |
f6a78bfc HX |
1610 | EXPORT_SYMBOL(skb_gso_segment); |
1611 | ||
fb286bb2 HX |
1612 | /* Take action when hardware reception checksum errors are detected. */ |
1613 | #ifdef CONFIG_BUG | |
1614 | void netdev_rx_csum_fault(struct net_device *dev) | |
1615 | { | |
1616 | if (net_ratelimit()) { | |
4ec93edb | 1617 | printk(KERN_ERR "%s: hw csum failure.\n", |
246a4212 | 1618 | dev ? dev->name : "<unknown>"); |
fb286bb2 HX |
1619 | dump_stack(); |
1620 | } | |
1621 | } | |
1622 | EXPORT_SYMBOL(netdev_rx_csum_fault); | |
1623 | #endif | |
1624 | ||
1da177e4 LT |
1625 | /* Actually, we should eliminate this check as soon as we know, that: |
1626 | * 1. IOMMU is present and allows to map all the memory. | |
1627 | * 2. No high memory really exists on this machine. | |
1628 | */ | |
1629 | ||
1630 | static inline int illegal_highdma(struct net_device *dev, struct sk_buff *skb) | |
1631 | { | |
3d3a8533 | 1632 | #ifdef CONFIG_HIGHMEM |
1da177e4 LT |
1633 | int i; |
1634 | ||
1635 | if (dev->features & NETIF_F_HIGHDMA) | |
1636 | return 0; | |
1637 | ||
1638 | for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) | |
1639 | if (PageHighMem(skb_shinfo(skb)->frags[i].page)) | |
1640 | return 1; | |
1641 | ||
3d3a8533 | 1642 | #endif |
1da177e4 LT |
1643 | return 0; |
1644 | } | |
1da177e4 | 1645 | |
f6a78bfc HX |
1646 | struct dev_gso_cb { |
1647 | void (*destructor)(struct sk_buff *skb); | |
1648 | }; | |
1649 | ||
1650 | #define DEV_GSO_CB(skb) ((struct dev_gso_cb *)(skb)->cb) | |
1651 | ||
1652 | static void dev_gso_skb_destructor(struct sk_buff *skb) | |
1653 | { | |
1654 | struct dev_gso_cb *cb; | |
1655 | ||
1656 | do { | |
1657 | struct sk_buff *nskb = skb->next; | |
1658 | ||
1659 | skb->next = nskb->next; | |
1660 | nskb->next = NULL; | |
1661 | kfree_skb(nskb); | |
1662 | } while (skb->next); | |
1663 | ||
1664 | cb = DEV_GSO_CB(skb); | |
1665 | if (cb->destructor) | |
1666 | cb->destructor(skb); | |
1667 | } | |
1668 | ||
1669 | /** | |
1670 | * dev_gso_segment - Perform emulated hardware segmentation on skb. | |
1671 | * @skb: buffer to segment | |
1672 | * | |
1673 | * This function segments the given skb and stores the list of segments | |
1674 | * in skb->next. | |
1675 | */ | |
1676 | static int dev_gso_segment(struct sk_buff *skb) | |
1677 | { | |
1678 | struct net_device *dev = skb->dev; | |
1679 | struct sk_buff *segs; | |
576a30eb HX |
1680 | int features = dev->features & ~(illegal_highdma(dev, skb) ? |
1681 | NETIF_F_SG : 0); | |
1682 | ||
1683 | segs = skb_gso_segment(skb, features); | |
1684 | ||
1685 | /* Verifying header integrity only. */ | |
1686 | if (!segs) | |
1687 | return 0; | |
f6a78bfc | 1688 | |
801678c5 | 1689 | if (IS_ERR(segs)) |
f6a78bfc HX |
1690 | return PTR_ERR(segs); |
1691 | ||
1692 | skb->next = segs; | |
1693 | DEV_GSO_CB(skb)->destructor = skb->destructor; | |
1694 | skb->destructor = dev_gso_skb_destructor; | |
1695 | ||
1696 | return 0; | |
1697 | } | |
1698 | ||
fd2ea0a7 DM |
1699 | int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, |
1700 | struct netdev_queue *txq) | |
f6a78bfc | 1701 | { |
00829823 | 1702 | const struct net_device_ops *ops = dev->netdev_ops; |
ac45f602 | 1703 | int rc; |
00829823 | 1704 | |
f6a78bfc | 1705 | if (likely(!skb->next)) { |
9be9a6b9 | 1706 | if (!list_empty(&ptype_all)) |
f6a78bfc HX |
1707 | dev_queue_xmit_nit(skb, dev); |
1708 | ||
576a30eb HX |
1709 | if (netif_needs_gso(dev, skb)) { |
1710 | if (unlikely(dev_gso_segment(skb))) | |
1711 | goto out_kfree_skb; | |
1712 | if (skb->next) | |
1713 | goto gso; | |
1714 | } | |
f6a78bfc | 1715 | |
93f154b5 ED |
1716 | /* |
1717 | * If device doesnt need skb->dst, release it right now while | |
1718 | * its hot in this cpu cache | |
1719 | */ | |
adf30907 ED |
1720 | if (dev->priv_flags & IFF_XMIT_DST_RELEASE) |
1721 | skb_dst_drop(skb); | |
1722 | ||
ac45f602 | 1723 | rc = ops->ndo_start_xmit(skb, dev); |
ec634fe3 | 1724 | if (rc == NETDEV_TX_OK) |
08baf561 | 1725 | txq_trans_update(txq); |
ac45f602 PO |
1726 | /* |
1727 | * TODO: if skb_orphan() was called by | |
1728 | * dev->hard_start_xmit() (for example, the unmodified | |
1729 | * igb driver does that; bnx2 doesn't), then | |
1730 | * skb_tx_software_timestamp() will be unable to send | |
1731 | * back the time stamp. | |
1732 | * | |
1733 | * How can this be prevented? Always create another | |
1734 | * reference to the socket before calling | |
1735 | * dev->hard_start_xmit()? Prevent that skb_orphan() | |
1736 | * does anything in dev->hard_start_xmit() by clearing | |
1737 | * the skb destructor before the call and restoring it | |
1738 | * afterwards, then doing the skb_orphan() ourselves? | |
1739 | */ | |
ac45f602 | 1740 | return rc; |
f6a78bfc HX |
1741 | } |
1742 | ||
576a30eb | 1743 | gso: |
f6a78bfc HX |
1744 | do { |
1745 | struct sk_buff *nskb = skb->next; | |
f6a78bfc HX |
1746 | |
1747 | skb->next = nskb->next; | |
1748 | nskb->next = NULL; | |
00829823 | 1749 | rc = ops->ndo_start_xmit(nskb, dev); |
ec634fe3 | 1750 | if (unlikely(rc != NETDEV_TX_OK)) { |
f54d9e8d | 1751 | nskb->next = skb->next; |
f6a78bfc HX |
1752 | skb->next = nskb; |
1753 | return rc; | |
1754 | } | |
08baf561 | 1755 | txq_trans_update(txq); |
fd2ea0a7 | 1756 | if (unlikely(netif_tx_queue_stopped(txq) && skb->next)) |
f54d9e8d | 1757 | return NETDEV_TX_BUSY; |
f6a78bfc | 1758 | } while (skb->next); |
4ec93edb | 1759 | |
f6a78bfc HX |
1760 | skb->destructor = DEV_GSO_CB(skb)->destructor; |
1761 | ||
1762 | out_kfree_skb: | |
1763 | kfree_skb(skb); | |
ec634fe3 | 1764 | return NETDEV_TX_OK; |
f6a78bfc HX |
1765 | } |
1766 | ||
7019298a | 1767 | static u32 skb_tx_hashrnd; |
b6b2fed1 | 1768 | |
9247744e | 1769 | u16 skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb) |
8f0f2223 | 1770 | { |
7019298a | 1771 | u32 hash; |
b6b2fed1 | 1772 | |
513de11b DM |
1773 | if (skb_rx_queue_recorded(skb)) { |
1774 | hash = skb_get_rx_queue(skb); | |
d1b19dff | 1775 | while (unlikely(hash >= dev->real_num_tx_queues)) |
513de11b DM |
1776 | hash -= dev->real_num_tx_queues; |
1777 | return hash; | |
1778 | } | |
ec581f6a ED |
1779 | |
1780 | if (skb->sk && skb->sk->sk_hash) | |
7019298a | 1781 | hash = skb->sk->sk_hash; |
ec581f6a | 1782 | else |
7019298a | 1783 | hash = skb->protocol; |
d5a9e24a | 1784 | |
7019298a | 1785 | hash = jhash_1word(hash, skb_tx_hashrnd); |
b6b2fed1 DM |
1786 | |
1787 | return (u16) (((u64) hash * dev->real_num_tx_queues) >> 32); | |
8f0f2223 | 1788 | } |
9247744e | 1789 | EXPORT_SYMBOL(skb_tx_hash); |
8f0f2223 | 1790 | |
e8a0464c DM |
1791 | static struct netdev_queue *dev_pick_tx(struct net_device *dev, |
1792 | struct sk_buff *skb) | |
1793 | { | |
00829823 | 1794 | const struct net_device_ops *ops = dev->netdev_ops; |
fd2ea0a7 DM |
1795 | u16 queue_index = 0; |
1796 | ||
00829823 SH |
1797 | if (ops->ndo_select_queue) |
1798 | queue_index = ops->ndo_select_queue(dev, skb); | |
8f0f2223 | 1799 | else if (dev->real_num_tx_queues > 1) |
7019298a | 1800 | queue_index = skb_tx_hash(dev, skb); |
eae792b7 | 1801 | |
fd2ea0a7 DM |
1802 | skb_set_queue_mapping(skb, queue_index); |
1803 | return netdev_get_tx_queue(dev, queue_index); | |
e8a0464c DM |
1804 | } |
1805 | ||
bbd8a0d3 KK |
1806 | static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q, |
1807 | struct net_device *dev, | |
1808 | struct netdev_queue *txq) | |
1809 | { | |
1810 | spinlock_t *root_lock = qdisc_lock(q); | |
1811 | int rc; | |
1812 | ||
1813 | spin_lock(root_lock); | |
1814 | if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) { | |
1815 | kfree_skb(skb); | |
1816 | rc = NET_XMIT_DROP; | |
1817 | } else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) && | |
1818 | !test_and_set_bit(__QDISC_STATE_RUNNING, &q->state)) { | |
1819 | /* | |
1820 | * This is a work-conserving queue; there are no old skbs | |
1821 | * waiting to be sent out; and the qdisc is not running - | |
1822 | * xmit the skb directly. | |
1823 | */ | |
1824 | __qdisc_update_bstats(q, skb->len); | |
1825 | if (sch_direct_xmit(skb, q, dev, txq, root_lock)) | |
1826 | __qdisc_run(q); | |
1827 | else | |
1828 | clear_bit(__QDISC_STATE_RUNNING, &q->state); | |
1829 | ||
1830 | rc = NET_XMIT_SUCCESS; | |
1831 | } else { | |
1832 | rc = qdisc_enqueue_root(skb, q); | |
1833 | qdisc_run(q); | |
1834 | } | |
1835 | spin_unlock(root_lock); | |
1836 | ||
1837 | return rc; | |
1838 | } | |
1839 | ||
d29f749e DJ |
1840 | /** |
1841 | * dev_queue_xmit - transmit a buffer | |
1842 | * @skb: buffer to transmit | |
1843 | * | |
1844 | * Queue a buffer for transmission to a network device. The caller must | |
1845 | * have set the device and priority and built the buffer before calling | |
1846 | * this function. The function can be called from an interrupt. | |
1847 | * | |
1848 | * A negative errno code is returned on a failure. A success does not | |
1849 | * guarantee the frame will be transmitted as it may be dropped due | |
1850 | * to congestion or traffic shaping. | |
1851 | * | |
1852 | * ----------------------------------------------------------------------------------- | |
1853 | * I notice this method can also return errors from the queue disciplines, | |
1854 | * including NET_XMIT_DROP, which is a positive value. So, errors can also | |
1855 | * be positive. | |
1856 | * | |
1857 | * Regardless of the return value, the skb is consumed, so it is currently | |
1858 | * difficult to retry a send to this method. (You can bump the ref count | |
1859 | * before sending to hold a reference for retry if you are careful.) | |
1860 | * | |
1861 | * When calling this method, interrupts MUST be enabled. This is because | |
1862 | * the BH enable code must have IRQs enabled so that it will not deadlock. | |
1863 | * --BLG | |
1864 | */ | |
1da177e4 LT |
1865 | int dev_queue_xmit(struct sk_buff *skb) |
1866 | { | |
1867 | struct net_device *dev = skb->dev; | |
dc2b4847 | 1868 | struct netdev_queue *txq; |
1da177e4 LT |
1869 | struct Qdisc *q; |
1870 | int rc = -ENOMEM; | |
1871 | ||
f6a78bfc HX |
1872 | /* GSO will handle the following emulations directly. */ |
1873 | if (netif_needs_gso(dev, skb)) | |
1874 | goto gso; | |
1875 | ||
4cf704fb | 1876 | if (skb_has_frags(skb) && |
1da177e4 | 1877 | !(dev->features & NETIF_F_FRAGLIST) && |
364c6bad | 1878 | __skb_linearize(skb)) |
1da177e4 LT |
1879 | goto out_kfree_skb; |
1880 | ||
1881 | /* Fragmented skb is linearized if device does not support SG, | |
1882 | * or if at least one of fragments is in highmem and device | |
1883 | * does not support DMA from it. | |
1884 | */ | |
1885 | if (skb_shinfo(skb)->nr_frags && | |
1886 | (!(dev->features & NETIF_F_SG) || illegal_highdma(dev, skb)) && | |
364c6bad | 1887 | __skb_linearize(skb)) |
1da177e4 LT |
1888 | goto out_kfree_skb; |
1889 | ||
1890 | /* If packet is not checksummed and device does not support | |
1891 | * checksumming for this protocol, complete checksumming here. | |
1892 | */ | |
663ead3b HX |
1893 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
1894 | skb_set_transport_header(skb, skb->csum_start - | |
1895 | skb_headroom(skb)); | |
6de329e2 BH |
1896 | if (!dev_can_checksum(dev, skb) && skb_checksum_help(skb)) |
1897 | goto out_kfree_skb; | |
663ead3b | 1898 | } |
1da177e4 | 1899 | |
f6a78bfc | 1900 | gso: |
4ec93edb YH |
1901 | /* Disable soft irqs for various locks below. Also |
1902 | * stops preemption for RCU. | |
1da177e4 | 1903 | */ |
4ec93edb | 1904 | rcu_read_lock_bh(); |
1da177e4 | 1905 | |
eae792b7 | 1906 | txq = dev_pick_tx(dev, skb); |
b0e1e646 | 1907 | q = rcu_dereference(txq->qdisc); |
37437bb2 | 1908 | |
1da177e4 | 1909 | #ifdef CONFIG_NET_CLS_ACT |
d1b19dff | 1910 | skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS); |
1da177e4 LT |
1911 | #endif |
1912 | if (q->enqueue) { | |
bbd8a0d3 | 1913 | rc = __dev_xmit_skb(skb, q, dev, txq); |
37437bb2 | 1914 | goto out; |
1da177e4 LT |
1915 | } |
1916 | ||
1917 | /* The device has no queue. Common case for software devices: | |
1918 | loopback, all the sorts of tunnels... | |
1919 | ||
932ff279 HX |
1920 | Really, it is unlikely that netif_tx_lock protection is necessary |
1921 | here. (f.e. loopback and IP tunnels are clean ignoring statistics | |
1da177e4 LT |
1922 | counters.) |
1923 | However, it is possible, that they rely on protection | |
1924 | made by us here. | |
1925 | ||
1926 | Check this and shot the lock. It is not prone from deadlocks. | |
1927 | Either shot noqueue qdisc, it is even simpler 8) | |
1928 | */ | |
1929 | if (dev->flags & IFF_UP) { | |
1930 | int cpu = smp_processor_id(); /* ok because BHs are off */ | |
1931 | ||
c773e847 | 1932 | if (txq->xmit_lock_owner != cpu) { |
1da177e4 | 1933 | |
c773e847 | 1934 | HARD_TX_LOCK(dev, txq, cpu); |
1da177e4 | 1935 | |
fd2ea0a7 | 1936 | if (!netif_tx_queue_stopped(txq)) { |
03a9a447 | 1937 | rc = NET_XMIT_SUCCESS; |
fd2ea0a7 | 1938 | if (!dev_hard_start_xmit(skb, dev, txq)) { |
c773e847 | 1939 | HARD_TX_UNLOCK(dev, txq); |
1da177e4 LT |
1940 | goto out; |
1941 | } | |
1942 | } | |
c773e847 | 1943 | HARD_TX_UNLOCK(dev, txq); |
1da177e4 LT |
1944 | if (net_ratelimit()) |
1945 | printk(KERN_CRIT "Virtual device %s asks to " | |
1946 | "queue packet!\n", dev->name); | |
1947 | } else { | |
1948 | /* Recursion is detected! It is possible, | |
1949 | * unfortunately */ | |
1950 | if (net_ratelimit()) | |
1951 | printk(KERN_CRIT "Dead loop on virtual device " | |
1952 | "%s, fix it urgently!\n", dev->name); | |
1953 | } | |
1954 | } | |
1955 | ||
1956 | rc = -ENETDOWN; | |
d4828d85 | 1957 | rcu_read_unlock_bh(); |
1da177e4 LT |
1958 | |
1959 | out_kfree_skb: | |
1960 | kfree_skb(skb); | |
1961 | return rc; | |
1962 | out: | |
d4828d85 | 1963 | rcu_read_unlock_bh(); |
1da177e4 LT |
1964 | return rc; |
1965 | } | |
d1b19dff | 1966 | EXPORT_SYMBOL(dev_queue_xmit); |
1da177e4 LT |
1967 | |
1968 | ||
1969 | /*======================================================================= | |
1970 | Receiver routines | |
1971 | =======================================================================*/ | |
1972 | ||
6b2bedc3 SH |
1973 | int netdev_max_backlog __read_mostly = 1000; |
1974 | int netdev_budget __read_mostly = 300; | |
1975 | int weight_p __read_mostly = 64; /* old backlog weight */ | |
1da177e4 LT |
1976 | |
1977 | DEFINE_PER_CPU(struct netif_rx_stats, netdev_rx_stat) = { 0, }; | |
1978 | ||
1979 | ||
1da177e4 LT |
1980 | /** |
1981 | * netif_rx - post buffer to the network code | |
1982 | * @skb: buffer to post | |
1983 | * | |
1984 | * This function receives a packet from a device driver and queues it for | |
1985 | * the upper (protocol) levels to process. It always succeeds. The buffer | |
1986 | * may be dropped during processing for congestion control or by the | |
1987 | * protocol layers. | |
1988 | * | |
1989 | * return values: | |
1990 | * NET_RX_SUCCESS (no congestion) | |
1da177e4 LT |
1991 | * NET_RX_DROP (packet was dropped) |
1992 | * | |
1993 | */ | |
1994 | ||
1995 | int netif_rx(struct sk_buff *skb) | |
1996 | { | |
1da177e4 LT |
1997 | struct softnet_data *queue; |
1998 | unsigned long flags; | |
1999 | ||
2000 | /* if netpoll wants it, pretend we never saw it */ | |
2001 | if (netpoll_rx(skb)) | |
2002 | return NET_RX_DROP; | |
2003 | ||
b7aa0bf7 | 2004 | if (!skb->tstamp.tv64) |
a61bbcf2 | 2005 | net_timestamp(skb); |
1da177e4 LT |
2006 | |
2007 | /* | |
2008 | * The code is rearranged so that the path is the most | |
2009 | * short when CPU is congested, but is still operating. | |
2010 | */ | |
2011 | local_irq_save(flags); | |
1da177e4 LT |
2012 | queue = &__get_cpu_var(softnet_data); |
2013 | ||
2014 | __get_cpu_var(netdev_rx_stat).total++; | |
2015 | if (queue->input_pkt_queue.qlen <= netdev_max_backlog) { | |
2016 | if (queue->input_pkt_queue.qlen) { | |
1da177e4 | 2017 | enqueue: |
1da177e4 | 2018 | __skb_queue_tail(&queue->input_pkt_queue, skb); |
1da177e4 | 2019 | local_irq_restore(flags); |
34008d8c | 2020 | return NET_RX_SUCCESS; |
1da177e4 LT |
2021 | } |
2022 | ||
bea3348e | 2023 | napi_schedule(&queue->backlog); |
1da177e4 LT |
2024 | goto enqueue; |
2025 | } | |
2026 | ||
1da177e4 LT |
2027 | __get_cpu_var(netdev_rx_stat).dropped++; |
2028 | local_irq_restore(flags); | |
2029 | ||
2030 | kfree_skb(skb); | |
2031 | return NET_RX_DROP; | |
2032 | } | |
d1b19dff | 2033 | EXPORT_SYMBOL(netif_rx); |
1da177e4 LT |
2034 | |
2035 | int netif_rx_ni(struct sk_buff *skb) | |
2036 | { | |
2037 | int err; | |
2038 | ||
2039 | preempt_disable(); | |
2040 | err = netif_rx(skb); | |
2041 | if (local_softirq_pending()) | |
2042 | do_softirq(); | |
2043 | preempt_enable(); | |
2044 | ||
2045 | return err; | |
2046 | } | |
1da177e4 LT |
2047 | EXPORT_SYMBOL(netif_rx_ni); |
2048 | ||
1da177e4 LT |
2049 | static void net_tx_action(struct softirq_action *h) |
2050 | { | |
2051 | struct softnet_data *sd = &__get_cpu_var(softnet_data); | |
2052 | ||
2053 | if (sd->completion_queue) { | |
2054 | struct sk_buff *clist; | |
2055 | ||
2056 | local_irq_disable(); | |
2057 | clist = sd->completion_queue; | |
2058 | sd->completion_queue = NULL; | |
2059 | local_irq_enable(); | |
2060 | ||
2061 | while (clist) { | |
2062 | struct sk_buff *skb = clist; | |
2063 | clist = clist->next; | |
2064 | ||
547b792c | 2065 | WARN_ON(atomic_read(&skb->users)); |
1da177e4 LT |
2066 | __kfree_skb(skb); |
2067 | } | |
2068 | } | |
2069 | ||
2070 | if (sd->output_queue) { | |
37437bb2 | 2071 | struct Qdisc *head; |
1da177e4 LT |
2072 | |
2073 | local_irq_disable(); | |
2074 | head = sd->output_queue; | |
2075 | sd->output_queue = NULL; | |
2076 | local_irq_enable(); | |
2077 | ||
2078 | while (head) { | |
37437bb2 DM |
2079 | struct Qdisc *q = head; |
2080 | spinlock_t *root_lock; | |
2081 | ||
1da177e4 LT |
2082 | head = head->next_sched; |
2083 | ||
5fb66229 | 2084 | root_lock = qdisc_lock(q); |
37437bb2 | 2085 | if (spin_trylock(root_lock)) { |
def82a1d JP |
2086 | smp_mb__before_clear_bit(); |
2087 | clear_bit(__QDISC_STATE_SCHED, | |
2088 | &q->state); | |
37437bb2 DM |
2089 | qdisc_run(q); |
2090 | spin_unlock(root_lock); | |
1da177e4 | 2091 | } else { |
195648bb | 2092 | if (!test_bit(__QDISC_STATE_DEACTIVATED, |
e8a83e10 | 2093 | &q->state)) { |
195648bb | 2094 | __netif_reschedule(q); |
e8a83e10 JP |
2095 | } else { |
2096 | smp_mb__before_clear_bit(); | |
2097 | clear_bit(__QDISC_STATE_SCHED, | |
2098 | &q->state); | |
2099 | } | |
1da177e4 LT |
2100 | } |
2101 | } | |
2102 | } | |
2103 | } | |
2104 | ||
6f05f629 SH |
2105 | static inline int deliver_skb(struct sk_buff *skb, |
2106 | struct packet_type *pt_prev, | |
2107 | struct net_device *orig_dev) | |
1da177e4 LT |
2108 | { |
2109 | atomic_inc(&skb->users); | |
f2ccd8fa | 2110 | return pt_prev->func(skb, skb->dev, pt_prev, orig_dev); |
1da177e4 LT |
2111 | } |
2112 | ||
2113 | #if defined(CONFIG_BRIDGE) || defined (CONFIG_BRIDGE_MODULE) | |
da678292 MM |
2114 | |
2115 | #if defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE) | |
2116 | /* This hook is defined here for ATM LANE */ | |
2117 | int (*br_fdb_test_addr_hook)(struct net_device *dev, | |
2118 | unsigned char *addr) __read_mostly; | |
4fb019a0 | 2119 | EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook); |
da678292 | 2120 | #endif |
1da177e4 | 2121 | |
6229e362 SH |
2122 | /* |
2123 | * If bridge module is loaded call bridging hook. | |
2124 | * returns NULL if packet was consumed. | |
2125 | */ | |
2126 | struct sk_buff *(*br_handle_frame_hook)(struct net_bridge_port *p, | |
2127 | struct sk_buff *skb) __read_mostly; | |
4fb019a0 | 2128 | EXPORT_SYMBOL_GPL(br_handle_frame_hook); |
da678292 | 2129 | |
6229e362 SH |
2130 | static inline struct sk_buff *handle_bridge(struct sk_buff *skb, |
2131 | struct packet_type **pt_prev, int *ret, | |
2132 | struct net_device *orig_dev) | |
1da177e4 LT |
2133 | { |
2134 | struct net_bridge_port *port; | |
2135 | ||
6229e362 SH |
2136 | if (skb->pkt_type == PACKET_LOOPBACK || |
2137 | (port = rcu_dereference(skb->dev->br_port)) == NULL) | |
2138 | return skb; | |
1da177e4 LT |
2139 | |
2140 | if (*pt_prev) { | |
6229e362 | 2141 | *ret = deliver_skb(skb, *pt_prev, orig_dev); |
1da177e4 | 2142 | *pt_prev = NULL; |
4ec93edb YH |
2143 | } |
2144 | ||
6229e362 | 2145 | return br_handle_frame_hook(port, skb); |
1da177e4 LT |
2146 | } |
2147 | #else | |
6229e362 | 2148 | #define handle_bridge(skb, pt_prev, ret, orig_dev) (skb) |
1da177e4 LT |
2149 | #endif |
2150 | ||
b863ceb7 PM |
2151 | #if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) |
2152 | struct sk_buff *(*macvlan_handle_frame_hook)(struct sk_buff *skb) __read_mostly; | |
2153 | EXPORT_SYMBOL_GPL(macvlan_handle_frame_hook); | |
2154 | ||
2155 | static inline struct sk_buff *handle_macvlan(struct sk_buff *skb, | |
2156 | struct packet_type **pt_prev, | |
2157 | int *ret, | |
2158 | struct net_device *orig_dev) | |
2159 | { | |
2160 | if (skb->dev->macvlan_port == NULL) | |
2161 | return skb; | |
2162 | ||
2163 | if (*pt_prev) { | |
2164 | *ret = deliver_skb(skb, *pt_prev, orig_dev); | |
2165 | *pt_prev = NULL; | |
2166 | } | |
2167 | return macvlan_handle_frame_hook(skb); | |
2168 | } | |
2169 | #else | |
2170 | #define handle_macvlan(skb, pt_prev, ret, orig_dev) (skb) | |
2171 | #endif | |
2172 | ||
1da177e4 LT |
2173 | #ifdef CONFIG_NET_CLS_ACT |
2174 | /* TODO: Maybe we should just force sch_ingress to be compiled in | |
2175 | * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions | |
2176 | * a compare and 2 stores extra right now if we dont have it on | |
2177 | * but have CONFIG_NET_CLS_ACT | |
4ec93edb | 2178 | * NOTE: This doesnt stop any functionality; if you dont have |
1da177e4 LT |
2179 | * the ingress scheduler, you just cant add policies on ingress. |
2180 | * | |
2181 | */ | |
4ec93edb | 2182 | static int ing_filter(struct sk_buff *skb) |
1da177e4 | 2183 | { |
1da177e4 | 2184 | struct net_device *dev = skb->dev; |
f697c3e8 | 2185 | u32 ttl = G_TC_RTTL(skb->tc_verd); |
555353cf DM |
2186 | struct netdev_queue *rxq; |
2187 | int result = TC_ACT_OK; | |
2188 | struct Qdisc *q; | |
4ec93edb | 2189 | |
f697c3e8 HX |
2190 | if (MAX_RED_LOOP < ttl++) { |
2191 | printk(KERN_WARNING | |
2192 | "Redir loop detected Dropping packet (%d->%d)\n", | |
2193 | skb->iif, dev->ifindex); | |
2194 | return TC_ACT_SHOT; | |
2195 | } | |
1da177e4 | 2196 | |
f697c3e8 HX |
2197 | skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl); |
2198 | skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS); | |
1da177e4 | 2199 | |
555353cf DM |
2200 | rxq = &dev->rx_queue; |
2201 | ||
83874000 | 2202 | q = rxq->qdisc; |
8d50b53d | 2203 | if (q != &noop_qdisc) { |
83874000 | 2204 | spin_lock(qdisc_lock(q)); |
a9312ae8 DM |
2205 | if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) |
2206 | result = qdisc_enqueue_root(skb, q); | |
83874000 DM |
2207 | spin_unlock(qdisc_lock(q)); |
2208 | } | |
f697c3e8 HX |
2209 | |
2210 | return result; | |
2211 | } | |
86e65da9 | 2212 | |
f697c3e8 HX |
2213 | static inline struct sk_buff *handle_ing(struct sk_buff *skb, |
2214 | struct packet_type **pt_prev, | |
2215 | int *ret, struct net_device *orig_dev) | |
2216 | { | |
8d50b53d | 2217 | if (skb->dev->rx_queue.qdisc == &noop_qdisc) |
f697c3e8 | 2218 | goto out; |
1da177e4 | 2219 | |
f697c3e8 HX |
2220 | if (*pt_prev) { |
2221 | *ret = deliver_skb(skb, *pt_prev, orig_dev); | |
2222 | *pt_prev = NULL; | |
2223 | } else { | |
2224 | /* Huh? Why does turning on AF_PACKET affect this? */ | |
2225 | skb->tc_verd = SET_TC_OK2MUNGE(skb->tc_verd); | |
1da177e4 LT |
2226 | } |
2227 | ||
f697c3e8 HX |
2228 | switch (ing_filter(skb)) { |
2229 | case TC_ACT_SHOT: | |
2230 | case TC_ACT_STOLEN: | |
2231 | kfree_skb(skb); | |
2232 | return NULL; | |
2233 | } | |
2234 | ||
2235 | out: | |
2236 | skb->tc_verd = 0; | |
2237 | return skb; | |
1da177e4 LT |
2238 | } |
2239 | #endif | |
2240 | ||
bc1d0411 PM |
2241 | /* |
2242 | * netif_nit_deliver - deliver received packets to network taps | |
2243 | * @skb: buffer | |
2244 | * | |
2245 | * This function is used to deliver incoming packets to network | |
2246 | * taps. It should be used when the normal netif_receive_skb path | |
2247 | * is bypassed, for example because of VLAN acceleration. | |
2248 | */ | |
2249 | void netif_nit_deliver(struct sk_buff *skb) | |
2250 | { | |
2251 | struct packet_type *ptype; | |
2252 | ||
2253 | if (list_empty(&ptype_all)) | |
2254 | return; | |
2255 | ||
2256 | skb_reset_network_header(skb); | |
2257 | skb_reset_transport_header(skb); | |
2258 | skb->mac_len = skb->network_header - skb->mac_header; | |
2259 | ||
2260 | rcu_read_lock(); | |
2261 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
2262 | if (!ptype->dev || ptype->dev == skb->dev) | |
2263 | deliver_skb(skb, ptype, skb->dev); | |
2264 | } | |
2265 | rcu_read_unlock(); | |
2266 | } | |
2267 | ||
3b582cc1 SH |
2268 | /** |
2269 | * netif_receive_skb - process receive buffer from network | |
2270 | * @skb: buffer to process | |
2271 | * | |
2272 | * netif_receive_skb() is the main receive data processing function. | |
2273 | * It always succeeds. The buffer may be dropped during processing | |
2274 | * for congestion control or by the protocol layers. | |
2275 | * | |
2276 | * This function may only be called from softirq context and interrupts | |
2277 | * should be enabled. | |
2278 | * | |
2279 | * Return values (usually ignored): | |
2280 | * NET_RX_SUCCESS: no congestion | |
2281 | * NET_RX_DROP: packet was dropped | |
2282 | */ | |
1da177e4 LT |
2283 | int netif_receive_skb(struct sk_buff *skb) |
2284 | { | |
2285 | struct packet_type *ptype, *pt_prev; | |
f2ccd8fa | 2286 | struct net_device *orig_dev; |
0d7a3681 | 2287 | struct net_device *null_or_orig; |
1da177e4 | 2288 | int ret = NET_RX_DROP; |
252e3346 | 2289 | __be16 type; |
1da177e4 | 2290 | |
9b22ea56 PM |
2291 | if (skb->vlan_tci && vlan_hwaccel_do_receive(skb)) |
2292 | return NET_RX_SUCCESS; | |
2293 | ||
1da177e4 | 2294 | /* if we've gotten here through NAPI, check netpoll */ |
bea3348e | 2295 | if (netpoll_receive_skb(skb)) |
1da177e4 LT |
2296 | return NET_RX_DROP; |
2297 | ||
b7aa0bf7 | 2298 | if (!skb->tstamp.tv64) |
a61bbcf2 | 2299 | net_timestamp(skb); |
1da177e4 | 2300 | |
c01003c2 PM |
2301 | if (!skb->iif) |
2302 | skb->iif = skb->dev->ifindex; | |
86e65da9 | 2303 | |
0d7a3681 | 2304 | null_or_orig = NULL; |
cc9bd5ce JE |
2305 | orig_dev = skb->dev; |
2306 | if (orig_dev->master) { | |
0d7a3681 JE |
2307 | if (skb_bond_should_drop(skb)) |
2308 | null_or_orig = orig_dev; /* deliver only exact match */ | |
2309 | else | |
2310 | skb->dev = orig_dev->master; | |
cc9bd5ce | 2311 | } |
8f903c70 | 2312 | |
1da177e4 LT |
2313 | __get_cpu_var(netdev_rx_stat).total++; |
2314 | ||
c1d2bbe1 | 2315 | skb_reset_network_header(skb); |
badff6d0 | 2316 | skb_reset_transport_header(skb); |
b0e380b1 | 2317 | skb->mac_len = skb->network_header - skb->mac_header; |
1da177e4 LT |
2318 | |
2319 | pt_prev = NULL; | |
2320 | ||
2321 | rcu_read_lock(); | |
2322 | ||
2323 | #ifdef CONFIG_NET_CLS_ACT | |
2324 | if (skb->tc_verd & TC_NCLS) { | |
2325 | skb->tc_verd = CLR_TC_NCLS(skb->tc_verd); | |
2326 | goto ncls; | |
2327 | } | |
2328 | #endif | |
2329 | ||
2330 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
f982307f JE |
2331 | if (ptype->dev == null_or_orig || ptype->dev == skb->dev || |
2332 | ptype->dev == orig_dev) { | |
4ec93edb | 2333 | if (pt_prev) |
f2ccd8fa | 2334 | ret = deliver_skb(skb, pt_prev, orig_dev); |
1da177e4 LT |
2335 | pt_prev = ptype; |
2336 | } | |
2337 | } | |
2338 | ||
2339 | #ifdef CONFIG_NET_CLS_ACT | |
f697c3e8 HX |
2340 | skb = handle_ing(skb, &pt_prev, &ret, orig_dev); |
2341 | if (!skb) | |
1da177e4 | 2342 | goto out; |
1da177e4 LT |
2343 | ncls: |
2344 | #endif | |
2345 | ||
6229e362 | 2346 | skb = handle_bridge(skb, &pt_prev, &ret, orig_dev); |
b863ceb7 PM |
2347 | if (!skb) |
2348 | goto out; | |
2349 | skb = handle_macvlan(skb, &pt_prev, &ret, orig_dev); | |
6229e362 | 2350 | if (!skb) |
1da177e4 LT |
2351 | goto out; |
2352 | ||
2353 | type = skb->protocol; | |
82d8a867 PE |
2354 | list_for_each_entry_rcu(ptype, |
2355 | &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { | |
1da177e4 | 2356 | if (ptype->type == type && |
f982307f JE |
2357 | (ptype->dev == null_or_orig || ptype->dev == skb->dev || |
2358 | ptype->dev == orig_dev)) { | |
4ec93edb | 2359 | if (pt_prev) |
f2ccd8fa | 2360 | ret = deliver_skb(skb, pt_prev, orig_dev); |
1da177e4 LT |
2361 | pt_prev = ptype; |
2362 | } | |
2363 | } | |
2364 | ||
2365 | if (pt_prev) { | |
f2ccd8fa | 2366 | ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev); |
1da177e4 LT |
2367 | } else { |
2368 | kfree_skb(skb); | |
2369 | /* Jamal, now you will not able to escape explaining | |
2370 | * me how you were going to use this. :-) | |
2371 | */ | |
2372 | ret = NET_RX_DROP; | |
2373 | } | |
2374 | ||
2375 | out: | |
2376 | rcu_read_unlock(); | |
2377 | return ret; | |
2378 | } | |
d1b19dff | 2379 | EXPORT_SYMBOL(netif_receive_skb); |
1da177e4 | 2380 | |
6e583ce5 SH |
2381 | /* Network device is going away, flush any packets still pending */ |
2382 | static void flush_backlog(void *arg) | |
2383 | { | |
2384 | struct net_device *dev = arg; | |
2385 | struct softnet_data *queue = &__get_cpu_var(softnet_data); | |
2386 | struct sk_buff *skb, *tmp; | |
2387 | ||
2388 | skb_queue_walk_safe(&queue->input_pkt_queue, skb, tmp) | |
2389 | if (skb->dev == dev) { | |
2390 | __skb_unlink(skb, &queue->input_pkt_queue); | |
2391 | kfree_skb(skb); | |
2392 | } | |
2393 | } | |
2394 | ||
d565b0a1 HX |
2395 | static int napi_gro_complete(struct sk_buff *skb) |
2396 | { | |
2397 | struct packet_type *ptype; | |
2398 | __be16 type = skb->protocol; | |
2399 | struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK]; | |
2400 | int err = -ENOENT; | |
2401 | ||
fc59f9a3 HX |
2402 | if (NAPI_GRO_CB(skb)->count == 1) { |
2403 | skb_shinfo(skb)->gso_size = 0; | |
d565b0a1 | 2404 | goto out; |
fc59f9a3 | 2405 | } |
d565b0a1 HX |
2406 | |
2407 | rcu_read_lock(); | |
2408 | list_for_each_entry_rcu(ptype, head, list) { | |
2409 | if (ptype->type != type || ptype->dev || !ptype->gro_complete) | |
2410 | continue; | |
2411 | ||
2412 | err = ptype->gro_complete(skb); | |
2413 | break; | |
2414 | } | |
2415 | rcu_read_unlock(); | |
2416 | ||
2417 | if (err) { | |
2418 | WARN_ON(&ptype->list == head); | |
2419 | kfree_skb(skb); | |
2420 | return NET_RX_SUCCESS; | |
2421 | } | |
2422 | ||
2423 | out: | |
d565b0a1 HX |
2424 | return netif_receive_skb(skb); |
2425 | } | |
2426 | ||
2427 | void napi_gro_flush(struct napi_struct *napi) | |
2428 | { | |
2429 | struct sk_buff *skb, *next; | |
2430 | ||
2431 | for (skb = napi->gro_list; skb; skb = next) { | |
2432 | next = skb->next; | |
2433 | skb->next = NULL; | |
2434 | napi_gro_complete(skb); | |
2435 | } | |
2436 | ||
4ae5544f | 2437 | napi->gro_count = 0; |
d565b0a1 HX |
2438 | napi->gro_list = NULL; |
2439 | } | |
2440 | EXPORT_SYMBOL(napi_gro_flush); | |
2441 | ||
96e93eab | 2442 | int dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb) |
d565b0a1 HX |
2443 | { |
2444 | struct sk_buff **pp = NULL; | |
2445 | struct packet_type *ptype; | |
2446 | __be16 type = skb->protocol; | |
2447 | struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK]; | |
0da2afd5 | 2448 | int same_flow; |
d565b0a1 | 2449 | int mac_len; |
5d0d9be8 | 2450 | int ret; |
d565b0a1 HX |
2451 | |
2452 | if (!(skb->dev->features & NETIF_F_GRO)) | |
2453 | goto normal; | |
2454 | ||
4cf704fb | 2455 | if (skb_is_gso(skb) || skb_has_frags(skb)) |
f17f5c91 HX |
2456 | goto normal; |
2457 | ||
d565b0a1 HX |
2458 | rcu_read_lock(); |
2459 | list_for_each_entry_rcu(ptype, head, list) { | |
d565b0a1 HX |
2460 | if (ptype->type != type || ptype->dev || !ptype->gro_receive) |
2461 | continue; | |
2462 | ||
86911732 | 2463 | skb_set_network_header(skb, skb_gro_offset(skb)); |
d565b0a1 HX |
2464 | mac_len = skb->network_header - skb->mac_header; |
2465 | skb->mac_len = mac_len; | |
2466 | NAPI_GRO_CB(skb)->same_flow = 0; | |
2467 | NAPI_GRO_CB(skb)->flush = 0; | |
5d38a079 | 2468 | NAPI_GRO_CB(skb)->free = 0; |
d565b0a1 | 2469 | |
d565b0a1 HX |
2470 | pp = ptype->gro_receive(&napi->gro_list, skb); |
2471 | break; | |
2472 | } | |
2473 | rcu_read_unlock(); | |
2474 | ||
2475 | if (&ptype->list == head) | |
2476 | goto normal; | |
2477 | ||
0da2afd5 | 2478 | same_flow = NAPI_GRO_CB(skb)->same_flow; |
5d0d9be8 | 2479 | ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED; |
0da2afd5 | 2480 | |
d565b0a1 HX |
2481 | if (pp) { |
2482 | struct sk_buff *nskb = *pp; | |
2483 | ||
2484 | *pp = nskb->next; | |
2485 | nskb->next = NULL; | |
2486 | napi_gro_complete(nskb); | |
4ae5544f | 2487 | napi->gro_count--; |
d565b0a1 HX |
2488 | } |
2489 | ||
0da2afd5 | 2490 | if (same_flow) |
d565b0a1 HX |
2491 | goto ok; |
2492 | ||
4ae5544f | 2493 | if (NAPI_GRO_CB(skb)->flush || napi->gro_count >= MAX_GRO_SKBS) |
d565b0a1 | 2494 | goto normal; |
d565b0a1 | 2495 | |
4ae5544f | 2496 | napi->gro_count++; |
d565b0a1 | 2497 | NAPI_GRO_CB(skb)->count = 1; |
86911732 | 2498 | skb_shinfo(skb)->gso_size = skb_gro_len(skb); |
d565b0a1 HX |
2499 | skb->next = napi->gro_list; |
2500 | napi->gro_list = skb; | |
5d0d9be8 | 2501 | ret = GRO_HELD; |
d565b0a1 | 2502 | |
ad0f9904 | 2503 | pull: |
cb18978c HX |
2504 | if (skb_headlen(skb) < skb_gro_offset(skb)) { |
2505 | int grow = skb_gro_offset(skb) - skb_headlen(skb); | |
2506 | ||
2507 | BUG_ON(skb->end - skb->tail < grow); | |
2508 | ||
2509 | memcpy(skb_tail_pointer(skb), NAPI_GRO_CB(skb)->frag0, grow); | |
2510 | ||
2511 | skb->tail += grow; | |
2512 | skb->data_len -= grow; | |
2513 | ||
2514 | skb_shinfo(skb)->frags[0].page_offset += grow; | |
2515 | skb_shinfo(skb)->frags[0].size -= grow; | |
2516 | ||
2517 | if (unlikely(!skb_shinfo(skb)->frags[0].size)) { | |
2518 | put_page(skb_shinfo(skb)->frags[0].page); | |
2519 | memmove(skb_shinfo(skb)->frags, | |
2520 | skb_shinfo(skb)->frags + 1, | |
2521 | --skb_shinfo(skb)->nr_frags); | |
2522 | } | |
ad0f9904 HX |
2523 | } |
2524 | ||
d565b0a1 | 2525 | ok: |
5d0d9be8 | 2526 | return ret; |
d565b0a1 HX |
2527 | |
2528 | normal: | |
ad0f9904 HX |
2529 | ret = GRO_NORMAL; |
2530 | goto pull; | |
5d38a079 | 2531 | } |
96e93eab HX |
2532 | EXPORT_SYMBOL(dev_gro_receive); |
2533 | ||
2534 | static int __napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb) | |
2535 | { | |
2536 | struct sk_buff *p; | |
2537 | ||
d1c76af9 HX |
2538 | if (netpoll_rx_on(skb)) |
2539 | return GRO_NORMAL; | |
2540 | ||
96e93eab | 2541 | for (p = napi->gro_list; p; p = p->next) { |
f2bde732 SH |
2542 | NAPI_GRO_CB(p)->same_flow = (p->dev == skb->dev) |
2543 | && !compare_ether_header(skb_mac_header(p), | |
2544 | skb_gro_mac_header(skb)); | |
96e93eab HX |
2545 | NAPI_GRO_CB(p)->flush = 0; |
2546 | } | |
2547 | ||
2548 | return dev_gro_receive(napi, skb); | |
2549 | } | |
5d38a079 | 2550 | |
5d0d9be8 | 2551 | int napi_skb_finish(int ret, struct sk_buff *skb) |
5d38a079 | 2552 | { |
5d0d9be8 HX |
2553 | int err = NET_RX_SUCCESS; |
2554 | ||
2555 | switch (ret) { | |
2556 | case GRO_NORMAL: | |
5d38a079 HX |
2557 | return netif_receive_skb(skb); |
2558 | ||
5d0d9be8 HX |
2559 | case GRO_DROP: |
2560 | err = NET_RX_DROP; | |
2561 | /* fall through */ | |
2562 | ||
2563 | case GRO_MERGED_FREE: | |
5d38a079 HX |
2564 | kfree_skb(skb); |
2565 | break; | |
2566 | } | |
2567 | ||
5d0d9be8 HX |
2568 | return err; |
2569 | } | |
2570 | EXPORT_SYMBOL(napi_skb_finish); | |
2571 | ||
78a478d0 HX |
2572 | void skb_gro_reset_offset(struct sk_buff *skb) |
2573 | { | |
2574 | NAPI_GRO_CB(skb)->data_offset = 0; | |
2575 | NAPI_GRO_CB(skb)->frag0 = NULL; | |
7489594c | 2576 | NAPI_GRO_CB(skb)->frag0_len = 0; |
78a478d0 | 2577 | |
78d3fd0b | 2578 | if (skb->mac_header == skb->tail && |
7489594c | 2579 | !PageHighMem(skb_shinfo(skb)->frags[0].page)) { |
78a478d0 HX |
2580 | NAPI_GRO_CB(skb)->frag0 = |
2581 | page_address(skb_shinfo(skb)->frags[0].page) + | |
2582 | skb_shinfo(skb)->frags[0].page_offset; | |
7489594c HX |
2583 | NAPI_GRO_CB(skb)->frag0_len = skb_shinfo(skb)->frags[0].size; |
2584 | } | |
78a478d0 HX |
2585 | } |
2586 | EXPORT_SYMBOL(skb_gro_reset_offset); | |
2587 | ||
5d0d9be8 HX |
2588 | int napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb) |
2589 | { | |
86911732 HX |
2590 | skb_gro_reset_offset(skb); |
2591 | ||
5d0d9be8 | 2592 | return napi_skb_finish(__napi_gro_receive(napi, skb), skb); |
d565b0a1 HX |
2593 | } |
2594 | EXPORT_SYMBOL(napi_gro_receive); | |
2595 | ||
96e93eab HX |
2596 | void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb) |
2597 | { | |
96e93eab HX |
2598 | __skb_pull(skb, skb_headlen(skb)); |
2599 | skb_reserve(skb, NET_IP_ALIGN - skb_headroom(skb)); | |
2600 | ||
2601 | napi->skb = skb; | |
2602 | } | |
2603 | EXPORT_SYMBOL(napi_reuse_skb); | |
2604 | ||
76620aaf | 2605 | struct sk_buff *napi_get_frags(struct napi_struct *napi) |
5d38a079 HX |
2606 | { |
2607 | struct net_device *dev = napi->dev; | |
2608 | struct sk_buff *skb = napi->skb; | |
5d38a079 HX |
2609 | |
2610 | if (!skb) { | |
2611 | skb = netdev_alloc_skb(dev, GRO_MAX_HEAD + NET_IP_ALIGN); | |
2612 | if (!skb) | |
2613 | goto out; | |
2614 | ||
2615 | skb_reserve(skb, NET_IP_ALIGN); | |
80595d59 | 2616 | |
76620aaf | 2617 | napi->skb = skb; |
80595d59 | 2618 | } |
5d38a079 | 2619 | |
96e93eab HX |
2620 | out: |
2621 | return skb; | |
2622 | } | |
76620aaf | 2623 | EXPORT_SYMBOL(napi_get_frags); |
96e93eab | 2624 | |
5d0d9be8 | 2625 | int napi_frags_finish(struct napi_struct *napi, struct sk_buff *skb, int ret) |
96e93eab | 2626 | { |
5d0d9be8 | 2627 | int err = NET_RX_SUCCESS; |
96e93eab | 2628 | |
5d0d9be8 HX |
2629 | switch (ret) { |
2630 | case GRO_NORMAL: | |
86911732 | 2631 | case GRO_HELD: |
86911732 HX |
2632 | skb->protocol = eth_type_trans(skb, napi->dev); |
2633 | ||
2634 | if (ret == GRO_NORMAL) | |
2635 | return netif_receive_skb(skb); | |
2636 | ||
2637 | skb_gro_pull(skb, -ETH_HLEN); | |
2638 | break; | |
5d38a079 | 2639 | |
5d0d9be8 HX |
2640 | case GRO_DROP: |
2641 | err = NET_RX_DROP; | |
2642 | /* fall through */ | |
5d38a079 | 2643 | |
5d0d9be8 HX |
2644 | case GRO_MERGED_FREE: |
2645 | napi_reuse_skb(napi, skb); | |
2646 | break; | |
2647 | } | |
5d38a079 | 2648 | |
5d38a079 HX |
2649 | return err; |
2650 | } | |
5d0d9be8 HX |
2651 | EXPORT_SYMBOL(napi_frags_finish); |
2652 | ||
76620aaf HX |
2653 | struct sk_buff *napi_frags_skb(struct napi_struct *napi) |
2654 | { | |
2655 | struct sk_buff *skb = napi->skb; | |
2656 | struct ethhdr *eth; | |
a5b1cf28 HX |
2657 | unsigned int hlen; |
2658 | unsigned int off; | |
76620aaf HX |
2659 | |
2660 | napi->skb = NULL; | |
2661 | ||
2662 | skb_reset_mac_header(skb); | |
2663 | skb_gro_reset_offset(skb); | |
2664 | ||
a5b1cf28 HX |
2665 | off = skb_gro_offset(skb); |
2666 | hlen = off + sizeof(*eth); | |
2667 | eth = skb_gro_header_fast(skb, off); | |
2668 | if (skb_gro_header_hard(skb, hlen)) { | |
2669 | eth = skb_gro_header_slow(skb, hlen, off); | |
2670 | if (unlikely(!eth)) { | |
2671 | napi_reuse_skb(napi, skb); | |
2672 | skb = NULL; | |
2673 | goto out; | |
2674 | } | |
76620aaf HX |
2675 | } |
2676 | ||
2677 | skb_gro_pull(skb, sizeof(*eth)); | |
2678 | ||
2679 | /* | |
2680 | * This works because the only protocols we care about don't require | |
2681 | * special handling. We'll fix it up properly at the end. | |
2682 | */ | |
2683 | skb->protocol = eth->h_proto; | |
2684 | ||
2685 | out: | |
2686 | return skb; | |
2687 | } | |
2688 | EXPORT_SYMBOL(napi_frags_skb); | |
2689 | ||
2690 | int napi_gro_frags(struct napi_struct *napi) | |
5d0d9be8 | 2691 | { |
76620aaf | 2692 | struct sk_buff *skb = napi_frags_skb(napi); |
5d0d9be8 HX |
2693 | |
2694 | if (!skb) | |
2695 | return NET_RX_DROP; | |
2696 | ||
2697 | return napi_frags_finish(napi, skb, __napi_gro_receive(napi, skb)); | |
2698 | } | |
5d38a079 HX |
2699 | EXPORT_SYMBOL(napi_gro_frags); |
2700 | ||
bea3348e | 2701 | static int process_backlog(struct napi_struct *napi, int quota) |
1da177e4 LT |
2702 | { |
2703 | int work = 0; | |
1da177e4 LT |
2704 | struct softnet_data *queue = &__get_cpu_var(softnet_data); |
2705 | unsigned long start_time = jiffies; | |
2706 | ||
bea3348e SH |
2707 | napi->weight = weight_p; |
2708 | do { | |
1da177e4 | 2709 | struct sk_buff *skb; |
1da177e4 LT |
2710 | |
2711 | local_irq_disable(); | |
2712 | skb = __skb_dequeue(&queue->input_pkt_queue); | |
bea3348e | 2713 | if (!skb) { |
8f1ead2d | 2714 | __napi_complete(napi); |
bea3348e | 2715 | local_irq_enable(); |
8f1ead2d | 2716 | break; |
bea3348e | 2717 | } |
1da177e4 LT |
2718 | local_irq_enable(); |
2719 | ||
8f1ead2d | 2720 | netif_receive_skb(skb); |
bea3348e | 2721 | } while (++work < quota && jiffies == start_time); |
1da177e4 | 2722 | |
bea3348e SH |
2723 | return work; |
2724 | } | |
1da177e4 | 2725 | |
bea3348e SH |
2726 | /** |
2727 | * __napi_schedule - schedule for receive | |
c4ea43c5 | 2728 | * @n: entry to schedule |
bea3348e SH |
2729 | * |
2730 | * The entry's receive function will be scheduled to run | |
2731 | */ | |
b5606c2d | 2732 | void __napi_schedule(struct napi_struct *n) |
bea3348e SH |
2733 | { |
2734 | unsigned long flags; | |
1da177e4 | 2735 | |
bea3348e SH |
2736 | local_irq_save(flags); |
2737 | list_add_tail(&n->poll_list, &__get_cpu_var(softnet_data).poll_list); | |
2738 | __raise_softirq_irqoff(NET_RX_SOFTIRQ); | |
2739 | local_irq_restore(flags); | |
1da177e4 | 2740 | } |
bea3348e SH |
2741 | EXPORT_SYMBOL(__napi_schedule); |
2742 | ||
d565b0a1 HX |
2743 | void __napi_complete(struct napi_struct *n) |
2744 | { | |
2745 | BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); | |
2746 | BUG_ON(n->gro_list); | |
2747 | ||
2748 | list_del(&n->poll_list); | |
2749 | smp_mb__before_clear_bit(); | |
2750 | clear_bit(NAPI_STATE_SCHED, &n->state); | |
2751 | } | |
2752 | EXPORT_SYMBOL(__napi_complete); | |
2753 | ||
2754 | void napi_complete(struct napi_struct *n) | |
2755 | { | |
2756 | unsigned long flags; | |
2757 | ||
2758 | /* | |
2759 | * don't let napi dequeue from the cpu poll list | |
2760 | * just in case its running on a different cpu | |
2761 | */ | |
2762 | if (unlikely(test_bit(NAPI_STATE_NPSVC, &n->state))) | |
2763 | return; | |
2764 | ||
2765 | napi_gro_flush(n); | |
2766 | local_irq_save(flags); | |
2767 | __napi_complete(n); | |
2768 | local_irq_restore(flags); | |
2769 | } | |
2770 | EXPORT_SYMBOL(napi_complete); | |
2771 | ||
2772 | void netif_napi_add(struct net_device *dev, struct napi_struct *napi, | |
2773 | int (*poll)(struct napi_struct *, int), int weight) | |
2774 | { | |
2775 | INIT_LIST_HEAD(&napi->poll_list); | |
4ae5544f | 2776 | napi->gro_count = 0; |
d565b0a1 | 2777 | napi->gro_list = NULL; |
5d38a079 | 2778 | napi->skb = NULL; |
d565b0a1 HX |
2779 | napi->poll = poll; |
2780 | napi->weight = weight; | |
2781 | list_add(&napi->dev_list, &dev->napi_list); | |
d565b0a1 | 2782 | napi->dev = dev; |
5d38a079 | 2783 | #ifdef CONFIG_NETPOLL |
d565b0a1 HX |
2784 | spin_lock_init(&napi->poll_lock); |
2785 | napi->poll_owner = -1; | |
2786 | #endif | |
2787 | set_bit(NAPI_STATE_SCHED, &napi->state); | |
2788 | } | |
2789 | EXPORT_SYMBOL(netif_napi_add); | |
2790 | ||
2791 | void netif_napi_del(struct napi_struct *napi) | |
2792 | { | |
2793 | struct sk_buff *skb, *next; | |
2794 | ||
d7b06636 | 2795 | list_del_init(&napi->dev_list); |
76620aaf | 2796 | napi_free_frags(napi); |
d565b0a1 HX |
2797 | |
2798 | for (skb = napi->gro_list; skb; skb = next) { | |
2799 | next = skb->next; | |
2800 | skb->next = NULL; | |
2801 | kfree_skb(skb); | |
2802 | } | |
2803 | ||
2804 | napi->gro_list = NULL; | |
4ae5544f | 2805 | napi->gro_count = 0; |
d565b0a1 HX |
2806 | } |
2807 | EXPORT_SYMBOL(netif_napi_del); | |
2808 | ||
1da177e4 LT |
2809 | |
2810 | static void net_rx_action(struct softirq_action *h) | |
2811 | { | |
bea3348e | 2812 | struct list_head *list = &__get_cpu_var(softnet_data).poll_list; |
24f8b238 | 2813 | unsigned long time_limit = jiffies + 2; |
51b0bded | 2814 | int budget = netdev_budget; |
53fb95d3 MM |
2815 | void *have; |
2816 | ||
1da177e4 LT |
2817 | local_irq_disable(); |
2818 | ||
bea3348e SH |
2819 | while (!list_empty(list)) { |
2820 | struct napi_struct *n; | |
2821 | int work, weight; | |
1da177e4 | 2822 | |
bea3348e | 2823 | /* If softirq window is exhuasted then punt. |
24f8b238 SH |
2824 | * Allow this to run for 2 jiffies since which will allow |
2825 | * an average latency of 1.5/HZ. | |
bea3348e | 2826 | */ |
24f8b238 | 2827 | if (unlikely(budget <= 0 || time_after(jiffies, time_limit))) |
1da177e4 LT |
2828 | goto softnet_break; |
2829 | ||
2830 | local_irq_enable(); | |
2831 | ||
bea3348e SH |
2832 | /* Even though interrupts have been re-enabled, this |
2833 | * access is safe because interrupts can only add new | |
2834 | * entries to the tail of this list, and only ->poll() | |
2835 | * calls can remove this head entry from the list. | |
2836 | */ | |
2837 | n = list_entry(list->next, struct napi_struct, poll_list); | |
1da177e4 | 2838 | |
bea3348e SH |
2839 | have = netpoll_poll_lock(n); |
2840 | ||
2841 | weight = n->weight; | |
2842 | ||
0a7606c1 DM |
2843 | /* This NAPI_STATE_SCHED test is for avoiding a race |
2844 | * with netpoll's poll_napi(). Only the entity which | |
2845 | * obtains the lock and sees NAPI_STATE_SCHED set will | |
2846 | * actually make the ->poll() call. Therefore we avoid | |
2847 | * accidently calling ->poll() when NAPI is not scheduled. | |
2848 | */ | |
2849 | work = 0; | |
4ea7e386 | 2850 | if (test_bit(NAPI_STATE_SCHED, &n->state)) { |
0a7606c1 | 2851 | work = n->poll(n, weight); |
4ea7e386 NH |
2852 | trace_napi_poll(n); |
2853 | } | |
bea3348e SH |
2854 | |
2855 | WARN_ON_ONCE(work > weight); | |
2856 | ||
2857 | budget -= work; | |
2858 | ||
2859 | local_irq_disable(); | |
2860 | ||
2861 | /* Drivers must not modify the NAPI state if they | |
2862 | * consume the entire weight. In such cases this code | |
2863 | * still "owns" the NAPI instance and therefore can | |
2864 | * move the instance around on the list at-will. | |
2865 | */ | |
fed17f30 | 2866 | if (unlikely(work == weight)) { |
ff780cd8 HX |
2867 | if (unlikely(napi_disable_pending(n))) { |
2868 | local_irq_enable(); | |
2869 | napi_complete(n); | |
2870 | local_irq_disable(); | |
2871 | } else | |
fed17f30 DM |
2872 | list_move_tail(&n->poll_list, list); |
2873 | } | |
bea3348e SH |
2874 | |
2875 | netpoll_poll_unlock(have); | |
1da177e4 LT |
2876 | } |
2877 | out: | |
515e06c4 | 2878 | local_irq_enable(); |
bea3348e | 2879 | |
db217334 CL |
2880 | #ifdef CONFIG_NET_DMA |
2881 | /* | |
2882 | * There may not be any more sk_buffs coming right now, so push | |
2883 | * any pending DMA copies to hardware | |
2884 | */ | |
2ba05622 | 2885 | dma_issue_pending_all(); |
db217334 | 2886 | #endif |
bea3348e | 2887 | |
1da177e4 LT |
2888 | return; |
2889 | ||
2890 | softnet_break: | |
2891 | __get_cpu_var(netdev_rx_stat).time_squeeze++; | |
2892 | __raise_softirq_irqoff(NET_RX_SOFTIRQ); | |
2893 | goto out; | |
2894 | } | |
2895 | ||
d1b19dff | 2896 | static gifconf_func_t *gifconf_list[NPROTO]; |
1da177e4 LT |
2897 | |
2898 | /** | |
2899 | * register_gifconf - register a SIOCGIF handler | |
2900 | * @family: Address family | |
2901 | * @gifconf: Function handler | |
2902 | * | |
2903 | * Register protocol dependent address dumping routines. The handler | |
2904 | * that is passed must not be freed or reused until it has been replaced | |
2905 | * by another handler. | |
2906 | */ | |
d1b19dff | 2907 | int register_gifconf(unsigned int family, gifconf_func_t *gifconf) |
1da177e4 LT |
2908 | { |
2909 | if (family >= NPROTO) | |
2910 | return -EINVAL; | |
2911 | gifconf_list[family] = gifconf; | |
2912 | return 0; | |
2913 | } | |
d1b19dff | 2914 | EXPORT_SYMBOL(register_gifconf); |
1da177e4 LT |
2915 | |
2916 | ||
2917 | /* | |
2918 | * Map an interface index to its name (SIOCGIFNAME) | |
2919 | */ | |
2920 | ||
2921 | /* | |
2922 | * We need this ioctl for efficient implementation of the | |
2923 | * if_indextoname() function required by the IPv6 API. Without | |
2924 | * it, we would have to search all the interfaces to find a | |
2925 | * match. --pb | |
2926 | */ | |
2927 | ||
881d966b | 2928 | static int dev_ifname(struct net *net, struct ifreq __user *arg) |
1da177e4 LT |
2929 | { |
2930 | struct net_device *dev; | |
2931 | struct ifreq ifr; | |
2932 | ||
2933 | /* | |
2934 | * Fetch the caller's info block. | |
2935 | */ | |
2936 | ||
2937 | if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) | |
2938 | return -EFAULT; | |
2939 | ||
2940 | read_lock(&dev_base_lock); | |
881d966b | 2941 | dev = __dev_get_by_index(net, ifr.ifr_ifindex); |
1da177e4 LT |
2942 | if (!dev) { |
2943 | read_unlock(&dev_base_lock); | |
2944 | return -ENODEV; | |
2945 | } | |
2946 | ||
2947 | strcpy(ifr.ifr_name, dev->name); | |
2948 | read_unlock(&dev_base_lock); | |
2949 | ||
2950 | if (copy_to_user(arg, &ifr, sizeof(struct ifreq))) | |
2951 | return -EFAULT; | |
2952 | return 0; | |
2953 | } | |
2954 | ||
2955 | /* | |
2956 | * Perform a SIOCGIFCONF call. This structure will change | |
2957 | * size eventually, and there is nothing I can do about it. | |
2958 | * Thus we will need a 'compatibility mode'. | |
2959 | */ | |
2960 | ||
881d966b | 2961 | static int dev_ifconf(struct net *net, char __user *arg) |
1da177e4 LT |
2962 | { |
2963 | struct ifconf ifc; | |
2964 | struct net_device *dev; | |
2965 | char __user *pos; | |
2966 | int len; | |
2967 | int total; | |
2968 | int i; | |
2969 | ||
2970 | /* | |
2971 | * Fetch the caller's info block. | |
2972 | */ | |
2973 | ||
2974 | if (copy_from_user(&ifc, arg, sizeof(struct ifconf))) | |
2975 | return -EFAULT; | |
2976 | ||
2977 | pos = ifc.ifc_buf; | |
2978 | len = ifc.ifc_len; | |
2979 | ||
2980 | /* | |
2981 | * Loop over the interfaces, and write an info block for each. | |
2982 | */ | |
2983 | ||
2984 | total = 0; | |
881d966b | 2985 | for_each_netdev(net, dev) { |
1da177e4 LT |
2986 | for (i = 0; i < NPROTO; i++) { |
2987 | if (gifconf_list[i]) { | |
2988 | int done; | |
2989 | if (!pos) | |
2990 | done = gifconf_list[i](dev, NULL, 0); | |
2991 | else | |
2992 | done = gifconf_list[i](dev, pos + total, | |
2993 | len - total); | |
2994 | if (done < 0) | |
2995 | return -EFAULT; | |
2996 | total += done; | |
2997 | } | |
2998 | } | |
4ec93edb | 2999 | } |
1da177e4 LT |
3000 | |
3001 | /* | |
3002 | * All done. Write the updated control block back to the caller. | |
3003 | */ | |
3004 | ifc.ifc_len = total; | |
3005 | ||
3006 | /* | |
3007 | * Both BSD and Solaris return 0 here, so we do too. | |
3008 | */ | |
3009 | return copy_to_user(arg, &ifc, sizeof(struct ifconf)) ? -EFAULT : 0; | |
3010 | } | |
3011 | ||
3012 | #ifdef CONFIG_PROC_FS | |
3013 | /* | |
3014 | * This is invoked by the /proc filesystem handler to display a device | |
3015 | * in detail. | |
3016 | */ | |
7562f876 | 3017 | void *dev_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 3018 | __acquires(dev_base_lock) |
1da177e4 | 3019 | { |
e372c414 | 3020 | struct net *net = seq_file_net(seq); |
7562f876 | 3021 | loff_t off; |
1da177e4 | 3022 | struct net_device *dev; |
1da177e4 | 3023 | |
7562f876 PE |
3024 | read_lock(&dev_base_lock); |
3025 | if (!*pos) | |
3026 | return SEQ_START_TOKEN; | |
1da177e4 | 3027 | |
7562f876 | 3028 | off = 1; |
881d966b | 3029 | for_each_netdev(net, dev) |
7562f876 PE |
3030 | if (off++ == *pos) |
3031 | return dev; | |
1da177e4 | 3032 | |
7562f876 | 3033 | return NULL; |
1da177e4 LT |
3034 | } |
3035 | ||
3036 | void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3037 | { | |
e372c414 | 3038 | struct net *net = seq_file_net(seq); |
1da177e4 | 3039 | ++*pos; |
7562f876 | 3040 | return v == SEQ_START_TOKEN ? |
881d966b | 3041 | first_net_device(net) : next_net_device((struct net_device *)v); |
1da177e4 LT |
3042 | } |
3043 | ||
3044 | void dev_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 3045 | __releases(dev_base_lock) |
1da177e4 LT |
3046 | { |
3047 | read_unlock(&dev_base_lock); | |
3048 | } | |
3049 | ||
3050 | static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev) | |
3051 | { | |
eeda3fd6 | 3052 | const struct net_device_stats *stats = dev_get_stats(dev); |
1da177e4 | 3053 | |
5a1b5898 RR |
3054 | seq_printf(seq, "%6s:%8lu %7lu %4lu %4lu %4lu %5lu %10lu %9lu " |
3055 | "%8lu %7lu %4lu %4lu %4lu %5lu %7lu %10lu\n", | |
3056 | dev->name, stats->rx_bytes, stats->rx_packets, | |
3057 | stats->rx_errors, | |
3058 | stats->rx_dropped + stats->rx_missed_errors, | |
3059 | stats->rx_fifo_errors, | |
3060 | stats->rx_length_errors + stats->rx_over_errors + | |
3061 | stats->rx_crc_errors + stats->rx_frame_errors, | |
3062 | stats->rx_compressed, stats->multicast, | |
3063 | stats->tx_bytes, stats->tx_packets, | |
3064 | stats->tx_errors, stats->tx_dropped, | |
3065 | stats->tx_fifo_errors, stats->collisions, | |
3066 | stats->tx_carrier_errors + | |
3067 | stats->tx_aborted_errors + | |
3068 | stats->tx_window_errors + | |
3069 | stats->tx_heartbeat_errors, | |
3070 | stats->tx_compressed); | |
1da177e4 LT |
3071 | } |
3072 | ||
3073 | /* | |
3074 | * Called from the PROCfs module. This now uses the new arbitrary sized | |
3075 | * /proc/net interface to create /proc/net/dev | |
3076 | */ | |
3077 | static int dev_seq_show(struct seq_file *seq, void *v) | |
3078 | { | |
3079 | if (v == SEQ_START_TOKEN) | |
3080 | seq_puts(seq, "Inter-| Receive " | |
3081 | " | Transmit\n" | |
3082 | " face |bytes packets errs drop fifo frame " | |
3083 | "compressed multicast|bytes packets errs " | |
3084 | "drop fifo colls carrier compressed\n"); | |
3085 | else | |
3086 | dev_seq_printf_stats(seq, v); | |
3087 | return 0; | |
3088 | } | |
3089 | ||
3090 | static struct netif_rx_stats *softnet_get_online(loff_t *pos) | |
3091 | { | |
3092 | struct netif_rx_stats *rc = NULL; | |
3093 | ||
0c0b0aca | 3094 | while (*pos < nr_cpu_ids) |
4ec93edb | 3095 | if (cpu_online(*pos)) { |
1da177e4 LT |
3096 | rc = &per_cpu(netdev_rx_stat, *pos); |
3097 | break; | |
3098 | } else | |
3099 | ++*pos; | |
3100 | return rc; | |
3101 | } | |
3102 | ||
3103 | static void *softnet_seq_start(struct seq_file *seq, loff_t *pos) | |
3104 | { | |
3105 | return softnet_get_online(pos); | |
3106 | } | |
3107 | ||
3108 | static void *softnet_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3109 | { | |
3110 | ++*pos; | |
3111 | return softnet_get_online(pos); | |
3112 | } | |
3113 | ||
3114 | static void softnet_seq_stop(struct seq_file *seq, void *v) | |
3115 | { | |
3116 | } | |
3117 | ||
3118 | static int softnet_seq_show(struct seq_file *seq, void *v) | |
3119 | { | |
3120 | struct netif_rx_stats *s = v; | |
3121 | ||
3122 | seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n", | |
31aa02c5 | 3123 | s->total, s->dropped, s->time_squeeze, 0, |
c1ebcdb8 | 3124 | 0, 0, 0, 0, /* was fastroute */ |
d1b19dff | 3125 | s->cpu_collision); |
1da177e4 LT |
3126 | return 0; |
3127 | } | |
3128 | ||
f690808e | 3129 | static const struct seq_operations dev_seq_ops = { |
1da177e4 LT |
3130 | .start = dev_seq_start, |
3131 | .next = dev_seq_next, | |
3132 | .stop = dev_seq_stop, | |
3133 | .show = dev_seq_show, | |
3134 | }; | |
3135 | ||
3136 | static int dev_seq_open(struct inode *inode, struct file *file) | |
3137 | { | |
e372c414 DL |
3138 | return seq_open_net(inode, file, &dev_seq_ops, |
3139 | sizeof(struct seq_net_private)); | |
1da177e4 LT |
3140 | } |
3141 | ||
9a32144e | 3142 | static const struct file_operations dev_seq_fops = { |
1da177e4 LT |
3143 | .owner = THIS_MODULE, |
3144 | .open = dev_seq_open, | |
3145 | .read = seq_read, | |
3146 | .llseek = seq_lseek, | |
e372c414 | 3147 | .release = seq_release_net, |
1da177e4 LT |
3148 | }; |
3149 | ||
f690808e | 3150 | static const struct seq_operations softnet_seq_ops = { |
1da177e4 LT |
3151 | .start = softnet_seq_start, |
3152 | .next = softnet_seq_next, | |
3153 | .stop = softnet_seq_stop, | |
3154 | .show = softnet_seq_show, | |
3155 | }; | |
3156 | ||
3157 | static int softnet_seq_open(struct inode *inode, struct file *file) | |
3158 | { | |
3159 | return seq_open(file, &softnet_seq_ops); | |
3160 | } | |
3161 | ||
9a32144e | 3162 | static const struct file_operations softnet_seq_fops = { |
1da177e4 LT |
3163 | .owner = THIS_MODULE, |
3164 | .open = softnet_seq_open, | |
3165 | .read = seq_read, | |
3166 | .llseek = seq_lseek, | |
3167 | .release = seq_release, | |
3168 | }; | |
3169 | ||
0e1256ff SH |
3170 | static void *ptype_get_idx(loff_t pos) |
3171 | { | |
3172 | struct packet_type *pt = NULL; | |
3173 | loff_t i = 0; | |
3174 | int t; | |
3175 | ||
3176 | list_for_each_entry_rcu(pt, &ptype_all, list) { | |
3177 | if (i == pos) | |
3178 | return pt; | |
3179 | ++i; | |
3180 | } | |
3181 | ||
82d8a867 | 3182 | for (t = 0; t < PTYPE_HASH_SIZE; t++) { |
0e1256ff SH |
3183 | list_for_each_entry_rcu(pt, &ptype_base[t], list) { |
3184 | if (i == pos) | |
3185 | return pt; | |
3186 | ++i; | |
3187 | } | |
3188 | } | |
3189 | return NULL; | |
3190 | } | |
3191 | ||
3192 | static void *ptype_seq_start(struct seq_file *seq, loff_t *pos) | |
72348a42 | 3193 | __acquires(RCU) |
0e1256ff SH |
3194 | { |
3195 | rcu_read_lock(); | |
3196 | return *pos ? ptype_get_idx(*pos - 1) : SEQ_START_TOKEN; | |
3197 | } | |
3198 | ||
3199 | static void *ptype_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3200 | { | |
3201 | struct packet_type *pt; | |
3202 | struct list_head *nxt; | |
3203 | int hash; | |
3204 | ||
3205 | ++*pos; | |
3206 | if (v == SEQ_START_TOKEN) | |
3207 | return ptype_get_idx(0); | |
3208 | ||
3209 | pt = v; | |
3210 | nxt = pt->list.next; | |
3211 | if (pt->type == htons(ETH_P_ALL)) { | |
3212 | if (nxt != &ptype_all) | |
3213 | goto found; | |
3214 | hash = 0; | |
3215 | nxt = ptype_base[0].next; | |
3216 | } else | |
82d8a867 | 3217 | hash = ntohs(pt->type) & PTYPE_HASH_MASK; |
0e1256ff SH |
3218 | |
3219 | while (nxt == &ptype_base[hash]) { | |
82d8a867 | 3220 | if (++hash >= PTYPE_HASH_SIZE) |
0e1256ff SH |
3221 | return NULL; |
3222 | nxt = ptype_base[hash].next; | |
3223 | } | |
3224 | found: | |
3225 | return list_entry(nxt, struct packet_type, list); | |
3226 | } | |
3227 | ||
3228 | static void ptype_seq_stop(struct seq_file *seq, void *v) | |
72348a42 | 3229 | __releases(RCU) |
0e1256ff SH |
3230 | { |
3231 | rcu_read_unlock(); | |
3232 | } | |
3233 | ||
0e1256ff SH |
3234 | static int ptype_seq_show(struct seq_file *seq, void *v) |
3235 | { | |
3236 | struct packet_type *pt = v; | |
3237 | ||
3238 | if (v == SEQ_START_TOKEN) | |
3239 | seq_puts(seq, "Type Device Function\n"); | |
c346dca1 | 3240 | else if (pt->dev == NULL || dev_net(pt->dev) == seq_file_net(seq)) { |
0e1256ff SH |
3241 | if (pt->type == htons(ETH_P_ALL)) |
3242 | seq_puts(seq, "ALL "); | |
3243 | else | |
3244 | seq_printf(seq, "%04x", ntohs(pt->type)); | |
3245 | ||
908cd2da AD |
3246 | seq_printf(seq, " %-8s %pF\n", |
3247 | pt->dev ? pt->dev->name : "", pt->func); | |
0e1256ff SH |
3248 | } |
3249 | ||
3250 | return 0; | |
3251 | } | |
3252 | ||
3253 | static const struct seq_operations ptype_seq_ops = { | |
3254 | .start = ptype_seq_start, | |
3255 | .next = ptype_seq_next, | |
3256 | .stop = ptype_seq_stop, | |
3257 | .show = ptype_seq_show, | |
3258 | }; | |
3259 | ||
3260 | static int ptype_seq_open(struct inode *inode, struct file *file) | |
3261 | { | |
2feb27db PE |
3262 | return seq_open_net(inode, file, &ptype_seq_ops, |
3263 | sizeof(struct seq_net_private)); | |
0e1256ff SH |
3264 | } |
3265 | ||
3266 | static const struct file_operations ptype_seq_fops = { | |
3267 | .owner = THIS_MODULE, | |
3268 | .open = ptype_seq_open, | |
3269 | .read = seq_read, | |
3270 | .llseek = seq_lseek, | |
2feb27db | 3271 | .release = seq_release_net, |
0e1256ff SH |
3272 | }; |
3273 | ||
3274 | ||
4665079c | 3275 | static int __net_init dev_proc_net_init(struct net *net) |
1da177e4 LT |
3276 | { |
3277 | int rc = -ENOMEM; | |
3278 | ||
881d966b | 3279 | if (!proc_net_fops_create(net, "dev", S_IRUGO, &dev_seq_fops)) |
1da177e4 | 3280 | goto out; |
881d966b | 3281 | if (!proc_net_fops_create(net, "softnet_stat", S_IRUGO, &softnet_seq_fops)) |
1da177e4 | 3282 | goto out_dev; |
881d966b | 3283 | if (!proc_net_fops_create(net, "ptype", S_IRUGO, &ptype_seq_fops)) |
457c4cbc | 3284 | goto out_softnet; |
0e1256ff | 3285 | |
881d966b | 3286 | if (wext_proc_init(net)) |
457c4cbc | 3287 | goto out_ptype; |
1da177e4 LT |
3288 | rc = 0; |
3289 | out: | |
3290 | return rc; | |
457c4cbc | 3291 | out_ptype: |
881d966b | 3292 | proc_net_remove(net, "ptype"); |
1da177e4 | 3293 | out_softnet: |
881d966b | 3294 | proc_net_remove(net, "softnet_stat"); |
1da177e4 | 3295 | out_dev: |
881d966b | 3296 | proc_net_remove(net, "dev"); |
1da177e4 LT |
3297 | goto out; |
3298 | } | |
881d966b | 3299 | |
4665079c | 3300 | static void __net_exit dev_proc_net_exit(struct net *net) |
881d966b EB |
3301 | { |
3302 | wext_proc_exit(net); | |
3303 | ||
3304 | proc_net_remove(net, "ptype"); | |
3305 | proc_net_remove(net, "softnet_stat"); | |
3306 | proc_net_remove(net, "dev"); | |
3307 | } | |
3308 | ||
022cbae6 | 3309 | static struct pernet_operations __net_initdata dev_proc_ops = { |
881d966b EB |
3310 | .init = dev_proc_net_init, |
3311 | .exit = dev_proc_net_exit, | |
3312 | }; | |
3313 | ||
3314 | static int __init dev_proc_init(void) | |
3315 | { | |
3316 | return register_pernet_subsys(&dev_proc_ops); | |
3317 | } | |
1da177e4 LT |
3318 | #else |
3319 | #define dev_proc_init() 0 | |
3320 | #endif /* CONFIG_PROC_FS */ | |
3321 | ||
3322 | ||
3323 | /** | |
3324 | * netdev_set_master - set up master/slave pair | |
3325 | * @slave: slave device | |
3326 | * @master: new master device | |
3327 | * | |
3328 | * Changes the master device of the slave. Pass %NULL to break the | |
3329 | * bonding. The caller must hold the RTNL semaphore. On a failure | |
3330 | * a negative errno code is returned. On success the reference counts | |
3331 | * are adjusted, %RTM_NEWLINK is sent to the routing socket and the | |
3332 | * function returns zero. | |
3333 | */ | |
3334 | int netdev_set_master(struct net_device *slave, struct net_device *master) | |
3335 | { | |
3336 | struct net_device *old = slave->master; | |
3337 | ||
3338 | ASSERT_RTNL(); | |
3339 | ||
3340 | if (master) { | |
3341 | if (old) | |
3342 | return -EBUSY; | |
3343 | dev_hold(master); | |
3344 | } | |
3345 | ||
3346 | slave->master = master; | |
4ec93edb | 3347 | |
1da177e4 LT |
3348 | synchronize_net(); |
3349 | ||
3350 | if (old) | |
3351 | dev_put(old); | |
3352 | ||
3353 | if (master) | |
3354 | slave->flags |= IFF_SLAVE; | |
3355 | else | |
3356 | slave->flags &= ~IFF_SLAVE; | |
3357 | ||
3358 | rtmsg_ifinfo(RTM_NEWLINK, slave, IFF_SLAVE); | |
3359 | return 0; | |
3360 | } | |
d1b19dff | 3361 | EXPORT_SYMBOL(netdev_set_master); |
1da177e4 | 3362 | |
b6c40d68 PM |
3363 | static void dev_change_rx_flags(struct net_device *dev, int flags) |
3364 | { | |
d314774c SH |
3365 | const struct net_device_ops *ops = dev->netdev_ops; |
3366 | ||
3367 | if ((dev->flags & IFF_UP) && ops->ndo_change_rx_flags) | |
3368 | ops->ndo_change_rx_flags(dev, flags); | |
b6c40d68 PM |
3369 | } |
3370 | ||
dad9b335 | 3371 | static int __dev_set_promiscuity(struct net_device *dev, int inc) |
1da177e4 LT |
3372 | { |
3373 | unsigned short old_flags = dev->flags; | |
8192b0c4 DH |
3374 | uid_t uid; |
3375 | gid_t gid; | |
1da177e4 | 3376 | |
24023451 PM |
3377 | ASSERT_RTNL(); |
3378 | ||
dad9b335 WC |
3379 | dev->flags |= IFF_PROMISC; |
3380 | dev->promiscuity += inc; | |
3381 | if (dev->promiscuity == 0) { | |
3382 | /* | |
3383 | * Avoid overflow. | |
3384 | * If inc causes overflow, untouch promisc and return error. | |
3385 | */ | |
3386 | if (inc < 0) | |
3387 | dev->flags &= ~IFF_PROMISC; | |
3388 | else { | |
3389 | dev->promiscuity -= inc; | |
3390 | printk(KERN_WARNING "%s: promiscuity touches roof, " | |
3391 | "set promiscuity failed, promiscuity feature " | |
3392 | "of device might be broken.\n", dev->name); | |
3393 | return -EOVERFLOW; | |
3394 | } | |
3395 | } | |
52609c0b | 3396 | if (dev->flags != old_flags) { |
1da177e4 LT |
3397 | printk(KERN_INFO "device %s %s promiscuous mode\n", |
3398 | dev->name, (dev->flags & IFF_PROMISC) ? "entered" : | |
4ec93edb | 3399 | "left"); |
8192b0c4 DH |
3400 | if (audit_enabled) { |
3401 | current_uid_gid(&uid, &gid); | |
7759db82 KHK |
3402 | audit_log(current->audit_context, GFP_ATOMIC, |
3403 | AUDIT_ANOM_PROMISCUOUS, | |
3404 | "dev=%s prom=%d old_prom=%d auid=%u uid=%u gid=%u ses=%u", | |
3405 | dev->name, (dev->flags & IFF_PROMISC), | |
3406 | (old_flags & IFF_PROMISC), | |
3407 | audit_get_loginuid(current), | |
8192b0c4 | 3408 | uid, gid, |
7759db82 | 3409 | audit_get_sessionid(current)); |
8192b0c4 | 3410 | } |
24023451 | 3411 | |
b6c40d68 | 3412 | dev_change_rx_flags(dev, IFF_PROMISC); |
1da177e4 | 3413 | } |
dad9b335 | 3414 | return 0; |
1da177e4 LT |
3415 | } |
3416 | ||
4417da66 PM |
3417 | /** |
3418 | * dev_set_promiscuity - update promiscuity count on a device | |
3419 | * @dev: device | |
3420 | * @inc: modifier | |
3421 | * | |
3422 | * Add or remove promiscuity from a device. While the count in the device | |
3423 | * remains above zero the interface remains promiscuous. Once it hits zero | |
3424 | * the device reverts back to normal filtering operation. A negative inc | |
3425 | * value is used to drop promiscuity on the device. | |
dad9b335 | 3426 | * Return 0 if successful or a negative errno code on error. |
4417da66 | 3427 | */ |
dad9b335 | 3428 | int dev_set_promiscuity(struct net_device *dev, int inc) |
4417da66 PM |
3429 | { |
3430 | unsigned short old_flags = dev->flags; | |
dad9b335 | 3431 | int err; |
4417da66 | 3432 | |
dad9b335 | 3433 | err = __dev_set_promiscuity(dev, inc); |
4b5a698e | 3434 | if (err < 0) |
dad9b335 | 3435 | return err; |
4417da66 PM |
3436 | if (dev->flags != old_flags) |
3437 | dev_set_rx_mode(dev); | |
dad9b335 | 3438 | return err; |
4417da66 | 3439 | } |
d1b19dff | 3440 | EXPORT_SYMBOL(dev_set_promiscuity); |
4417da66 | 3441 | |
1da177e4 LT |
3442 | /** |
3443 | * dev_set_allmulti - update allmulti count on a device | |
3444 | * @dev: device | |
3445 | * @inc: modifier | |
3446 | * | |
3447 | * Add or remove reception of all multicast frames to a device. While the | |
3448 | * count in the device remains above zero the interface remains listening | |
3449 | * to all interfaces. Once it hits zero the device reverts back to normal | |
3450 | * filtering operation. A negative @inc value is used to drop the counter | |
3451 | * when releasing a resource needing all multicasts. | |
dad9b335 | 3452 | * Return 0 if successful or a negative errno code on error. |
1da177e4 LT |
3453 | */ |
3454 | ||
dad9b335 | 3455 | int dev_set_allmulti(struct net_device *dev, int inc) |
1da177e4 LT |
3456 | { |
3457 | unsigned short old_flags = dev->flags; | |
3458 | ||
24023451 PM |
3459 | ASSERT_RTNL(); |
3460 | ||
1da177e4 | 3461 | dev->flags |= IFF_ALLMULTI; |
dad9b335 WC |
3462 | dev->allmulti += inc; |
3463 | if (dev->allmulti == 0) { | |
3464 | /* | |
3465 | * Avoid overflow. | |
3466 | * If inc causes overflow, untouch allmulti and return error. | |
3467 | */ | |
3468 | if (inc < 0) | |
3469 | dev->flags &= ~IFF_ALLMULTI; | |
3470 | else { | |
3471 | dev->allmulti -= inc; | |
3472 | printk(KERN_WARNING "%s: allmulti touches roof, " | |
3473 | "set allmulti failed, allmulti feature of " | |
3474 | "device might be broken.\n", dev->name); | |
3475 | return -EOVERFLOW; | |
3476 | } | |
3477 | } | |
24023451 | 3478 | if (dev->flags ^ old_flags) { |
b6c40d68 | 3479 | dev_change_rx_flags(dev, IFF_ALLMULTI); |
4417da66 | 3480 | dev_set_rx_mode(dev); |
24023451 | 3481 | } |
dad9b335 | 3482 | return 0; |
4417da66 | 3483 | } |
d1b19dff | 3484 | EXPORT_SYMBOL(dev_set_allmulti); |
4417da66 PM |
3485 | |
3486 | /* | |
3487 | * Upload unicast and multicast address lists to device and | |
3488 | * configure RX filtering. When the device doesn't support unicast | |
53ccaae1 | 3489 | * filtering it is put in promiscuous mode while unicast addresses |
4417da66 PM |
3490 | * are present. |
3491 | */ | |
3492 | void __dev_set_rx_mode(struct net_device *dev) | |
3493 | { | |
d314774c SH |
3494 | const struct net_device_ops *ops = dev->netdev_ops; |
3495 | ||
4417da66 PM |
3496 | /* dev_open will call this function so the list will stay sane. */ |
3497 | if (!(dev->flags&IFF_UP)) | |
3498 | return; | |
3499 | ||
3500 | if (!netif_device_present(dev)) | |
40b77c94 | 3501 | return; |
4417da66 | 3502 | |
d314774c SH |
3503 | if (ops->ndo_set_rx_mode) |
3504 | ops->ndo_set_rx_mode(dev); | |
4417da66 PM |
3505 | else { |
3506 | /* Unicast addresses changes may only happen under the rtnl, | |
3507 | * therefore calling __dev_set_promiscuity here is safe. | |
3508 | */ | |
31278e71 | 3509 | if (dev->uc.count > 0 && !dev->uc_promisc) { |
4417da66 PM |
3510 | __dev_set_promiscuity(dev, 1); |
3511 | dev->uc_promisc = 1; | |
31278e71 | 3512 | } else if (dev->uc.count == 0 && dev->uc_promisc) { |
4417da66 PM |
3513 | __dev_set_promiscuity(dev, -1); |
3514 | dev->uc_promisc = 0; | |
3515 | } | |
3516 | ||
d314774c SH |
3517 | if (ops->ndo_set_multicast_list) |
3518 | ops->ndo_set_multicast_list(dev); | |
4417da66 PM |
3519 | } |
3520 | } | |
3521 | ||
3522 | void dev_set_rx_mode(struct net_device *dev) | |
3523 | { | |
b9e40857 | 3524 | netif_addr_lock_bh(dev); |
4417da66 | 3525 | __dev_set_rx_mode(dev); |
b9e40857 | 3526 | netif_addr_unlock_bh(dev); |
1da177e4 LT |
3527 | } |
3528 | ||
f001fde5 JP |
3529 | /* hw addresses list handling functions */ |
3530 | ||
31278e71 JP |
3531 | static int __hw_addr_add(struct netdev_hw_addr_list *list, unsigned char *addr, |
3532 | int addr_len, unsigned char addr_type) | |
f001fde5 JP |
3533 | { |
3534 | struct netdev_hw_addr *ha; | |
3535 | int alloc_size; | |
3536 | ||
3537 | if (addr_len > MAX_ADDR_LEN) | |
3538 | return -EINVAL; | |
3539 | ||
31278e71 | 3540 | list_for_each_entry(ha, &list->list, list) { |
ccffad25 JP |
3541 | if (!memcmp(ha->addr, addr, addr_len) && |
3542 | ha->type == addr_type) { | |
3543 | ha->refcount++; | |
3544 | return 0; | |
3545 | } | |
3546 | } | |
3547 | ||
3548 | ||
f001fde5 JP |
3549 | alloc_size = sizeof(*ha); |
3550 | if (alloc_size < L1_CACHE_BYTES) | |
3551 | alloc_size = L1_CACHE_BYTES; | |
3552 | ha = kmalloc(alloc_size, GFP_ATOMIC); | |
3553 | if (!ha) | |
3554 | return -ENOMEM; | |
3555 | memcpy(ha->addr, addr, addr_len); | |
3556 | ha->type = addr_type; | |
ccffad25 JP |
3557 | ha->refcount = 1; |
3558 | ha->synced = false; | |
31278e71 JP |
3559 | list_add_tail_rcu(&ha->list, &list->list); |
3560 | list->count++; | |
f001fde5 JP |
3561 | return 0; |
3562 | } | |
3563 | ||
3564 | static void ha_rcu_free(struct rcu_head *head) | |
3565 | { | |
3566 | struct netdev_hw_addr *ha; | |
3567 | ||
3568 | ha = container_of(head, struct netdev_hw_addr, rcu_head); | |
3569 | kfree(ha); | |
3570 | } | |
3571 | ||
31278e71 JP |
3572 | static int __hw_addr_del(struct netdev_hw_addr_list *list, unsigned char *addr, |
3573 | int addr_len, unsigned char addr_type) | |
f001fde5 JP |
3574 | { |
3575 | struct netdev_hw_addr *ha; | |
f001fde5 | 3576 | |
31278e71 | 3577 | list_for_each_entry(ha, &list->list, list) { |
ccffad25 | 3578 | if (!memcmp(ha->addr, addr, addr_len) && |
f001fde5 | 3579 | (ha->type == addr_type || !addr_type)) { |
ccffad25 JP |
3580 | if (--ha->refcount) |
3581 | return 0; | |
f001fde5 JP |
3582 | list_del_rcu(&ha->list); |
3583 | call_rcu(&ha->rcu_head, ha_rcu_free); | |
31278e71 | 3584 | list->count--; |
f001fde5 JP |
3585 | return 0; |
3586 | } | |
3587 | } | |
3588 | return -ENOENT; | |
3589 | } | |
3590 | ||
31278e71 JP |
3591 | static int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list, |
3592 | struct netdev_hw_addr_list *from_list, | |
3593 | int addr_len, | |
ccffad25 | 3594 | unsigned char addr_type) |
f001fde5 JP |
3595 | { |
3596 | int err; | |
3597 | struct netdev_hw_addr *ha, *ha2; | |
3598 | unsigned char type; | |
3599 | ||
31278e71 | 3600 | list_for_each_entry(ha, &from_list->list, list) { |
f001fde5 | 3601 | type = addr_type ? addr_type : ha->type; |
31278e71 | 3602 | err = __hw_addr_add(to_list, ha->addr, addr_len, type); |
f001fde5 JP |
3603 | if (err) |
3604 | goto unroll; | |
3605 | } | |
3606 | return 0; | |
3607 | ||
3608 | unroll: | |
31278e71 | 3609 | list_for_each_entry(ha2, &from_list->list, list) { |
f001fde5 JP |
3610 | if (ha2 == ha) |
3611 | break; | |
3612 | type = addr_type ? addr_type : ha2->type; | |
31278e71 | 3613 | __hw_addr_del(to_list, ha2->addr, addr_len, type); |
f001fde5 JP |
3614 | } |
3615 | return err; | |
3616 | } | |
3617 | ||
31278e71 JP |
3618 | static void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list, |
3619 | struct netdev_hw_addr_list *from_list, | |
3620 | int addr_len, | |
ccffad25 | 3621 | unsigned char addr_type) |
f001fde5 JP |
3622 | { |
3623 | struct netdev_hw_addr *ha; | |
3624 | unsigned char type; | |
3625 | ||
31278e71 | 3626 | list_for_each_entry(ha, &from_list->list, list) { |
f001fde5 | 3627 | type = addr_type ? addr_type : ha->type; |
31278e71 | 3628 | __hw_addr_del(to_list, ha->addr, addr_len, addr_type); |
ccffad25 JP |
3629 | } |
3630 | } | |
3631 | ||
31278e71 JP |
3632 | static int __hw_addr_sync(struct netdev_hw_addr_list *to_list, |
3633 | struct netdev_hw_addr_list *from_list, | |
ccffad25 JP |
3634 | int addr_len) |
3635 | { | |
3636 | int err = 0; | |
3637 | struct netdev_hw_addr *ha, *tmp; | |
3638 | ||
31278e71 | 3639 | list_for_each_entry_safe(ha, tmp, &from_list->list, list) { |
ccffad25 | 3640 | if (!ha->synced) { |
31278e71 | 3641 | err = __hw_addr_add(to_list, ha->addr, |
ccffad25 JP |
3642 | addr_len, ha->type); |
3643 | if (err) | |
3644 | break; | |
3645 | ha->synced = true; | |
3646 | ha->refcount++; | |
3647 | } else if (ha->refcount == 1) { | |
31278e71 JP |
3648 | __hw_addr_del(to_list, ha->addr, addr_len, ha->type); |
3649 | __hw_addr_del(from_list, ha->addr, addr_len, ha->type); | |
ccffad25 | 3650 | } |
f001fde5 | 3651 | } |
ccffad25 | 3652 | return err; |
f001fde5 JP |
3653 | } |
3654 | ||
31278e71 JP |
3655 | static void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, |
3656 | struct netdev_hw_addr_list *from_list, | |
ccffad25 JP |
3657 | int addr_len) |
3658 | { | |
3659 | struct netdev_hw_addr *ha, *tmp; | |
3660 | ||
31278e71 | 3661 | list_for_each_entry_safe(ha, tmp, &from_list->list, list) { |
ccffad25 | 3662 | if (ha->synced) { |
31278e71 | 3663 | __hw_addr_del(to_list, ha->addr, |
ccffad25 JP |
3664 | addr_len, ha->type); |
3665 | ha->synced = false; | |
31278e71 | 3666 | __hw_addr_del(from_list, ha->addr, |
ccffad25 JP |
3667 | addr_len, ha->type); |
3668 | } | |
3669 | } | |
3670 | } | |
3671 | ||
31278e71 | 3672 | static void __hw_addr_flush(struct netdev_hw_addr_list *list) |
f001fde5 JP |
3673 | { |
3674 | struct netdev_hw_addr *ha, *tmp; | |
3675 | ||
31278e71 | 3676 | list_for_each_entry_safe(ha, tmp, &list->list, list) { |
f001fde5 JP |
3677 | list_del_rcu(&ha->list); |
3678 | call_rcu(&ha->rcu_head, ha_rcu_free); | |
3679 | } | |
31278e71 JP |
3680 | list->count = 0; |
3681 | } | |
3682 | ||
3683 | static void __hw_addr_init(struct netdev_hw_addr_list *list) | |
3684 | { | |
3685 | INIT_LIST_HEAD(&list->list); | |
3686 | list->count = 0; | |
f001fde5 JP |
3687 | } |
3688 | ||
3689 | /* Device addresses handling functions */ | |
3690 | ||
3691 | static void dev_addr_flush(struct net_device *dev) | |
3692 | { | |
3693 | /* rtnl_mutex must be held here */ | |
3694 | ||
31278e71 | 3695 | __hw_addr_flush(&dev->dev_addrs); |
f001fde5 JP |
3696 | dev->dev_addr = NULL; |
3697 | } | |
3698 | ||
3699 | static int dev_addr_init(struct net_device *dev) | |
3700 | { | |
3701 | unsigned char addr[MAX_ADDR_LEN]; | |
3702 | struct netdev_hw_addr *ha; | |
3703 | int err; | |
3704 | ||
3705 | /* rtnl_mutex must be held here */ | |
3706 | ||
31278e71 | 3707 | __hw_addr_init(&dev->dev_addrs); |
0c27922e | 3708 | memset(addr, 0, sizeof(addr)); |
31278e71 | 3709 | err = __hw_addr_add(&dev->dev_addrs, addr, sizeof(addr), |
f001fde5 JP |
3710 | NETDEV_HW_ADDR_T_LAN); |
3711 | if (!err) { | |
3712 | /* | |
3713 | * Get the first (previously created) address from the list | |
3714 | * and set dev_addr pointer to this location. | |
3715 | */ | |
31278e71 | 3716 | ha = list_first_entry(&dev->dev_addrs.list, |
f001fde5 JP |
3717 | struct netdev_hw_addr, list); |
3718 | dev->dev_addr = ha->addr; | |
3719 | } | |
3720 | return err; | |
3721 | } | |
3722 | ||
3723 | /** | |
3724 | * dev_addr_add - Add a device address | |
3725 | * @dev: device | |
3726 | * @addr: address to add | |
3727 | * @addr_type: address type | |
3728 | * | |
3729 | * Add a device address to the device or increase the reference count if | |
3730 | * it already exists. | |
3731 | * | |
3732 | * The caller must hold the rtnl_mutex. | |
3733 | */ | |
3734 | int dev_addr_add(struct net_device *dev, unsigned char *addr, | |
3735 | unsigned char addr_type) | |
3736 | { | |
3737 | int err; | |
3738 | ||
3739 | ASSERT_RTNL(); | |
3740 | ||
31278e71 | 3741 | err = __hw_addr_add(&dev->dev_addrs, addr, dev->addr_len, addr_type); |
f001fde5 JP |
3742 | if (!err) |
3743 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); | |
3744 | return err; | |
3745 | } | |
3746 | EXPORT_SYMBOL(dev_addr_add); | |
3747 | ||
3748 | /** | |
3749 | * dev_addr_del - Release a device address. | |
3750 | * @dev: device | |
3751 | * @addr: address to delete | |
3752 | * @addr_type: address type | |
3753 | * | |
3754 | * Release reference to a device address and remove it from the device | |
3755 | * if the reference count drops to zero. | |
3756 | * | |
3757 | * The caller must hold the rtnl_mutex. | |
3758 | */ | |
3759 | int dev_addr_del(struct net_device *dev, unsigned char *addr, | |
3760 | unsigned char addr_type) | |
3761 | { | |
3762 | int err; | |
ccffad25 | 3763 | struct netdev_hw_addr *ha; |
f001fde5 JP |
3764 | |
3765 | ASSERT_RTNL(); | |
3766 | ||
ccffad25 JP |
3767 | /* |
3768 | * We can not remove the first address from the list because | |
3769 | * dev->dev_addr points to that. | |
3770 | */ | |
31278e71 JP |
3771 | ha = list_first_entry(&dev->dev_addrs.list, |
3772 | struct netdev_hw_addr, list); | |
ccffad25 JP |
3773 | if (ha->addr == dev->dev_addr && ha->refcount == 1) |
3774 | return -ENOENT; | |
3775 | ||
31278e71 | 3776 | err = __hw_addr_del(&dev->dev_addrs, addr, dev->addr_len, |
ccffad25 | 3777 | addr_type); |
f001fde5 JP |
3778 | if (!err) |
3779 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); | |
3780 | return err; | |
3781 | } | |
3782 | EXPORT_SYMBOL(dev_addr_del); | |
3783 | ||
3784 | /** | |
3785 | * dev_addr_add_multiple - Add device addresses from another device | |
3786 | * @to_dev: device to which addresses will be added | |
3787 | * @from_dev: device from which addresses will be added | |
3788 | * @addr_type: address type - 0 means type will be used from from_dev | |
3789 | * | |
3790 | * Add device addresses of the one device to another. | |
3791 | ** | |
3792 | * The caller must hold the rtnl_mutex. | |
3793 | */ | |
3794 | int dev_addr_add_multiple(struct net_device *to_dev, | |
3795 | struct net_device *from_dev, | |
3796 | unsigned char addr_type) | |
3797 | { | |
3798 | int err; | |
3799 | ||
3800 | ASSERT_RTNL(); | |
3801 | ||
3802 | if (from_dev->addr_len != to_dev->addr_len) | |
3803 | return -EINVAL; | |
31278e71 | 3804 | err = __hw_addr_add_multiple(&to_dev->dev_addrs, &from_dev->dev_addrs, |
ccffad25 | 3805 | to_dev->addr_len, addr_type); |
f001fde5 JP |
3806 | if (!err) |
3807 | call_netdevice_notifiers(NETDEV_CHANGEADDR, to_dev); | |
3808 | return err; | |
3809 | } | |
3810 | EXPORT_SYMBOL(dev_addr_add_multiple); | |
3811 | ||
3812 | /** | |
3813 | * dev_addr_del_multiple - Delete device addresses by another device | |
3814 | * @to_dev: device where the addresses will be deleted | |
3815 | * @from_dev: device by which addresses the addresses will be deleted | |
3816 | * @addr_type: address type - 0 means type will used from from_dev | |
3817 | * | |
3818 | * Deletes addresses in to device by the list of addresses in from device. | |
3819 | * | |
3820 | * The caller must hold the rtnl_mutex. | |
3821 | */ | |
3822 | int dev_addr_del_multiple(struct net_device *to_dev, | |
3823 | struct net_device *from_dev, | |
3824 | unsigned char addr_type) | |
3825 | { | |
3826 | ASSERT_RTNL(); | |
3827 | ||
3828 | if (from_dev->addr_len != to_dev->addr_len) | |
3829 | return -EINVAL; | |
31278e71 | 3830 | __hw_addr_del_multiple(&to_dev->dev_addrs, &from_dev->dev_addrs, |
ccffad25 | 3831 | to_dev->addr_len, addr_type); |
f001fde5 JP |
3832 | call_netdevice_notifiers(NETDEV_CHANGEADDR, to_dev); |
3833 | return 0; | |
3834 | } | |
3835 | EXPORT_SYMBOL(dev_addr_del_multiple); | |
3836 | ||
31278e71 | 3837 | /* multicast addresses handling functions */ |
f001fde5 | 3838 | |
61cbc2fc PM |
3839 | int __dev_addr_delete(struct dev_addr_list **list, int *count, |
3840 | void *addr, int alen, int glbl) | |
bf742482 PM |
3841 | { |
3842 | struct dev_addr_list *da; | |
3843 | ||
3844 | for (; (da = *list) != NULL; list = &da->next) { | |
3845 | if (memcmp(da->da_addr, addr, da->da_addrlen) == 0 && | |
3846 | alen == da->da_addrlen) { | |
3847 | if (glbl) { | |
3848 | int old_glbl = da->da_gusers; | |
3849 | da->da_gusers = 0; | |
3850 | if (old_glbl == 0) | |
3851 | break; | |
3852 | } | |
3853 | if (--da->da_users) | |
3854 | return 0; | |
3855 | ||
3856 | *list = da->next; | |
3857 | kfree(da); | |
61cbc2fc | 3858 | (*count)--; |
bf742482 PM |
3859 | return 0; |
3860 | } | |
3861 | } | |
3862 | return -ENOENT; | |
3863 | } | |
3864 | ||
61cbc2fc PM |
3865 | int __dev_addr_add(struct dev_addr_list **list, int *count, |
3866 | void *addr, int alen, int glbl) | |
bf742482 PM |
3867 | { |
3868 | struct dev_addr_list *da; | |
3869 | ||
3870 | for (da = *list; da != NULL; da = da->next) { | |
3871 | if (memcmp(da->da_addr, addr, da->da_addrlen) == 0 && | |
3872 | da->da_addrlen == alen) { | |
3873 | if (glbl) { | |
3874 | int old_glbl = da->da_gusers; | |
3875 | da->da_gusers = 1; | |
3876 | if (old_glbl) | |
3877 | return 0; | |
3878 | } | |
3879 | da->da_users++; | |
3880 | return 0; | |
3881 | } | |
3882 | } | |
3883 | ||
12aa343a | 3884 | da = kzalloc(sizeof(*da), GFP_ATOMIC); |
bf742482 PM |
3885 | if (da == NULL) |
3886 | return -ENOMEM; | |
3887 | memcpy(da->da_addr, addr, alen); | |
3888 | da->da_addrlen = alen; | |
3889 | da->da_users = 1; | |
3890 | da->da_gusers = glbl ? 1 : 0; | |
3891 | da->next = *list; | |
3892 | *list = da; | |
61cbc2fc | 3893 | (*count)++; |
bf742482 PM |
3894 | return 0; |
3895 | } | |
3896 | ||
4417da66 PM |
3897 | /** |
3898 | * dev_unicast_delete - Release secondary unicast address. | |
3899 | * @dev: device | |
0ed72ec4 | 3900 | * @addr: address to delete |
4417da66 PM |
3901 | * |
3902 | * Release reference to a secondary unicast address and remove it | |
0ed72ec4 | 3903 | * from the device if the reference count drops to zero. |
4417da66 PM |
3904 | * |
3905 | * The caller must hold the rtnl_mutex. | |
3906 | */ | |
ccffad25 | 3907 | int dev_unicast_delete(struct net_device *dev, void *addr) |
4417da66 PM |
3908 | { |
3909 | int err; | |
3910 | ||
3911 | ASSERT_RTNL(); | |
3912 | ||
a6ac65db | 3913 | netif_addr_lock_bh(dev); |
31278e71 JP |
3914 | err = __hw_addr_del(&dev->uc, addr, dev->addr_len, |
3915 | NETDEV_HW_ADDR_T_UNICAST); | |
61cbc2fc | 3916 | if (!err) |
4417da66 | 3917 | __dev_set_rx_mode(dev); |
a6ac65db | 3918 | netif_addr_unlock_bh(dev); |
4417da66 PM |
3919 | return err; |
3920 | } | |
3921 | EXPORT_SYMBOL(dev_unicast_delete); | |
3922 | ||
3923 | /** | |
3924 | * dev_unicast_add - add a secondary unicast address | |
3925 | * @dev: device | |
5dbaec5d | 3926 | * @addr: address to add |
4417da66 PM |
3927 | * |
3928 | * Add a secondary unicast address to the device or increase | |
3929 | * the reference count if it already exists. | |
3930 | * | |
3931 | * The caller must hold the rtnl_mutex. | |
3932 | */ | |
ccffad25 | 3933 | int dev_unicast_add(struct net_device *dev, void *addr) |
4417da66 PM |
3934 | { |
3935 | int err; | |
3936 | ||
3937 | ASSERT_RTNL(); | |
3938 | ||
a6ac65db | 3939 | netif_addr_lock_bh(dev); |
31278e71 JP |
3940 | err = __hw_addr_add(&dev->uc, addr, dev->addr_len, |
3941 | NETDEV_HW_ADDR_T_UNICAST); | |
61cbc2fc | 3942 | if (!err) |
4417da66 | 3943 | __dev_set_rx_mode(dev); |
a6ac65db | 3944 | netif_addr_unlock_bh(dev); |
4417da66 PM |
3945 | return err; |
3946 | } | |
3947 | EXPORT_SYMBOL(dev_unicast_add); | |
3948 | ||
e83a2ea8 CL |
3949 | int __dev_addr_sync(struct dev_addr_list **to, int *to_count, |
3950 | struct dev_addr_list **from, int *from_count) | |
3951 | { | |
3952 | struct dev_addr_list *da, *next; | |
3953 | int err = 0; | |
3954 | ||
3955 | da = *from; | |
3956 | while (da != NULL) { | |
3957 | next = da->next; | |
3958 | if (!da->da_synced) { | |
3959 | err = __dev_addr_add(to, to_count, | |
3960 | da->da_addr, da->da_addrlen, 0); | |
3961 | if (err < 0) | |
3962 | break; | |
3963 | da->da_synced = 1; | |
3964 | da->da_users++; | |
3965 | } else if (da->da_users == 1) { | |
3966 | __dev_addr_delete(to, to_count, | |
3967 | da->da_addr, da->da_addrlen, 0); | |
3968 | __dev_addr_delete(from, from_count, | |
3969 | da->da_addr, da->da_addrlen, 0); | |
3970 | } | |
3971 | da = next; | |
3972 | } | |
3973 | return err; | |
3974 | } | |
c4029083 | 3975 | EXPORT_SYMBOL_GPL(__dev_addr_sync); |
e83a2ea8 CL |
3976 | |
3977 | void __dev_addr_unsync(struct dev_addr_list **to, int *to_count, | |
3978 | struct dev_addr_list **from, int *from_count) | |
3979 | { | |
3980 | struct dev_addr_list *da, *next; | |
3981 | ||
3982 | da = *from; | |
3983 | while (da != NULL) { | |
3984 | next = da->next; | |
3985 | if (da->da_synced) { | |
3986 | __dev_addr_delete(to, to_count, | |
3987 | da->da_addr, da->da_addrlen, 0); | |
3988 | da->da_synced = 0; | |
3989 | __dev_addr_delete(from, from_count, | |
3990 | da->da_addr, da->da_addrlen, 0); | |
3991 | } | |
3992 | da = next; | |
3993 | } | |
3994 | } | |
c4029083 | 3995 | EXPORT_SYMBOL_GPL(__dev_addr_unsync); |
e83a2ea8 CL |
3996 | |
3997 | /** | |
3998 | * dev_unicast_sync - Synchronize device's unicast list to another device | |
3999 | * @to: destination device | |
4000 | * @from: source device | |
4001 | * | |
4002 | * Add newly added addresses to the destination device and release | |
a6ac65db JP |
4003 | * addresses that have no users left. The source device must be |
4004 | * locked by netif_tx_lock_bh. | |
e83a2ea8 CL |
4005 | * |
4006 | * This function is intended to be called from the dev->set_rx_mode | |
4007 | * function of layered software devices. | |
4008 | */ | |
4009 | int dev_unicast_sync(struct net_device *to, struct net_device *from) | |
4010 | { | |
4011 | int err = 0; | |
4012 | ||
ccffad25 JP |
4013 | if (to->addr_len != from->addr_len) |
4014 | return -EINVAL; | |
4015 | ||
a6ac65db | 4016 | netif_addr_lock_bh(to); |
31278e71 | 4017 | err = __hw_addr_sync(&to->uc, &from->uc, to->addr_len); |
e83a2ea8 CL |
4018 | if (!err) |
4019 | __dev_set_rx_mode(to); | |
a6ac65db | 4020 | netif_addr_unlock_bh(to); |
e83a2ea8 CL |
4021 | return err; |
4022 | } | |
4023 | EXPORT_SYMBOL(dev_unicast_sync); | |
4024 | ||
4025 | /** | |
bc2cda1e | 4026 | * dev_unicast_unsync - Remove synchronized addresses from the destination device |
e83a2ea8 CL |
4027 | * @to: destination device |
4028 | * @from: source device | |
4029 | * | |
4030 | * Remove all addresses that were added to the destination device by | |
4031 | * dev_unicast_sync(). This function is intended to be called from the | |
4032 | * dev->stop function of layered software devices. | |
4033 | */ | |
4034 | void dev_unicast_unsync(struct net_device *to, struct net_device *from) | |
4035 | { | |
ccffad25 JP |
4036 | if (to->addr_len != from->addr_len) |
4037 | return; | |
e83a2ea8 | 4038 | |
a6ac65db JP |
4039 | netif_addr_lock_bh(from); |
4040 | netif_addr_lock(to); | |
31278e71 | 4041 | __hw_addr_unsync(&to->uc, &from->uc, to->addr_len); |
ccffad25 | 4042 | __dev_set_rx_mode(to); |
a6ac65db JP |
4043 | netif_addr_unlock(to); |
4044 | netif_addr_unlock_bh(from); | |
e83a2ea8 CL |
4045 | } |
4046 | EXPORT_SYMBOL(dev_unicast_unsync); | |
4047 | ||
ccffad25 JP |
4048 | static void dev_unicast_flush(struct net_device *dev) |
4049 | { | |
a6ac65db | 4050 | netif_addr_lock_bh(dev); |
31278e71 | 4051 | __hw_addr_flush(&dev->uc); |
a6ac65db | 4052 | netif_addr_unlock_bh(dev); |
ccffad25 JP |
4053 | } |
4054 | ||
4055 | static void dev_unicast_init(struct net_device *dev) | |
4056 | { | |
31278e71 | 4057 | __hw_addr_init(&dev->uc); |
ccffad25 JP |
4058 | } |
4059 | ||
4060 | ||
12972621 DC |
4061 | static void __dev_addr_discard(struct dev_addr_list **list) |
4062 | { | |
4063 | struct dev_addr_list *tmp; | |
4064 | ||
4065 | while (*list != NULL) { | |
4066 | tmp = *list; | |
4067 | *list = tmp->next; | |
4068 | if (tmp->da_users > tmp->da_gusers) | |
4069 | printk("__dev_addr_discard: address leakage! " | |
4070 | "da_users=%d\n", tmp->da_users); | |
4071 | kfree(tmp); | |
4072 | } | |
4073 | } | |
4074 | ||
26cc2522 | 4075 | static void dev_addr_discard(struct net_device *dev) |
4417da66 | 4076 | { |
b9e40857 | 4077 | netif_addr_lock_bh(dev); |
26cc2522 | 4078 | |
456ad75c DC |
4079 | __dev_addr_discard(&dev->mc_list); |
4080 | dev->mc_count = 0; | |
26cc2522 | 4081 | |
b9e40857 | 4082 | netif_addr_unlock_bh(dev); |
456ad75c DC |
4083 | } |
4084 | ||
f0db275a SH |
4085 | /** |
4086 | * dev_get_flags - get flags reported to userspace | |
4087 | * @dev: device | |
4088 | * | |
4089 | * Get the combination of flag bits exported through APIs to userspace. | |
4090 | */ | |
1da177e4 LT |
4091 | unsigned dev_get_flags(const struct net_device *dev) |
4092 | { | |
4093 | unsigned flags; | |
4094 | ||
4095 | flags = (dev->flags & ~(IFF_PROMISC | | |
4096 | IFF_ALLMULTI | | |
b00055aa SR |
4097 | IFF_RUNNING | |
4098 | IFF_LOWER_UP | | |
4099 | IFF_DORMANT)) | | |
1da177e4 LT |
4100 | (dev->gflags & (IFF_PROMISC | |
4101 | IFF_ALLMULTI)); | |
4102 | ||
b00055aa SR |
4103 | if (netif_running(dev)) { |
4104 | if (netif_oper_up(dev)) | |
4105 | flags |= IFF_RUNNING; | |
4106 | if (netif_carrier_ok(dev)) | |
4107 | flags |= IFF_LOWER_UP; | |
4108 | if (netif_dormant(dev)) | |
4109 | flags |= IFF_DORMANT; | |
4110 | } | |
1da177e4 LT |
4111 | |
4112 | return flags; | |
4113 | } | |
d1b19dff | 4114 | EXPORT_SYMBOL(dev_get_flags); |
1da177e4 | 4115 | |
f0db275a SH |
4116 | /** |
4117 | * dev_change_flags - change device settings | |
4118 | * @dev: device | |
4119 | * @flags: device state flags | |
4120 | * | |
4121 | * Change settings on device based state flags. The flags are | |
4122 | * in the userspace exported format. | |
4123 | */ | |
1da177e4 LT |
4124 | int dev_change_flags(struct net_device *dev, unsigned flags) |
4125 | { | |
7c355f53 | 4126 | int ret, changes; |
1da177e4 LT |
4127 | int old_flags = dev->flags; |
4128 | ||
24023451 PM |
4129 | ASSERT_RTNL(); |
4130 | ||
1da177e4 LT |
4131 | /* |
4132 | * Set the flags on our device. | |
4133 | */ | |
4134 | ||
4135 | dev->flags = (flags & (IFF_DEBUG | IFF_NOTRAILERS | IFF_NOARP | | |
4136 | IFF_DYNAMIC | IFF_MULTICAST | IFF_PORTSEL | | |
4137 | IFF_AUTOMEDIA)) | | |
4138 | (dev->flags & (IFF_UP | IFF_VOLATILE | IFF_PROMISC | | |
4139 | IFF_ALLMULTI)); | |
4140 | ||
4141 | /* | |
4142 | * Load in the correct multicast list now the flags have changed. | |
4143 | */ | |
4144 | ||
b6c40d68 PM |
4145 | if ((old_flags ^ flags) & IFF_MULTICAST) |
4146 | dev_change_rx_flags(dev, IFF_MULTICAST); | |
24023451 | 4147 | |
4417da66 | 4148 | dev_set_rx_mode(dev); |
1da177e4 LT |
4149 | |
4150 | /* | |
4151 | * Have we downed the interface. We handle IFF_UP ourselves | |
4152 | * according to user attempts to set it, rather than blindly | |
4153 | * setting it. | |
4154 | */ | |
4155 | ||
4156 | ret = 0; | |
4157 | if ((old_flags ^ flags) & IFF_UP) { /* Bit is different ? */ | |
4158 | ret = ((old_flags & IFF_UP) ? dev_close : dev_open)(dev); | |
4159 | ||
4160 | if (!ret) | |
4417da66 | 4161 | dev_set_rx_mode(dev); |
1da177e4 LT |
4162 | } |
4163 | ||
4164 | if (dev->flags & IFF_UP && | |
d1b19dff | 4165 | ((old_flags ^ dev->flags) & ~(IFF_UP | IFF_PROMISC | IFF_ALLMULTI | |
1da177e4 | 4166 | IFF_VOLATILE))) |
056925ab | 4167 | call_netdevice_notifiers(NETDEV_CHANGE, dev); |
1da177e4 LT |
4168 | |
4169 | if ((flags ^ dev->gflags) & IFF_PROMISC) { | |
d1b19dff ED |
4170 | int inc = (flags & IFF_PROMISC) ? 1 : -1; |
4171 | ||
1da177e4 LT |
4172 | dev->gflags ^= IFF_PROMISC; |
4173 | dev_set_promiscuity(dev, inc); | |
4174 | } | |
4175 | ||
4176 | /* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI | |
4177 | is important. Some (broken) drivers set IFF_PROMISC, when | |
4178 | IFF_ALLMULTI is requested not asking us and not reporting. | |
4179 | */ | |
4180 | if ((flags ^ dev->gflags) & IFF_ALLMULTI) { | |
d1b19dff ED |
4181 | int inc = (flags & IFF_ALLMULTI) ? 1 : -1; |
4182 | ||
1da177e4 LT |
4183 | dev->gflags ^= IFF_ALLMULTI; |
4184 | dev_set_allmulti(dev, inc); | |
4185 | } | |
4186 | ||
7c355f53 TG |
4187 | /* Exclude state transition flags, already notified */ |
4188 | changes = (old_flags ^ dev->flags) & ~(IFF_UP | IFF_RUNNING); | |
4189 | if (changes) | |
4190 | rtmsg_ifinfo(RTM_NEWLINK, dev, changes); | |
1da177e4 LT |
4191 | |
4192 | return ret; | |
4193 | } | |
d1b19dff | 4194 | EXPORT_SYMBOL(dev_change_flags); |
1da177e4 | 4195 | |
f0db275a SH |
4196 | /** |
4197 | * dev_set_mtu - Change maximum transfer unit | |
4198 | * @dev: device | |
4199 | * @new_mtu: new transfer unit | |
4200 | * | |
4201 | * Change the maximum transfer size of the network device. | |
4202 | */ | |
1da177e4 LT |
4203 | int dev_set_mtu(struct net_device *dev, int new_mtu) |
4204 | { | |
d314774c | 4205 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
4206 | int err; |
4207 | ||
4208 | if (new_mtu == dev->mtu) | |
4209 | return 0; | |
4210 | ||
4211 | /* MTU must be positive. */ | |
4212 | if (new_mtu < 0) | |
4213 | return -EINVAL; | |
4214 | ||
4215 | if (!netif_device_present(dev)) | |
4216 | return -ENODEV; | |
4217 | ||
4218 | err = 0; | |
d314774c SH |
4219 | if (ops->ndo_change_mtu) |
4220 | err = ops->ndo_change_mtu(dev, new_mtu); | |
1da177e4 LT |
4221 | else |
4222 | dev->mtu = new_mtu; | |
d314774c | 4223 | |
1da177e4 | 4224 | if (!err && dev->flags & IFF_UP) |
056925ab | 4225 | call_netdevice_notifiers(NETDEV_CHANGEMTU, dev); |
1da177e4 LT |
4226 | return err; |
4227 | } | |
d1b19dff | 4228 | EXPORT_SYMBOL(dev_set_mtu); |
1da177e4 | 4229 | |
f0db275a SH |
4230 | /** |
4231 | * dev_set_mac_address - Change Media Access Control Address | |
4232 | * @dev: device | |
4233 | * @sa: new address | |
4234 | * | |
4235 | * Change the hardware (MAC) address of the device | |
4236 | */ | |
1da177e4 LT |
4237 | int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa) |
4238 | { | |
d314774c | 4239 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
4240 | int err; |
4241 | ||
d314774c | 4242 | if (!ops->ndo_set_mac_address) |
1da177e4 LT |
4243 | return -EOPNOTSUPP; |
4244 | if (sa->sa_family != dev->type) | |
4245 | return -EINVAL; | |
4246 | if (!netif_device_present(dev)) | |
4247 | return -ENODEV; | |
d314774c | 4248 | err = ops->ndo_set_mac_address(dev, sa); |
1da177e4 | 4249 | if (!err) |
056925ab | 4250 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); |
1da177e4 LT |
4251 | return err; |
4252 | } | |
d1b19dff | 4253 | EXPORT_SYMBOL(dev_set_mac_address); |
1da177e4 LT |
4254 | |
4255 | /* | |
14e3e079 | 4256 | * Perform the SIOCxIFxxx calls, inside read_lock(dev_base_lock) |
1da177e4 | 4257 | */ |
14e3e079 | 4258 | static int dev_ifsioc_locked(struct net *net, struct ifreq *ifr, unsigned int cmd) |
1da177e4 LT |
4259 | { |
4260 | int err; | |
881d966b | 4261 | struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name); |
1da177e4 LT |
4262 | |
4263 | if (!dev) | |
4264 | return -ENODEV; | |
4265 | ||
4266 | switch (cmd) { | |
d1b19dff ED |
4267 | case SIOCGIFFLAGS: /* Get interface flags */ |
4268 | ifr->ifr_flags = (short) dev_get_flags(dev); | |
4269 | return 0; | |
1da177e4 | 4270 | |
d1b19dff ED |
4271 | case SIOCGIFMETRIC: /* Get the metric on the interface |
4272 | (currently unused) */ | |
4273 | ifr->ifr_metric = 0; | |
4274 | return 0; | |
1da177e4 | 4275 | |
d1b19dff ED |
4276 | case SIOCGIFMTU: /* Get the MTU of a device */ |
4277 | ifr->ifr_mtu = dev->mtu; | |
4278 | return 0; | |
1da177e4 | 4279 | |
d1b19dff ED |
4280 | case SIOCGIFHWADDR: |
4281 | if (!dev->addr_len) | |
4282 | memset(ifr->ifr_hwaddr.sa_data, 0, sizeof ifr->ifr_hwaddr.sa_data); | |
4283 | else | |
4284 | memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr, | |
4285 | min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); | |
4286 | ifr->ifr_hwaddr.sa_family = dev->type; | |
4287 | return 0; | |
1da177e4 | 4288 | |
d1b19dff ED |
4289 | case SIOCGIFSLAVE: |
4290 | err = -EINVAL; | |
4291 | break; | |
14e3e079 | 4292 | |
d1b19dff ED |
4293 | case SIOCGIFMAP: |
4294 | ifr->ifr_map.mem_start = dev->mem_start; | |
4295 | ifr->ifr_map.mem_end = dev->mem_end; | |
4296 | ifr->ifr_map.base_addr = dev->base_addr; | |
4297 | ifr->ifr_map.irq = dev->irq; | |
4298 | ifr->ifr_map.dma = dev->dma; | |
4299 | ifr->ifr_map.port = dev->if_port; | |
4300 | return 0; | |
14e3e079 | 4301 | |
d1b19dff ED |
4302 | case SIOCGIFINDEX: |
4303 | ifr->ifr_ifindex = dev->ifindex; | |
4304 | return 0; | |
14e3e079 | 4305 | |
d1b19dff ED |
4306 | case SIOCGIFTXQLEN: |
4307 | ifr->ifr_qlen = dev->tx_queue_len; | |
4308 | return 0; | |
14e3e079 | 4309 | |
d1b19dff ED |
4310 | default: |
4311 | /* dev_ioctl() should ensure this case | |
4312 | * is never reached | |
4313 | */ | |
4314 | WARN_ON(1); | |
4315 | err = -EINVAL; | |
4316 | break; | |
14e3e079 JG |
4317 | |
4318 | } | |
4319 | return err; | |
4320 | } | |
4321 | ||
4322 | /* | |
4323 | * Perform the SIOCxIFxxx calls, inside rtnl_lock() | |
4324 | */ | |
4325 | static int dev_ifsioc(struct net *net, struct ifreq *ifr, unsigned int cmd) | |
4326 | { | |
4327 | int err; | |
4328 | struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name); | |
5f2f6da7 | 4329 | const struct net_device_ops *ops; |
14e3e079 JG |
4330 | |
4331 | if (!dev) | |
4332 | return -ENODEV; | |
4333 | ||
5f2f6da7 JP |
4334 | ops = dev->netdev_ops; |
4335 | ||
14e3e079 | 4336 | switch (cmd) { |
d1b19dff ED |
4337 | case SIOCSIFFLAGS: /* Set interface flags */ |
4338 | return dev_change_flags(dev, ifr->ifr_flags); | |
14e3e079 | 4339 | |
d1b19dff ED |
4340 | case SIOCSIFMETRIC: /* Set the metric on the interface |
4341 | (currently unused) */ | |
4342 | return -EOPNOTSUPP; | |
14e3e079 | 4343 | |
d1b19dff ED |
4344 | case SIOCSIFMTU: /* Set the MTU of a device */ |
4345 | return dev_set_mtu(dev, ifr->ifr_mtu); | |
1da177e4 | 4346 | |
d1b19dff ED |
4347 | case SIOCSIFHWADDR: |
4348 | return dev_set_mac_address(dev, &ifr->ifr_hwaddr); | |
1da177e4 | 4349 | |
d1b19dff ED |
4350 | case SIOCSIFHWBROADCAST: |
4351 | if (ifr->ifr_hwaddr.sa_family != dev->type) | |
4352 | return -EINVAL; | |
4353 | memcpy(dev->broadcast, ifr->ifr_hwaddr.sa_data, | |
4354 | min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); | |
4355 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); | |
4356 | return 0; | |
1da177e4 | 4357 | |
d1b19dff ED |
4358 | case SIOCSIFMAP: |
4359 | if (ops->ndo_set_config) { | |
1da177e4 LT |
4360 | if (!netif_device_present(dev)) |
4361 | return -ENODEV; | |
d1b19dff ED |
4362 | return ops->ndo_set_config(dev, &ifr->ifr_map); |
4363 | } | |
4364 | return -EOPNOTSUPP; | |
1da177e4 | 4365 | |
d1b19dff ED |
4366 | case SIOCADDMULTI: |
4367 | if ((!ops->ndo_set_multicast_list && !ops->ndo_set_rx_mode) || | |
4368 | ifr->ifr_hwaddr.sa_family != AF_UNSPEC) | |
4369 | return -EINVAL; | |
4370 | if (!netif_device_present(dev)) | |
4371 | return -ENODEV; | |
4372 | return dev_mc_add(dev, ifr->ifr_hwaddr.sa_data, | |
4373 | dev->addr_len, 1); | |
4374 | ||
4375 | case SIOCDELMULTI: | |
4376 | if ((!ops->ndo_set_multicast_list && !ops->ndo_set_rx_mode) || | |
4377 | ifr->ifr_hwaddr.sa_family != AF_UNSPEC) | |
4378 | return -EINVAL; | |
4379 | if (!netif_device_present(dev)) | |
4380 | return -ENODEV; | |
4381 | return dev_mc_delete(dev, ifr->ifr_hwaddr.sa_data, | |
4382 | dev->addr_len, 1); | |
1da177e4 | 4383 | |
d1b19dff ED |
4384 | case SIOCSIFTXQLEN: |
4385 | if (ifr->ifr_qlen < 0) | |
4386 | return -EINVAL; | |
4387 | dev->tx_queue_len = ifr->ifr_qlen; | |
4388 | return 0; | |
1da177e4 | 4389 | |
d1b19dff ED |
4390 | case SIOCSIFNAME: |
4391 | ifr->ifr_newname[IFNAMSIZ-1] = '\0'; | |
4392 | return dev_change_name(dev, ifr->ifr_newname); | |
1da177e4 | 4393 | |
d1b19dff ED |
4394 | /* |
4395 | * Unknown or private ioctl | |
4396 | */ | |
4397 | default: | |
4398 | if ((cmd >= SIOCDEVPRIVATE && | |
4399 | cmd <= SIOCDEVPRIVATE + 15) || | |
4400 | cmd == SIOCBONDENSLAVE || | |
4401 | cmd == SIOCBONDRELEASE || | |
4402 | cmd == SIOCBONDSETHWADDR || | |
4403 | cmd == SIOCBONDSLAVEINFOQUERY || | |
4404 | cmd == SIOCBONDINFOQUERY || | |
4405 | cmd == SIOCBONDCHANGEACTIVE || | |
4406 | cmd == SIOCGMIIPHY || | |
4407 | cmd == SIOCGMIIREG || | |
4408 | cmd == SIOCSMIIREG || | |
4409 | cmd == SIOCBRADDIF || | |
4410 | cmd == SIOCBRDELIF || | |
4411 | cmd == SIOCSHWTSTAMP || | |
4412 | cmd == SIOCWANDEV) { | |
4413 | err = -EOPNOTSUPP; | |
4414 | if (ops->ndo_do_ioctl) { | |
4415 | if (netif_device_present(dev)) | |
4416 | err = ops->ndo_do_ioctl(dev, ifr, cmd); | |
4417 | else | |
4418 | err = -ENODEV; | |
4419 | } | |
4420 | } else | |
4421 | err = -EINVAL; | |
1da177e4 LT |
4422 | |
4423 | } | |
4424 | return err; | |
4425 | } | |
4426 | ||
4427 | /* | |
4428 | * This function handles all "interface"-type I/O control requests. The actual | |
4429 | * 'doing' part of this is dev_ifsioc above. | |
4430 | */ | |
4431 | ||
4432 | /** | |
4433 | * dev_ioctl - network device ioctl | |
c4ea43c5 | 4434 | * @net: the applicable net namespace |
1da177e4 LT |
4435 | * @cmd: command to issue |
4436 | * @arg: pointer to a struct ifreq in user space | |
4437 | * | |
4438 | * Issue ioctl functions to devices. This is normally called by the | |
4439 | * user space syscall interfaces but can sometimes be useful for | |
4440 | * other purposes. The return value is the return from the syscall if | |
4441 | * positive or a negative errno code on error. | |
4442 | */ | |
4443 | ||
881d966b | 4444 | int dev_ioctl(struct net *net, unsigned int cmd, void __user *arg) |
1da177e4 LT |
4445 | { |
4446 | struct ifreq ifr; | |
4447 | int ret; | |
4448 | char *colon; | |
4449 | ||
4450 | /* One special case: SIOCGIFCONF takes ifconf argument | |
4451 | and requires shared lock, because it sleeps writing | |
4452 | to user space. | |
4453 | */ | |
4454 | ||
4455 | if (cmd == SIOCGIFCONF) { | |
6756ae4b | 4456 | rtnl_lock(); |
881d966b | 4457 | ret = dev_ifconf(net, (char __user *) arg); |
6756ae4b | 4458 | rtnl_unlock(); |
1da177e4 LT |
4459 | return ret; |
4460 | } | |
4461 | if (cmd == SIOCGIFNAME) | |
881d966b | 4462 | return dev_ifname(net, (struct ifreq __user *)arg); |
1da177e4 LT |
4463 | |
4464 | if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) | |
4465 | return -EFAULT; | |
4466 | ||
4467 | ifr.ifr_name[IFNAMSIZ-1] = 0; | |
4468 | ||
4469 | colon = strchr(ifr.ifr_name, ':'); | |
4470 | if (colon) | |
4471 | *colon = 0; | |
4472 | ||
4473 | /* | |
4474 | * See which interface the caller is talking about. | |
4475 | */ | |
4476 | ||
4477 | switch (cmd) { | |
d1b19dff ED |
4478 | /* |
4479 | * These ioctl calls: | |
4480 | * - can be done by all. | |
4481 | * - atomic and do not require locking. | |
4482 | * - return a value | |
4483 | */ | |
4484 | case SIOCGIFFLAGS: | |
4485 | case SIOCGIFMETRIC: | |
4486 | case SIOCGIFMTU: | |
4487 | case SIOCGIFHWADDR: | |
4488 | case SIOCGIFSLAVE: | |
4489 | case SIOCGIFMAP: | |
4490 | case SIOCGIFINDEX: | |
4491 | case SIOCGIFTXQLEN: | |
4492 | dev_load(net, ifr.ifr_name); | |
4493 | read_lock(&dev_base_lock); | |
4494 | ret = dev_ifsioc_locked(net, &ifr, cmd); | |
4495 | read_unlock(&dev_base_lock); | |
4496 | if (!ret) { | |
4497 | if (colon) | |
4498 | *colon = ':'; | |
4499 | if (copy_to_user(arg, &ifr, | |
4500 | sizeof(struct ifreq))) | |
4501 | ret = -EFAULT; | |
4502 | } | |
4503 | return ret; | |
1da177e4 | 4504 | |
d1b19dff ED |
4505 | case SIOCETHTOOL: |
4506 | dev_load(net, ifr.ifr_name); | |
4507 | rtnl_lock(); | |
4508 | ret = dev_ethtool(net, &ifr); | |
4509 | rtnl_unlock(); | |
4510 | if (!ret) { | |
4511 | if (colon) | |
4512 | *colon = ':'; | |
4513 | if (copy_to_user(arg, &ifr, | |
4514 | sizeof(struct ifreq))) | |
4515 | ret = -EFAULT; | |
4516 | } | |
4517 | return ret; | |
1da177e4 | 4518 | |
d1b19dff ED |
4519 | /* |
4520 | * These ioctl calls: | |
4521 | * - require superuser power. | |
4522 | * - require strict serialization. | |
4523 | * - return a value | |
4524 | */ | |
4525 | case SIOCGMIIPHY: | |
4526 | case SIOCGMIIREG: | |
4527 | case SIOCSIFNAME: | |
4528 | if (!capable(CAP_NET_ADMIN)) | |
4529 | return -EPERM; | |
4530 | dev_load(net, ifr.ifr_name); | |
4531 | rtnl_lock(); | |
4532 | ret = dev_ifsioc(net, &ifr, cmd); | |
4533 | rtnl_unlock(); | |
4534 | if (!ret) { | |
4535 | if (colon) | |
4536 | *colon = ':'; | |
4537 | if (copy_to_user(arg, &ifr, | |
4538 | sizeof(struct ifreq))) | |
4539 | ret = -EFAULT; | |
4540 | } | |
4541 | return ret; | |
1da177e4 | 4542 | |
d1b19dff ED |
4543 | /* |
4544 | * These ioctl calls: | |
4545 | * - require superuser power. | |
4546 | * - require strict serialization. | |
4547 | * - do not return a value | |
4548 | */ | |
4549 | case SIOCSIFFLAGS: | |
4550 | case SIOCSIFMETRIC: | |
4551 | case SIOCSIFMTU: | |
4552 | case SIOCSIFMAP: | |
4553 | case SIOCSIFHWADDR: | |
4554 | case SIOCSIFSLAVE: | |
4555 | case SIOCADDMULTI: | |
4556 | case SIOCDELMULTI: | |
4557 | case SIOCSIFHWBROADCAST: | |
4558 | case SIOCSIFTXQLEN: | |
4559 | case SIOCSMIIREG: | |
4560 | case SIOCBONDENSLAVE: | |
4561 | case SIOCBONDRELEASE: | |
4562 | case SIOCBONDSETHWADDR: | |
4563 | case SIOCBONDCHANGEACTIVE: | |
4564 | case SIOCBRADDIF: | |
4565 | case SIOCBRDELIF: | |
4566 | case SIOCSHWTSTAMP: | |
4567 | if (!capable(CAP_NET_ADMIN)) | |
4568 | return -EPERM; | |
4569 | /* fall through */ | |
4570 | case SIOCBONDSLAVEINFOQUERY: | |
4571 | case SIOCBONDINFOQUERY: | |
4572 | dev_load(net, ifr.ifr_name); | |
4573 | rtnl_lock(); | |
4574 | ret = dev_ifsioc(net, &ifr, cmd); | |
4575 | rtnl_unlock(); | |
4576 | return ret; | |
4577 | ||
4578 | case SIOCGIFMEM: | |
4579 | /* Get the per device memory space. We can add this but | |
4580 | * currently do not support it */ | |
4581 | case SIOCSIFMEM: | |
4582 | /* Set the per device memory buffer space. | |
4583 | * Not applicable in our case */ | |
4584 | case SIOCSIFLINK: | |
4585 | return -EINVAL; | |
4586 | ||
4587 | /* | |
4588 | * Unknown or private ioctl. | |
4589 | */ | |
4590 | default: | |
4591 | if (cmd == SIOCWANDEV || | |
4592 | (cmd >= SIOCDEVPRIVATE && | |
4593 | cmd <= SIOCDEVPRIVATE + 15)) { | |
881d966b | 4594 | dev_load(net, ifr.ifr_name); |
1da177e4 | 4595 | rtnl_lock(); |
881d966b | 4596 | ret = dev_ifsioc(net, &ifr, cmd); |
1da177e4 | 4597 | rtnl_unlock(); |
d1b19dff ED |
4598 | if (!ret && copy_to_user(arg, &ifr, |
4599 | sizeof(struct ifreq))) | |
4600 | ret = -EFAULT; | |
1da177e4 | 4601 | return ret; |
d1b19dff ED |
4602 | } |
4603 | /* Take care of Wireless Extensions */ | |
4604 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) | |
4605 | return wext_handle_ioctl(net, &ifr, cmd, arg); | |
4606 | return -EINVAL; | |
1da177e4 LT |
4607 | } |
4608 | } | |
4609 | ||
4610 | ||
4611 | /** | |
4612 | * dev_new_index - allocate an ifindex | |
c4ea43c5 | 4613 | * @net: the applicable net namespace |
1da177e4 LT |
4614 | * |
4615 | * Returns a suitable unique value for a new device interface | |
4616 | * number. The caller must hold the rtnl semaphore or the | |
4617 | * dev_base_lock to be sure it remains unique. | |
4618 | */ | |
881d966b | 4619 | static int dev_new_index(struct net *net) |
1da177e4 LT |
4620 | { |
4621 | static int ifindex; | |
4622 | for (;;) { | |
4623 | if (++ifindex <= 0) | |
4624 | ifindex = 1; | |
881d966b | 4625 | if (!__dev_get_by_index(net, ifindex)) |
1da177e4 LT |
4626 | return ifindex; |
4627 | } | |
4628 | } | |
4629 | ||
1da177e4 | 4630 | /* Delayed registration/unregisteration */ |
3b5b34fd | 4631 | static LIST_HEAD(net_todo_list); |
1da177e4 | 4632 | |
6f05f629 | 4633 | static void net_set_todo(struct net_device *dev) |
1da177e4 | 4634 | { |
1da177e4 | 4635 | list_add_tail(&dev->todo_list, &net_todo_list); |
1da177e4 LT |
4636 | } |
4637 | ||
93ee31f1 DL |
4638 | static void rollback_registered(struct net_device *dev) |
4639 | { | |
4640 | BUG_ON(dev_boot_phase); | |
4641 | ASSERT_RTNL(); | |
4642 | ||
4643 | /* Some devices call without registering for initialization unwind. */ | |
4644 | if (dev->reg_state == NETREG_UNINITIALIZED) { | |
4645 | printk(KERN_DEBUG "unregister_netdevice: device %s/%p never " | |
4646 | "was registered\n", dev->name, dev); | |
4647 | ||
4648 | WARN_ON(1); | |
4649 | return; | |
4650 | } | |
4651 | ||
4652 | BUG_ON(dev->reg_state != NETREG_REGISTERED); | |
4653 | ||
4654 | /* If device is running, close it first. */ | |
4655 | dev_close(dev); | |
4656 | ||
4657 | /* And unlink it from device chain. */ | |
4658 | unlist_netdevice(dev); | |
4659 | ||
4660 | dev->reg_state = NETREG_UNREGISTERING; | |
4661 | ||
4662 | synchronize_net(); | |
4663 | ||
4664 | /* Shutdown queueing discipline. */ | |
4665 | dev_shutdown(dev); | |
4666 | ||
4667 | ||
4668 | /* Notify protocols, that we are about to destroy | |
4669 | this device. They should clean all the things. | |
4670 | */ | |
4671 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); | |
4672 | ||
4673 | /* | |
4674 | * Flush the unicast and multicast chains | |
4675 | */ | |
ccffad25 | 4676 | dev_unicast_flush(dev); |
93ee31f1 DL |
4677 | dev_addr_discard(dev); |
4678 | ||
d314774c SH |
4679 | if (dev->netdev_ops->ndo_uninit) |
4680 | dev->netdev_ops->ndo_uninit(dev); | |
93ee31f1 DL |
4681 | |
4682 | /* Notifier chain MUST detach us from master device. */ | |
547b792c | 4683 | WARN_ON(dev->master); |
93ee31f1 DL |
4684 | |
4685 | /* Remove entries from kobject tree */ | |
4686 | netdev_unregister_kobject(dev); | |
4687 | ||
4688 | synchronize_net(); | |
4689 | ||
4690 | dev_put(dev); | |
4691 | } | |
4692 | ||
e8a0464c DM |
4693 | static void __netdev_init_queue_locks_one(struct net_device *dev, |
4694 | struct netdev_queue *dev_queue, | |
4695 | void *_unused) | |
c773e847 DM |
4696 | { |
4697 | spin_lock_init(&dev_queue->_xmit_lock); | |
cf508b12 | 4698 | netdev_set_xmit_lockdep_class(&dev_queue->_xmit_lock, dev->type); |
c773e847 DM |
4699 | dev_queue->xmit_lock_owner = -1; |
4700 | } | |
4701 | ||
4702 | static void netdev_init_queue_locks(struct net_device *dev) | |
4703 | { | |
e8a0464c DM |
4704 | netdev_for_each_tx_queue(dev, __netdev_init_queue_locks_one, NULL); |
4705 | __netdev_init_queue_locks_one(dev, &dev->rx_queue, NULL); | |
c773e847 DM |
4706 | } |
4707 | ||
b63365a2 HX |
4708 | unsigned long netdev_fix_features(unsigned long features, const char *name) |
4709 | { | |
4710 | /* Fix illegal SG+CSUM combinations. */ | |
4711 | if ((features & NETIF_F_SG) && | |
4712 | !(features & NETIF_F_ALL_CSUM)) { | |
4713 | if (name) | |
4714 | printk(KERN_NOTICE "%s: Dropping NETIF_F_SG since no " | |
4715 | "checksum feature.\n", name); | |
4716 | features &= ~NETIF_F_SG; | |
4717 | } | |
4718 | ||
4719 | /* TSO requires that SG is present as well. */ | |
4720 | if ((features & NETIF_F_TSO) && !(features & NETIF_F_SG)) { | |
4721 | if (name) | |
4722 | printk(KERN_NOTICE "%s: Dropping NETIF_F_TSO since no " | |
4723 | "SG feature.\n", name); | |
4724 | features &= ~NETIF_F_TSO; | |
4725 | } | |
4726 | ||
4727 | if (features & NETIF_F_UFO) { | |
4728 | if (!(features & NETIF_F_GEN_CSUM)) { | |
4729 | if (name) | |
4730 | printk(KERN_ERR "%s: Dropping NETIF_F_UFO " | |
4731 | "since no NETIF_F_HW_CSUM feature.\n", | |
4732 | name); | |
4733 | features &= ~NETIF_F_UFO; | |
4734 | } | |
4735 | ||
4736 | if (!(features & NETIF_F_SG)) { | |
4737 | if (name) | |
4738 | printk(KERN_ERR "%s: Dropping NETIF_F_UFO " | |
4739 | "since no NETIF_F_SG feature.\n", name); | |
4740 | features &= ~NETIF_F_UFO; | |
4741 | } | |
4742 | } | |
4743 | ||
4744 | return features; | |
4745 | } | |
4746 | EXPORT_SYMBOL(netdev_fix_features); | |
4747 | ||
1da177e4 LT |
4748 | /** |
4749 | * register_netdevice - register a network device | |
4750 | * @dev: device to register | |
4751 | * | |
4752 | * Take a completed network device structure and add it to the kernel | |
4753 | * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier | |
4754 | * chain. 0 is returned on success. A negative errno code is returned | |
4755 | * on a failure to set up the device, or if the name is a duplicate. | |
4756 | * | |
4757 | * Callers must hold the rtnl semaphore. You may want | |
4758 | * register_netdev() instead of this. | |
4759 | * | |
4760 | * BUGS: | |
4761 | * The locking appears insufficient to guarantee two parallel registers | |
4762 | * will not get the same name. | |
4763 | */ | |
4764 | ||
4765 | int register_netdevice(struct net_device *dev) | |
4766 | { | |
4767 | struct hlist_head *head; | |
4768 | struct hlist_node *p; | |
4769 | int ret; | |
d314774c | 4770 | struct net *net = dev_net(dev); |
1da177e4 LT |
4771 | |
4772 | BUG_ON(dev_boot_phase); | |
4773 | ASSERT_RTNL(); | |
4774 | ||
b17a7c17 SH |
4775 | might_sleep(); |
4776 | ||
1da177e4 LT |
4777 | /* When net_device's are persistent, this will be fatal. */ |
4778 | BUG_ON(dev->reg_state != NETREG_UNINITIALIZED); | |
d314774c | 4779 | BUG_ON(!net); |
1da177e4 | 4780 | |
f1f28aa3 | 4781 | spin_lock_init(&dev->addr_list_lock); |
cf508b12 | 4782 | netdev_set_addr_lockdep_class(dev); |
c773e847 | 4783 | netdev_init_queue_locks(dev); |
1da177e4 | 4784 | |
1da177e4 LT |
4785 | dev->iflink = -1; |
4786 | ||
4787 | /* Init, if this function is available */ | |
d314774c SH |
4788 | if (dev->netdev_ops->ndo_init) { |
4789 | ret = dev->netdev_ops->ndo_init(dev); | |
1da177e4 LT |
4790 | if (ret) { |
4791 | if (ret > 0) | |
4792 | ret = -EIO; | |
90833aa4 | 4793 | goto out; |
1da177e4 LT |
4794 | } |
4795 | } | |
4ec93edb | 4796 | |
1da177e4 LT |
4797 | if (!dev_valid_name(dev->name)) { |
4798 | ret = -EINVAL; | |
7ce1b0ed | 4799 | goto err_uninit; |
1da177e4 LT |
4800 | } |
4801 | ||
881d966b | 4802 | dev->ifindex = dev_new_index(net); |
1da177e4 LT |
4803 | if (dev->iflink == -1) |
4804 | dev->iflink = dev->ifindex; | |
4805 | ||
4806 | /* Check for existence of name */ | |
881d966b | 4807 | head = dev_name_hash(net, dev->name); |
1da177e4 LT |
4808 | hlist_for_each(p, head) { |
4809 | struct net_device *d | |
4810 | = hlist_entry(p, struct net_device, name_hlist); | |
4811 | if (!strncmp(d->name, dev->name, IFNAMSIZ)) { | |
4812 | ret = -EEXIST; | |
7ce1b0ed | 4813 | goto err_uninit; |
1da177e4 | 4814 | } |
4ec93edb | 4815 | } |
1da177e4 | 4816 | |
d212f87b SH |
4817 | /* Fix illegal checksum combinations */ |
4818 | if ((dev->features & NETIF_F_HW_CSUM) && | |
4819 | (dev->features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { | |
4820 | printk(KERN_NOTICE "%s: mixed HW and IP checksum settings.\n", | |
4821 | dev->name); | |
4822 | dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); | |
4823 | } | |
4824 | ||
4825 | if ((dev->features & NETIF_F_NO_CSUM) && | |
4826 | (dev->features & (NETIF_F_HW_CSUM|NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { | |
4827 | printk(KERN_NOTICE "%s: mixed no checksumming and other settings.\n", | |
4828 | dev->name); | |
4829 | dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM|NETIF_F_HW_CSUM); | |
4830 | } | |
4831 | ||
b63365a2 | 4832 | dev->features = netdev_fix_features(dev->features, dev->name); |
1da177e4 | 4833 | |
e5a4a72d LB |
4834 | /* Enable software GSO if SG is supported. */ |
4835 | if (dev->features & NETIF_F_SG) | |
4836 | dev->features |= NETIF_F_GSO; | |
4837 | ||
aaf8cdc3 | 4838 | netdev_initialize_kobject(dev); |
8b41d188 | 4839 | ret = netdev_register_kobject(dev); |
b17a7c17 | 4840 | if (ret) |
7ce1b0ed | 4841 | goto err_uninit; |
b17a7c17 SH |
4842 | dev->reg_state = NETREG_REGISTERED; |
4843 | ||
1da177e4 LT |
4844 | /* |
4845 | * Default initial state at registry is that the | |
4846 | * device is present. | |
4847 | */ | |
4848 | ||
4849 | set_bit(__LINK_STATE_PRESENT, &dev->state); | |
4850 | ||
1da177e4 | 4851 | dev_init_scheduler(dev); |
1da177e4 | 4852 | dev_hold(dev); |
ce286d32 | 4853 | list_netdevice(dev); |
1da177e4 LT |
4854 | |
4855 | /* Notify protocols, that a new device appeared. */ | |
056925ab | 4856 | ret = call_netdevice_notifiers(NETDEV_REGISTER, dev); |
fcc5a03a | 4857 | ret = notifier_to_errno(ret); |
93ee31f1 DL |
4858 | if (ret) { |
4859 | rollback_registered(dev); | |
4860 | dev->reg_state = NETREG_UNREGISTERED; | |
4861 | } | |
1da177e4 LT |
4862 | |
4863 | out: | |
4864 | return ret; | |
7ce1b0ed HX |
4865 | |
4866 | err_uninit: | |
d314774c SH |
4867 | if (dev->netdev_ops->ndo_uninit) |
4868 | dev->netdev_ops->ndo_uninit(dev); | |
7ce1b0ed | 4869 | goto out; |
1da177e4 | 4870 | } |
d1b19dff | 4871 | EXPORT_SYMBOL(register_netdevice); |
1da177e4 | 4872 | |
937f1ba5 BH |
4873 | /** |
4874 | * init_dummy_netdev - init a dummy network device for NAPI | |
4875 | * @dev: device to init | |
4876 | * | |
4877 | * This takes a network device structure and initialize the minimum | |
4878 | * amount of fields so it can be used to schedule NAPI polls without | |
4879 | * registering a full blown interface. This is to be used by drivers | |
4880 | * that need to tie several hardware interfaces to a single NAPI | |
4881 | * poll scheduler due to HW limitations. | |
4882 | */ | |
4883 | int init_dummy_netdev(struct net_device *dev) | |
4884 | { | |
4885 | /* Clear everything. Note we don't initialize spinlocks | |
4886 | * are they aren't supposed to be taken by any of the | |
4887 | * NAPI code and this dummy netdev is supposed to be | |
4888 | * only ever used for NAPI polls | |
4889 | */ | |
4890 | memset(dev, 0, sizeof(struct net_device)); | |
4891 | ||
4892 | /* make sure we BUG if trying to hit standard | |
4893 | * register/unregister code path | |
4894 | */ | |
4895 | dev->reg_state = NETREG_DUMMY; | |
4896 | ||
4897 | /* initialize the ref count */ | |
4898 | atomic_set(&dev->refcnt, 1); | |
4899 | ||
4900 | /* NAPI wants this */ | |
4901 | INIT_LIST_HEAD(&dev->napi_list); | |
4902 | ||
4903 | /* a dummy interface is started by default */ | |
4904 | set_bit(__LINK_STATE_PRESENT, &dev->state); | |
4905 | set_bit(__LINK_STATE_START, &dev->state); | |
4906 | ||
4907 | return 0; | |
4908 | } | |
4909 | EXPORT_SYMBOL_GPL(init_dummy_netdev); | |
4910 | ||
4911 | ||
1da177e4 LT |
4912 | /** |
4913 | * register_netdev - register a network device | |
4914 | * @dev: device to register | |
4915 | * | |
4916 | * Take a completed network device structure and add it to the kernel | |
4917 | * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier | |
4918 | * chain. 0 is returned on success. A negative errno code is returned | |
4919 | * on a failure to set up the device, or if the name is a duplicate. | |
4920 | * | |
38b4da38 | 4921 | * This is a wrapper around register_netdevice that takes the rtnl semaphore |
1da177e4 LT |
4922 | * and expands the device name if you passed a format string to |
4923 | * alloc_netdev. | |
4924 | */ | |
4925 | int register_netdev(struct net_device *dev) | |
4926 | { | |
4927 | int err; | |
4928 | ||
4929 | rtnl_lock(); | |
4930 | ||
4931 | /* | |
4932 | * If the name is a format string the caller wants us to do a | |
4933 | * name allocation. | |
4934 | */ | |
4935 | if (strchr(dev->name, '%')) { | |
4936 | err = dev_alloc_name(dev, dev->name); | |
4937 | if (err < 0) | |
4938 | goto out; | |
4939 | } | |
4ec93edb | 4940 | |
1da177e4 LT |
4941 | err = register_netdevice(dev); |
4942 | out: | |
4943 | rtnl_unlock(); | |
4944 | return err; | |
4945 | } | |
4946 | EXPORT_SYMBOL(register_netdev); | |
4947 | ||
4948 | /* | |
4949 | * netdev_wait_allrefs - wait until all references are gone. | |
4950 | * | |
4951 | * This is called when unregistering network devices. | |
4952 | * | |
4953 | * Any protocol or device that holds a reference should register | |
4954 | * for netdevice notification, and cleanup and put back the | |
4955 | * reference if they receive an UNREGISTER event. | |
4956 | * We can get stuck here if buggy protocols don't correctly | |
4ec93edb | 4957 | * call dev_put. |
1da177e4 LT |
4958 | */ |
4959 | static void netdev_wait_allrefs(struct net_device *dev) | |
4960 | { | |
4961 | unsigned long rebroadcast_time, warning_time; | |
4962 | ||
4963 | rebroadcast_time = warning_time = jiffies; | |
4964 | while (atomic_read(&dev->refcnt) != 0) { | |
4965 | if (time_after(jiffies, rebroadcast_time + 1 * HZ)) { | |
6756ae4b | 4966 | rtnl_lock(); |
1da177e4 LT |
4967 | |
4968 | /* Rebroadcast unregister notification */ | |
056925ab | 4969 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); |
1da177e4 LT |
4970 | |
4971 | if (test_bit(__LINK_STATE_LINKWATCH_PENDING, | |
4972 | &dev->state)) { | |
4973 | /* We must not have linkwatch events | |
4974 | * pending on unregister. If this | |
4975 | * happens, we simply run the queue | |
4976 | * unscheduled, resulting in a noop | |
4977 | * for this device. | |
4978 | */ | |
4979 | linkwatch_run_queue(); | |
4980 | } | |
4981 | ||
6756ae4b | 4982 | __rtnl_unlock(); |
1da177e4 LT |
4983 | |
4984 | rebroadcast_time = jiffies; | |
4985 | } | |
4986 | ||
4987 | msleep(250); | |
4988 | ||
4989 | if (time_after(jiffies, warning_time + 10 * HZ)) { | |
4990 | printk(KERN_EMERG "unregister_netdevice: " | |
4991 | "waiting for %s to become free. Usage " | |
4992 | "count = %d\n", | |
4993 | dev->name, atomic_read(&dev->refcnt)); | |
4994 | warning_time = jiffies; | |
4995 | } | |
4996 | } | |
4997 | } | |
4998 | ||
4999 | /* The sequence is: | |
5000 | * | |
5001 | * rtnl_lock(); | |
5002 | * ... | |
5003 | * register_netdevice(x1); | |
5004 | * register_netdevice(x2); | |
5005 | * ... | |
5006 | * unregister_netdevice(y1); | |
5007 | * unregister_netdevice(y2); | |
5008 | * ... | |
5009 | * rtnl_unlock(); | |
5010 | * free_netdev(y1); | |
5011 | * free_netdev(y2); | |
5012 | * | |
58ec3b4d | 5013 | * We are invoked by rtnl_unlock(). |
1da177e4 | 5014 | * This allows us to deal with problems: |
b17a7c17 | 5015 | * 1) We can delete sysfs objects which invoke hotplug |
1da177e4 LT |
5016 | * without deadlocking with linkwatch via keventd. |
5017 | * 2) Since we run with the RTNL semaphore not held, we can sleep | |
5018 | * safely in order to wait for the netdev refcnt to drop to zero. | |
58ec3b4d HX |
5019 | * |
5020 | * We must not return until all unregister events added during | |
5021 | * the interval the lock was held have been completed. | |
1da177e4 | 5022 | */ |
1da177e4 LT |
5023 | void netdev_run_todo(void) |
5024 | { | |
626ab0e6 | 5025 | struct list_head list; |
1da177e4 | 5026 | |
1da177e4 | 5027 | /* Snapshot list, allow later requests */ |
626ab0e6 | 5028 | list_replace_init(&net_todo_list, &list); |
58ec3b4d HX |
5029 | |
5030 | __rtnl_unlock(); | |
626ab0e6 | 5031 | |
1da177e4 LT |
5032 | while (!list_empty(&list)) { |
5033 | struct net_device *dev | |
5034 | = list_entry(list.next, struct net_device, todo_list); | |
5035 | list_del(&dev->todo_list); | |
5036 | ||
b17a7c17 SH |
5037 | if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) { |
5038 | printk(KERN_ERR "network todo '%s' but state %d\n", | |
5039 | dev->name, dev->reg_state); | |
5040 | dump_stack(); | |
5041 | continue; | |
5042 | } | |
1da177e4 | 5043 | |
b17a7c17 | 5044 | dev->reg_state = NETREG_UNREGISTERED; |
1da177e4 | 5045 | |
6e583ce5 SH |
5046 | on_each_cpu(flush_backlog, dev, 1); |
5047 | ||
b17a7c17 | 5048 | netdev_wait_allrefs(dev); |
1da177e4 | 5049 | |
b17a7c17 SH |
5050 | /* paranoia */ |
5051 | BUG_ON(atomic_read(&dev->refcnt)); | |
547b792c IJ |
5052 | WARN_ON(dev->ip_ptr); |
5053 | WARN_ON(dev->ip6_ptr); | |
5054 | WARN_ON(dev->dn_ptr); | |
1da177e4 | 5055 | |
b17a7c17 SH |
5056 | if (dev->destructor) |
5057 | dev->destructor(dev); | |
9093bbb2 SH |
5058 | |
5059 | /* Free network device */ | |
5060 | kobject_put(&dev->dev.kobj); | |
1da177e4 | 5061 | } |
1da177e4 LT |
5062 | } |
5063 | ||
eeda3fd6 SH |
5064 | /** |
5065 | * dev_get_stats - get network device statistics | |
5066 | * @dev: device to get statistics from | |
5067 | * | |
5068 | * Get network statistics from device. The device driver may provide | |
5069 | * its own method by setting dev->netdev_ops->get_stats; otherwise | |
5070 | * the internal statistics structure is used. | |
5071 | */ | |
5072 | const struct net_device_stats *dev_get_stats(struct net_device *dev) | |
7004bf25 | 5073 | { |
eeda3fd6 SH |
5074 | const struct net_device_ops *ops = dev->netdev_ops; |
5075 | ||
5076 | if (ops->ndo_get_stats) | |
5077 | return ops->ndo_get_stats(dev); | |
7004bf25 ED |
5078 | else { |
5079 | unsigned long tx_bytes = 0, tx_packets = 0, tx_dropped = 0; | |
5080 | struct net_device_stats *stats = &dev->stats; | |
5081 | unsigned int i; | |
5082 | struct netdev_queue *txq; | |
5083 | ||
5084 | for (i = 0; i < dev->num_tx_queues; i++) { | |
5085 | txq = netdev_get_tx_queue(dev, i); | |
5086 | tx_bytes += txq->tx_bytes; | |
5087 | tx_packets += txq->tx_packets; | |
5088 | tx_dropped += txq->tx_dropped; | |
5089 | } | |
5090 | if (tx_bytes || tx_packets || tx_dropped) { | |
5091 | stats->tx_bytes = tx_bytes; | |
5092 | stats->tx_packets = tx_packets; | |
5093 | stats->tx_dropped = tx_dropped; | |
5094 | } | |
5095 | return stats; | |
5096 | } | |
c45d286e | 5097 | } |
eeda3fd6 | 5098 | EXPORT_SYMBOL(dev_get_stats); |
c45d286e | 5099 | |
dc2b4847 | 5100 | static void netdev_init_one_queue(struct net_device *dev, |
e8a0464c DM |
5101 | struct netdev_queue *queue, |
5102 | void *_unused) | |
dc2b4847 | 5103 | { |
dc2b4847 DM |
5104 | queue->dev = dev; |
5105 | } | |
5106 | ||
bb949fbd DM |
5107 | static void netdev_init_queues(struct net_device *dev) |
5108 | { | |
e8a0464c DM |
5109 | netdev_init_one_queue(dev, &dev->rx_queue, NULL); |
5110 | netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL); | |
c3f26a26 | 5111 | spin_lock_init(&dev->tx_global_lock); |
bb949fbd DM |
5112 | } |
5113 | ||
1da177e4 | 5114 | /** |
f25f4e44 | 5115 | * alloc_netdev_mq - allocate network device |
1da177e4 LT |
5116 | * @sizeof_priv: size of private data to allocate space for |
5117 | * @name: device name format string | |
5118 | * @setup: callback to initialize device | |
f25f4e44 | 5119 | * @queue_count: the number of subqueues to allocate |
1da177e4 LT |
5120 | * |
5121 | * Allocates a struct net_device with private data area for driver use | |
f25f4e44 PWJ |
5122 | * and performs basic initialization. Also allocates subquue structs |
5123 | * for each queue on the device at the end of the netdevice. | |
1da177e4 | 5124 | */ |
f25f4e44 PWJ |
5125 | struct net_device *alloc_netdev_mq(int sizeof_priv, const char *name, |
5126 | void (*setup)(struct net_device *), unsigned int queue_count) | |
1da177e4 | 5127 | { |
e8a0464c | 5128 | struct netdev_queue *tx; |
1da177e4 | 5129 | struct net_device *dev; |
7943986c | 5130 | size_t alloc_size; |
1ce8e7b5 | 5131 | struct net_device *p; |
1da177e4 | 5132 | |
b6fe17d6 SH |
5133 | BUG_ON(strlen(name) >= sizeof(dev->name)); |
5134 | ||
fd2ea0a7 | 5135 | alloc_size = sizeof(struct net_device); |
d1643d24 AD |
5136 | if (sizeof_priv) { |
5137 | /* ensure 32-byte alignment of private area */ | |
1ce8e7b5 | 5138 | alloc_size = ALIGN(alloc_size, NETDEV_ALIGN); |
d1643d24 AD |
5139 | alloc_size += sizeof_priv; |
5140 | } | |
5141 | /* ensure 32-byte alignment of whole construct */ | |
1ce8e7b5 | 5142 | alloc_size += NETDEV_ALIGN - 1; |
1da177e4 | 5143 | |
31380de9 | 5144 | p = kzalloc(alloc_size, GFP_KERNEL); |
1da177e4 | 5145 | if (!p) { |
b6fe17d6 | 5146 | printk(KERN_ERR "alloc_netdev: Unable to allocate device.\n"); |
1da177e4 LT |
5147 | return NULL; |
5148 | } | |
1da177e4 | 5149 | |
7943986c | 5150 | tx = kcalloc(queue_count, sizeof(struct netdev_queue), GFP_KERNEL); |
e8a0464c DM |
5151 | if (!tx) { |
5152 | printk(KERN_ERR "alloc_netdev: Unable to allocate " | |
5153 | "tx qdiscs.\n"); | |
ab9c73cc | 5154 | goto free_p; |
e8a0464c DM |
5155 | } |
5156 | ||
1ce8e7b5 | 5157 | dev = PTR_ALIGN(p, NETDEV_ALIGN); |
1da177e4 | 5158 | dev->padded = (char *)dev - (char *)p; |
ab9c73cc JP |
5159 | |
5160 | if (dev_addr_init(dev)) | |
5161 | goto free_tx; | |
5162 | ||
ccffad25 JP |
5163 | dev_unicast_init(dev); |
5164 | ||
c346dca1 | 5165 | dev_net_set(dev, &init_net); |
1da177e4 | 5166 | |
e8a0464c DM |
5167 | dev->_tx = tx; |
5168 | dev->num_tx_queues = queue_count; | |
fd2ea0a7 | 5169 | dev->real_num_tx_queues = queue_count; |
e8a0464c | 5170 | |
82cc1a7a | 5171 | dev->gso_max_size = GSO_MAX_SIZE; |
1da177e4 | 5172 | |
bb949fbd DM |
5173 | netdev_init_queues(dev); |
5174 | ||
d565b0a1 | 5175 | INIT_LIST_HEAD(&dev->napi_list); |
93f154b5 | 5176 | dev->priv_flags = IFF_XMIT_DST_RELEASE; |
1da177e4 LT |
5177 | setup(dev); |
5178 | strcpy(dev->name, name); | |
5179 | return dev; | |
ab9c73cc JP |
5180 | |
5181 | free_tx: | |
5182 | kfree(tx); | |
5183 | ||
5184 | free_p: | |
5185 | kfree(p); | |
5186 | return NULL; | |
1da177e4 | 5187 | } |
f25f4e44 | 5188 | EXPORT_SYMBOL(alloc_netdev_mq); |
1da177e4 LT |
5189 | |
5190 | /** | |
5191 | * free_netdev - free network device | |
5192 | * @dev: device | |
5193 | * | |
4ec93edb YH |
5194 | * This function does the last stage of destroying an allocated device |
5195 | * interface. The reference to the device object is released. | |
1da177e4 LT |
5196 | * If this is the last reference then it will be freed. |
5197 | */ | |
5198 | void free_netdev(struct net_device *dev) | |
5199 | { | |
d565b0a1 HX |
5200 | struct napi_struct *p, *n; |
5201 | ||
f3005d7f DL |
5202 | release_net(dev_net(dev)); |
5203 | ||
e8a0464c DM |
5204 | kfree(dev->_tx); |
5205 | ||
f001fde5 JP |
5206 | /* Flush device addresses */ |
5207 | dev_addr_flush(dev); | |
5208 | ||
d565b0a1 HX |
5209 | list_for_each_entry_safe(p, n, &dev->napi_list, dev_list) |
5210 | netif_napi_del(p); | |
5211 | ||
3041a069 | 5212 | /* Compatibility with error handling in drivers */ |
1da177e4 LT |
5213 | if (dev->reg_state == NETREG_UNINITIALIZED) { |
5214 | kfree((char *)dev - dev->padded); | |
5215 | return; | |
5216 | } | |
5217 | ||
5218 | BUG_ON(dev->reg_state != NETREG_UNREGISTERED); | |
5219 | dev->reg_state = NETREG_RELEASED; | |
5220 | ||
43cb76d9 GKH |
5221 | /* will free via device release */ |
5222 | put_device(&dev->dev); | |
1da177e4 | 5223 | } |
d1b19dff | 5224 | EXPORT_SYMBOL(free_netdev); |
4ec93edb | 5225 | |
f0db275a SH |
5226 | /** |
5227 | * synchronize_net - Synchronize with packet receive processing | |
5228 | * | |
5229 | * Wait for packets currently being received to be done. | |
5230 | * Does not block later packets from starting. | |
5231 | */ | |
4ec93edb | 5232 | void synchronize_net(void) |
1da177e4 LT |
5233 | { |
5234 | might_sleep(); | |
fbd568a3 | 5235 | synchronize_rcu(); |
1da177e4 | 5236 | } |
d1b19dff | 5237 | EXPORT_SYMBOL(synchronize_net); |
1da177e4 LT |
5238 | |
5239 | /** | |
5240 | * unregister_netdevice - remove device from the kernel | |
5241 | * @dev: device | |
5242 | * | |
5243 | * This function shuts down a device interface and removes it | |
d59b54b1 | 5244 | * from the kernel tables. |
1da177e4 LT |
5245 | * |
5246 | * Callers must hold the rtnl semaphore. You may want | |
5247 | * unregister_netdev() instead of this. | |
5248 | */ | |
5249 | ||
22f8cde5 | 5250 | void unregister_netdevice(struct net_device *dev) |
1da177e4 | 5251 | { |
a6620712 HX |
5252 | ASSERT_RTNL(); |
5253 | ||
93ee31f1 | 5254 | rollback_registered(dev); |
1da177e4 LT |
5255 | /* Finish processing unregister after unlock */ |
5256 | net_set_todo(dev); | |
1da177e4 | 5257 | } |
d1b19dff | 5258 | EXPORT_SYMBOL(unregister_netdevice); |
1da177e4 LT |
5259 | |
5260 | /** | |
5261 | * unregister_netdev - remove device from the kernel | |
5262 | * @dev: device | |
5263 | * | |
5264 | * This function shuts down a device interface and removes it | |
d59b54b1 | 5265 | * from the kernel tables. |
1da177e4 LT |
5266 | * |
5267 | * This is just a wrapper for unregister_netdevice that takes | |
5268 | * the rtnl semaphore. In general you want to use this and not | |
5269 | * unregister_netdevice. | |
5270 | */ | |
5271 | void unregister_netdev(struct net_device *dev) | |
5272 | { | |
5273 | rtnl_lock(); | |
5274 | unregister_netdevice(dev); | |
5275 | rtnl_unlock(); | |
5276 | } | |
1da177e4 LT |
5277 | EXPORT_SYMBOL(unregister_netdev); |
5278 | ||
ce286d32 EB |
5279 | /** |
5280 | * dev_change_net_namespace - move device to different nethost namespace | |
5281 | * @dev: device | |
5282 | * @net: network namespace | |
5283 | * @pat: If not NULL name pattern to try if the current device name | |
5284 | * is already taken in the destination network namespace. | |
5285 | * | |
5286 | * This function shuts down a device interface and moves it | |
5287 | * to a new network namespace. On success 0 is returned, on | |
5288 | * a failure a netagive errno code is returned. | |
5289 | * | |
5290 | * Callers must hold the rtnl semaphore. | |
5291 | */ | |
5292 | ||
5293 | int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat) | |
5294 | { | |
5295 | char buf[IFNAMSIZ]; | |
5296 | const char *destname; | |
5297 | int err; | |
5298 | ||
5299 | ASSERT_RTNL(); | |
5300 | ||
5301 | /* Don't allow namespace local devices to be moved. */ | |
5302 | err = -EINVAL; | |
5303 | if (dev->features & NETIF_F_NETNS_LOCAL) | |
5304 | goto out; | |
5305 | ||
3891845e EB |
5306 | #ifdef CONFIG_SYSFS |
5307 | /* Don't allow real devices to be moved when sysfs | |
5308 | * is enabled. | |
5309 | */ | |
5310 | err = -EINVAL; | |
5311 | if (dev->dev.parent) | |
5312 | goto out; | |
5313 | #endif | |
5314 | ||
ce286d32 EB |
5315 | /* Ensure the device has been registrered */ |
5316 | err = -EINVAL; | |
5317 | if (dev->reg_state != NETREG_REGISTERED) | |
5318 | goto out; | |
5319 | ||
5320 | /* Get out if there is nothing todo */ | |
5321 | err = 0; | |
878628fb | 5322 | if (net_eq(dev_net(dev), net)) |
ce286d32 EB |
5323 | goto out; |
5324 | ||
5325 | /* Pick the destination device name, and ensure | |
5326 | * we can use it in the destination network namespace. | |
5327 | */ | |
5328 | err = -EEXIST; | |
5329 | destname = dev->name; | |
5330 | if (__dev_get_by_name(net, destname)) { | |
5331 | /* We get here if we can't use the current device name */ | |
5332 | if (!pat) | |
5333 | goto out; | |
5334 | if (!dev_valid_name(pat)) | |
5335 | goto out; | |
5336 | if (strchr(pat, '%')) { | |
5337 | if (__dev_alloc_name(net, pat, buf) < 0) | |
5338 | goto out; | |
5339 | destname = buf; | |
5340 | } else | |
5341 | destname = pat; | |
5342 | if (__dev_get_by_name(net, destname)) | |
5343 | goto out; | |
5344 | } | |
5345 | ||
5346 | /* | |
5347 | * And now a mini version of register_netdevice unregister_netdevice. | |
5348 | */ | |
5349 | ||
5350 | /* If device is running close it first. */ | |
9b772652 | 5351 | dev_close(dev); |
ce286d32 EB |
5352 | |
5353 | /* And unlink it from device chain */ | |
5354 | err = -ENODEV; | |
5355 | unlist_netdevice(dev); | |
5356 | ||
5357 | synchronize_net(); | |
5358 | ||
5359 | /* Shutdown queueing discipline. */ | |
5360 | dev_shutdown(dev); | |
5361 | ||
5362 | /* Notify protocols, that we are about to destroy | |
5363 | this device. They should clean all the things. | |
5364 | */ | |
5365 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); | |
5366 | ||
5367 | /* | |
5368 | * Flush the unicast and multicast chains | |
5369 | */ | |
ccffad25 | 5370 | dev_unicast_flush(dev); |
ce286d32 EB |
5371 | dev_addr_discard(dev); |
5372 | ||
3891845e EB |
5373 | netdev_unregister_kobject(dev); |
5374 | ||
ce286d32 | 5375 | /* Actually switch the network namespace */ |
c346dca1 | 5376 | dev_net_set(dev, net); |
ce286d32 EB |
5377 | |
5378 | /* Assign the new device name */ | |
5379 | if (destname != dev->name) | |
5380 | strcpy(dev->name, destname); | |
5381 | ||
5382 | /* If there is an ifindex conflict assign a new one */ | |
5383 | if (__dev_get_by_index(net, dev->ifindex)) { | |
5384 | int iflink = (dev->iflink == dev->ifindex); | |
5385 | dev->ifindex = dev_new_index(net); | |
5386 | if (iflink) | |
5387 | dev->iflink = dev->ifindex; | |
5388 | } | |
5389 | ||
8b41d188 | 5390 | /* Fixup kobjects */ |
aaf8cdc3 | 5391 | err = netdev_register_kobject(dev); |
8b41d188 | 5392 | WARN_ON(err); |
ce286d32 EB |
5393 | |
5394 | /* Add the device back in the hashes */ | |
5395 | list_netdevice(dev); | |
5396 | ||
5397 | /* Notify protocols, that a new device appeared. */ | |
5398 | call_netdevice_notifiers(NETDEV_REGISTER, dev); | |
5399 | ||
5400 | synchronize_net(); | |
5401 | err = 0; | |
5402 | out: | |
5403 | return err; | |
5404 | } | |
463d0183 | 5405 | EXPORT_SYMBOL_GPL(dev_change_net_namespace); |
ce286d32 | 5406 | |
1da177e4 LT |
5407 | static int dev_cpu_callback(struct notifier_block *nfb, |
5408 | unsigned long action, | |
5409 | void *ocpu) | |
5410 | { | |
5411 | struct sk_buff **list_skb; | |
37437bb2 | 5412 | struct Qdisc **list_net; |
1da177e4 LT |
5413 | struct sk_buff *skb; |
5414 | unsigned int cpu, oldcpu = (unsigned long)ocpu; | |
5415 | struct softnet_data *sd, *oldsd; | |
5416 | ||
8bb78442 | 5417 | if (action != CPU_DEAD && action != CPU_DEAD_FROZEN) |
1da177e4 LT |
5418 | return NOTIFY_OK; |
5419 | ||
5420 | local_irq_disable(); | |
5421 | cpu = smp_processor_id(); | |
5422 | sd = &per_cpu(softnet_data, cpu); | |
5423 | oldsd = &per_cpu(softnet_data, oldcpu); | |
5424 | ||
5425 | /* Find end of our completion_queue. */ | |
5426 | list_skb = &sd->completion_queue; | |
5427 | while (*list_skb) | |
5428 | list_skb = &(*list_skb)->next; | |
5429 | /* Append completion queue from offline CPU. */ | |
5430 | *list_skb = oldsd->completion_queue; | |
5431 | oldsd->completion_queue = NULL; | |
5432 | ||
5433 | /* Find end of our output_queue. */ | |
5434 | list_net = &sd->output_queue; | |
5435 | while (*list_net) | |
5436 | list_net = &(*list_net)->next_sched; | |
5437 | /* Append output queue from offline CPU. */ | |
5438 | *list_net = oldsd->output_queue; | |
5439 | oldsd->output_queue = NULL; | |
5440 | ||
5441 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
5442 | local_irq_enable(); | |
5443 | ||
5444 | /* Process offline CPU's input_pkt_queue */ | |
5445 | while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) | |
5446 | netif_rx(skb); | |
5447 | ||
5448 | return NOTIFY_OK; | |
5449 | } | |
1da177e4 LT |
5450 | |
5451 | ||
7f353bf2 | 5452 | /** |
b63365a2 HX |
5453 | * netdev_increment_features - increment feature set by one |
5454 | * @all: current feature set | |
5455 | * @one: new feature set | |
5456 | * @mask: mask feature set | |
7f353bf2 HX |
5457 | * |
5458 | * Computes a new feature set after adding a device with feature set | |
b63365a2 HX |
5459 | * @one to the master device with current feature set @all. Will not |
5460 | * enable anything that is off in @mask. Returns the new feature set. | |
7f353bf2 | 5461 | */ |
b63365a2 HX |
5462 | unsigned long netdev_increment_features(unsigned long all, unsigned long one, |
5463 | unsigned long mask) | |
5464 | { | |
5465 | /* If device needs checksumming, downgrade to it. */ | |
d1b19dff | 5466 | if (all & NETIF_F_NO_CSUM && !(one & NETIF_F_NO_CSUM)) |
b63365a2 HX |
5467 | all ^= NETIF_F_NO_CSUM | (one & NETIF_F_ALL_CSUM); |
5468 | else if (mask & NETIF_F_ALL_CSUM) { | |
5469 | /* If one device supports v4/v6 checksumming, set for all. */ | |
5470 | if (one & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM) && | |
5471 | !(all & NETIF_F_GEN_CSUM)) { | |
5472 | all &= ~NETIF_F_ALL_CSUM; | |
5473 | all |= one & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM); | |
5474 | } | |
e2a6b852 | 5475 | |
b63365a2 HX |
5476 | /* If one device supports hw checksumming, set for all. */ |
5477 | if (one & NETIF_F_GEN_CSUM && !(all & NETIF_F_GEN_CSUM)) { | |
5478 | all &= ~NETIF_F_ALL_CSUM; | |
5479 | all |= NETIF_F_HW_CSUM; | |
5480 | } | |
5481 | } | |
7f353bf2 | 5482 | |
b63365a2 | 5483 | one |= NETIF_F_ALL_CSUM; |
7f353bf2 | 5484 | |
b63365a2 HX |
5485 | one |= all & NETIF_F_ONE_FOR_ALL; |
5486 | all &= one | NETIF_F_LLTX | NETIF_F_GSO; | |
5487 | all |= one & mask & NETIF_F_ONE_FOR_ALL; | |
7f353bf2 HX |
5488 | |
5489 | return all; | |
5490 | } | |
b63365a2 | 5491 | EXPORT_SYMBOL(netdev_increment_features); |
7f353bf2 | 5492 | |
30d97d35 PE |
5493 | static struct hlist_head *netdev_create_hash(void) |
5494 | { | |
5495 | int i; | |
5496 | struct hlist_head *hash; | |
5497 | ||
5498 | hash = kmalloc(sizeof(*hash) * NETDEV_HASHENTRIES, GFP_KERNEL); | |
5499 | if (hash != NULL) | |
5500 | for (i = 0; i < NETDEV_HASHENTRIES; i++) | |
5501 | INIT_HLIST_HEAD(&hash[i]); | |
5502 | ||
5503 | return hash; | |
5504 | } | |
5505 | ||
881d966b | 5506 | /* Initialize per network namespace state */ |
4665079c | 5507 | static int __net_init netdev_init(struct net *net) |
881d966b | 5508 | { |
881d966b | 5509 | INIT_LIST_HEAD(&net->dev_base_head); |
881d966b | 5510 | |
30d97d35 PE |
5511 | net->dev_name_head = netdev_create_hash(); |
5512 | if (net->dev_name_head == NULL) | |
5513 | goto err_name; | |
881d966b | 5514 | |
30d97d35 PE |
5515 | net->dev_index_head = netdev_create_hash(); |
5516 | if (net->dev_index_head == NULL) | |
5517 | goto err_idx; | |
881d966b EB |
5518 | |
5519 | return 0; | |
30d97d35 PE |
5520 | |
5521 | err_idx: | |
5522 | kfree(net->dev_name_head); | |
5523 | err_name: | |
5524 | return -ENOMEM; | |
881d966b EB |
5525 | } |
5526 | ||
f0db275a SH |
5527 | /** |
5528 | * netdev_drivername - network driver for the device | |
5529 | * @dev: network device | |
5530 | * @buffer: buffer for resulting name | |
5531 | * @len: size of buffer | |
5532 | * | |
5533 | * Determine network driver for device. | |
5534 | */ | |
cf04a4c7 | 5535 | char *netdev_drivername(const struct net_device *dev, char *buffer, int len) |
6579e57b | 5536 | { |
cf04a4c7 SH |
5537 | const struct device_driver *driver; |
5538 | const struct device *parent; | |
6579e57b AV |
5539 | |
5540 | if (len <= 0 || !buffer) | |
5541 | return buffer; | |
5542 | buffer[0] = 0; | |
5543 | ||
5544 | parent = dev->dev.parent; | |
5545 | ||
5546 | if (!parent) | |
5547 | return buffer; | |
5548 | ||
5549 | driver = parent->driver; | |
5550 | if (driver && driver->name) | |
5551 | strlcpy(buffer, driver->name, len); | |
5552 | return buffer; | |
5553 | } | |
5554 | ||
4665079c | 5555 | static void __net_exit netdev_exit(struct net *net) |
881d966b EB |
5556 | { |
5557 | kfree(net->dev_name_head); | |
5558 | kfree(net->dev_index_head); | |
5559 | } | |
5560 | ||
022cbae6 | 5561 | static struct pernet_operations __net_initdata netdev_net_ops = { |
881d966b EB |
5562 | .init = netdev_init, |
5563 | .exit = netdev_exit, | |
5564 | }; | |
5565 | ||
4665079c | 5566 | static void __net_exit default_device_exit(struct net *net) |
ce286d32 | 5567 | { |
8eb79863 | 5568 | struct net_device *dev; |
ce286d32 EB |
5569 | /* |
5570 | * Push all migratable of the network devices back to the | |
5571 | * initial network namespace | |
5572 | */ | |
5573 | rtnl_lock(); | |
8eb79863 EB |
5574 | restart: |
5575 | for_each_netdev(net, dev) { | |
ce286d32 | 5576 | int err; |
aca51397 | 5577 | char fb_name[IFNAMSIZ]; |
ce286d32 EB |
5578 | |
5579 | /* Ignore unmoveable devices (i.e. loopback) */ | |
5580 | if (dev->features & NETIF_F_NETNS_LOCAL) | |
5581 | continue; | |
5582 | ||
d0c082ce EB |
5583 | /* Delete virtual devices */ |
5584 | if (dev->rtnl_link_ops && dev->rtnl_link_ops->dellink) { | |
5585 | dev->rtnl_link_ops->dellink(dev); | |
8eb79863 | 5586 | goto restart; |
d0c082ce EB |
5587 | } |
5588 | ||
ce286d32 | 5589 | /* Push remaing network devices to init_net */ |
aca51397 PE |
5590 | snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex); |
5591 | err = dev_change_net_namespace(dev, &init_net, fb_name); | |
ce286d32 | 5592 | if (err) { |
aca51397 | 5593 | printk(KERN_EMERG "%s: failed to move %s to init_net: %d\n", |
ce286d32 | 5594 | __func__, dev->name, err); |
aca51397 | 5595 | BUG(); |
ce286d32 | 5596 | } |
8eb79863 | 5597 | goto restart; |
ce286d32 EB |
5598 | } |
5599 | rtnl_unlock(); | |
5600 | } | |
5601 | ||
022cbae6 | 5602 | static struct pernet_operations __net_initdata default_device_ops = { |
ce286d32 EB |
5603 | .exit = default_device_exit, |
5604 | }; | |
5605 | ||
1da177e4 LT |
5606 | /* |
5607 | * Initialize the DEV module. At boot time this walks the device list and | |
5608 | * unhooks any devices that fail to initialise (normally hardware not | |
5609 | * present) and leaves us with a valid list of present and active devices. | |
5610 | * | |
5611 | */ | |
5612 | ||
5613 | /* | |
5614 | * This is called single threaded during boot, so no need | |
5615 | * to take the rtnl semaphore. | |
5616 | */ | |
5617 | static int __init net_dev_init(void) | |
5618 | { | |
5619 | int i, rc = -ENOMEM; | |
5620 | ||
5621 | BUG_ON(!dev_boot_phase); | |
5622 | ||
1da177e4 LT |
5623 | if (dev_proc_init()) |
5624 | goto out; | |
5625 | ||
8b41d188 | 5626 | if (netdev_kobject_init()) |
1da177e4 LT |
5627 | goto out; |
5628 | ||
5629 | INIT_LIST_HEAD(&ptype_all); | |
82d8a867 | 5630 | for (i = 0; i < PTYPE_HASH_SIZE; i++) |
1da177e4 LT |
5631 | INIT_LIST_HEAD(&ptype_base[i]); |
5632 | ||
881d966b EB |
5633 | if (register_pernet_subsys(&netdev_net_ops)) |
5634 | goto out; | |
1da177e4 LT |
5635 | |
5636 | /* | |
5637 | * Initialise the packet receive queues. | |
5638 | */ | |
5639 | ||
6f912042 | 5640 | for_each_possible_cpu(i) { |
1da177e4 LT |
5641 | struct softnet_data *queue; |
5642 | ||
5643 | queue = &per_cpu(softnet_data, i); | |
5644 | skb_queue_head_init(&queue->input_pkt_queue); | |
1da177e4 LT |
5645 | queue->completion_queue = NULL; |
5646 | INIT_LIST_HEAD(&queue->poll_list); | |
bea3348e SH |
5647 | |
5648 | queue->backlog.poll = process_backlog; | |
5649 | queue->backlog.weight = weight_p; | |
d565b0a1 | 5650 | queue->backlog.gro_list = NULL; |
4ae5544f | 5651 | queue->backlog.gro_count = 0; |
1da177e4 LT |
5652 | } |
5653 | ||
1da177e4 LT |
5654 | dev_boot_phase = 0; |
5655 | ||
505d4f73 EB |
5656 | /* The loopback device is special if any other network devices |
5657 | * is present in a network namespace the loopback device must | |
5658 | * be present. Since we now dynamically allocate and free the | |
5659 | * loopback device ensure this invariant is maintained by | |
5660 | * keeping the loopback device as the first device on the | |
5661 | * list of network devices. Ensuring the loopback devices | |
5662 | * is the first device that appears and the last network device | |
5663 | * that disappears. | |
5664 | */ | |
5665 | if (register_pernet_device(&loopback_net_ops)) | |
5666 | goto out; | |
5667 | ||
5668 | if (register_pernet_device(&default_device_ops)) | |
5669 | goto out; | |
5670 | ||
962cf36c CM |
5671 | open_softirq(NET_TX_SOFTIRQ, net_tx_action); |
5672 | open_softirq(NET_RX_SOFTIRQ, net_rx_action); | |
1da177e4 LT |
5673 | |
5674 | hotcpu_notifier(dev_cpu_callback, 0); | |
5675 | dst_init(); | |
5676 | dev_mcast_init(); | |
5677 | rc = 0; | |
5678 | out: | |
5679 | return rc; | |
5680 | } | |
5681 | ||
5682 | subsys_initcall(net_dev_init); | |
5683 | ||
e88721f8 KK |
5684 | static int __init initialize_hashrnd(void) |
5685 | { | |
5686 | get_random_bytes(&skb_tx_hashrnd, sizeof(skb_tx_hashrnd)); | |
5687 | return 0; | |
5688 | } | |
5689 | ||
5690 | late_initcall_sync(initialize_hashrnd); | |
5691 |