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> | |
111 | #include <linux/kallsyms.h> | |
112 | #include <linux/netpoll.h> | |
113 | #include <linux/rcupdate.h> | |
114 | #include <linux/delay.h> | |
295f4a1f | 115 | #include <net/wext.h> |
1da177e4 | 116 | #include <net/iw_handler.h> |
1da177e4 | 117 | #include <asm/current.h> |
5bdb9886 | 118 | #include <linux/audit.h> |
db217334 | 119 | #include <linux/dmaengine.h> |
f6a78bfc | 120 | #include <linux/err.h> |
c7fa9d18 | 121 | #include <linux/ctype.h> |
723e98b7 | 122 | #include <linux/if_arp.h> |
6de329e2 | 123 | #include <linux/if_vlan.h> |
1da177e4 | 124 | |
342709ef PE |
125 | #include "net-sysfs.h" |
126 | ||
1da177e4 LT |
127 | /* |
128 | * The list of packet types we will receive (as opposed to discard) | |
129 | * and the routines to invoke. | |
130 | * | |
131 | * Why 16. Because with 16 the only overlap we get on a hash of the | |
132 | * low nibble of the protocol value is RARP/SNAP/X.25. | |
133 | * | |
134 | * NOTE: That is no longer true with the addition of VLAN tags. Not | |
135 | * sure which should go first, but I bet it won't make much | |
136 | * difference if we are running VLANs. The good news is that | |
137 | * this protocol won't be in the list unless compiled in, so | |
3041a069 | 138 | * the average user (w/out VLANs) will not be adversely affected. |
1da177e4 LT |
139 | * --BLG |
140 | * | |
141 | * 0800 IP | |
142 | * 8100 802.1Q VLAN | |
143 | * 0001 802.3 | |
144 | * 0002 AX.25 | |
145 | * 0004 802.2 | |
146 | * 8035 RARP | |
147 | * 0005 SNAP | |
148 | * 0805 X.25 | |
149 | * 0806 ARP | |
150 | * 8137 IPX | |
151 | * 0009 Localtalk | |
152 | * 86DD IPv6 | |
153 | */ | |
154 | ||
82d8a867 PE |
155 | #define PTYPE_HASH_SIZE (16) |
156 | #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) | |
157 | ||
1da177e4 | 158 | static DEFINE_SPINLOCK(ptype_lock); |
82d8a867 | 159 | static struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; |
6b2bedc3 | 160 | static struct list_head ptype_all __read_mostly; /* Taps */ |
1da177e4 | 161 | |
db217334 | 162 | #ifdef CONFIG_NET_DMA |
d379b01e DW |
163 | struct net_dma { |
164 | struct dma_client client; | |
165 | spinlock_t lock; | |
166 | cpumask_t channel_mask; | |
0c0b0aca | 167 | struct dma_chan **channels; |
d379b01e DW |
168 | }; |
169 | ||
170 | static enum dma_state_client | |
171 | netdev_dma_event(struct dma_client *client, struct dma_chan *chan, | |
172 | enum dma_state state); | |
173 | ||
174 | static struct net_dma net_dma = { | |
175 | .client = { | |
176 | .event_callback = netdev_dma_event, | |
177 | }, | |
178 | }; | |
db217334 CL |
179 | #endif |
180 | ||
1da177e4 | 181 | /* |
7562f876 | 182 | * The @dev_base_head list is protected by @dev_base_lock and the rtnl |
1da177e4 LT |
183 | * semaphore. |
184 | * | |
185 | * Pure readers hold dev_base_lock for reading. | |
186 | * | |
187 | * Writers must hold the rtnl semaphore while they loop through the | |
7562f876 | 188 | * dev_base_head list, and hold dev_base_lock for writing when they do the |
1da177e4 LT |
189 | * actual updates. This allows pure readers to access the list even |
190 | * while a writer is preparing to update it. | |
191 | * | |
192 | * To put it another way, dev_base_lock is held for writing only to | |
193 | * protect against pure readers; the rtnl semaphore provides the | |
194 | * protection against other writers. | |
195 | * | |
196 | * See, for example usages, register_netdevice() and | |
197 | * unregister_netdevice(), which must be called with the rtnl | |
198 | * semaphore held. | |
199 | */ | |
1da177e4 LT |
200 | DEFINE_RWLOCK(dev_base_lock); |
201 | ||
1da177e4 LT |
202 | EXPORT_SYMBOL(dev_base_lock); |
203 | ||
204 | #define NETDEV_HASHBITS 8 | |
881d966b | 205 | #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS) |
1da177e4 | 206 | |
881d966b | 207 | static inline struct hlist_head *dev_name_hash(struct net *net, const char *name) |
1da177e4 LT |
208 | { |
209 | unsigned hash = full_name_hash(name, strnlen(name, IFNAMSIZ)); | |
881d966b | 210 | return &net->dev_name_head[hash & ((1 << NETDEV_HASHBITS) - 1)]; |
1da177e4 LT |
211 | } |
212 | ||
881d966b | 213 | static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex) |
1da177e4 | 214 | { |
881d966b | 215 | return &net->dev_index_head[ifindex & ((1 << NETDEV_HASHBITS) - 1)]; |
1da177e4 LT |
216 | } |
217 | ||
ce286d32 EB |
218 | /* Device list insertion */ |
219 | static int list_netdevice(struct net_device *dev) | |
220 | { | |
c346dca1 | 221 | struct net *net = dev_net(dev); |
ce286d32 EB |
222 | |
223 | ASSERT_RTNL(); | |
224 | ||
225 | write_lock_bh(&dev_base_lock); | |
226 | list_add_tail(&dev->dev_list, &net->dev_base_head); | |
227 | hlist_add_head(&dev->name_hlist, dev_name_hash(net, dev->name)); | |
228 | hlist_add_head(&dev->index_hlist, dev_index_hash(net, dev->ifindex)); | |
229 | write_unlock_bh(&dev_base_lock); | |
230 | return 0; | |
231 | } | |
232 | ||
233 | /* Device list removal */ | |
234 | static void unlist_netdevice(struct net_device *dev) | |
235 | { | |
236 | ASSERT_RTNL(); | |
237 | ||
238 | /* Unlink dev from the device chain */ | |
239 | write_lock_bh(&dev_base_lock); | |
240 | list_del(&dev->dev_list); | |
241 | hlist_del(&dev->name_hlist); | |
242 | hlist_del(&dev->index_hlist); | |
243 | write_unlock_bh(&dev_base_lock); | |
244 | } | |
245 | ||
1da177e4 LT |
246 | /* |
247 | * Our notifier list | |
248 | */ | |
249 | ||
f07d5b94 | 250 | static RAW_NOTIFIER_HEAD(netdev_chain); |
1da177e4 LT |
251 | |
252 | /* | |
253 | * Device drivers call our routines to queue packets here. We empty the | |
254 | * queue in the local softnet handler. | |
255 | */ | |
bea3348e SH |
256 | |
257 | DEFINE_PER_CPU(struct softnet_data, softnet_data); | |
1da177e4 | 258 | |
723e98b7 JP |
259 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
260 | /* | |
c773e847 | 261 | * register_netdevice() inits txq->_xmit_lock and sets lockdep class |
723e98b7 JP |
262 | * according to dev->type |
263 | */ | |
264 | static const unsigned short netdev_lock_type[] = | |
265 | {ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25, | |
266 | ARPHRD_PRONET, ARPHRD_CHAOS, ARPHRD_IEEE802, ARPHRD_ARCNET, | |
267 | ARPHRD_APPLETLK, ARPHRD_DLCI, ARPHRD_ATM, ARPHRD_METRICOM, | |
268 | ARPHRD_IEEE1394, ARPHRD_EUI64, ARPHRD_INFINIBAND, ARPHRD_SLIP, | |
269 | ARPHRD_CSLIP, ARPHRD_SLIP6, ARPHRD_CSLIP6, ARPHRD_RSRVD, | |
270 | ARPHRD_ADAPT, ARPHRD_ROSE, ARPHRD_X25, ARPHRD_HWX25, | |
271 | ARPHRD_PPP, ARPHRD_CISCO, ARPHRD_LAPB, ARPHRD_DDCMP, | |
272 | ARPHRD_RAWHDLC, ARPHRD_TUNNEL, ARPHRD_TUNNEL6, ARPHRD_FRAD, | |
273 | ARPHRD_SKIP, ARPHRD_LOOPBACK, ARPHRD_LOCALTLK, ARPHRD_FDDI, | |
274 | ARPHRD_BIF, ARPHRD_SIT, ARPHRD_IPDDP, ARPHRD_IPGRE, | |
275 | ARPHRD_PIMREG, ARPHRD_HIPPI, ARPHRD_ASH, ARPHRD_ECONET, | |
276 | ARPHRD_IRDA, ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL, | |
277 | ARPHRD_FCFABRIC, ARPHRD_IEEE802_TR, ARPHRD_IEEE80211, | |
278 | ARPHRD_IEEE80211_PRISM, ARPHRD_IEEE80211_RADIOTAP, ARPHRD_VOID, | |
279 | ARPHRD_NONE}; | |
280 | ||
281 | static const char *netdev_lock_name[] = | |
282 | {"_xmit_NETROM", "_xmit_ETHER", "_xmit_EETHER", "_xmit_AX25", | |
283 | "_xmit_PRONET", "_xmit_CHAOS", "_xmit_IEEE802", "_xmit_ARCNET", | |
284 | "_xmit_APPLETLK", "_xmit_DLCI", "_xmit_ATM", "_xmit_METRICOM", | |
285 | "_xmit_IEEE1394", "_xmit_EUI64", "_xmit_INFINIBAND", "_xmit_SLIP", | |
286 | "_xmit_CSLIP", "_xmit_SLIP6", "_xmit_CSLIP6", "_xmit_RSRVD", | |
287 | "_xmit_ADAPT", "_xmit_ROSE", "_xmit_X25", "_xmit_HWX25", | |
288 | "_xmit_PPP", "_xmit_CISCO", "_xmit_LAPB", "_xmit_DDCMP", | |
289 | "_xmit_RAWHDLC", "_xmit_TUNNEL", "_xmit_TUNNEL6", "_xmit_FRAD", | |
290 | "_xmit_SKIP", "_xmit_LOOPBACK", "_xmit_LOCALTLK", "_xmit_FDDI", | |
291 | "_xmit_BIF", "_xmit_SIT", "_xmit_IPDDP", "_xmit_IPGRE", | |
292 | "_xmit_PIMREG", "_xmit_HIPPI", "_xmit_ASH", "_xmit_ECONET", | |
293 | "_xmit_IRDA", "_xmit_FCPP", "_xmit_FCAL", "_xmit_FCPL", | |
294 | "_xmit_FCFABRIC", "_xmit_IEEE802_TR", "_xmit_IEEE80211", | |
295 | "_xmit_IEEE80211_PRISM", "_xmit_IEEE80211_RADIOTAP", "_xmit_VOID", | |
296 | "_xmit_NONE"}; | |
297 | ||
298 | static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)]; | |
299 | ||
300 | static inline unsigned short netdev_lock_pos(unsigned short dev_type) | |
301 | { | |
302 | int i; | |
303 | ||
304 | for (i = 0; i < ARRAY_SIZE(netdev_lock_type); i++) | |
305 | if (netdev_lock_type[i] == dev_type) | |
306 | return i; | |
307 | /* the last key is used by default */ | |
308 | return ARRAY_SIZE(netdev_lock_type) - 1; | |
309 | } | |
310 | ||
311 | static inline void netdev_set_lockdep_class(spinlock_t *lock, | |
312 | unsigned short dev_type) | |
313 | { | |
314 | int i; | |
315 | ||
316 | i = netdev_lock_pos(dev_type); | |
317 | lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i], | |
318 | netdev_lock_name[i]); | |
319 | } | |
320 | #else | |
321 | static inline void netdev_set_lockdep_class(spinlock_t *lock, | |
322 | unsigned short dev_type) | |
323 | { | |
324 | } | |
325 | #endif | |
1da177e4 LT |
326 | |
327 | /******************************************************************************* | |
328 | ||
329 | Protocol management and registration routines | |
330 | ||
331 | *******************************************************************************/ | |
332 | ||
1da177e4 LT |
333 | /* |
334 | * Add a protocol ID to the list. Now that the input handler is | |
335 | * smarter we can dispense with all the messy stuff that used to be | |
336 | * here. | |
337 | * | |
338 | * BEWARE!!! Protocol handlers, mangling input packets, | |
339 | * MUST BE last in hash buckets and checking protocol handlers | |
340 | * MUST start from promiscuous ptype_all chain in net_bh. | |
341 | * It is true now, do not change it. | |
342 | * Explanation follows: if protocol handler, mangling packet, will | |
343 | * be the first on list, it is not able to sense, that packet | |
344 | * is cloned and should be copied-on-write, so that it will | |
345 | * change it and subsequent readers will get broken packet. | |
346 | * --ANK (980803) | |
347 | */ | |
348 | ||
349 | /** | |
350 | * dev_add_pack - add packet handler | |
351 | * @pt: packet type declaration | |
352 | * | |
353 | * Add a protocol handler to the networking stack. The passed &packet_type | |
354 | * is linked into kernel lists and may not be freed until it has been | |
355 | * removed from the kernel lists. | |
356 | * | |
4ec93edb | 357 | * This call does not sleep therefore it can not |
1da177e4 LT |
358 | * guarantee all CPU's that are in middle of receiving packets |
359 | * will see the new packet type (until the next received packet). | |
360 | */ | |
361 | ||
362 | void dev_add_pack(struct packet_type *pt) | |
363 | { | |
364 | int hash; | |
365 | ||
366 | spin_lock_bh(&ptype_lock); | |
9be9a6b9 | 367 | if (pt->type == htons(ETH_P_ALL)) |
1da177e4 | 368 | list_add_rcu(&pt->list, &ptype_all); |
9be9a6b9 | 369 | else { |
82d8a867 | 370 | hash = ntohs(pt->type) & PTYPE_HASH_MASK; |
1da177e4 LT |
371 | list_add_rcu(&pt->list, &ptype_base[hash]); |
372 | } | |
373 | spin_unlock_bh(&ptype_lock); | |
374 | } | |
375 | ||
1da177e4 LT |
376 | /** |
377 | * __dev_remove_pack - remove packet handler | |
378 | * @pt: packet type declaration | |
379 | * | |
380 | * Remove a protocol handler that was previously added to the kernel | |
381 | * protocol handlers by dev_add_pack(). The passed &packet_type is removed | |
382 | * from the kernel lists and can be freed or reused once this function | |
4ec93edb | 383 | * returns. |
1da177e4 LT |
384 | * |
385 | * The packet type might still be in use by receivers | |
386 | * and must not be freed until after all the CPU's have gone | |
387 | * through a quiescent state. | |
388 | */ | |
389 | void __dev_remove_pack(struct packet_type *pt) | |
390 | { | |
391 | struct list_head *head; | |
392 | struct packet_type *pt1; | |
393 | ||
394 | spin_lock_bh(&ptype_lock); | |
395 | ||
9be9a6b9 | 396 | if (pt->type == htons(ETH_P_ALL)) |
1da177e4 | 397 | head = &ptype_all; |
9be9a6b9 | 398 | else |
82d8a867 | 399 | head = &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK]; |
1da177e4 LT |
400 | |
401 | list_for_each_entry(pt1, head, list) { | |
402 | if (pt == pt1) { | |
403 | list_del_rcu(&pt->list); | |
404 | goto out; | |
405 | } | |
406 | } | |
407 | ||
408 | printk(KERN_WARNING "dev_remove_pack: %p not found.\n", pt); | |
409 | out: | |
410 | spin_unlock_bh(&ptype_lock); | |
411 | } | |
412 | /** | |
413 | * dev_remove_pack - remove packet handler | |
414 | * @pt: packet type declaration | |
415 | * | |
416 | * Remove a protocol handler that was previously added to the kernel | |
417 | * protocol handlers by dev_add_pack(). The passed &packet_type is removed | |
418 | * from the kernel lists and can be freed or reused once this function | |
419 | * returns. | |
420 | * | |
421 | * This call sleeps to guarantee that no CPU is looking at the packet | |
422 | * type after return. | |
423 | */ | |
424 | void dev_remove_pack(struct packet_type *pt) | |
425 | { | |
426 | __dev_remove_pack(pt); | |
4ec93edb | 427 | |
1da177e4 LT |
428 | synchronize_net(); |
429 | } | |
430 | ||
431 | /****************************************************************************** | |
432 | ||
433 | Device Boot-time Settings Routines | |
434 | ||
435 | *******************************************************************************/ | |
436 | ||
437 | /* Boot time configuration table */ | |
438 | static struct netdev_boot_setup dev_boot_setup[NETDEV_BOOT_SETUP_MAX]; | |
439 | ||
440 | /** | |
441 | * netdev_boot_setup_add - add new setup entry | |
442 | * @name: name of the device | |
443 | * @map: configured settings for the device | |
444 | * | |
445 | * Adds new setup entry to the dev_boot_setup list. The function | |
446 | * returns 0 on error and 1 on success. This is a generic routine to | |
447 | * all netdevices. | |
448 | */ | |
449 | static int netdev_boot_setup_add(char *name, struct ifmap *map) | |
450 | { | |
451 | struct netdev_boot_setup *s; | |
452 | int i; | |
453 | ||
454 | s = dev_boot_setup; | |
455 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { | |
456 | if (s[i].name[0] == '\0' || s[i].name[0] == ' ') { | |
457 | memset(s[i].name, 0, sizeof(s[i].name)); | |
93b3cff9 | 458 | strlcpy(s[i].name, name, IFNAMSIZ); |
1da177e4 LT |
459 | memcpy(&s[i].map, map, sizeof(s[i].map)); |
460 | break; | |
461 | } | |
462 | } | |
463 | ||
464 | return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1; | |
465 | } | |
466 | ||
467 | /** | |
468 | * netdev_boot_setup_check - check boot time settings | |
469 | * @dev: the netdevice | |
470 | * | |
471 | * Check boot time settings for the device. | |
472 | * The found settings are set for the device to be used | |
473 | * later in the device probing. | |
474 | * Returns 0 if no settings found, 1 if they are. | |
475 | */ | |
476 | int netdev_boot_setup_check(struct net_device *dev) | |
477 | { | |
478 | struct netdev_boot_setup *s = dev_boot_setup; | |
479 | int i; | |
480 | ||
481 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { | |
482 | if (s[i].name[0] != '\0' && s[i].name[0] != ' ' && | |
93b3cff9 | 483 | !strcmp(dev->name, s[i].name)) { |
1da177e4 LT |
484 | dev->irq = s[i].map.irq; |
485 | dev->base_addr = s[i].map.base_addr; | |
486 | dev->mem_start = s[i].map.mem_start; | |
487 | dev->mem_end = s[i].map.mem_end; | |
488 | return 1; | |
489 | } | |
490 | } | |
491 | return 0; | |
492 | } | |
493 | ||
494 | ||
495 | /** | |
496 | * netdev_boot_base - get address from boot time settings | |
497 | * @prefix: prefix for network device | |
498 | * @unit: id for network device | |
499 | * | |
500 | * Check boot time settings for the base address of device. | |
501 | * The found settings are set for the device to be used | |
502 | * later in the device probing. | |
503 | * Returns 0 if no settings found. | |
504 | */ | |
505 | unsigned long netdev_boot_base(const char *prefix, int unit) | |
506 | { | |
507 | const struct netdev_boot_setup *s = dev_boot_setup; | |
508 | char name[IFNAMSIZ]; | |
509 | int i; | |
510 | ||
511 | sprintf(name, "%s%d", prefix, unit); | |
512 | ||
513 | /* | |
514 | * If device already registered then return base of 1 | |
515 | * to indicate not to probe for this interface | |
516 | */ | |
881d966b | 517 | if (__dev_get_by_name(&init_net, name)) |
1da177e4 LT |
518 | return 1; |
519 | ||
520 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) | |
521 | if (!strcmp(name, s[i].name)) | |
522 | return s[i].map.base_addr; | |
523 | return 0; | |
524 | } | |
525 | ||
526 | /* | |
527 | * Saves at boot time configured settings for any netdevice. | |
528 | */ | |
529 | int __init netdev_boot_setup(char *str) | |
530 | { | |
531 | int ints[5]; | |
532 | struct ifmap map; | |
533 | ||
534 | str = get_options(str, ARRAY_SIZE(ints), ints); | |
535 | if (!str || !*str) | |
536 | return 0; | |
537 | ||
538 | /* Save settings */ | |
539 | memset(&map, 0, sizeof(map)); | |
540 | if (ints[0] > 0) | |
541 | map.irq = ints[1]; | |
542 | if (ints[0] > 1) | |
543 | map.base_addr = ints[2]; | |
544 | if (ints[0] > 2) | |
545 | map.mem_start = ints[3]; | |
546 | if (ints[0] > 3) | |
547 | map.mem_end = ints[4]; | |
548 | ||
549 | /* Add new entry to the list */ | |
550 | return netdev_boot_setup_add(str, &map); | |
551 | } | |
552 | ||
553 | __setup("netdev=", netdev_boot_setup); | |
554 | ||
555 | /******************************************************************************* | |
556 | ||
557 | Device Interface Subroutines | |
558 | ||
559 | *******************************************************************************/ | |
560 | ||
561 | /** | |
562 | * __dev_get_by_name - find a device by its name | |
c4ea43c5 | 563 | * @net: the applicable net namespace |
1da177e4 LT |
564 | * @name: name to find |
565 | * | |
566 | * Find an interface by name. Must be called under RTNL semaphore | |
567 | * or @dev_base_lock. If the name is found a pointer to the device | |
568 | * is returned. If the name is not found then %NULL is returned. The | |
569 | * reference counters are not incremented so the caller must be | |
570 | * careful with locks. | |
571 | */ | |
572 | ||
881d966b | 573 | struct net_device *__dev_get_by_name(struct net *net, const char *name) |
1da177e4 LT |
574 | { |
575 | struct hlist_node *p; | |
576 | ||
881d966b | 577 | hlist_for_each(p, dev_name_hash(net, name)) { |
1da177e4 LT |
578 | struct net_device *dev |
579 | = hlist_entry(p, struct net_device, name_hlist); | |
580 | if (!strncmp(dev->name, name, IFNAMSIZ)) | |
581 | return dev; | |
582 | } | |
583 | return NULL; | |
584 | } | |
585 | ||
586 | /** | |
587 | * dev_get_by_name - find a device by its name | |
c4ea43c5 | 588 | * @net: the applicable net namespace |
1da177e4 LT |
589 | * @name: name to find |
590 | * | |
591 | * Find an interface by name. This can be called from any | |
592 | * context and does its own locking. The returned handle has | |
593 | * the usage count incremented and the caller must use dev_put() to | |
594 | * release it when it is no longer needed. %NULL is returned if no | |
595 | * matching device is found. | |
596 | */ | |
597 | ||
881d966b | 598 | struct net_device *dev_get_by_name(struct net *net, const char *name) |
1da177e4 LT |
599 | { |
600 | struct net_device *dev; | |
601 | ||
602 | read_lock(&dev_base_lock); | |
881d966b | 603 | dev = __dev_get_by_name(net, name); |
1da177e4 LT |
604 | if (dev) |
605 | dev_hold(dev); | |
606 | read_unlock(&dev_base_lock); | |
607 | return dev; | |
608 | } | |
609 | ||
610 | /** | |
611 | * __dev_get_by_index - find a device by its ifindex | |
c4ea43c5 | 612 | * @net: the applicable net namespace |
1da177e4 LT |
613 | * @ifindex: index of device |
614 | * | |
615 | * Search for an interface by index. Returns %NULL if the device | |
616 | * is not found or a pointer to the device. The device has not | |
617 | * had its reference counter increased so the caller must be careful | |
618 | * about locking. The caller must hold either the RTNL semaphore | |
619 | * or @dev_base_lock. | |
620 | */ | |
621 | ||
881d966b | 622 | struct net_device *__dev_get_by_index(struct net *net, int ifindex) |
1da177e4 LT |
623 | { |
624 | struct hlist_node *p; | |
625 | ||
881d966b | 626 | hlist_for_each(p, dev_index_hash(net, ifindex)) { |
1da177e4 LT |
627 | struct net_device *dev |
628 | = hlist_entry(p, struct net_device, index_hlist); | |
629 | if (dev->ifindex == ifindex) | |
630 | return dev; | |
631 | } | |
632 | return NULL; | |
633 | } | |
634 | ||
635 | ||
636 | /** | |
637 | * dev_get_by_index - find a device by its ifindex | |
c4ea43c5 | 638 | * @net: the applicable net namespace |
1da177e4 LT |
639 | * @ifindex: index of device |
640 | * | |
641 | * Search for an interface by index. Returns NULL if the device | |
642 | * is not found or a pointer to the device. The device returned has | |
643 | * had a reference added and the pointer is safe until the user calls | |
644 | * dev_put to indicate they have finished with it. | |
645 | */ | |
646 | ||
881d966b | 647 | struct net_device *dev_get_by_index(struct net *net, int ifindex) |
1da177e4 LT |
648 | { |
649 | struct net_device *dev; | |
650 | ||
651 | read_lock(&dev_base_lock); | |
881d966b | 652 | dev = __dev_get_by_index(net, ifindex); |
1da177e4 LT |
653 | if (dev) |
654 | dev_hold(dev); | |
655 | read_unlock(&dev_base_lock); | |
656 | return dev; | |
657 | } | |
658 | ||
659 | /** | |
660 | * dev_getbyhwaddr - find a device by its hardware address | |
c4ea43c5 | 661 | * @net: the applicable net namespace |
1da177e4 LT |
662 | * @type: media type of device |
663 | * @ha: hardware address | |
664 | * | |
665 | * Search for an interface by MAC address. Returns NULL if the device | |
666 | * is not found or a pointer to the device. The caller must hold the | |
667 | * rtnl semaphore. The returned device has not had its ref count increased | |
668 | * and the caller must therefore be careful about locking | |
669 | * | |
670 | * BUGS: | |
671 | * If the API was consistent this would be __dev_get_by_hwaddr | |
672 | */ | |
673 | ||
881d966b | 674 | struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, char *ha) |
1da177e4 LT |
675 | { |
676 | struct net_device *dev; | |
677 | ||
678 | ASSERT_RTNL(); | |
679 | ||
81103a52 | 680 | for_each_netdev(net, dev) |
1da177e4 LT |
681 | if (dev->type == type && |
682 | !memcmp(dev->dev_addr, ha, dev->addr_len)) | |
7562f876 PE |
683 | return dev; |
684 | ||
685 | return NULL; | |
1da177e4 LT |
686 | } |
687 | ||
cf309e3f JF |
688 | EXPORT_SYMBOL(dev_getbyhwaddr); |
689 | ||
881d966b | 690 | struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type) |
1da177e4 LT |
691 | { |
692 | struct net_device *dev; | |
693 | ||
4e9cac2b | 694 | ASSERT_RTNL(); |
881d966b | 695 | for_each_netdev(net, dev) |
4e9cac2b | 696 | if (dev->type == type) |
7562f876 PE |
697 | return dev; |
698 | ||
699 | return NULL; | |
4e9cac2b PM |
700 | } |
701 | ||
702 | EXPORT_SYMBOL(__dev_getfirstbyhwtype); | |
703 | ||
881d966b | 704 | struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type) |
4e9cac2b PM |
705 | { |
706 | struct net_device *dev; | |
707 | ||
708 | rtnl_lock(); | |
881d966b | 709 | dev = __dev_getfirstbyhwtype(net, type); |
4e9cac2b PM |
710 | if (dev) |
711 | dev_hold(dev); | |
1da177e4 LT |
712 | rtnl_unlock(); |
713 | return dev; | |
714 | } | |
715 | ||
716 | EXPORT_SYMBOL(dev_getfirstbyhwtype); | |
717 | ||
718 | /** | |
719 | * dev_get_by_flags - find any device with given flags | |
c4ea43c5 | 720 | * @net: the applicable net namespace |
1da177e4 LT |
721 | * @if_flags: IFF_* values |
722 | * @mask: bitmask of bits in if_flags to check | |
723 | * | |
724 | * Search for any interface with the given flags. Returns NULL if a device | |
4ec93edb | 725 | * is not found or a pointer to the device. The device returned has |
1da177e4 LT |
726 | * had a reference added and the pointer is safe until the user calls |
727 | * dev_put to indicate they have finished with it. | |
728 | */ | |
729 | ||
881d966b | 730 | struct net_device * dev_get_by_flags(struct net *net, unsigned short if_flags, unsigned short mask) |
1da177e4 | 731 | { |
7562f876 | 732 | struct net_device *dev, *ret; |
1da177e4 | 733 | |
7562f876 | 734 | ret = NULL; |
1da177e4 | 735 | read_lock(&dev_base_lock); |
881d966b | 736 | for_each_netdev(net, dev) { |
1da177e4 LT |
737 | if (((dev->flags ^ if_flags) & mask) == 0) { |
738 | dev_hold(dev); | |
7562f876 | 739 | ret = dev; |
1da177e4 LT |
740 | break; |
741 | } | |
742 | } | |
743 | read_unlock(&dev_base_lock); | |
7562f876 | 744 | return ret; |
1da177e4 LT |
745 | } |
746 | ||
747 | /** | |
748 | * dev_valid_name - check if name is okay for network device | |
749 | * @name: name string | |
750 | * | |
751 | * Network device names need to be valid file names to | |
c7fa9d18 DM |
752 | * to allow sysfs to work. We also disallow any kind of |
753 | * whitespace. | |
1da177e4 | 754 | */ |
c2373ee9 | 755 | int dev_valid_name(const char *name) |
1da177e4 | 756 | { |
c7fa9d18 DM |
757 | if (*name == '\0') |
758 | return 0; | |
b6fe17d6 SH |
759 | if (strlen(name) >= IFNAMSIZ) |
760 | return 0; | |
c7fa9d18 DM |
761 | if (!strcmp(name, ".") || !strcmp(name, "..")) |
762 | return 0; | |
763 | ||
764 | while (*name) { | |
765 | if (*name == '/' || isspace(*name)) | |
766 | return 0; | |
767 | name++; | |
768 | } | |
769 | return 1; | |
1da177e4 LT |
770 | } |
771 | ||
772 | /** | |
b267b179 EB |
773 | * __dev_alloc_name - allocate a name for a device |
774 | * @net: network namespace to allocate the device name in | |
1da177e4 | 775 | * @name: name format string |
b267b179 | 776 | * @buf: scratch buffer and result name string |
1da177e4 LT |
777 | * |
778 | * Passed a format string - eg "lt%d" it will try and find a suitable | |
3041a069 SH |
779 | * id. It scans list of devices to build up a free map, then chooses |
780 | * the first empty slot. The caller must hold the dev_base or rtnl lock | |
781 | * while allocating the name and adding the device in order to avoid | |
782 | * duplicates. | |
783 | * Limited to bits_per_byte * page size devices (ie 32K on most platforms). | |
784 | * Returns the number of the unit assigned or a negative errno code. | |
1da177e4 LT |
785 | */ |
786 | ||
b267b179 | 787 | static int __dev_alloc_name(struct net *net, const char *name, char *buf) |
1da177e4 LT |
788 | { |
789 | int i = 0; | |
1da177e4 LT |
790 | const char *p; |
791 | const int max_netdevices = 8*PAGE_SIZE; | |
cfcabdcc | 792 | unsigned long *inuse; |
1da177e4 LT |
793 | struct net_device *d; |
794 | ||
795 | p = strnchr(name, IFNAMSIZ-1, '%'); | |
796 | if (p) { | |
797 | /* | |
798 | * Verify the string as this thing may have come from | |
799 | * the user. There must be either one "%d" and no other "%" | |
800 | * characters. | |
801 | */ | |
802 | if (p[1] != 'd' || strchr(p + 2, '%')) | |
803 | return -EINVAL; | |
804 | ||
805 | /* Use one page as a bit array of possible slots */ | |
cfcabdcc | 806 | inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC); |
1da177e4 LT |
807 | if (!inuse) |
808 | return -ENOMEM; | |
809 | ||
881d966b | 810 | for_each_netdev(net, d) { |
1da177e4 LT |
811 | if (!sscanf(d->name, name, &i)) |
812 | continue; | |
813 | if (i < 0 || i >= max_netdevices) | |
814 | continue; | |
815 | ||
816 | /* avoid cases where sscanf is not exact inverse of printf */ | |
b267b179 | 817 | snprintf(buf, IFNAMSIZ, name, i); |
1da177e4 LT |
818 | if (!strncmp(buf, d->name, IFNAMSIZ)) |
819 | set_bit(i, inuse); | |
820 | } | |
821 | ||
822 | i = find_first_zero_bit(inuse, max_netdevices); | |
823 | free_page((unsigned long) inuse); | |
824 | } | |
825 | ||
b267b179 EB |
826 | snprintf(buf, IFNAMSIZ, name, i); |
827 | if (!__dev_get_by_name(net, buf)) | |
1da177e4 | 828 | return i; |
1da177e4 LT |
829 | |
830 | /* It is possible to run out of possible slots | |
831 | * when the name is long and there isn't enough space left | |
832 | * for the digits, or if all bits are used. | |
833 | */ | |
834 | return -ENFILE; | |
835 | } | |
836 | ||
b267b179 EB |
837 | /** |
838 | * dev_alloc_name - allocate a name for a device | |
839 | * @dev: device | |
840 | * @name: name format string | |
841 | * | |
842 | * Passed a format string - eg "lt%d" it will try and find a suitable | |
843 | * id. It scans list of devices to build up a free map, then chooses | |
844 | * the first empty slot. The caller must hold the dev_base or rtnl lock | |
845 | * while allocating the name and adding the device in order to avoid | |
846 | * duplicates. | |
847 | * Limited to bits_per_byte * page size devices (ie 32K on most platforms). | |
848 | * Returns the number of the unit assigned or a negative errno code. | |
849 | */ | |
850 | ||
851 | int dev_alloc_name(struct net_device *dev, const char *name) | |
852 | { | |
853 | char buf[IFNAMSIZ]; | |
854 | struct net *net; | |
855 | int ret; | |
856 | ||
c346dca1 YH |
857 | BUG_ON(!dev_net(dev)); |
858 | net = dev_net(dev); | |
b267b179 EB |
859 | ret = __dev_alloc_name(net, name, buf); |
860 | if (ret >= 0) | |
861 | strlcpy(dev->name, buf, IFNAMSIZ); | |
862 | return ret; | |
863 | } | |
864 | ||
1da177e4 LT |
865 | |
866 | /** | |
867 | * dev_change_name - change name of a device | |
868 | * @dev: device | |
869 | * @newname: name (or format string) must be at least IFNAMSIZ | |
870 | * | |
871 | * Change name of a device, can pass format strings "eth%d". | |
872 | * for wildcarding. | |
873 | */ | |
874 | int dev_change_name(struct net_device *dev, char *newname) | |
875 | { | |
fcc5a03a | 876 | char oldname[IFNAMSIZ]; |
1da177e4 | 877 | int err = 0; |
fcc5a03a | 878 | int ret; |
881d966b | 879 | struct net *net; |
1da177e4 LT |
880 | |
881 | ASSERT_RTNL(); | |
c346dca1 | 882 | BUG_ON(!dev_net(dev)); |
1da177e4 | 883 | |
c346dca1 | 884 | net = dev_net(dev); |
1da177e4 LT |
885 | if (dev->flags & IFF_UP) |
886 | return -EBUSY; | |
887 | ||
888 | if (!dev_valid_name(newname)) | |
889 | return -EINVAL; | |
890 | ||
c8d90dca SH |
891 | if (strncmp(newname, dev->name, IFNAMSIZ) == 0) |
892 | return 0; | |
893 | ||
fcc5a03a HX |
894 | memcpy(oldname, dev->name, IFNAMSIZ); |
895 | ||
1da177e4 LT |
896 | if (strchr(newname, '%')) { |
897 | err = dev_alloc_name(dev, newname); | |
898 | if (err < 0) | |
899 | return err; | |
900 | strcpy(newname, dev->name); | |
901 | } | |
881d966b | 902 | else if (__dev_get_by_name(net, newname)) |
1da177e4 LT |
903 | return -EEXIST; |
904 | else | |
905 | strlcpy(dev->name, newname, IFNAMSIZ); | |
906 | ||
fcc5a03a | 907 | rollback: |
dcc99773 SH |
908 | err = device_rename(&dev->dev, dev->name); |
909 | if (err) { | |
910 | memcpy(dev->name, oldname, IFNAMSIZ); | |
911 | return err; | |
912 | } | |
7f988eab HX |
913 | |
914 | write_lock_bh(&dev_base_lock); | |
92749821 | 915 | hlist_del(&dev->name_hlist); |
881d966b | 916 | hlist_add_head(&dev->name_hlist, dev_name_hash(net, dev->name)); |
7f988eab HX |
917 | write_unlock_bh(&dev_base_lock); |
918 | ||
056925ab | 919 | ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev); |
fcc5a03a HX |
920 | ret = notifier_to_errno(ret); |
921 | ||
922 | if (ret) { | |
923 | if (err) { | |
924 | printk(KERN_ERR | |
925 | "%s: name change rollback failed: %d.\n", | |
926 | dev->name, ret); | |
927 | } else { | |
928 | err = ret; | |
929 | memcpy(dev->name, oldname, IFNAMSIZ); | |
930 | goto rollback; | |
931 | } | |
932 | } | |
1da177e4 LT |
933 | |
934 | return err; | |
935 | } | |
936 | ||
d8a33ac4 | 937 | /** |
3041a069 | 938 | * netdev_features_change - device changes features |
d8a33ac4 SH |
939 | * @dev: device to cause notification |
940 | * | |
941 | * Called to indicate a device has changed features. | |
942 | */ | |
943 | void netdev_features_change(struct net_device *dev) | |
944 | { | |
056925ab | 945 | call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev); |
d8a33ac4 SH |
946 | } |
947 | EXPORT_SYMBOL(netdev_features_change); | |
948 | ||
1da177e4 LT |
949 | /** |
950 | * netdev_state_change - device changes state | |
951 | * @dev: device to cause notification | |
952 | * | |
953 | * Called to indicate a device has changed state. This function calls | |
954 | * the notifier chains for netdev_chain and sends a NEWLINK message | |
955 | * to the routing socket. | |
956 | */ | |
957 | void netdev_state_change(struct net_device *dev) | |
958 | { | |
959 | if (dev->flags & IFF_UP) { | |
056925ab | 960 | call_netdevice_notifiers(NETDEV_CHANGE, dev); |
1da177e4 LT |
961 | rtmsg_ifinfo(RTM_NEWLINK, dev, 0); |
962 | } | |
963 | } | |
964 | ||
c1da4ac7 OG |
965 | void netdev_bonding_change(struct net_device *dev) |
966 | { | |
967 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, dev); | |
968 | } | |
969 | EXPORT_SYMBOL(netdev_bonding_change); | |
970 | ||
1da177e4 LT |
971 | /** |
972 | * dev_load - load a network module | |
c4ea43c5 | 973 | * @net: the applicable net namespace |
1da177e4 LT |
974 | * @name: name of interface |
975 | * | |
976 | * If a network interface is not present and the process has suitable | |
977 | * privileges this function loads the module. If module loading is not | |
978 | * available in this kernel then it becomes a nop. | |
979 | */ | |
980 | ||
881d966b | 981 | void dev_load(struct net *net, const char *name) |
1da177e4 | 982 | { |
4ec93edb | 983 | struct net_device *dev; |
1da177e4 LT |
984 | |
985 | read_lock(&dev_base_lock); | |
881d966b | 986 | dev = __dev_get_by_name(net, name); |
1da177e4 LT |
987 | read_unlock(&dev_base_lock); |
988 | ||
989 | if (!dev && capable(CAP_SYS_MODULE)) | |
990 | request_module("%s", name); | |
991 | } | |
992 | ||
1da177e4 LT |
993 | /** |
994 | * dev_open - prepare an interface for use. | |
995 | * @dev: device to open | |
996 | * | |
997 | * Takes a device from down to up state. The device's private open | |
998 | * function is invoked and then the multicast lists are loaded. Finally | |
999 | * the device is moved into the up state and a %NETDEV_UP message is | |
1000 | * sent to the netdev notifier chain. | |
1001 | * | |
1002 | * Calling this function on an active interface is a nop. On a failure | |
1003 | * a negative errno code is returned. | |
1004 | */ | |
1005 | int dev_open(struct net_device *dev) | |
1006 | { | |
1007 | int ret = 0; | |
1008 | ||
e46b66bc BH |
1009 | ASSERT_RTNL(); |
1010 | ||
1da177e4 LT |
1011 | /* |
1012 | * Is it already up? | |
1013 | */ | |
1014 | ||
1015 | if (dev->flags & IFF_UP) | |
1016 | return 0; | |
1017 | ||
1018 | /* | |
1019 | * Is it even present? | |
1020 | */ | |
1021 | if (!netif_device_present(dev)) | |
1022 | return -ENODEV; | |
1023 | ||
1024 | /* | |
1025 | * Call device private open method | |
1026 | */ | |
1027 | set_bit(__LINK_STATE_START, &dev->state); | |
bada339b JG |
1028 | |
1029 | if (dev->validate_addr) | |
1030 | ret = dev->validate_addr(dev); | |
1031 | ||
1032 | if (!ret && dev->open) | |
1da177e4 | 1033 | ret = dev->open(dev); |
1da177e4 | 1034 | |
4ec93edb | 1035 | /* |
1da177e4 LT |
1036 | * If it went open OK then: |
1037 | */ | |
1038 | ||
bada339b JG |
1039 | if (ret) |
1040 | clear_bit(__LINK_STATE_START, &dev->state); | |
1041 | else { | |
1da177e4 LT |
1042 | /* |
1043 | * Set the flags. | |
1044 | */ | |
1045 | dev->flags |= IFF_UP; | |
1046 | ||
1047 | /* | |
1048 | * Initialize multicasting status | |
1049 | */ | |
4417da66 | 1050 | dev_set_rx_mode(dev); |
1da177e4 LT |
1051 | |
1052 | /* | |
1053 | * Wakeup transmit queue engine | |
1054 | */ | |
1055 | dev_activate(dev); | |
1056 | ||
1057 | /* | |
1058 | * ... and announce new interface. | |
1059 | */ | |
056925ab | 1060 | call_netdevice_notifiers(NETDEV_UP, dev); |
1da177e4 | 1061 | } |
bada339b | 1062 | |
1da177e4 LT |
1063 | return ret; |
1064 | } | |
1065 | ||
1066 | /** | |
1067 | * dev_close - shutdown an interface. | |
1068 | * @dev: device to shutdown | |
1069 | * | |
1070 | * This function moves an active device into down state. A | |
1071 | * %NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device | |
1072 | * is then deactivated and finally a %NETDEV_DOWN is sent to the notifier | |
1073 | * chain. | |
1074 | */ | |
1075 | int dev_close(struct net_device *dev) | |
1076 | { | |
e46b66bc BH |
1077 | ASSERT_RTNL(); |
1078 | ||
9d5010db DM |
1079 | might_sleep(); |
1080 | ||
1da177e4 LT |
1081 | if (!(dev->flags & IFF_UP)) |
1082 | return 0; | |
1083 | ||
1084 | /* | |
1085 | * Tell people we are going down, so that they can | |
1086 | * prepare to death, when device is still operating. | |
1087 | */ | |
056925ab | 1088 | call_netdevice_notifiers(NETDEV_GOING_DOWN, dev); |
1da177e4 | 1089 | |
1da177e4 LT |
1090 | clear_bit(__LINK_STATE_START, &dev->state); |
1091 | ||
1092 | /* Synchronize to scheduled poll. We cannot touch poll list, | |
bea3348e SH |
1093 | * it can be even on different cpu. So just clear netif_running(). |
1094 | * | |
1095 | * dev->stop() will invoke napi_disable() on all of it's | |
1096 | * napi_struct instances on this device. | |
1097 | */ | |
1da177e4 | 1098 | smp_mb__after_clear_bit(); /* Commit netif_running(). */ |
1da177e4 | 1099 | |
d8b2a4d2 ML |
1100 | dev_deactivate(dev); |
1101 | ||
1da177e4 LT |
1102 | /* |
1103 | * Call the device specific close. This cannot fail. | |
1104 | * Only if device is UP | |
1105 | * | |
1106 | * We allow it to be called even after a DETACH hot-plug | |
1107 | * event. | |
1108 | */ | |
1109 | if (dev->stop) | |
1110 | dev->stop(dev); | |
1111 | ||
1112 | /* | |
1113 | * Device is now down. | |
1114 | */ | |
1115 | ||
1116 | dev->flags &= ~IFF_UP; | |
1117 | ||
1118 | /* | |
1119 | * Tell people we are down | |
1120 | */ | |
056925ab | 1121 | call_netdevice_notifiers(NETDEV_DOWN, dev); |
1da177e4 LT |
1122 | |
1123 | return 0; | |
1124 | } | |
1125 | ||
1126 | ||
0187bdfb BH |
1127 | /** |
1128 | * dev_disable_lro - disable Large Receive Offload on a device | |
1129 | * @dev: device | |
1130 | * | |
1131 | * Disable Large Receive Offload (LRO) on a net device. Must be | |
1132 | * called under RTNL. This is needed if received packets may be | |
1133 | * forwarded to another interface. | |
1134 | */ | |
1135 | void dev_disable_lro(struct net_device *dev) | |
1136 | { | |
1137 | if (dev->ethtool_ops && dev->ethtool_ops->get_flags && | |
1138 | dev->ethtool_ops->set_flags) { | |
1139 | u32 flags = dev->ethtool_ops->get_flags(dev); | |
1140 | if (flags & ETH_FLAG_LRO) { | |
1141 | flags &= ~ETH_FLAG_LRO; | |
1142 | dev->ethtool_ops->set_flags(dev, flags); | |
1143 | } | |
1144 | } | |
1145 | WARN_ON(dev->features & NETIF_F_LRO); | |
1146 | } | |
1147 | EXPORT_SYMBOL(dev_disable_lro); | |
1148 | ||
1149 | ||
881d966b EB |
1150 | static int dev_boot_phase = 1; |
1151 | ||
1da177e4 LT |
1152 | /* |
1153 | * Device change register/unregister. These are not inline or static | |
1154 | * as we export them to the world. | |
1155 | */ | |
1156 | ||
1157 | /** | |
1158 | * register_netdevice_notifier - register a network notifier block | |
1159 | * @nb: notifier | |
1160 | * | |
1161 | * Register a notifier to be called when network device events occur. | |
1162 | * The notifier passed is linked into the kernel structures and must | |
1163 | * not be reused until it has been unregistered. A negative errno code | |
1164 | * is returned on a failure. | |
1165 | * | |
1166 | * When registered all registration and up events are replayed | |
4ec93edb | 1167 | * to the new notifier to allow device to have a race free |
1da177e4 LT |
1168 | * view of the network device list. |
1169 | */ | |
1170 | ||
1171 | int register_netdevice_notifier(struct notifier_block *nb) | |
1172 | { | |
1173 | struct net_device *dev; | |
fcc5a03a | 1174 | struct net_device *last; |
881d966b | 1175 | struct net *net; |
1da177e4 LT |
1176 | int err; |
1177 | ||
1178 | rtnl_lock(); | |
f07d5b94 | 1179 | err = raw_notifier_chain_register(&netdev_chain, nb); |
fcc5a03a HX |
1180 | if (err) |
1181 | goto unlock; | |
881d966b EB |
1182 | if (dev_boot_phase) |
1183 | goto unlock; | |
1184 | for_each_net(net) { | |
1185 | for_each_netdev(net, dev) { | |
1186 | err = nb->notifier_call(nb, NETDEV_REGISTER, dev); | |
1187 | err = notifier_to_errno(err); | |
1188 | if (err) | |
1189 | goto rollback; | |
1190 | ||
1191 | if (!(dev->flags & IFF_UP)) | |
1192 | continue; | |
1da177e4 | 1193 | |
881d966b EB |
1194 | nb->notifier_call(nb, NETDEV_UP, dev); |
1195 | } | |
1da177e4 | 1196 | } |
fcc5a03a HX |
1197 | |
1198 | unlock: | |
1da177e4 LT |
1199 | rtnl_unlock(); |
1200 | return err; | |
fcc5a03a HX |
1201 | |
1202 | rollback: | |
1203 | last = dev; | |
881d966b EB |
1204 | for_each_net(net) { |
1205 | for_each_netdev(net, dev) { | |
1206 | if (dev == last) | |
1207 | break; | |
fcc5a03a | 1208 | |
881d966b EB |
1209 | if (dev->flags & IFF_UP) { |
1210 | nb->notifier_call(nb, NETDEV_GOING_DOWN, dev); | |
1211 | nb->notifier_call(nb, NETDEV_DOWN, dev); | |
1212 | } | |
1213 | nb->notifier_call(nb, NETDEV_UNREGISTER, dev); | |
fcc5a03a | 1214 | } |
fcc5a03a | 1215 | } |
c67625a1 PE |
1216 | |
1217 | raw_notifier_chain_unregister(&netdev_chain, nb); | |
fcc5a03a | 1218 | goto unlock; |
1da177e4 LT |
1219 | } |
1220 | ||
1221 | /** | |
1222 | * unregister_netdevice_notifier - unregister a network notifier block | |
1223 | * @nb: notifier | |
1224 | * | |
1225 | * Unregister a notifier previously registered by | |
1226 | * register_netdevice_notifier(). The notifier is unlinked into the | |
1227 | * kernel structures and may then be reused. A negative errno code | |
1228 | * is returned on a failure. | |
1229 | */ | |
1230 | ||
1231 | int unregister_netdevice_notifier(struct notifier_block *nb) | |
1232 | { | |
9f514950 HX |
1233 | int err; |
1234 | ||
1235 | rtnl_lock(); | |
f07d5b94 | 1236 | err = raw_notifier_chain_unregister(&netdev_chain, nb); |
9f514950 HX |
1237 | rtnl_unlock(); |
1238 | return err; | |
1da177e4 LT |
1239 | } |
1240 | ||
1241 | /** | |
1242 | * call_netdevice_notifiers - call all network notifier blocks | |
1243 | * @val: value passed unmodified to notifier function | |
c4ea43c5 | 1244 | * @dev: net_device pointer passed unmodified to notifier function |
1da177e4 LT |
1245 | * |
1246 | * Call all network notifier blocks. Parameters and return value | |
f07d5b94 | 1247 | * are as for raw_notifier_call_chain(). |
1da177e4 LT |
1248 | */ |
1249 | ||
ad7379d4 | 1250 | int call_netdevice_notifiers(unsigned long val, struct net_device *dev) |
1da177e4 | 1251 | { |
ad7379d4 | 1252 | return raw_notifier_call_chain(&netdev_chain, val, dev); |
1da177e4 LT |
1253 | } |
1254 | ||
1255 | /* When > 0 there are consumers of rx skb time stamps */ | |
1256 | static atomic_t netstamp_needed = ATOMIC_INIT(0); | |
1257 | ||
1258 | void net_enable_timestamp(void) | |
1259 | { | |
1260 | atomic_inc(&netstamp_needed); | |
1261 | } | |
1262 | ||
1263 | void net_disable_timestamp(void) | |
1264 | { | |
1265 | atomic_dec(&netstamp_needed); | |
1266 | } | |
1267 | ||
a61bbcf2 | 1268 | static inline void net_timestamp(struct sk_buff *skb) |
1da177e4 LT |
1269 | { |
1270 | if (atomic_read(&netstamp_needed)) | |
a61bbcf2 | 1271 | __net_timestamp(skb); |
b7aa0bf7 ED |
1272 | else |
1273 | skb->tstamp.tv64 = 0; | |
1da177e4 LT |
1274 | } |
1275 | ||
1276 | /* | |
1277 | * Support routine. Sends outgoing frames to any network | |
1278 | * taps currently in use. | |
1279 | */ | |
1280 | ||
f6a78bfc | 1281 | static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
1282 | { |
1283 | struct packet_type *ptype; | |
a61bbcf2 PM |
1284 | |
1285 | net_timestamp(skb); | |
1da177e4 LT |
1286 | |
1287 | rcu_read_lock(); | |
1288 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
1289 | /* Never send packets back to the socket | |
1290 | * they originated from - MvS (miquels@drinkel.ow.org) | |
1291 | */ | |
1292 | if ((ptype->dev == dev || !ptype->dev) && | |
1293 | (ptype->af_packet_priv == NULL || | |
1294 | (struct sock *)ptype->af_packet_priv != skb->sk)) { | |
1295 | struct sk_buff *skb2= skb_clone(skb, GFP_ATOMIC); | |
1296 | if (!skb2) | |
1297 | break; | |
1298 | ||
1299 | /* skb->nh should be correctly | |
1300 | set by sender, so that the second statement is | |
1301 | just protection against buggy protocols. | |
1302 | */ | |
459a98ed | 1303 | skb_reset_mac_header(skb2); |
1da177e4 | 1304 | |
d56f90a7 | 1305 | if (skb_network_header(skb2) < skb2->data || |
27a884dc | 1306 | skb2->network_header > skb2->tail) { |
1da177e4 LT |
1307 | if (net_ratelimit()) |
1308 | printk(KERN_CRIT "protocol %04x is " | |
1309 | "buggy, dev %s\n", | |
1310 | skb2->protocol, dev->name); | |
c1d2bbe1 | 1311 | skb_reset_network_header(skb2); |
1da177e4 LT |
1312 | } |
1313 | ||
b0e380b1 | 1314 | skb2->transport_header = skb2->network_header; |
1da177e4 | 1315 | skb2->pkt_type = PACKET_OUTGOING; |
f2ccd8fa | 1316 | ptype->func(skb2, skb->dev, ptype, skb->dev); |
1da177e4 LT |
1317 | } |
1318 | } | |
1319 | rcu_read_unlock(); | |
1320 | } | |
1321 | ||
56079431 | 1322 | |
86d804e1 | 1323 | void __netif_schedule(struct netdev_queue *txq) |
56079431 | 1324 | { |
86d804e1 DM |
1325 | struct net_device *dev = txq->dev; |
1326 | ||
56079431 | 1327 | if (!test_and_set_bit(__LINK_STATE_SCHED, &dev->state)) { |
56079431 | 1328 | struct softnet_data *sd; |
86d804e1 | 1329 | unsigned long flags; |
56079431 DV |
1330 | |
1331 | local_irq_save(flags); | |
1332 | sd = &__get_cpu_var(softnet_data); | |
ee609cb3 DM |
1333 | txq->next_sched = sd->output_queue; |
1334 | sd->output_queue = txq; | |
56079431 DV |
1335 | raise_softirq_irqoff(NET_TX_SOFTIRQ); |
1336 | local_irq_restore(flags); | |
1337 | } | |
1338 | } | |
1339 | EXPORT_SYMBOL(__netif_schedule); | |
1340 | ||
bea3348e | 1341 | void dev_kfree_skb_irq(struct sk_buff *skb) |
56079431 | 1342 | { |
bea3348e SH |
1343 | if (atomic_dec_and_test(&skb->users)) { |
1344 | struct softnet_data *sd; | |
1345 | unsigned long flags; | |
56079431 | 1346 | |
bea3348e SH |
1347 | local_irq_save(flags); |
1348 | sd = &__get_cpu_var(softnet_data); | |
1349 | skb->next = sd->completion_queue; | |
1350 | sd->completion_queue = skb; | |
1351 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
1352 | local_irq_restore(flags); | |
1353 | } | |
56079431 | 1354 | } |
bea3348e | 1355 | EXPORT_SYMBOL(dev_kfree_skb_irq); |
56079431 DV |
1356 | |
1357 | void dev_kfree_skb_any(struct sk_buff *skb) | |
1358 | { | |
1359 | if (in_irq() || irqs_disabled()) | |
1360 | dev_kfree_skb_irq(skb); | |
1361 | else | |
1362 | dev_kfree_skb(skb); | |
1363 | } | |
1364 | EXPORT_SYMBOL(dev_kfree_skb_any); | |
1365 | ||
1366 | ||
bea3348e SH |
1367 | /** |
1368 | * netif_device_detach - mark device as removed | |
1369 | * @dev: network device | |
1370 | * | |
1371 | * Mark device as removed from system and therefore no longer available. | |
1372 | */ | |
56079431 DV |
1373 | void netif_device_detach(struct net_device *dev) |
1374 | { | |
1375 | if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) && | |
1376 | netif_running(dev)) { | |
1377 | netif_stop_queue(dev); | |
1378 | } | |
1379 | } | |
1380 | EXPORT_SYMBOL(netif_device_detach); | |
1381 | ||
bea3348e SH |
1382 | /** |
1383 | * netif_device_attach - mark device as attached | |
1384 | * @dev: network device | |
1385 | * | |
1386 | * Mark device as attached from system and restart if needed. | |
1387 | */ | |
56079431 DV |
1388 | void netif_device_attach(struct net_device *dev) |
1389 | { | |
1390 | if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) && | |
1391 | netif_running(dev)) { | |
1392 | netif_wake_queue(dev); | |
4ec93edb | 1393 | __netdev_watchdog_up(dev); |
56079431 DV |
1394 | } |
1395 | } | |
1396 | EXPORT_SYMBOL(netif_device_attach); | |
1397 | ||
6de329e2 BH |
1398 | static bool can_checksum_protocol(unsigned long features, __be16 protocol) |
1399 | { | |
1400 | return ((features & NETIF_F_GEN_CSUM) || | |
1401 | ((features & NETIF_F_IP_CSUM) && | |
1402 | protocol == htons(ETH_P_IP)) || | |
1403 | ((features & NETIF_F_IPV6_CSUM) && | |
1404 | protocol == htons(ETH_P_IPV6))); | |
1405 | } | |
1406 | ||
1407 | static bool dev_can_checksum(struct net_device *dev, struct sk_buff *skb) | |
1408 | { | |
1409 | if (can_checksum_protocol(dev->features, skb->protocol)) | |
1410 | return true; | |
1411 | ||
1412 | if (skb->protocol == htons(ETH_P_8021Q)) { | |
1413 | struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data; | |
1414 | if (can_checksum_protocol(dev->features & dev->vlan_features, | |
1415 | veh->h_vlan_encapsulated_proto)) | |
1416 | return true; | |
1417 | } | |
1418 | ||
1419 | return false; | |
1420 | } | |
56079431 | 1421 | |
1da177e4 LT |
1422 | /* |
1423 | * Invalidate hardware checksum when packet is to be mangled, and | |
1424 | * complete checksum manually on outgoing path. | |
1425 | */ | |
84fa7933 | 1426 | int skb_checksum_help(struct sk_buff *skb) |
1da177e4 | 1427 | { |
d3bc23e7 | 1428 | __wsum csum; |
663ead3b | 1429 | int ret = 0, offset; |
1da177e4 | 1430 | |
84fa7933 | 1431 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
a430a43d HX |
1432 | goto out_set_summed; |
1433 | ||
1434 | if (unlikely(skb_shinfo(skb)->gso_size)) { | |
a430a43d HX |
1435 | /* Let GSO fix up the checksum. */ |
1436 | goto out_set_summed; | |
1da177e4 LT |
1437 | } |
1438 | ||
a030847e HX |
1439 | offset = skb->csum_start - skb_headroom(skb); |
1440 | BUG_ON(offset >= skb_headlen(skb)); | |
1441 | csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
1442 | ||
1443 | offset += skb->csum_offset; | |
1444 | BUG_ON(offset + sizeof(__sum16) > skb_headlen(skb)); | |
1445 | ||
1446 | if (skb_cloned(skb) && | |
1447 | !skb_clone_writable(skb, offset + sizeof(__sum16))) { | |
1da177e4 LT |
1448 | ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC); |
1449 | if (ret) | |
1450 | goto out; | |
1451 | } | |
1452 | ||
a030847e | 1453 | *(__sum16 *)(skb->data + offset) = csum_fold(csum); |
a430a43d | 1454 | out_set_summed: |
1da177e4 | 1455 | skb->ip_summed = CHECKSUM_NONE; |
4ec93edb | 1456 | out: |
1da177e4 LT |
1457 | return ret; |
1458 | } | |
1459 | ||
f6a78bfc HX |
1460 | /** |
1461 | * skb_gso_segment - Perform segmentation on skb. | |
1462 | * @skb: buffer to segment | |
576a30eb | 1463 | * @features: features for the output path (see dev->features) |
f6a78bfc HX |
1464 | * |
1465 | * This function segments the given skb and returns a list of segments. | |
576a30eb HX |
1466 | * |
1467 | * It may return NULL if the skb requires no segmentation. This is | |
1468 | * only possible when GSO is used for verifying header integrity. | |
f6a78bfc | 1469 | */ |
576a30eb | 1470 | struct sk_buff *skb_gso_segment(struct sk_buff *skb, int features) |
f6a78bfc HX |
1471 | { |
1472 | struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT); | |
1473 | struct packet_type *ptype; | |
252e3346 | 1474 | __be16 type = skb->protocol; |
a430a43d | 1475 | int err; |
f6a78bfc HX |
1476 | |
1477 | BUG_ON(skb_shinfo(skb)->frag_list); | |
f6a78bfc | 1478 | |
459a98ed | 1479 | skb_reset_mac_header(skb); |
b0e380b1 | 1480 | skb->mac_len = skb->network_header - skb->mac_header; |
f6a78bfc HX |
1481 | __skb_pull(skb, skb->mac_len); |
1482 | ||
f9d106a6 | 1483 | if (WARN_ON(skb->ip_summed != CHECKSUM_PARTIAL)) { |
a430a43d HX |
1484 | if (skb_header_cloned(skb) && |
1485 | (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) | |
1486 | return ERR_PTR(err); | |
1487 | } | |
1488 | ||
f6a78bfc | 1489 | rcu_read_lock(); |
82d8a867 PE |
1490 | list_for_each_entry_rcu(ptype, |
1491 | &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { | |
f6a78bfc | 1492 | if (ptype->type == type && !ptype->dev && ptype->gso_segment) { |
84fa7933 | 1493 | if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { |
a430a43d HX |
1494 | err = ptype->gso_send_check(skb); |
1495 | segs = ERR_PTR(err); | |
1496 | if (err || skb_gso_ok(skb, features)) | |
1497 | break; | |
d56f90a7 ACM |
1498 | __skb_push(skb, (skb->data - |
1499 | skb_network_header(skb))); | |
a430a43d | 1500 | } |
576a30eb | 1501 | segs = ptype->gso_segment(skb, features); |
f6a78bfc HX |
1502 | break; |
1503 | } | |
1504 | } | |
1505 | rcu_read_unlock(); | |
1506 | ||
98e399f8 | 1507 | __skb_push(skb, skb->data - skb_mac_header(skb)); |
576a30eb | 1508 | |
f6a78bfc HX |
1509 | return segs; |
1510 | } | |
1511 | ||
1512 | EXPORT_SYMBOL(skb_gso_segment); | |
1513 | ||
fb286bb2 HX |
1514 | /* Take action when hardware reception checksum errors are detected. */ |
1515 | #ifdef CONFIG_BUG | |
1516 | void netdev_rx_csum_fault(struct net_device *dev) | |
1517 | { | |
1518 | if (net_ratelimit()) { | |
4ec93edb | 1519 | printk(KERN_ERR "%s: hw csum failure.\n", |
246a4212 | 1520 | dev ? dev->name : "<unknown>"); |
fb286bb2 HX |
1521 | dump_stack(); |
1522 | } | |
1523 | } | |
1524 | EXPORT_SYMBOL(netdev_rx_csum_fault); | |
1525 | #endif | |
1526 | ||
1da177e4 LT |
1527 | /* Actually, we should eliminate this check as soon as we know, that: |
1528 | * 1. IOMMU is present and allows to map all the memory. | |
1529 | * 2. No high memory really exists on this machine. | |
1530 | */ | |
1531 | ||
1532 | static inline int illegal_highdma(struct net_device *dev, struct sk_buff *skb) | |
1533 | { | |
3d3a8533 | 1534 | #ifdef CONFIG_HIGHMEM |
1da177e4 LT |
1535 | int i; |
1536 | ||
1537 | if (dev->features & NETIF_F_HIGHDMA) | |
1538 | return 0; | |
1539 | ||
1540 | for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) | |
1541 | if (PageHighMem(skb_shinfo(skb)->frags[i].page)) | |
1542 | return 1; | |
1543 | ||
3d3a8533 | 1544 | #endif |
1da177e4 LT |
1545 | return 0; |
1546 | } | |
1da177e4 | 1547 | |
f6a78bfc HX |
1548 | struct dev_gso_cb { |
1549 | void (*destructor)(struct sk_buff *skb); | |
1550 | }; | |
1551 | ||
1552 | #define DEV_GSO_CB(skb) ((struct dev_gso_cb *)(skb)->cb) | |
1553 | ||
1554 | static void dev_gso_skb_destructor(struct sk_buff *skb) | |
1555 | { | |
1556 | struct dev_gso_cb *cb; | |
1557 | ||
1558 | do { | |
1559 | struct sk_buff *nskb = skb->next; | |
1560 | ||
1561 | skb->next = nskb->next; | |
1562 | nskb->next = NULL; | |
1563 | kfree_skb(nskb); | |
1564 | } while (skb->next); | |
1565 | ||
1566 | cb = DEV_GSO_CB(skb); | |
1567 | if (cb->destructor) | |
1568 | cb->destructor(skb); | |
1569 | } | |
1570 | ||
1571 | /** | |
1572 | * dev_gso_segment - Perform emulated hardware segmentation on skb. | |
1573 | * @skb: buffer to segment | |
1574 | * | |
1575 | * This function segments the given skb and stores the list of segments | |
1576 | * in skb->next. | |
1577 | */ | |
1578 | static int dev_gso_segment(struct sk_buff *skb) | |
1579 | { | |
1580 | struct net_device *dev = skb->dev; | |
1581 | struct sk_buff *segs; | |
576a30eb HX |
1582 | int features = dev->features & ~(illegal_highdma(dev, skb) ? |
1583 | NETIF_F_SG : 0); | |
1584 | ||
1585 | segs = skb_gso_segment(skb, features); | |
1586 | ||
1587 | /* Verifying header integrity only. */ | |
1588 | if (!segs) | |
1589 | return 0; | |
f6a78bfc | 1590 | |
801678c5 | 1591 | if (IS_ERR(segs)) |
f6a78bfc HX |
1592 | return PTR_ERR(segs); |
1593 | ||
1594 | skb->next = segs; | |
1595 | DEV_GSO_CB(skb)->destructor = skb->destructor; | |
1596 | skb->destructor = dev_gso_skb_destructor; | |
1597 | ||
1598 | return 0; | |
1599 | } | |
1600 | ||
1601 | int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev) | |
1602 | { | |
1603 | if (likely(!skb->next)) { | |
9be9a6b9 | 1604 | if (!list_empty(&ptype_all)) |
f6a78bfc HX |
1605 | dev_queue_xmit_nit(skb, dev); |
1606 | ||
576a30eb HX |
1607 | if (netif_needs_gso(dev, skb)) { |
1608 | if (unlikely(dev_gso_segment(skb))) | |
1609 | goto out_kfree_skb; | |
1610 | if (skb->next) | |
1611 | goto gso; | |
1612 | } | |
f6a78bfc | 1613 | |
576a30eb | 1614 | return dev->hard_start_xmit(skb, dev); |
f6a78bfc HX |
1615 | } |
1616 | ||
576a30eb | 1617 | gso: |
f6a78bfc HX |
1618 | do { |
1619 | struct sk_buff *nskb = skb->next; | |
1620 | int rc; | |
1621 | ||
1622 | skb->next = nskb->next; | |
1623 | nskb->next = NULL; | |
1624 | rc = dev->hard_start_xmit(nskb, dev); | |
1625 | if (unlikely(rc)) { | |
f54d9e8d | 1626 | nskb->next = skb->next; |
f6a78bfc HX |
1627 | skb->next = nskb; |
1628 | return rc; | |
1629 | } | |
f25f4e44 | 1630 | if (unlikely((netif_queue_stopped(dev) || |
668f895a | 1631 | netif_subqueue_stopped(dev, skb)) && |
f25f4e44 | 1632 | skb->next)) |
f54d9e8d | 1633 | return NETDEV_TX_BUSY; |
f6a78bfc | 1634 | } while (skb->next); |
4ec93edb | 1635 | |
f6a78bfc HX |
1636 | skb->destructor = DEV_GSO_CB(skb)->destructor; |
1637 | ||
1638 | out_kfree_skb: | |
1639 | kfree_skb(skb); | |
1640 | return 0; | |
1641 | } | |
1642 | ||
1da177e4 LT |
1643 | /** |
1644 | * dev_queue_xmit - transmit a buffer | |
1645 | * @skb: buffer to transmit | |
1646 | * | |
1647 | * Queue a buffer for transmission to a network device. The caller must | |
1648 | * have set the device and priority and built the buffer before calling | |
1649 | * this function. The function can be called from an interrupt. | |
1650 | * | |
1651 | * A negative errno code is returned on a failure. A success does not | |
1652 | * guarantee the frame will be transmitted as it may be dropped due | |
1653 | * to congestion or traffic shaping. | |
af191367 BG |
1654 | * |
1655 | * ----------------------------------------------------------------------------------- | |
1656 | * I notice this method can also return errors from the queue disciplines, | |
1657 | * including NET_XMIT_DROP, which is a positive value. So, errors can also | |
1658 | * be positive. | |
1659 | * | |
1660 | * Regardless of the return value, the skb is consumed, so it is currently | |
1661 | * difficult to retry a send to this method. (You can bump the ref count | |
1662 | * before sending to hold a reference for retry if you are careful.) | |
1663 | * | |
1664 | * When calling this method, interrupts MUST be enabled. This is because | |
1665 | * the BH enable code must have IRQs enabled so that it will not deadlock. | |
1666 | * --BLG | |
1da177e4 LT |
1667 | */ |
1668 | ||
e8a0464c DM |
1669 | static struct netdev_queue *dev_pick_tx(struct net_device *dev, |
1670 | struct sk_buff *skb) | |
1671 | { | |
1672 | return netdev_get_tx_queue(dev, 0); | |
1673 | } | |
1674 | ||
1da177e4 LT |
1675 | int dev_queue_xmit(struct sk_buff *skb) |
1676 | { | |
1677 | struct net_device *dev = skb->dev; | |
dc2b4847 | 1678 | struct netdev_queue *txq; |
1da177e4 LT |
1679 | struct Qdisc *q; |
1680 | int rc = -ENOMEM; | |
1681 | ||
f6a78bfc HX |
1682 | /* GSO will handle the following emulations directly. */ |
1683 | if (netif_needs_gso(dev, skb)) | |
1684 | goto gso; | |
1685 | ||
1da177e4 LT |
1686 | if (skb_shinfo(skb)->frag_list && |
1687 | !(dev->features & NETIF_F_FRAGLIST) && | |
364c6bad | 1688 | __skb_linearize(skb)) |
1da177e4 LT |
1689 | goto out_kfree_skb; |
1690 | ||
1691 | /* Fragmented skb is linearized if device does not support SG, | |
1692 | * or if at least one of fragments is in highmem and device | |
1693 | * does not support DMA from it. | |
1694 | */ | |
1695 | if (skb_shinfo(skb)->nr_frags && | |
1696 | (!(dev->features & NETIF_F_SG) || illegal_highdma(dev, skb)) && | |
364c6bad | 1697 | __skb_linearize(skb)) |
1da177e4 LT |
1698 | goto out_kfree_skb; |
1699 | ||
1700 | /* If packet is not checksummed and device does not support | |
1701 | * checksumming for this protocol, complete checksumming here. | |
1702 | */ | |
663ead3b HX |
1703 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
1704 | skb_set_transport_header(skb, skb->csum_start - | |
1705 | skb_headroom(skb)); | |
6de329e2 BH |
1706 | if (!dev_can_checksum(dev, skb) && skb_checksum_help(skb)) |
1707 | goto out_kfree_skb; | |
663ead3b | 1708 | } |
1da177e4 | 1709 | |
f6a78bfc | 1710 | gso: |
e8a0464c | 1711 | txq = dev_pick_tx(dev, skb); |
dc2b4847 | 1712 | spin_lock_prefetch(&txq->lock); |
2d7ceece | 1713 | |
4ec93edb YH |
1714 | /* Disable soft irqs for various locks below. Also |
1715 | * stops preemption for RCU. | |
1da177e4 | 1716 | */ |
4ec93edb | 1717 | rcu_read_lock_bh(); |
1da177e4 | 1718 | |
dc2b4847 | 1719 | /* Updates of qdisc are serialized by queue->lock. |
4ec93edb YH |
1720 | * The struct Qdisc which is pointed to by qdisc is now a |
1721 | * rcu structure - it may be accessed without acquiring | |
1da177e4 | 1722 | * a lock (but the structure may be stale.) The freeing of the |
4ec93edb | 1723 | * qdisc will be deferred until it's known that there are no |
1da177e4 | 1724 | * more references to it. |
4ec93edb YH |
1725 | * |
1726 | * If the qdisc has an enqueue function, we still need to | |
dc2b4847 | 1727 | * hold the queue->lock before calling it, since queue->lock |
1da177e4 LT |
1728 | * also serializes access to the device queue. |
1729 | */ | |
1730 | ||
b0e1e646 | 1731 | q = rcu_dereference(txq->qdisc); |
1da177e4 LT |
1732 | #ifdef CONFIG_NET_CLS_ACT |
1733 | skb->tc_verd = SET_TC_AT(skb->tc_verd,AT_EGRESS); | |
1734 | #endif | |
1735 | if (q->enqueue) { | |
1736 | /* Grab device queue */ | |
dc2b4847 | 1737 | spin_lock(&txq->lock); |
b0e1e646 | 1738 | q = txq->qdisc; |
85670cc1 | 1739 | if (q->enqueue) { |
f25f4e44 | 1740 | /* reset queue_mapping to zero */ |
dfa40911 | 1741 | skb_set_queue_mapping(skb, 0); |
85670cc1 | 1742 | rc = q->enqueue(skb, q); |
eb6aafe3 | 1743 | qdisc_run(txq); |
dc2b4847 | 1744 | spin_unlock(&txq->lock); |
1da177e4 | 1745 | |
85670cc1 PM |
1746 | rc = rc == NET_XMIT_BYPASS ? NET_XMIT_SUCCESS : rc; |
1747 | goto out; | |
1748 | } | |
dc2b4847 | 1749 | spin_unlock(&txq->lock); |
1da177e4 LT |
1750 | } |
1751 | ||
1752 | /* The device has no queue. Common case for software devices: | |
1753 | loopback, all the sorts of tunnels... | |
1754 | ||
932ff279 HX |
1755 | Really, it is unlikely that netif_tx_lock protection is necessary |
1756 | here. (f.e. loopback and IP tunnels are clean ignoring statistics | |
1da177e4 LT |
1757 | counters.) |
1758 | However, it is possible, that they rely on protection | |
1759 | made by us here. | |
1760 | ||
1761 | Check this and shot the lock. It is not prone from deadlocks. | |
1762 | Either shot noqueue qdisc, it is even simpler 8) | |
1763 | */ | |
1764 | if (dev->flags & IFF_UP) { | |
1765 | int cpu = smp_processor_id(); /* ok because BHs are off */ | |
1766 | ||
c773e847 | 1767 | if (txq->xmit_lock_owner != cpu) { |
1da177e4 | 1768 | |
c773e847 | 1769 | HARD_TX_LOCK(dev, txq, cpu); |
1da177e4 | 1770 | |
f25f4e44 | 1771 | if (!netif_queue_stopped(dev) && |
668f895a | 1772 | !netif_subqueue_stopped(dev, skb)) { |
1da177e4 | 1773 | rc = 0; |
f6a78bfc | 1774 | if (!dev_hard_start_xmit(skb, dev)) { |
c773e847 | 1775 | HARD_TX_UNLOCK(dev, txq); |
1da177e4 LT |
1776 | goto out; |
1777 | } | |
1778 | } | |
c773e847 | 1779 | HARD_TX_UNLOCK(dev, txq); |
1da177e4 LT |
1780 | if (net_ratelimit()) |
1781 | printk(KERN_CRIT "Virtual device %s asks to " | |
1782 | "queue packet!\n", dev->name); | |
1783 | } else { | |
1784 | /* Recursion is detected! It is possible, | |
1785 | * unfortunately */ | |
1786 | if (net_ratelimit()) | |
1787 | printk(KERN_CRIT "Dead loop on virtual device " | |
1788 | "%s, fix it urgently!\n", dev->name); | |
1789 | } | |
1790 | } | |
1791 | ||
1792 | rc = -ENETDOWN; | |
d4828d85 | 1793 | rcu_read_unlock_bh(); |
1da177e4 LT |
1794 | |
1795 | out_kfree_skb: | |
1796 | kfree_skb(skb); | |
1797 | return rc; | |
1798 | out: | |
d4828d85 | 1799 | rcu_read_unlock_bh(); |
1da177e4 LT |
1800 | return rc; |
1801 | } | |
1802 | ||
1803 | ||
1804 | /*======================================================================= | |
1805 | Receiver routines | |
1806 | =======================================================================*/ | |
1807 | ||
6b2bedc3 SH |
1808 | int netdev_max_backlog __read_mostly = 1000; |
1809 | int netdev_budget __read_mostly = 300; | |
1810 | int weight_p __read_mostly = 64; /* old backlog weight */ | |
1da177e4 LT |
1811 | |
1812 | DEFINE_PER_CPU(struct netif_rx_stats, netdev_rx_stat) = { 0, }; | |
1813 | ||
1814 | ||
1da177e4 LT |
1815 | /** |
1816 | * netif_rx - post buffer to the network code | |
1817 | * @skb: buffer to post | |
1818 | * | |
1819 | * This function receives a packet from a device driver and queues it for | |
1820 | * the upper (protocol) levels to process. It always succeeds. The buffer | |
1821 | * may be dropped during processing for congestion control or by the | |
1822 | * protocol layers. | |
1823 | * | |
1824 | * return values: | |
1825 | * NET_RX_SUCCESS (no congestion) | |
1da177e4 LT |
1826 | * NET_RX_DROP (packet was dropped) |
1827 | * | |
1828 | */ | |
1829 | ||
1830 | int netif_rx(struct sk_buff *skb) | |
1831 | { | |
1da177e4 LT |
1832 | struct softnet_data *queue; |
1833 | unsigned long flags; | |
1834 | ||
1835 | /* if netpoll wants it, pretend we never saw it */ | |
1836 | if (netpoll_rx(skb)) | |
1837 | return NET_RX_DROP; | |
1838 | ||
b7aa0bf7 | 1839 | if (!skb->tstamp.tv64) |
a61bbcf2 | 1840 | net_timestamp(skb); |
1da177e4 LT |
1841 | |
1842 | /* | |
1843 | * The code is rearranged so that the path is the most | |
1844 | * short when CPU is congested, but is still operating. | |
1845 | */ | |
1846 | local_irq_save(flags); | |
1da177e4 LT |
1847 | queue = &__get_cpu_var(softnet_data); |
1848 | ||
1849 | __get_cpu_var(netdev_rx_stat).total++; | |
1850 | if (queue->input_pkt_queue.qlen <= netdev_max_backlog) { | |
1851 | if (queue->input_pkt_queue.qlen) { | |
1da177e4 LT |
1852 | enqueue: |
1853 | dev_hold(skb->dev); | |
1854 | __skb_queue_tail(&queue->input_pkt_queue, skb); | |
1da177e4 | 1855 | local_irq_restore(flags); |
34008d8c | 1856 | return NET_RX_SUCCESS; |
1da177e4 LT |
1857 | } |
1858 | ||
bea3348e | 1859 | napi_schedule(&queue->backlog); |
1da177e4 LT |
1860 | goto enqueue; |
1861 | } | |
1862 | ||
1da177e4 LT |
1863 | __get_cpu_var(netdev_rx_stat).dropped++; |
1864 | local_irq_restore(flags); | |
1865 | ||
1866 | kfree_skb(skb); | |
1867 | return NET_RX_DROP; | |
1868 | } | |
1869 | ||
1870 | int netif_rx_ni(struct sk_buff *skb) | |
1871 | { | |
1872 | int err; | |
1873 | ||
1874 | preempt_disable(); | |
1875 | err = netif_rx(skb); | |
1876 | if (local_softirq_pending()) | |
1877 | do_softirq(); | |
1878 | preempt_enable(); | |
1879 | ||
1880 | return err; | |
1881 | } | |
1882 | ||
1883 | EXPORT_SYMBOL(netif_rx_ni); | |
1884 | ||
f2ccd8fa | 1885 | static inline struct net_device *skb_bond(struct sk_buff *skb) |
1da177e4 LT |
1886 | { |
1887 | struct net_device *dev = skb->dev; | |
1888 | ||
8f903c70 | 1889 | if (dev->master) { |
7ea49ed7 | 1890 | if (skb_bond_should_drop(skb)) { |
8f903c70 JV |
1891 | kfree_skb(skb); |
1892 | return NULL; | |
1893 | } | |
1da177e4 | 1894 | skb->dev = dev->master; |
8f903c70 | 1895 | } |
f2ccd8fa DM |
1896 | |
1897 | return dev; | |
1da177e4 LT |
1898 | } |
1899 | ||
bea3348e | 1900 | |
1da177e4 LT |
1901 | static void net_tx_action(struct softirq_action *h) |
1902 | { | |
1903 | struct softnet_data *sd = &__get_cpu_var(softnet_data); | |
1904 | ||
1905 | if (sd->completion_queue) { | |
1906 | struct sk_buff *clist; | |
1907 | ||
1908 | local_irq_disable(); | |
1909 | clist = sd->completion_queue; | |
1910 | sd->completion_queue = NULL; | |
1911 | local_irq_enable(); | |
1912 | ||
1913 | while (clist) { | |
1914 | struct sk_buff *skb = clist; | |
1915 | clist = clist->next; | |
1916 | ||
1917 | BUG_TRAP(!atomic_read(&skb->users)); | |
1918 | __kfree_skb(skb); | |
1919 | } | |
1920 | } | |
1921 | ||
1922 | if (sd->output_queue) { | |
ee609cb3 | 1923 | struct netdev_queue *head; |
1da177e4 LT |
1924 | |
1925 | local_irq_disable(); | |
1926 | head = sd->output_queue; | |
1927 | sd->output_queue = NULL; | |
1928 | local_irq_enable(); | |
1929 | ||
1930 | while (head) { | |
ee609cb3 DM |
1931 | struct netdev_queue *txq = head; |
1932 | struct net_device *dev = txq->dev; | |
1da177e4 LT |
1933 | head = head->next_sched; |
1934 | ||
1935 | smp_mb__before_clear_bit(); | |
1936 | clear_bit(__LINK_STATE_SCHED, &dev->state); | |
1937 | ||
dc2b4847 | 1938 | if (spin_trylock(&txq->lock)) { |
eb6aafe3 | 1939 | qdisc_run(txq); |
dc2b4847 | 1940 | spin_unlock(&txq->lock); |
1da177e4 | 1941 | } else { |
86d804e1 | 1942 | netif_schedule_queue(txq); |
1da177e4 LT |
1943 | } |
1944 | } | |
1945 | } | |
1946 | } | |
1947 | ||
6f05f629 SH |
1948 | static inline int deliver_skb(struct sk_buff *skb, |
1949 | struct packet_type *pt_prev, | |
1950 | struct net_device *orig_dev) | |
1da177e4 LT |
1951 | { |
1952 | atomic_inc(&skb->users); | |
f2ccd8fa | 1953 | return pt_prev->func(skb, skb->dev, pt_prev, orig_dev); |
1da177e4 LT |
1954 | } |
1955 | ||
1956 | #if defined(CONFIG_BRIDGE) || defined (CONFIG_BRIDGE_MODULE) | |
6229e362 | 1957 | /* These hooks defined here for ATM */ |
1da177e4 LT |
1958 | struct net_bridge; |
1959 | struct net_bridge_fdb_entry *(*br_fdb_get_hook)(struct net_bridge *br, | |
1960 | unsigned char *addr); | |
6229e362 | 1961 | void (*br_fdb_put_hook)(struct net_bridge_fdb_entry *ent) __read_mostly; |
1da177e4 | 1962 | |
6229e362 SH |
1963 | /* |
1964 | * If bridge module is loaded call bridging hook. | |
1965 | * returns NULL if packet was consumed. | |
1966 | */ | |
1967 | struct sk_buff *(*br_handle_frame_hook)(struct net_bridge_port *p, | |
1968 | struct sk_buff *skb) __read_mostly; | |
1969 | static inline struct sk_buff *handle_bridge(struct sk_buff *skb, | |
1970 | struct packet_type **pt_prev, int *ret, | |
1971 | struct net_device *orig_dev) | |
1da177e4 LT |
1972 | { |
1973 | struct net_bridge_port *port; | |
1974 | ||
6229e362 SH |
1975 | if (skb->pkt_type == PACKET_LOOPBACK || |
1976 | (port = rcu_dereference(skb->dev->br_port)) == NULL) | |
1977 | return skb; | |
1da177e4 LT |
1978 | |
1979 | if (*pt_prev) { | |
6229e362 | 1980 | *ret = deliver_skb(skb, *pt_prev, orig_dev); |
1da177e4 | 1981 | *pt_prev = NULL; |
4ec93edb YH |
1982 | } |
1983 | ||
6229e362 | 1984 | return br_handle_frame_hook(port, skb); |
1da177e4 LT |
1985 | } |
1986 | #else | |
6229e362 | 1987 | #define handle_bridge(skb, pt_prev, ret, orig_dev) (skb) |
1da177e4 LT |
1988 | #endif |
1989 | ||
b863ceb7 PM |
1990 | #if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) |
1991 | struct sk_buff *(*macvlan_handle_frame_hook)(struct sk_buff *skb) __read_mostly; | |
1992 | EXPORT_SYMBOL_GPL(macvlan_handle_frame_hook); | |
1993 | ||
1994 | static inline struct sk_buff *handle_macvlan(struct sk_buff *skb, | |
1995 | struct packet_type **pt_prev, | |
1996 | int *ret, | |
1997 | struct net_device *orig_dev) | |
1998 | { | |
1999 | if (skb->dev->macvlan_port == NULL) | |
2000 | return skb; | |
2001 | ||
2002 | if (*pt_prev) { | |
2003 | *ret = deliver_skb(skb, *pt_prev, orig_dev); | |
2004 | *pt_prev = NULL; | |
2005 | } | |
2006 | return macvlan_handle_frame_hook(skb); | |
2007 | } | |
2008 | #else | |
2009 | #define handle_macvlan(skb, pt_prev, ret, orig_dev) (skb) | |
2010 | #endif | |
2011 | ||
1da177e4 LT |
2012 | #ifdef CONFIG_NET_CLS_ACT |
2013 | /* TODO: Maybe we should just force sch_ingress to be compiled in | |
2014 | * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions | |
2015 | * a compare and 2 stores extra right now if we dont have it on | |
2016 | * but have CONFIG_NET_CLS_ACT | |
4ec93edb | 2017 | * NOTE: This doesnt stop any functionality; if you dont have |
1da177e4 LT |
2018 | * the ingress scheduler, you just cant add policies on ingress. |
2019 | * | |
2020 | */ | |
4ec93edb | 2021 | static int ing_filter(struct sk_buff *skb) |
1da177e4 | 2022 | { |
1da177e4 | 2023 | struct net_device *dev = skb->dev; |
f697c3e8 | 2024 | u32 ttl = G_TC_RTTL(skb->tc_verd); |
555353cf DM |
2025 | struct netdev_queue *rxq; |
2026 | int result = TC_ACT_OK; | |
2027 | struct Qdisc *q; | |
4ec93edb | 2028 | |
f697c3e8 HX |
2029 | if (MAX_RED_LOOP < ttl++) { |
2030 | printk(KERN_WARNING | |
2031 | "Redir loop detected Dropping packet (%d->%d)\n", | |
2032 | skb->iif, dev->ifindex); | |
2033 | return TC_ACT_SHOT; | |
2034 | } | |
1da177e4 | 2035 | |
f697c3e8 HX |
2036 | skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl); |
2037 | skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS); | |
1da177e4 | 2038 | |
555353cf DM |
2039 | rxq = &dev->rx_queue; |
2040 | ||
2041 | spin_lock(&rxq->lock); | |
816f3258 | 2042 | if ((q = rxq->qdisc) != NULL) |
f697c3e8 | 2043 | result = q->enqueue(skb, q); |
555353cf | 2044 | spin_unlock(&rxq->lock); |
f697c3e8 HX |
2045 | |
2046 | return result; | |
2047 | } | |
86e65da9 | 2048 | |
f697c3e8 HX |
2049 | static inline struct sk_buff *handle_ing(struct sk_buff *skb, |
2050 | struct packet_type **pt_prev, | |
2051 | int *ret, struct net_device *orig_dev) | |
2052 | { | |
816f3258 | 2053 | if (!skb->dev->rx_queue.qdisc) |
f697c3e8 | 2054 | goto out; |
1da177e4 | 2055 | |
f697c3e8 HX |
2056 | if (*pt_prev) { |
2057 | *ret = deliver_skb(skb, *pt_prev, orig_dev); | |
2058 | *pt_prev = NULL; | |
2059 | } else { | |
2060 | /* Huh? Why does turning on AF_PACKET affect this? */ | |
2061 | skb->tc_verd = SET_TC_OK2MUNGE(skb->tc_verd); | |
1da177e4 LT |
2062 | } |
2063 | ||
f697c3e8 HX |
2064 | switch (ing_filter(skb)) { |
2065 | case TC_ACT_SHOT: | |
2066 | case TC_ACT_STOLEN: | |
2067 | kfree_skb(skb); | |
2068 | return NULL; | |
2069 | } | |
2070 | ||
2071 | out: | |
2072 | skb->tc_verd = 0; | |
2073 | return skb; | |
1da177e4 LT |
2074 | } |
2075 | #endif | |
2076 | ||
bc1d0411 PM |
2077 | /* |
2078 | * netif_nit_deliver - deliver received packets to network taps | |
2079 | * @skb: buffer | |
2080 | * | |
2081 | * This function is used to deliver incoming packets to network | |
2082 | * taps. It should be used when the normal netif_receive_skb path | |
2083 | * is bypassed, for example because of VLAN acceleration. | |
2084 | */ | |
2085 | void netif_nit_deliver(struct sk_buff *skb) | |
2086 | { | |
2087 | struct packet_type *ptype; | |
2088 | ||
2089 | if (list_empty(&ptype_all)) | |
2090 | return; | |
2091 | ||
2092 | skb_reset_network_header(skb); | |
2093 | skb_reset_transport_header(skb); | |
2094 | skb->mac_len = skb->network_header - skb->mac_header; | |
2095 | ||
2096 | rcu_read_lock(); | |
2097 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
2098 | if (!ptype->dev || ptype->dev == skb->dev) | |
2099 | deliver_skb(skb, ptype, skb->dev); | |
2100 | } | |
2101 | rcu_read_unlock(); | |
2102 | } | |
2103 | ||
3b582cc1 SH |
2104 | /** |
2105 | * netif_receive_skb - process receive buffer from network | |
2106 | * @skb: buffer to process | |
2107 | * | |
2108 | * netif_receive_skb() is the main receive data processing function. | |
2109 | * It always succeeds. The buffer may be dropped during processing | |
2110 | * for congestion control or by the protocol layers. | |
2111 | * | |
2112 | * This function may only be called from softirq context and interrupts | |
2113 | * should be enabled. | |
2114 | * | |
2115 | * Return values (usually ignored): | |
2116 | * NET_RX_SUCCESS: no congestion | |
2117 | * NET_RX_DROP: packet was dropped | |
2118 | */ | |
1da177e4 LT |
2119 | int netif_receive_skb(struct sk_buff *skb) |
2120 | { | |
2121 | struct packet_type *ptype, *pt_prev; | |
f2ccd8fa | 2122 | struct net_device *orig_dev; |
1da177e4 | 2123 | int ret = NET_RX_DROP; |
252e3346 | 2124 | __be16 type; |
1da177e4 LT |
2125 | |
2126 | /* if we've gotten here through NAPI, check netpoll */ | |
bea3348e | 2127 | if (netpoll_receive_skb(skb)) |
1da177e4 LT |
2128 | return NET_RX_DROP; |
2129 | ||
b7aa0bf7 | 2130 | if (!skb->tstamp.tv64) |
a61bbcf2 | 2131 | net_timestamp(skb); |
1da177e4 | 2132 | |
c01003c2 PM |
2133 | if (!skb->iif) |
2134 | skb->iif = skb->dev->ifindex; | |
86e65da9 | 2135 | |
f2ccd8fa | 2136 | orig_dev = skb_bond(skb); |
1da177e4 | 2137 | |
8f903c70 JV |
2138 | if (!orig_dev) |
2139 | return NET_RX_DROP; | |
2140 | ||
1da177e4 LT |
2141 | __get_cpu_var(netdev_rx_stat).total++; |
2142 | ||
c1d2bbe1 | 2143 | skb_reset_network_header(skb); |
badff6d0 | 2144 | skb_reset_transport_header(skb); |
b0e380b1 | 2145 | skb->mac_len = skb->network_header - skb->mac_header; |
1da177e4 LT |
2146 | |
2147 | pt_prev = NULL; | |
2148 | ||
2149 | rcu_read_lock(); | |
2150 | ||
b9f75f45 EB |
2151 | /* Don't receive packets in an exiting network namespace */ |
2152 | if (!net_alive(dev_net(skb->dev))) | |
2153 | goto out; | |
2154 | ||
1da177e4 LT |
2155 | #ifdef CONFIG_NET_CLS_ACT |
2156 | if (skb->tc_verd & TC_NCLS) { | |
2157 | skb->tc_verd = CLR_TC_NCLS(skb->tc_verd); | |
2158 | goto ncls; | |
2159 | } | |
2160 | #endif | |
2161 | ||
2162 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
2163 | if (!ptype->dev || ptype->dev == skb->dev) { | |
4ec93edb | 2164 | if (pt_prev) |
f2ccd8fa | 2165 | ret = deliver_skb(skb, pt_prev, orig_dev); |
1da177e4 LT |
2166 | pt_prev = ptype; |
2167 | } | |
2168 | } | |
2169 | ||
2170 | #ifdef CONFIG_NET_CLS_ACT | |
f697c3e8 HX |
2171 | skb = handle_ing(skb, &pt_prev, &ret, orig_dev); |
2172 | if (!skb) | |
1da177e4 | 2173 | goto out; |
1da177e4 LT |
2174 | ncls: |
2175 | #endif | |
2176 | ||
6229e362 | 2177 | skb = handle_bridge(skb, &pt_prev, &ret, orig_dev); |
b863ceb7 PM |
2178 | if (!skb) |
2179 | goto out; | |
2180 | skb = handle_macvlan(skb, &pt_prev, &ret, orig_dev); | |
6229e362 | 2181 | if (!skb) |
1da177e4 LT |
2182 | goto out; |
2183 | ||
2184 | type = skb->protocol; | |
82d8a867 PE |
2185 | list_for_each_entry_rcu(ptype, |
2186 | &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { | |
1da177e4 LT |
2187 | if (ptype->type == type && |
2188 | (!ptype->dev || ptype->dev == skb->dev)) { | |
4ec93edb | 2189 | if (pt_prev) |
f2ccd8fa | 2190 | ret = deliver_skb(skb, pt_prev, orig_dev); |
1da177e4 LT |
2191 | pt_prev = ptype; |
2192 | } | |
2193 | } | |
2194 | ||
2195 | if (pt_prev) { | |
f2ccd8fa | 2196 | ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev); |
1da177e4 LT |
2197 | } else { |
2198 | kfree_skb(skb); | |
2199 | /* Jamal, now you will not able to escape explaining | |
2200 | * me how you were going to use this. :-) | |
2201 | */ | |
2202 | ret = NET_RX_DROP; | |
2203 | } | |
2204 | ||
2205 | out: | |
2206 | rcu_read_unlock(); | |
2207 | return ret; | |
2208 | } | |
2209 | ||
bea3348e | 2210 | static int process_backlog(struct napi_struct *napi, int quota) |
1da177e4 LT |
2211 | { |
2212 | int work = 0; | |
1da177e4 LT |
2213 | struct softnet_data *queue = &__get_cpu_var(softnet_data); |
2214 | unsigned long start_time = jiffies; | |
2215 | ||
bea3348e SH |
2216 | napi->weight = weight_p; |
2217 | do { | |
1da177e4 LT |
2218 | struct sk_buff *skb; |
2219 | struct net_device *dev; | |
2220 | ||
2221 | local_irq_disable(); | |
2222 | skb = __skb_dequeue(&queue->input_pkt_queue); | |
bea3348e SH |
2223 | if (!skb) { |
2224 | __napi_complete(napi); | |
2225 | local_irq_enable(); | |
2226 | break; | |
2227 | } | |
2228 | ||
1da177e4 LT |
2229 | local_irq_enable(); |
2230 | ||
2231 | dev = skb->dev; | |
2232 | ||
2233 | netif_receive_skb(skb); | |
2234 | ||
2235 | dev_put(dev); | |
bea3348e | 2236 | } while (++work < quota && jiffies == start_time); |
1da177e4 | 2237 | |
bea3348e SH |
2238 | return work; |
2239 | } | |
1da177e4 | 2240 | |
bea3348e SH |
2241 | /** |
2242 | * __napi_schedule - schedule for receive | |
c4ea43c5 | 2243 | * @n: entry to schedule |
bea3348e SH |
2244 | * |
2245 | * The entry's receive function will be scheduled to run | |
2246 | */ | |
b5606c2d | 2247 | void __napi_schedule(struct napi_struct *n) |
bea3348e SH |
2248 | { |
2249 | unsigned long flags; | |
1da177e4 | 2250 | |
bea3348e SH |
2251 | local_irq_save(flags); |
2252 | list_add_tail(&n->poll_list, &__get_cpu_var(softnet_data).poll_list); | |
2253 | __raise_softirq_irqoff(NET_RX_SOFTIRQ); | |
2254 | local_irq_restore(flags); | |
1da177e4 | 2255 | } |
bea3348e SH |
2256 | EXPORT_SYMBOL(__napi_schedule); |
2257 | ||
1da177e4 LT |
2258 | |
2259 | static void net_rx_action(struct softirq_action *h) | |
2260 | { | |
bea3348e | 2261 | struct list_head *list = &__get_cpu_var(softnet_data).poll_list; |
1da177e4 | 2262 | unsigned long start_time = jiffies; |
51b0bded | 2263 | int budget = netdev_budget; |
53fb95d3 MM |
2264 | void *have; |
2265 | ||
1da177e4 LT |
2266 | local_irq_disable(); |
2267 | ||
bea3348e SH |
2268 | while (!list_empty(list)) { |
2269 | struct napi_struct *n; | |
2270 | int work, weight; | |
1da177e4 | 2271 | |
bea3348e SH |
2272 | /* If softirq window is exhuasted then punt. |
2273 | * | |
2274 | * Note that this is a slight policy change from the | |
2275 | * previous NAPI code, which would allow up to 2 | |
2276 | * jiffies to pass before breaking out. The test | |
2277 | * used to be "jiffies - start_time > 1". | |
2278 | */ | |
2279 | if (unlikely(budget <= 0 || jiffies != start_time)) | |
1da177e4 LT |
2280 | goto softnet_break; |
2281 | ||
2282 | local_irq_enable(); | |
2283 | ||
bea3348e SH |
2284 | /* Even though interrupts have been re-enabled, this |
2285 | * access is safe because interrupts can only add new | |
2286 | * entries to the tail of this list, and only ->poll() | |
2287 | * calls can remove this head entry from the list. | |
2288 | */ | |
2289 | n = list_entry(list->next, struct napi_struct, poll_list); | |
1da177e4 | 2290 | |
bea3348e SH |
2291 | have = netpoll_poll_lock(n); |
2292 | ||
2293 | weight = n->weight; | |
2294 | ||
0a7606c1 DM |
2295 | /* This NAPI_STATE_SCHED test is for avoiding a race |
2296 | * with netpoll's poll_napi(). Only the entity which | |
2297 | * obtains the lock and sees NAPI_STATE_SCHED set will | |
2298 | * actually make the ->poll() call. Therefore we avoid | |
2299 | * accidently calling ->poll() when NAPI is not scheduled. | |
2300 | */ | |
2301 | work = 0; | |
2302 | if (test_bit(NAPI_STATE_SCHED, &n->state)) | |
2303 | work = n->poll(n, weight); | |
bea3348e SH |
2304 | |
2305 | WARN_ON_ONCE(work > weight); | |
2306 | ||
2307 | budget -= work; | |
2308 | ||
2309 | local_irq_disable(); | |
2310 | ||
2311 | /* Drivers must not modify the NAPI state if they | |
2312 | * consume the entire weight. In such cases this code | |
2313 | * still "owns" the NAPI instance and therefore can | |
2314 | * move the instance around on the list at-will. | |
2315 | */ | |
fed17f30 DM |
2316 | if (unlikely(work == weight)) { |
2317 | if (unlikely(napi_disable_pending(n))) | |
2318 | __napi_complete(n); | |
2319 | else | |
2320 | list_move_tail(&n->poll_list, list); | |
2321 | } | |
bea3348e SH |
2322 | |
2323 | netpoll_poll_unlock(have); | |
1da177e4 LT |
2324 | } |
2325 | out: | |
515e06c4 | 2326 | local_irq_enable(); |
bea3348e | 2327 | |
db217334 CL |
2328 | #ifdef CONFIG_NET_DMA |
2329 | /* | |
2330 | * There may not be any more sk_buffs coming right now, so push | |
2331 | * any pending DMA copies to hardware | |
2332 | */ | |
d379b01e DW |
2333 | if (!cpus_empty(net_dma.channel_mask)) { |
2334 | int chan_idx; | |
2335 | for_each_cpu_mask(chan_idx, net_dma.channel_mask) { | |
2336 | struct dma_chan *chan = net_dma.channels[chan_idx]; | |
2337 | if (chan) | |
2338 | dma_async_memcpy_issue_pending(chan); | |
2339 | } | |
db217334 CL |
2340 | } |
2341 | #endif | |
bea3348e | 2342 | |
1da177e4 LT |
2343 | return; |
2344 | ||
2345 | softnet_break: | |
2346 | __get_cpu_var(netdev_rx_stat).time_squeeze++; | |
2347 | __raise_softirq_irqoff(NET_RX_SOFTIRQ); | |
2348 | goto out; | |
2349 | } | |
2350 | ||
2351 | static gifconf_func_t * gifconf_list [NPROTO]; | |
2352 | ||
2353 | /** | |
2354 | * register_gifconf - register a SIOCGIF handler | |
2355 | * @family: Address family | |
2356 | * @gifconf: Function handler | |
2357 | * | |
2358 | * Register protocol dependent address dumping routines. The handler | |
2359 | * that is passed must not be freed or reused until it has been replaced | |
2360 | * by another handler. | |
2361 | */ | |
2362 | int register_gifconf(unsigned int family, gifconf_func_t * gifconf) | |
2363 | { | |
2364 | if (family >= NPROTO) | |
2365 | return -EINVAL; | |
2366 | gifconf_list[family] = gifconf; | |
2367 | return 0; | |
2368 | } | |
2369 | ||
2370 | ||
2371 | /* | |
2372 | * Map an interface index to its name (SIOCGIFNAME) | |
2373 | */ | |
2374 | ||
2375 | /* | |
2376 | * We need this ioctl for efficient implementation of the | |
2377 | * if_indextoname() function required by the IPv6 API. Without | |
2378 | * it, we would have to search all the interfaces to find a | |
2379 | * match. --pb | |
2380 | */ | |
2381 | ||
881d966b | 2382 | static int dev_ifname(struct net *net, struct ifreq __user *arg) |
1da177e4 LT |
2383 | { |
2384 | struct net_device *dev; | |
2385 | struct ifreq ifr; | |
2386 | ||
2387 | /* | |
2388 | * Fetch the caller's info block. | |
2389 | */ | |
2390 | ||
2391 | if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) | |
2392 | return -EFAULT; | |
2393 | ||
2394 | read_lock(&dev_base_lock); | |
881d966b | 2395 | dev = __dev_get_by_index(net, ifr.ifr_ifindex); |
1da177e4 LT |
2396 | if (!dev) { |
2397 | read_unlock(&dev_base_lock); | |
2398 | return -ENODEV; | |
2399 | } | |
2400 | ||
2401 | strcpy(ifr.ifr_name, dev->name); | |
2402 | read_unlock(&dev_base_lock); | |
2403 | ||
2404 | if (copy_to_user(arg, &ifr, sizeof(struct ifreq))) | |
2405 | return -EFAULT; | |
2406 | return 0; | |
2407 | } | |
2408 | ||
2409 | /* | |
2410 | * Perform a SIOCGIFCONF call. This structure will change | |
2411 | * size eventually, and there is nothing I can do about it. | |
2412 | * Thus we will need a 'compatibility mode'. | |
2413 | */ | |
2414 | ||
881d966b | 2415 | static int dev_ifconf(struct net *net, char __user *arg) |
1da177e4 LT |
2416 | { |
2417 | struct ifconf ifc; | |
2418 | struct net_device *dev; | |
2419 | char __user *pos; | |
2420 | int len; | |
2421 | int total; | |
2422 | int i; | |
2423 | ||
2424 | /* | |
2425 | * Fetch the caller's info block. | |
2426 | */ | |
2427 | ||
2428 | if (copy_from_user(&ifc, arg, sizeof(struct ifconf))) | |
2429 | return -EFAULT; | |
2430 | ||
2431 | pos = ifc.ifc_buf; | |
2432 | len = ifc.ifc_len; | |
2433 | ||
2434 | /* | |
2435 | * Loop over the interfaces, and write an info block for each. | |
2436 | */ | |
2437 | ||
2438 | total = 0; | |
881d966b | 2439 | for_each_netdev(net, dev) { |
1da177e4 LT |
2440 | for (i = 0; i < NPROTO; i++) { |
2441 | if (gifconf_list[i]) { | |
2442 | int done; | |
2443 | if (!pos) | |
2444 | done = gifconf_list[i](dev, NULL, 0); | |
2445 | else | |
2446 | done = gifconf_list[i](dev, pos + total, | |
2447 | len - total); | |
2448 | if (done < 0) | |
2449 | return -EFAULT; | |
2450 | total += done; | |
2451 | } | |
2452 | } | |
4ec93edb | 2453 | } |
1da177e4 LT |
2454 | |
2455 | /* | |
2456 | * All done. Write the updated control block back to the caller. | |
2457 | */ | |
2458 | ifc.ifc_len = total; | |
2459 | ||
2460 | /* | |
2461 | * Both BSD and Solaris return 0 here, so we do too. | |
2462 | */ | |
2463 | return copy_to_user(arg, &ifc, sizeof(struct ifconf)) ? -EFAULT : 0; | |
2464 | } | |
2465 | ||
2466 | #ifdef CONFIG_PROC_FS | |
2467 | /* | |
2468 | * This is invoked by the /proc filesystem handler to display a device | |
2469 | * in detail. | |
2470 | */ | |
7562f876 | 2471 | void *dev_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2472 | __acquires(dev_base_lock) |
1da177e4 | 2473 | { |
e372c414 | 2474 | struct net *net = seq_file_net(seq); |
7562f876 | 2475 | loff_t off; |
1da177e4 | 2476 | struct net_device *dev; |
1da177e4 | 2477 | |
7562f876 PE |
2478 | read_lock(&dev_base_lock); |
2479 | if (!*pos) | |
2480 | return SEQ_START_TOKEN; | |
1da177e4 | 2481 | |
7562f876 | 2482 | off = 1; |
881d966b | 2483 | for_each_netdev(net, dev) |
7562f876 PE |
2484 | if (off++ == *pos) |
2485 | return dev; | |
1da177e4 | 2486 | |
7562f876 | 2487 | return NULL; |
1da177e4 LT |
2488 | } |
2489 | ||
2490 | void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2491 | { | |
e372c414 | 2492 | struct net *net = seq_file_net(seq); |
1da177e4 | 2493 | ++*pos; |
7562f876 | 2494 | return v == SEQ_START_TOKEN ? |
881d966b | 2495 | first_net_device(net) : next_net_device((struct net_device *)v); |
1da177e4 LT |
2496 | } |
2497 | ||
2498 | void dev_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2499 | __releases(dev_base_lock) |
1da177e4 LT |
2500 | { |
2501 | read_unlock(&dev_base_lock); | |
2502 | } | |
2503 | ||
2504 | static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev) | |
2505 | { | |
c45d286e | 2506 | struct net_device_stats *stats = dev->get_stats(dev); |
1da177e4 | 2507 | |
5a1b5898 RR |
2508 | seq_printf(seq, "%6s:%8lu %7lu %4lu %4lu %4lu %5lu %10lu %9lu " |
2509 | "%8lu %7lu %4lu %4lu %4lu %5lu %7lu %10lu\n", | |
2510 | dev->name, stats->rx_bytes, stats->rx_packets, | |
2511 | stats->rx_errors, | |
2512 | stats->rx_dropped + stats->rx_missed_errors, | |
2513 | stats->rx_fifo_errors, | |
2514 | stats->rx_length_errors + stats->rx_over_errors + | |
2515 | stats->rx_crc_errors + stats->rx_frame_errors, | |
2516 | stats->rx_compressed, stats->multicast, | |
2517 | stats->tx_bytes, stats->tx_packets, | |
2518 | stats->tx_errors, stats->tx_dropped, | |
2519 | stats->tx_fifo_errors, stats->collisions, | |
2520 | stats->tx_carrier_errors + | |
2521 | stats->tx_aborted_errors + | |
2522 | stats->tx_window_errors + | |
2523 | stats->tx_heartbeat_errors, | |
2524 | stats->tx_compressed); | |
1da177e4 LT |
2525 | } |
2526 | ||
2527 | /* | |
2528 | * Called from the PROCfs module. This now uses the new arbitrary sized | |
2529 | * /proc/net interface to create /proc/net/dev | |
2530 | */ | |
2531 | static int dev_seq_show(struct seq_file *seq, void *v) | |
2532 | { | |
2533 | if (v == SEQ_START_TOKEN) | |
2534 | seq_puts(seq, "Inter-| Receive " | |
2535 | " | Transmit\n" | |
2536 | " face |bytes packets errs drop fifo frame " | |
2537 | "compressed multicast|bytes packets errs " | |
2538 | "drop fifo colls carrier compressed\n"); | |
2539 | else | |
2540 | dev_seq_printf_stats(seq, v); | |
2541 | return 0; | |
2542 | } | |
2543 | ||
2544 | static struct netif_rx_stats *softnet_get_online(loff_t *pos) | |
2545 | { | |
2546 | struct netif_rx_stats *rc = NULL; | |
2547 | ||
0c0b0aca | 2548 | while (*pos < nr_cpu_ids) |
4ec93edb | 2549 | if (cpu_online(*pos)) { |
1da177e4 LT |
2550 | rc = &per_cpu(netdev_rx_stat, *pos); |
2551 | break; | |
2552 | } else | |
2553 | ++*pos; | |
2554 | return rc; | |
2555 | } | |
2556 | ||
2557 | static void *softnet_seq_start(struct seq_file *seq, loff_t *pos) | |
2558 | { | |
2559 | return softnet_get_online(pos); | |
2560 | } | |
2561 | ||
2562 | static void *softnet_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2563 | { | |
2564 | ++*pos; | |
2565 | return softnet_get_online(pos); | |
2566 | } | |
2567 | ||
2568 | static void softnet_seq_stop(struct seq_file *seq, void *v) | |
2569 | { | |
2570 | } | |
2571 | ||
2572 | static int softnet_seq_show(struct seq_file *seq, void *v) | |
2573 | { | |
2574 | struct netif_rx_stats *s = v; | |
2575 | ||
2576 | seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n", | |
31aa02c5 | 2577 | s->total, s->dropped, s->time_squeeze, 0, |
c1ebcdb8 SH |
2578 | 0, 0, 0, 0, /* was fastroute */ |
2579 | s->cpu_collision ); | |
1da177e4 LT |
2580 | return 0; |
2581 | } | |
2582 | ||
f690808e | 2583 | static const struct seq_operations dev_seq_ops = { |
1da177e4 LT |
2584 | .start = dev_seq_start, |
2585 | .next = dev_seq_next, | |
2586 | .stop = dev_seq_stop, | |
2587 | .show = dev_seq_show, | |
2588 | }; | |
2589 | ||
2590 | static int dev_seq_open(struct inode *inode, struct file *file) | |
2591 | { | |
e372c414 DL |
2592 | return seq_open_net(inode, file, &dev_seq_ops, |
2593 | sizeof(struct seq_net_private)); | |
1da177e4 LT |
2594 | } |
2595 | ||
9a32144e | 2596 | static const struct file_operations dev_seq_fops = { |
1da177e4 LT |
2597 | .owner = THIS_MODULE, |
2598 | .open = dev_seq_open, | |
2599 | .read = seq_read, | |
2600 | .llseek = seq_lseek, | |
e372c414 | 2601 | .release = seq_release_net, |
1da177e4 LT |
2602 | }; |
2603 | ||
f690808e | 2604 | static const struct seq_operations softnet_seq_ops = { |
1da177e4 LT |
2605 | .start = softnet_seq_start, |
2606 | .next = softnet_seq_next, | |
2607 | .stop = softnet_seq_stop, | |
2608 | .show = softnet_seq_show, | |
2609 | }; | |
2610 | ||
2611 | static int softnet_seq_open(struct inode *inode, struct file *file) | |
2612 | { | |
2613 | return seq_open(file, &softnet_seq_ops); | |
2614 | } | |
2615 | ||
9a32144e | 2616 | static const struct file_operations softnet_seq_fops = { |
1da177e4 LT |
2617 | .owner = THIS_MODULE, |
2618 | .open = softnet_seq_open, | |
2619 | .read = seq_read, | |
2620 | .llseek = seq_lseek, | |
2621 | .release = seq_release, | |
2622 | }; | |
2623 | ||
0e1256ff SH |
2624 | static void *ptype_get_idx(loff_t pos) |
2625 | { | |
2626 | struct packet_type *pt = NULL; | |
2627 | loff_t i = 0; | |
2628 | int t; | |
2629 | ||
2630 | list_for_each_entry_rcu(pt, &ptype_all, list) { | |
2631 | if (i == pos) | |
2632 | return pt; | |
2633 | ++i; | |
2634 | } | |
2635 | ||
82d8a867 | 2636 | for (t = 0; t < PTYPE_HASH_SIZE; t++) { |
0e1256ff SH |
2637 | list_for_each_entry_rcu(pt, &ptype_base[t], list) { |
2638 | if (i == pos) | |
2639 | return pt; | |
2640 | ++i; | |
2641 | } | |
2642 | } | |
2643 | return NULL; | |
2644 | } | |
2645 | ||
2646 | static void *ptype_seq_start(struct seq_file *seq, loff_t *pos) | |
72348a42 | 2647 | __acquires(RCU) |
0e1256ff SH |
2648 | { |
2649 | rcu_read_lock(); | |
2650 | return *pos ? ptype_get_idx(*pos - 1) : SEQ_START_TOKEN; | |
2651 | } | |
2652 | ||
2653 | static void *ptype_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2654 | { | |
2655 | struct packet_type *pt; | |
2656 | struct list_head *nxt; | |
2657 | int hash; | |
2658 | ||
2659 | ++*pos; | |
2660 | if (v == SEQ_START_TOKEN) | |
2661 | return ptype_get_idx(0); | |
2662 | ||
2663 | pt = v; | |
2664 | nxt = pt->list.next; | |
2665 | if (pt->type == htons(ETH_P_ALL)) { | |
2666 | if (nxt != &ptype_all) | |
2667 | goto found; | |
2668 | hash = 0; | |
2669 | nxt = ptype_base[0].next; | |
2670 | } else | |
82d8a867 | 2671 | hash = ntohs(pt->type) & PTYPE_HASH_MASK; |
0e1256ff SH |
2672 | |
2673 | while (nxt == &ptype_base[hash]) { | |
82d8a867 | 2674 | if (++hash >= PTYPE_HASH_SIZE) |
0e1256ff SH |
2675 | return NULL; |
2676 | nxt = ptype_base[hash].next; | |
2677 | } | |
2678 | found: | |
2679 | return list_entry(nxt, struct packet_type, list); | |
2680 | } | |
2681 | ||
2682 | static void ptype_seq_stop(struct seq_file *seq, void *v) | |
72348a42 | 2683 | __releases(RCU) |
0e1256ff SH |
2684 | { |
2685 | rcu_read_unlock(); | |
2686 | } | |
2687 | ||
2688 | static void ptype_seq_decode(struct seq_file *seq, void *sym) | |
2689 | { | |
2690 | #ifdef CONFIG_KALLSYMS | |
2691 | unsigned long offset = 0, symsize; | |
2692 | const char *symname; | |
2693 | char *modname; | |
2694 | char namebuf[128]; | |
2695 | ||
2696 | symname = kallsyms_lookup((unsigned long)sym, &symsize, &offset, | |
2697 | &modname, namebuf); | |
2698 | ||
2699 | if (symname) { | |
2700 | char *delim = ":"; | |
2701 | ||
2702 | if (!modname) | |
2703 | modname = delim = ""; | |
2704 | seq_printf(seq, "%s%s%s%s+0x%lx", delim, modname, delim, | |
2705 | symname, offset); | |
2706 | return; | |
2707 | } | |
2708 | #endif | |
2709 | ||
2710 | seq_printf(seq, "[%p]", sym); | |
2711 | } | |
2712 | ||
2713 | static int ptype_seq_show(struct seq_file *seq, void *v) | |
2714 | { | |
2715 | struct packet_type *pt = v; | |
2716 | ||
2717 | if (v == SEQ_START_TOKEN) | |
2718 | seq_puts(seq, "Type Device Function\n"); | |
c346dca1 | 2719 | else if (pt->dev == NULL || dev_net(pt->dev) == seq_file_net(seq)) { |
0e1256ff SH |
2720 | if (pt->type == htons(ETH_P_ALL)) |
2721 | seq_puts(seq, "ALL "); | |
2722 | else | |
2723 | seq_printf(seq, "%04x", ntohs(pt->type)); | |
2724 | ||
2725 | seq_printf(seq, " %-8s ", | |
2726 | pt->dev ? pt->dev->name : ""); | |
2727 | ptype_seq_decode(seq, pt->func); | |
2728 | seq_putc(seq, '\n'); | |
2729 | } | |
2730 | ||
2731 | return 0; | |
2732 | } | |
2733 | ||
2734 | static const struct seq_operations ptype_seq_ops = { | |
2735 | .start = ptype_seq_start, | |
2736 | .next = ptype_seq_next, | |
2737 | .stop = ptype_seq_stop, | |
2738 | .show = ptype_seq_show, | |
2739 | }; | |
2740 | ||
2741 | static int ptype_seq_open(struct inode *inode, struct file *file) | |
2742 | { | |
2feb27db PE |
2743 | return seq_open_net(inode, file, &ptype_seq_ops, |
2744 | sizeof(struct seq_net_private)); | |
0e1256ff SH |
2745 | } |
2746 | ||
2747 | static const struct file_operations ptype_seq_fops = { | |
2748 | .owner = THIS_MODULE, | |
2749 | .open = ptype_seq_open, | |
2750 | .read = seq_read, | |
2751 | .llseek = seq_lseek, | |
2feb27db | 2752 | .release = seq_release_net, |
0e1256ff SH |
2753 | }; |
2754 | ||
2755 | ||
4665079c | 2756 | static int __net_init dev_proc_net_init(struct net *net) |
1da177e4 LT |
2757 | { |
2758 | int rc = -ENOMEM; | |
2759 | ||
881d966b | 2760 | if (!proc_net_fops_create(net, "dev", S_IRUGO, &dev_seq_fops)) |
1da177e4 | 2761 | goto out; |
881d966b | 2762 | if (!proc_net_fops_create(net, "softnet_stat", S_IRUGO, &softnet_seq_fops)) |
1da177e4 | 2763 | goto out_dev; |
881d966b | 2764 | if (!proc_net_fops_create(net, "ptype", S_IRUGO, &ptype_seq_fops)) |
457c4cbc | 2765 | goto out_softnet; |
0e1256ff | 2766 | |
881d966b | 2767 | if (wext_proc_init(net)) |
457c4cbc | 2768 | goto out_ptype; |
1da177e4 LT |
2769 | rc = 0; |
2770 | out: | |
2771 | return rc; | |
457c4cbc | 2772 | out_ptype: |
881d966b | 2773 | proc_net_remove(net, "ptype"); |
1da177e4 | 2774 | out_softnet: |
881d966b | 2775 | proc_net_remove(net, "softnet_stat"); |
1da177e4 | 2776 | out_dev: |
881d966b | 2777 | proc_net_remove(net, "dev"); |
1da177e4 LT |
2778 | goto out; |
2779 | } | |
881d966b | 2780 | |
4665079c | 2781 | static void __net_exit dev_proc_net_exit(struct net *net) |
881d966b EB |
2782 | { |
2783 | wext_proc_exit(net); | |
2784 | ||
2785 | proc_net_remove(net, "ptype"); | |
2786 | proc_net_remove(net, "softnet_stat"); | |
2787 | proc_net_remove(net, "dev"); | |
2788 | } | |
2789 | ||
022cbae6 | 2790 | static struct pernet_operations __net_initdata dev_proc_ops = { |
881d966b EB |
2791 | .init = dev_proc_net_init, |
2792 | .exit = dev_proc_net_exit, | |
2793 | }; | |
2794 | ||
2795 | static int __init dev_proc_init(void) | |
2796 | { | |
2797 | return register_pernet_subsys(&dev_proc_ops); | |
2798 | } | |
1da177e4 LT |
2799 | #else |
2800 | #define dev_proc_init() 0 | |
2801 | #endif /* CONFIG_PROC_FS */ | |
2802 | ||
2803 | ||
2804 | /** | |
2805 | * netdev_set_master - set up master/slave pair | |
2806 | * @slave: slave device | |
2807 | * @master: new master device | |
2808 | * | |
2809 | * Changes the master device of the slave. Pass %NULL to break the | |
2810 | * bonding. The caller must hold the RTNL semaphore. On a failure | |
2811 | * a negative errno code is returned. On success the reference counts | |
2812 | * are adjusted, %RTM_NEWLINK is sent to the routing socket and the | |
2813 | * function returns zero. | |
2814 | */ | |
2815 | int netdev_set_master(struct net_device *slave, struct net_device *master) | |
2816 | { | |
2817 | struct net_device *old = slave->master; | |
2818 | ||
2819 | ASSERT_RTNL(); | |
2820 | ||
2821 | if (master) { | |
2822 | if (old) | |
2823 | return -EBUSY; | |
2824 | dev_hold(master); | |
2825 | } | |
2826 | ||
2827 | slave->master = master; | |
4ec93edb | 2828 | |
1da177e4 LT |
2829 | synchronize_net(); |
2830 | ||
2831 | if (old) | |
2832 | dev_put(old); | |
2833 | ||
2834 | if (master) | |
2835 | slave->flags |= IFF_SLAVE; | |
2836 | else | |
2837 | slave->flags &= ~IFF_SLAVE; | |
2838 | ||
2839 | rtmsg_ifinfo(RTM_NEWLINK, slave, IFF_SLAVE); | |
2840 | return 0; | |
2841 | } | |
2842 | ||
dad9b335 | 2843 | static int __dev_set_promiscuity(struct net_device *dev, int inc) |
1da177e4 LT |
2844 | { |
2845 | unsigned short old_flags = dev->flags; | |
2846 | ||
24023451 PM |
2847 | ASSERT_RTNL(); |
2848 | ||
dad9b335 WC |
2849 | dev->flags |= IFF_PROMISC; |
2850 | dev->promiscuity += inc; | |
2851 | if (dev->promiscuity == 0) { | |
2852 | /* | |
2853 | * Avoid overflow. | |
2854 | * If inc causes overflow, untouch promisc and return error. | |
2855 | */ | |
2856 | if (inc < 0) | |
2857 | dev->flags &= ~IFF_PROMISC; | |
2858 | else { | |
2859 | dev->promiscuity -= inc; | |
2860 | printk(KERN_WARNING "%s: promiscuity touches roof, " | |
2861 | "set promiscuity failed, promiscuity feature " | |
2862 | "of device might be broken.\n", dev->name); | |
2863 | return -EOVERFLOW; | |
2864 | } | |
2865 | } | |
52609c0b | 2866 | if (dev->flags != old_flags) { |
1da177e4 LT |
2867 | printk(KERN_INFO "device %s %s promiscuous mode\n", |
2868 | dev->name, (dev->flags & IFF_PROMISC) ? "entered" : | |
4ec93edb | 2869 | "left"); |
7759db82 KHK |
2870 | if (audit_enabled) |
2871 | audit_log(current->audit_context, GFP_ATOMIC, | |
2872 | AUDIT_ANOM_PROMISCUOUS, | |
2873 | "dev=%s prom=%d old_prom=%d auid=%u uid=%u gid=%u ses=%u", | |
2874 | dev->name, (dev->flags & IFF_PROMISC), | |
2875 | (old_flags & IFF_PROMISC), | |
2876 | audit_get_loginuid(current), | |
2877 | current->uid, current->gid, | |
2878 | audit_get_sessionid(current)); | |
24023451 PM |
2879 | |
2880 | if (dev->change_rx_flags) | |
2881 | dev->change_rx_flags(dev, IFF_PROMISC); | |
1da177e4 | 2882 | } |
dad9b335 | 2883 | return 0; |
1da177e4 LT |
2884 | } |
2885 | ||
4417da66 PM |
2886 | /** |
2887 | * dev_set_promiscuity - update promiscuity count on a device | |
2888 | * @dev: device | |
2889 | * @inc: modifier | |
2890 | * | |
2891 | * Add or remove promiscuity from a device. While the count in the device | |
2892 | * remains above zero the interface remains promiscuous. Once it hits zero | |
2893 | * the device reverts back to normal filtering operation. A negative inc | |
2894 | * value is used to drop promiscuity on the device. | |
dad9b335 | 2895 | * Return 0 if successful or a negative errno code on error. |
4417da66 | 2896 | */ |
dad9b335 | 2897 | int dev_set_promiscuity(struct net_device *dev, int inc) |
4417da66 PM |
2898 | { |
2899 | unsigned short old_flags = dev->flags; | |
dad9b335 | 2900 | int err; |
4417da66 | 2901 | |
dad9b335 | 2902 | err = __dev_set_promiscuity(dev, inc); |
4b5a698e | 2903 | if (err < 0) |
dad9b335 | 2904 | return err; |
4417da66 PM |
2905 | if (dev->flags != old_flags) |
2906 | dev_set_rx_mode(dev); | |
dad9b335 | 2907 | return err; |
4417da66 PM |
2908 | } |
2909 | ||
1da177e4 LT |
2910 | /** |
2911 | * dev_set_allmulti - update allmulti count on a device | |
2912 | * @dev: device | |
2913 | * @inc: modifier | |
2914 | * | |
2915 | * Add or remove reception of all multicast frames to a device. While the | |
2916 | * count in the device remains above zero the interface remains listening | |
2917 | * to all interfaces. Once it hits zero the device reverts back to normal | |
2918 | * filtering operation. A negative @inc value is used to drop the counter | |
2919 | * when releasing a resource needing all multicasts. | |
dad9b335 | 2920 | * Return 0 if successful or a negative errno code on error. |
1da177e4 LT |
2921 | */ |
2922 | ||
dad9b335 | 2923 | int dev_set_allmulti(struct net_device *dev, int inc) |
1da177e4 LT |
2924 | { |
2925 | unsigned short old_flags = dev->flags; | |
2926 | ||
24023451 PM |
2927 | ASSERT_RTNL(); |
2928 | ||
1da177e4 | 2929 | dev->flags |= IFF_ALLMULTI; |
dad9b335 WC |
2930 | dev->allmulti += inc; |
2931 | if (dev->allmulti == 0) { | |
2932 | /* | |
2933 | * Avoid overflow. | |
2934 | * If inc causes overflow, untouch allmulti and return error. | |
2935 | */ | |
2936 | if (inc < 0) | |
2937 | dev->flags &= ~IFF_ALLMULTI; | |
2938 | else { | |
2939 | dev->allmulti -= inc; | |
2940 | printk(KERN_WARNING "%s: allmulti touches roof, " | |
2941 | "set allmulti failed, allmulti feature of " | |
2942 | "device might be broken.\n", dev->name); | |
2943 | return -EOVERFLOW; | |
2944 | } | |
2945 | } | |
24023451 PM |
2946 | if (dev->flags ^ old_flags) { |
2947 | if (dev->change_rx_flags) | |
2948 | dev->change_rx_flags(dev, IFF_ALLMULTI); | |
4417da66 | 2949 | dev_set_rx_mode(dev); |
24023451 | 2950 | } |
dad9b335 | 2951 | return 0; |
4417da66 PM |
2952 | } |
2953 | ||
2954 | /* | |
2955 | * Upload unicast and multicast address lists to device and | |
2956 | * configure RX filtering. When the device doesn't support unicast | |
53ccaae1 | 2957 | * filtering it is put in promiscuous mode while unicast addresses |
4417da66 PM |
2958 | * are present. |
2959 | */ | |
2960 | void __dev_set_rx_mode(struct net_device *dev) | |
2961 | { | |
2962 | /* dev_open will call this function so the list will stay sane. */ | |
2963 | if (!(dev->flags&IFF_UP)) | |
2964 | return; | |
2965 | ||
2966 | if (!netif_device_present(dev)) | |
40b77c94 | 2967 | return; |
4417da66 PM |
2968 | |
2969 | if (dev->set_rx_mode) | |
2970 | dev->set_rx_mode(dev); | |
2971 | else { | |
2972 | /* Unicast addresses changes may only happen under the rtnl, | |
2973 | * therefore calling __dev_set_promiscuity here is safe. | |
2974 | */ | |
2975 | if (dev->uc_count > 0 && !dev->uc_promisc) { | |
2976 | __dev_set_promiscuity(dev, 1); | |
2977 | dev->uc_promisc = 1; | |
2978 | } else if (dev->uc_count == 0 && dev->uc_promisc) { | |
2979 | __dev_set_promiscuity(dev, -1); | |
2980 | dev->uc_promisc = 0; | |
2981 | } | |
2982 | ||
2983 | if (dev->set_multicast_list) | |
2984 | dev->set_multicast_list(dev); | |
2985 | } | |
2986 | } | |
2987 | ||
2988 | void dev_set_rx_mode(struct net_device *dev) | |
2989 | { | |
b9e40857 | 2990 | netif_addr_lock_bh(dev); |
4417da66 | 2991 | __dev_set_rx_mode(dev); |
b9e40857 | 2992 | netif_addr_unlock_bh(dev); |
1da177e4 LT |
2993 | } |
2994 | ||
61cbc2fc PM |
2995 | int __dev_addr_delete(struct dev_addr_list **list, int *count, |
2996 | void *addr, int alen, int glbl) | |
bf742482 PM |
2997 | { |
2998 | struct dev_addr_list *da; | |
2999 | ||
3000 | for (; (da = *list) != NULL; list = &da->next) { | |
3001 | if (memcmp(da->da_addr, addr, da->da_addrlen) == 0 && | |
3002 | alen == da->da_addrlen) { | |
3003 | if (glbl) { | |
3004 | int old_glbl = da->da_gusers; | |
3005 | da->da_gusers = 0; | |
3006 | if (old_glbl == 0) | |
3007 | break; | |
3008 | } | |
3009 | if (--da->da_users) | |
3010 | return 0; | |
3011 | ||
3012 | *list = da->next; | |
3013 | kfree(da); | |
61cbc2fc | 3014 | (*count)--; |
bf742482 PM |
3015 | return 0; |
3016 | } | |
3017 | } | |
3018 | return -ENOENT; | |
3019 | } | |
3020 | ||
61cbc2fc PM |
3021 | int __dev_addr_add(struct dev_addr_list **list, int *count, |
3022 | void *addr, int alen, int glbl) | |
bf742482 PM |
3023 | { |
3024 | struct dev_addr_list *da; | |
3025 | ||
3026 | for (da = *list; da != NULL; da = da->next) { | |
3027 | if (memcmp(da->da_addr, addr, da->da_addrlen) == 0 && | |
3028 | da->da_addrlen == alen) { | |
3029 | if (glbl) { | |
3030 | int old_glbl = da->da_gusers; | |
3031 | da->da_gusers = 1; | |
3032 | if (old_glbl) | |
3033 | return 0; | |
3034 | } | |
3035 | da->da_users++; | |
3036 | return 0; | |
3037 | } | |
3038 | } | |
3039 | ||
12aa343a | 3040 | da = kzalloc(sizeof(*da), GFP_ATOMIC); |
bf742482 PM |
3041 | if (da == NULL) |
3042 | return -ENOMEM; | |
3043 | memcpy(da->da_addr, addr, alen); | |
3044 | da->da_addrlen = alen; | |
3045 | da->da_users = 1; | |
3046 | da->da_gusers = glbl ? 1 : 0; | |
3047 | da->next = *list; | |
3048 | *list = da; | |
61cbc2fc | 3049 | (*count)++; |
bf742482 PM |
3050 | return 0; |
3051 | } | |
3052 | ||
4417da66 PM |
3053 | /** |
3054 | * dev_unicast_delete - Release secondary unicast address. | |
3055 | * @dev: device | |
0ed72ec4 RD |
3056 | * @addr: address to delete |
3057 | * @alen: length of @addr | |
4417da66 PM |
3058 | * |
3059 | * Release reference to a secondary unicast address and remove it | |
0ed72ec4 | 3060 | * from the device if the reference count drops to zero. |
4417da66 PM |
3061 | * |
3062 | * The caller must hold the rtnl_mutex. | |
3063 | */ | |
3064 | int dev_unicast_delete(struct net_device *dev, void *addr, int alen) | |
3065 | { | |
3066 | int err; | |
3067 | ||
3068 | ASSERT_RTNL(); | |
3069 | ||
b9e40857 | 3070 | netif_addr_lock_bh(dev); |
61cbc2fc PM |
3071 | err = __dev_addr_delete(&dev->uc_list, &dev->uc_count, addr, alen, 0); |
3072 | if (!err) | |
4417da66 | 3073 | __dev_set_rx_mode(dev); |
b9e40857 | 3074 | netif_addr_unlock_bh(dev); |
4417da66 PM |
3075 | return err; |
3076 | } | |
3077 | EXPORT_SYMBOL(dev_unicast_delete); | |
3078 | ||
3079 | /** | |
3080 | * dev_unicast_add - add a secondary unicast address | |
3081 | * @dev: device | |
5dbaec5d | 3082 | * @addr: address to add |
0ed72ec4 | 3083 | * @alen: length of @addr |
4417da66 PM |
3084 | * |
3085 | * Add a secondary unicast address to the device or increase | |
3086 | * the reference count if it already exists. | |
3087 | * | |
3088 | * The caller must hold the rtnl_mutex. | |
3089 | */ | |
3090 | int dev_unicast_add(struct net_device *dev, void *addr, int alen) | |
3091 | { | |
3092 | int err; | |
3093 | ||
3094 | ASSERT_RTNL(); | |
3095 | ||
b9e40857 | 3096 | netif_addr_lock_bh(dev); |
61cbc2fc PM |
3097 | err = __dev_addr_add(&dev->uc_list, &dev->uc_count, addr, alen, 0); |
3098 | if (!err) | |
4417da66 | 3099 | __dev_set_rx_mode(dev); |
b9e40857 | 3100 | netif_addr_unlock_bh(dev); |
4417da66 PM |
3101 | return err; |
3102 | } | |
3103 | EXPORT_SYMBOL(dev_unicast_add); | |
3104 | ||
e83a2ea8 CL |
3105 | int __dev_addr_sync(struct dev_addr_list **to, int *to_count, |
3106 | struct dev_addr_list **from, int *from_count) | |
3107 | { | |
3108 | struct dev_addr_list *da, *next; | |
3109 | int err = 0; | |
3110 | ||
3111 | da = *from; | |
3112 | while (da != NULL) { | |
3113 | next = da->next; | |
3114 | if (!da->da_synced) { | |
3115 | err = __dev_addr_add(to, to_count, | |
3116 | da->da_addr, da->da_addrlen, 0); | |
3117 | if (err < 0) | |
3118 | break; | |
3119 | da->da_synced = 1; | |
3120 | da->da_users++; | |
3121 | } else if (da->da_users == 1) { | |
3122 | __dev_addr_delete(to, to_count, | |
3123 | da->da_addr, da->da_addrlen, 0); | |
3124 | __dev_addr_delete(from, from_count, | |
3125 | da->da_addr, da->da_addrlen, 0); | |
3126 | } | |
3127 | da = next; | |
3128 | } | |
3129 | return err; | |
3130 | } | |
3131 | ||
3132 | void __dev_addr_unsync(struct dev_addr_list **to, int *to_count, | |
3133 | struct dev_addr_list **from, int *from_count) | |
3134 | { | |
3135 | struct dev_addr_list *da, *next; | |
3136 | ||
3137 | da = *from; | |
3138 | while (da != NULL) { | |
3139 | next = da->next; | |
3140 | if (da->da_synced) { | |
3141 | __dev_addr_delete(to, to_count, | |
3142 | da->da_addr, da->da_addrlen, 0); | |
3143 | da->da_synced = 0; | |
3144 | __dev_addr_delete(from, from_count, | |
3145 | da->da_addr, da->da_addrlen, 0); | |
3146 | } | |
3147 | da = next; | |
3148 | } | |
3149 | } | |
3150 | ||
3151 | /** | |
3152 | * dev_unicast_sync - Synchronize device's unicast list to another device | |
3153 | * @to: destination device | |
3154 | * @from: source device | |
3155 | * | |
3156 | * Add newly added addresses to the destination device and release | |
3157 | * addresses that have no users left. The source device must be | |
3158 | * locked by netif_tx_lock_bh. | |
3159 | * | |
3160 | * This function is intended to be called from the dev->set_rx_mode | |
3161 | * function of layered software devices. | |
3162 | */ | |
3163 | int dev_unicast_sync(struct net_device *to, struct net_device *from) | |
3164 | { | |
3165 | int err = 0; | |
3166 | ||
b9e40857 | 3167 | netif_addr_lock_bh(to); |
e83a2ea8 CL |
3168 | err = __dev_addr_sync(&to->uc_list, &to->uc_count, |
3169 | &from->uc_list, &from->uc_count); | |
3170 | if (!err) | |
3171 | __dev_set_rx_mode(to); | |
b9e40857 | 3172 | netif_addr_unlock_bh(to); |
e83a2ea8 CL |
3173 | return err; |
3174 | } | |
3175 | EXPORT_SYMBOL(dev_unicast_sync); | |
3176 | ||
3177 | /** | |
bc2cda1e | 3178 | * dev_unicast_unsync - Remove synchronized addresses from the destination device |
e83a2ea8 CL |
3179 | * @to: destination device |
3180 | * @from: source device | |
3181 | * | |
3182 | * Remove all addresses that were added to the destination device by | |
3183 | * dev_unicast_sync(). This function is intended to be called from the | |
3184 | * dev->stop function of layered software devices. | |
3185 | */ | |
3186 | void dev_unicast_unsync(struct net_device *to, struct net_device *from) | |
3187 | { | |
b9e40857 | 3188 | netif_addr_lock_bh(from); |
e308a5d8 | 3189 | netif_addr_lock(to); |
e83a2ea8 CL |
3190 | |
3191 | __dev_addr_unsync(&to->uc_list, &to->uc_count, | |
3192 | &from->uc_list, &from->uc_count); | |
3193 | __dev_set_rx_mode(to); | |
3194 | ||
e308a5d8 | 3195 | netif_addr_unlock(to); |
b9e40857 | 3196 | netif_addr_unlock_bh(from); |
e83a2ea8 CL |
3197 | } |
3198 | EXPORT_SYMBOL(dev_unicast_unsync); | |
3199 | ||
12972621 DC |
3200 | static void __dev_addr_discard(struct dev_addr_list **list) |
3201 | { | |
3202 | struct dev_addr_list *tmp; | |
3203 | ||
3204 | while (*list != NULL) { | |
3205 | tmp = *list; | |
3206 | *list = tmp->next; | |
3207 | if (tmp->da_users > tmp->da_gusers) | |
3208 | printk("__dev_addr_discard: address leakage! " | |
3209 | "da_users=%d\n", tmp->da_users); | |
3210 | kfree(tmp); | |
3211 | } | |
3212 | } | |
3213 | ||
26cc2522 | 3214 | static void dev_addr_discard(struct net_device *dev) |
4417da66 | 3215 | { |
b9e40857 | 3216 | netif_addr_lock_bh(dev); |
26cc2522 | 3217 | |
4417da66 PM |
3218 | __dev_addr_discard(&dev->uc_list); |
3219 | dev->uc_count = 0; | |
4417da66 | 3220 | |
456ad75c DC |
3221 | __dev_addr_discard(&dev->mc_list); |
3222 | dev->mc_count = 0; | |
26cc2522 | 3223 | |
b9e40857 | 3224 | netif_addr_unlock_bh(dev); |
456ad75c DC |
3225 | } |
3226 | ||
1da177e4 LT |
3227 | unsigned dev_get_flags(const struct net_device *dev) |
3228 | { | |
3229 | unsigned flags; | |
3230 | ||
3231 | flags = (dev->flags & ~(IFF_PROMISC | | |
3232 | IFF_ALLMULTI | | |
b00055aa SR |
3233 | IFF_RUNNING | |
3234 | IFF_LOWER_UP | | |
3235 | IFF_DORMANT)) | | |
1da177e4 LT |
3236 | (dev->gflags & (IFF_PROMISC | |
3237 | IFF_ALLMULTI)); | |
3238 | ||
b00055aa SR |
3239 | if (netif_running(dev)) { |
3240 | if (netif_oper_up(dev)) | |
3241 | flags |= IFF_RUNNING; | |
3242 | if (netif_carrier_ok(dev)) | |
3243 | flags |= IFF_LOWER_UP; | |
3244 | if (netif_dormant(dev)) | |
3245 | flags |= IFF_DORMANT; | |
3246 | } | |
1da177e4 LT |
3247 | |
3248 | return flags; | |
3249 | } | |
3250 | ||
3251 | int dev_change_flags(struct net_device *dev, unsigned flags) | |
3252 | { | |
7c355f53 | 3253 | int ret, changes; |
1da177e4 LT |
3254 | int old_flags = dev->flags; |
3255 | ||
24023451 PM |
3256 | ASSERT_RTNL(); |
3257 | ||
1da177e4 LT |
3258 | /* |
3259 | * Set the flags on our device. | |
3260 | */ | |
3261 | ||
3262 | dev->flags = (flags & (IFF_DEBUG | IFF_NOTRAILERS | IFF_NOARP | | |
3263 | IFF_DYNAMIC | IFF_MULTICAST | IFF_PORTSEL | | |
3264 | IFF_AUTOMEDIA)) | | |
3265 | (dev->flags & (IFF_UP | IFF_VOLATILE | IFF_PROMISC | | |
3266 | IFF_ALLMULTI)); | |
3267 | ||
3268 | /* | |
3269 | * Load in the correct multicast list now the flags have changed. | |
3270 | */ | |
3271 | ||
0e91796e | 3272 | if (dev->change_rx_flags && (old_flags ^ flags) & IFF_MULTICAST) |
24023451 PM |
3273 | dev->change_rx_flags(dev, IFF_MULTICAST); |
3274 | ||
4417da66 | 3275 | dev_set_rx_mode(dev); |
1da177e4 LT |
3276 | |
3277 | /* | |
3278 | * Have we downed the interface. We handle IFF_UP ourselves | |
3279 | * according to user attempts to set it, rather than blindly | |
3280 | * setting it. | |
3281 | */ | |
3282 | ||
3283 | ret = 0; | |
3284 | if ((old_flags ^ flags) & IFF_UP) { /* Bit is different ? */ | |
3285 | ret = ((old_flags & IFF_UP) ? dev_close : dev_open)(dev); | |
3286 | ||
3287 | if (!ret) | |
4417da66 | 3288 | dev_set_rx_mode(dev); |
1da177e4 LT |
3289 | } |
3290 | ||
3291 | if (dev->flags & IFF_UP && | |
3292 | ((old_flags ^ dev->flags) &~ (IFF_UP | IFF_PROMISC | IFF_ALLMULTI | | |
3293 | IFF_VOLATILE))) | |
056925ab | 3294 | call_netdevice_notifiers(NETDEV_CHANGE, dev); |
1da177e4 LT |
3295 | |
3296 | if ((flags ^ dev->gflags) & IFF_PROMISC) { | |
3297 | int inc = (flags & IFF_PROMISC) ? +1 : -1; | |
3298 | dev->gflags ^= IFF_PROMISC; | |
3299 | dev_set_promiscuity(dev, inc); | |
3300 | } | |
3301 | ||
3302 | /* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI | |
3303 | is important. Some (broken) drivers set IFF_PROMISC, when | |
3304 | IFF_ALLMULTI is requested not asking us and not reporting. | |
3305 | */ | |
3306 | if ((flags ^ dev->gflags) & IFF_ALLMULTI) { | |
3307 | int inc = (flags & IFF_ALLMULTI) ? +1 : -1; | |
3308 | dev->gflags ^= IFF_ALLMULTI; | |
3309 | dev_set_allmulti(dev, inc); | |
3310 | } | |
3311 | ||
7c355f53 TG |
3312 | /* Exclude state transition flags, already notified */ |
3313 | changes = (old_flags ^ dev->flags) & ~(IFF_UP | IFF_RUNNING); | |
3314 | if (changes) | |
3315 | rtmsg_ifinfo(RTM_NEWLINK, dev, changes); | |
1da177e4 LT |
3316 | |
3317 | return ret; | |
3318 | } | |
3319 | ||
3320 | int dev_set_mtu(struct net_device *dev, int new_mtu) | |
3321 | { | |
3322 | int err; | |
3323 | ||
3324 | if (new_mtu == dev->mtu) | |
3325 | return 0; | |
3326 | ||
3327 | /* MTU must be positive. */ | |
3328 | if (new_mtu < 0) | |
3329 | return -EINVAL; | |
3330 | ||
3331 | if (!netif_device_present(dev)) | |
3332 | return -ENODEV; | |
3333 | ||
3334 | err = 0; | |
3335 | if (dev->change_mtu) | |
3336 | err = dev->change_mtu(dev, new_mtu); | |
3337 | else | |
3338 | dev->mtu = new_mtu; | |
3339 | if (!err && dev->flags & IFF_UP) | |
056925ab | 3340 | call_netdevice_notifiers(NETDEV_CHANGEMTU, dev); |
1da177e4 LT |
3341 | return err; |
3342 | } | |
3343 | ||
3344 | int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa) | |
3345 | { | |
3346 | int err; | |
3347 | ||
3348 | if (!dev->set_mac_address) | |
3349 | return -EOPNOTSUPP; | |
3350 | if (sa->sa_family != dev->type) | |
3351 | return -EINVAL; | |
3352 | if (!netif_device_present(dev)) | |
3353 | return -ENODEV; | |
3354 | err = dev->set_mac_address(dev, sa); | |
3355 | if (!err) | |
056925ab | 3356 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); |
1da177e4 LT |
3357 | return err; |
3358 | } | |
3359 | ||
3360 | /* | |
14e3e079 | 3361 | * Perform the SIOCxIFxxx calls, inside read_lock(dev_base_lock) |
1da177e4 | 3362 | */ |
14e3e079 | 3363 | static int dev_ifsioc_locked(struct net *net, struct ifreq *ifr, unsigned int cmd) |
1da177e4 LT |
3364 | { |
3365 | int err; | |
881d966b | 3366 | struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name); |
1da177e4 LT |
3367 | |
3368 | if (!dev) | |
3369 | return -ENODEV; | |
3370 | ||
3371 | switch (cmd) { | |
3372 | case SIOCGIFFLAGS: /* Get interface flags */ | |
3373 | ifr->ifr_flags = dev_get_flags(dev); | |
3374 | return 0; | |
3375 | ||
1da177e4 LT |
3376 | case SIOCGIFMETRIC: /* Get the metric on the interface |
3377 | (currently unused) */ | |
3378 | ifr->ifr_metric = 0; | |
3379 | return 0; | |
3380 | ||
1da177e4 LT |
3381 | case SIOCGIFMTU: /* Get the MTU of a device */ |
3382 | ifr->ifr_mtu = dev->mtu; | |
3383 | return 0; | |
3384 | ||
1da177e4 LT |
3385 | case SIOCGIFHWADDR: |
3386 | if (!dev->addr_len) | |
3387 | memset(ifr->ifr_hwaddr.sa_data, 0, sizeof ifr->ifr_hwaddr.sa_data); | |
3388 | else | |
3389 | memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr, | |
3390 | min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); | |
3391 | ifr->ifr_hwaddr.sa_family = dev->type; | |
3392 | return 0; | |
3393 | ||
14e3e079 JG |
3394 | case SIOCGIFSLAVE: |
3395 | err = -EINVAL; | |
3396 | break; | |
3397 | ||
3398 | case SIOCGIFMAP: | |
3399 | ifr->ifr_map.mem_start = dev->mem_start; | |
3400 | ifr->ifr_map.mem_end = dev->mem_end; | |
3401 | ifr->ifr_map.base_addr = dev->base_addr; | |
3402 | ifr->ifr_map.irq = dev->irq; | |
3403 | ifr->ifr_map.dma = dev->dma; | |
3404 | ifr->ifr_map.port = dev->if_port; | |
3405 | return 0; | |
3406 | ||
3407 | case SIOCGIFINDEX: | |
3408 | ifr->ifr_ifindex = dev->ifindex; | |
3409 | return 0; | |
3410 | ||
3411 | case SIOCGIFTXQLEN: | |
3412 | ifr->ifr_qlen = dev->tx_queue_len; | |
3413 | return 0; | |
3414 | ||
3415 | default: | |
3416 | /* dev_ioctl() should ensure this case | |
3417 | * is never reached | |
3418 | */ | |
3419 | WARN_ON(1); | |
3420 | err = -EINVAL; | |
3421 | break; | |
3422 | ||
3423 | } | |
3424 | return err; | |
3425 | } | |
3426 | ||
3427 | /* | |
3428 | * Perform the SIOCxIFxxx calls, inside rtnl_lock() | |
3429 | */ | |
3430 | static int dev_ifsioc(struct net *net, struct ifreq *ifr, unsigned int cmd) | |
3431 | { | |
3432 | int err; | |
3433 | struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name); | |
3434 | ||
3435 | if (!dev) | |
3436 | return -ENODEV; | |
3437 | ||
3438 | switch (cmd) { | |
3439 | case SIOCSIFFLAGS: /* Set interface flags */ | |
3440 | return dev_change_flags(dev, ifr->ifr_flags); | |
3441 | ||
3442 | case SIOCSIFMETRIC: /* Set the metric on the interface | |
3443 | (currently unused) */ | |
3444 | return -EOPNOTSUPP; | |
3445 | ||
3446 | case SIOCSIFMTU: /* Set the MTU of a device */ | |
3447 | return dev_set_mtu(dev, ifr->ifr_mtu); | |
3448 | ||
1da177e4 LT |
3449 | case SIOCSIFHWADDR: |
3450 | return dev_set_mac_address(dev, &ifr->ifr_hwaddr); | |
3451 | ||
3452 | case SIOCSIFHWBROADCAST: | |
3453 | if (ifr->ifr_hwaddr.sa_family != dev->type) | |
3454 | return -EINVAL; | |
3455 | memcpy(dev->broadcast, ifr->ifr_hwaddr.sa_data, | |
3456 | min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); | |
056925ab | 3457 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); |
1da177e4 LT |
3458 | return 0; |
3459 | ||
1da177e4 LT |
3460 | case SIOCSIFMAP: |
3461 | if (dev->set_config) { | |
3462 | if (!netif_device_present(dev)) | |
3463 | return -ENODEV; | |
3464 | return dev->set_config(dev, &ifr->ifr_map); | |
3465 | } | |
3466 | return -EOPNOTSUPP; | |
3467 | ||
3468 | case SIOCADDMULTI: | |
61ee6bd4 | 3469 | if ((!dev->set_multicast_list && !dev->set_rx_mode) || |
1da177e4 LT |
3470 | ifr->ifr_hwaddr.sa_family != AF_UNSPEC) |
3471 | return -EINVAL; | |
3472 | if (!netif_device_present(dev)) | |
3473 | return -ENODEV; | |
3474 | return dev_mc_add(dev, ifr->ifr_hwaddr.sa_data, | |
3475 | dev->addr_len, 1); | |
3476 | ||
3477 | case SIOCDELMULTI: | |
61ee6bd4 | 3478 | if ((!dev->set_multicast_list && !dev->set_rx_mode) || |
1da177e4 LT |
3479 | ifr->ifr_hwaddr.sa_family != AF_UNSPEC) |
3480 | return -EINVAL; | |
3481 | if (!netif_device_present(dev)) | |
3482 | return -ENODEV; | |
3483 | return dev_mc_delete(dev, ifr->ifr_hwaddr.sa_data, | |
3484 | dev->addr_len, 1); | |
3485 | ||
1da177e4 LT |
3486 | case SIOCSIFTXQLEN: |
3487 | if (ifr->ifr_qlen < 0) | |
3488 | return -EINVAL; | |
3489 | dev->tx_queue_len = ifr->ifr_qlen; | |
3490 | return 0; | |
3491 | ||
3492 | case SIOCSIFNAME: | |
3493 | ifr->ifr_newname[IFNAMSIZ-1] = '\0'; | |
3494 | return dev_change_name(dev, ifr->ifr_newname); | |
3495 | ||
3496 | /* | |
3497 | * Unknown or private ioctl | |
3498 | */ | |
3499 | ||
3500 | default: | |
3501 | if ((cmd >= SIOCDEVPRIVATE && | |
3502 | cmd <= SIOCDEVPRIVATE + 15) || | |
3503 | cmd == SIOCBONDENSLAVE || | |
3504 | cmd == SIOCBONDRELEASE || | |
3505 | cmd == SIOCBONDSETHWADDR || | |
3506 | cmd == SIOCBONDSLAVEINFOQUERY || | |
3507 | cmd == SIOCBONDINFOQUERY || | |
3508 | cmd == SIOCBONDCHANGEACTIVE || | |
3509 | cmd == SIOCGMIIPHY || | |
3510 | cmd == SIOCGMIIREG || | |
3511 | cmd == SIOCSMIIREG || | |
3512 | cmd == SIOCBRADDIF || | |
3513 | cmd == SIOCBRDELIF || | |
3514 | cmd == SIOCWANDEV) { | |
3515 | err = -EOPNOTSUPP; | |
3516 | if (dev->do_ioctl) { | |
3517 | if (netif_device_present(dev)) | |
3518 | err = dev->do_ioctl(dev, ifr, | |
3519 | cmd); | |
3520 | else | |
3521 | err = -ENODEV; | |
3522 | } | |
3523 | } else | |
3524 | err = -EINVAL; | |
3525 | ||
3526 | } | |
3527 | return err; | |
3528 | } | |
3529 | ||
3530 | /* | |
3531 | * This function handles all "interface"-type I/O control requests. The actual | |
3532 | * 'doing' part of this is dev_ifsioc above. | |
3533 | */ | |
3534 | ||
3535 | /** | |
3536 | * dev_ioctl - network device ioctl | |
c4ea43c5 | 3537 | * @net: the applicable net namespace |
1da177e4 LT |
3538 | * @cmd: command to issue |
3539 | * @arg: pointer to a struct ifreq in user space | |
3540 | * | |
3541 | * Issue ioctl functions to devices. This is normally called by the | |
3542 | * user space syscall interfaces but can sometimes be useful for | |
3543 | * other purposes. The return value is the return from the syscall if | |
3544 | * positive or a negative errno code on error. | |
3545 | */ | |
3546 | ||
881d966b | 3547 | int dev_ioctl(struct net *net, unsigned int cmd, void __user *arg) |
1da177e4 LT |
3548 | { |
3549 | struct ifreq ifr; | |
3550 | int ret; | |
3551 | char *colon; | |
3552 | ||
3553 | /* One special case: SIOCGIFCONF takes ifconf argument | |
3554 | and requires shared lock, because it sleeps writing | |
3555 | to user space. | |
3556 | */ | |
3557 | ||
3558 | if (cmd == SIOCGIFCONF) { | |
6756ae4b | 3559 | rtnl_lock(); |
881d966b | 3560 | ret = dev_ifconf(net, (char __user *) arg); |
6756ae4b | 3561 | rtnl_unlock(); |
1da177e4 LT |
3562 | return ret; |
3563 | } | |
3564 | if (cmd == SIOCGIFNAME) | |
881d966b | 3565 | return dev_ifname(net, (struct ifreq __user *)arg); |
1da177e4 LT |
3566 | |
3567 | if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) | |
3568 | return -EFAULT; | |
3569 | ||
3570 | ifr.ifr_name[IFNAMSIZ-1] = 0; | |
3571 | ||
3572 | colon = strchr(ifr.ifr_name, ':'); | |
3573 | if (colon) | |
3574 | *colon = 0; | |
3575 | ||
3576 | /* | |
3577 | * See which interface the caller is talking about. | |
3578 | */ | |
3579 | ||
3580 | switch (cmd) { | |
3581 | /* | |
3582 | * These ioctl calls: | |
3583 | * - can be done by all. | |
3584 | * - atomic and do not require locking. | |
3585 | * - return a value | |
3586 | */ | |
3587 | case SIOCGIFFLAGS: | |
3588 | case SIOCGIFMETRIC: | |
3589 | case SIOCGIFMTU: | |
3590 | case SIOCGIFHWADDR: | |
3591 | case SIOCGIFSLAVE: | |
3592 | case SIOCGIFMAP: | |
3593 | case SIOCGIFINDEX: | |
3594 | case SIOCGIFTXQLEN: | |
881d966b | 3595 | dev_load(net, ifr.ifr_name); |
1da177e4 | 3596 | read_lock(&dev_base_lock); |
14e3e079 | 3597 | ret = dev_ifsioc_locked(net, &ifr, cmd); |
1da177e4 LT |
3598 | read_unlock(&dev_base_lock); |
3599 | if (!ret) { | |
3600 | if (colon) | |
3601 | *colon = ':'; | |
3602 | if (copy_to_user(arg, &ifr, | |
3603 | sizeof(struct ifreq))) | |
3604 | ret = -EFAULT; | |
3605 | } | |
3606 | return ret; | |
3607 | ||
3608 | case SIOCETHTOOL: | |
881d966b | 3609 | dev_load(net, ifr.ifr_name); |
1da177e4 | 3610 | rtnl_lock(); |
881d966b | 3611 | ret = dev_ethtool(net, &ifr); |
1da177e4 LT |
3612 | rtnl_unlock(); |
3613 | if (!ret) { | |
3614 | if (colon) | |
3615 | *colon = ':'; | |
3616 | if (copy_to_user(arg, &ifr, | |
3617 | sizeof(struct ifreq))) | |
3618 | ret = -EFAULT; | |
3619 | } | |
3620 | return ret; | |
3621 | ||
3622 | /* | |
3623 | * These ioctl calls: | |
3624 | * - require superuser power. | |
3625 | * - require strict serialization. | |
3626 | * - return a value | |
3627 | */ | |
3628 | case SIOCGMIIPHY: | |
3629 | case SIOCGMIIREG: | |
3630 | case SIOCSIFNAME: | |
3631 | if (!capable(CAP_NET_ADMIN)) | |
3632 | return -EPERM; | |
881d966b | 3633 | dev_load(net, ifr.ifr_name); |
1da177e4 | 3634 | rtnl_lock(); |
881d966b | 3635 | ret = dev_ifsioc(net, &ifr, cmd); |
1da177e4 LT |
3636 | rtnl_unlock(); |
3637 | if (!ret) { | |
3638 | if (colon) | |
3639 | *colon = ':'; | |
3640 | if (copy_to_user(arg, &ifr, | |
3641 | sizeof(struct ifreq))) | |
3642 | ret = -EFAULT; | |
3643 | } | |
3644 | return ret; | |
3645 | ||
3646 | /* | |
3647 | * These ioctl calls: | |
3648 | * - require superuser power. | |
3649 | * - require strict serialization. | |
3650 | * - do not return a value | |
3651 | */ | |
3652 | case SIOCSIFFLAGS: | |
3653 | case SIOCSIFMETRIC: | |
3654 | case SIOCSIFMTU: | |
3655 | case SIOCSIFMAP: | |
3656 | case SIOCSIFHWADDR: | |
3657 | case SIOCSIFSLAVE: | |
3658 | case SIOCADDMULTI: | |
3659 | case SIOCDELMULTI: | |
3660 | case SIOCSIFHWBROADCAST: | |
3661 | case SIOCSIFTXQLEN: | |
3662 | case SIOCSMIIREG: | |
3663 | case SIOCBONDENSLAVE: | |
3664 | case SIOCBONDRELEASE: | |
3665 | case SIOCBONDSETHWADDR: | |
1da177e4 LT |
3666 | case SIOCBONDCHANGEACTIVE: |
3667 | case SIOCBRADDIF: | |
3668 | case SIOCBRDELIF: | |
3669 | if (!capable(CAP_NET_ADMIN)) | |
3670 | return -EPERM; | |
cabcac0b TG |
3671 | /* fall through */ |
3672 | case SIOCBONDSLAVEINFOQUERY: | |
3673 | case SIOCBONDINFOQUERY: | |
881d966b | 3674 | dev_load(net, ifr.ifr_name); |
1da177e4 | 3675 | rtnl_lock(); |
881d966b | 3676 | ret = dev_ifsioc(net, &ifr, cmd); |
1da177e4 LT |
3677 | rtnl_unlock(); |
3678 | return ret; | |
3679 | ||
3680 | case SIOCGIFMEM: | |
3681 | /* Get the per device memory space. We can add this but | |
3682 | * currently do not support it */ | |
3683 | case SIOCSIFMEM: | |
3684 | /* Set the per device memory buffer space. | |
3685 | * Not applicable in our case */ | |
3686 | case SIOCSIFLINK: | |
3687 | return -EINVAL; | |
3688 | ||
3689 | /* | |
3690 | * Unknown or private ioctl. | |
3691 | */ | |
3692 | default: | |
3693 | if (cmd == SIOCWANDEV || | |
3694 | (cmd >= SIOCDEVPRIVATE && | |
3695 | cmd <= SIOCDEVPRIVATE + 15)) { | |
881d966b | 3696 | dev_load(net, ifr.ifr_name); |
1da177e4 | 3697 | rtnl_lock(); |
881d966b | 3698 | ret = dev_ifsioc(net, &ifr, cmd); |
1da177e4 LT |
3699 | rtnl_unlock(); |
3700 | if (!ret && copy_to_user(arg, &ifr, | |
3701 | sizeof(struct ifreq))) | |
3702 | ret = -EFAULT; | |
3703 | return ret; | |
3704 | } | |
1da177e4 | 3705 | /* Take care of Wireless Extensions */ |
295f4a1f | 3706 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) |
881d966b | 3707 | return wext_handle_ioctl(net, &ifr, cmd, arg); |
1da177e4 LT |
3708 | return -EINVAL; |
3709 | } | |
3710 | } | |
3711 | ||
3712 | ||
3713 | /** | |
3714 | * dev_new_index - allocate an ifindex | |
c4ea43c5 | 3715 | * @net: the applicable net namespace |
1da177e4 LT |
3716 | * |
3717 | * Returns a suitable unique value for a new device interface | |
3718 | * number. The caller must hold the rtnl semaphore or the | |
3719 | * dev_base_lock to be sure it remains unique. | |
3720 | */ | |
881d966b | 3721 | static int dev_new_index(struct net *net) |
1da177e4 LT |
3722 | { |
3723 | static int ifindex; | |
3724 | for (;;) { | |
3725 | if (++ifindex <= 0) | |
3726 | ifindex = 1; | |
881d966b | 3727 | if (!__dev_get_by_index(net, ifindex)) |
1da177e4 LT |
3728 | return ifindex; |
3729 | } | |
3730 | } | |
3731 | ||
1da177e4 LT |
3732 | /* Delayed registration/unregisteration */ |
3733 | static DEFINE_SPINLOCK(net_todo_list_lock); | |
3b5b34fd | 3734 | static LIST_HEAD(net_todo_list); |
1da177e4 | 3735 | |
6f05f629 | 3736 | static void net_set_todo(struct net_device *dev) |
1da177e4 LT |
3737 | { |
3738 | spin_lock(&net_todo_list_lock); | |
3739 | list_add_tail(&dev->todo_list, &net_todo_list); | |
3740 | spin_unlock(&net_todo_list_lock); | |
3741 | } | |
3742 | ||
93ee31f1 DL |
3743 | static void rollback_registered(struct net_device *dev) |
3744 | { | |
3745 | BUG_ON(dev_boot_phase); | |
3746 | ASSERT_RTNL(); | |
3747 | ||
3748 | /* Some devices call without registering for initialization unwind. */ | |
3749 | if (dev->reg_state == NETREG_UNINITIALIZED) { | |
3750 | printk(KERN_DEBUG "unregister_netdevice: device %s/%p never " | |
3751 | "was registered\n", dev->name, dev); | |
3752 | ||
3753 | WARN_ON(1); | |
3754 | return; | |
3755 | } | |
3756 | ||
3757 | BUG_ON(dev->reg_state != NETREG_REGISTERED); | |
3758 | ||
3759 | /* If device is running, close it first. */ | |
3760 | dev_close(dev); | |
3761 | ||
3762 | /* And unlink it from device chain. */ | |
3763 | unlist_netdevice(dev); | |
3764 | ||
3765 | dev->reg_state = NETREG_UNREGISTERING; | |
3766 | ||
3767 | synchronize_net(); | |
3768 | ||
3769 | /* Shutdown queueing discipline. */ | |
3770 | dev_shutdown(dev); | |
3771 | ||
3772 | ||
3773 | /* Notify protocols, that we are about to destroy | |
3774 | this device. They should clean all the things. | |
3775 | */ | |
3776 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); | |
3777 | ||
3778 | /* | |
3779 | * Flush the unicast and multicast chains | |
3780 | */ | |
3781 | dev_addr_discard(dev); | |
3782 | ||
3783 | if (dev->uninit) | |
3784 | dev->uninit(dev); | |
3785 | ||
3786 | /* Notifier chain MUST detach us from master device. */ | |
3787 | BUG_TRAP(!dev->master); | |
3788 | ||
3789 | /* Remove entries from kobject tree */ | |
3790 | netdev_unregister_kobject(dev); | |
3791 | ||
3792 | synchronize_net(); | |
3793 | ||
3794 | dev_put(dev); | |
3795 | } | |
3796 | ||
e8a0464c DM |
3797 | static void __netdev_init_queue_locks_one(struct net_device *dev, |
3798 | struct netdev_queue *dev_queue, | |
3799 | void *_unused) | |
c773e847 DM |
3800 | { |
3801 | spin_lock_init(&dev_queue->_xmit_lock); | |
3802 | netdev_set_lockdep_class(&dev_queue->_xmit_lock, dev->type); | |
3803 | dev_queue->xmit_lock_owner = -1; | |
3804 | } | |
3805 | ||
3806 | static void netdev_init_queue_locks(struct net_device *dev) | |
3807 | { | |
e8a0464c DM |
3808 | netdev_for_each_tx_queue(dev, __netdev_init_queue_locks_one, NULL); |
3809 | __netdev_init_queue_locks_one(dev, &dev->rx_queue, NULL); | |
c773e847 DM |
3810 | } |
3811 | ||
1da177e4 LT |
3812 | /** |
3813 | * register_netdevice - register a network device | |
3814 | * @dev: device to register | |
3815 | * | |
3816 | * Take a completed network device structure and add it to the kernel | |
3817 | * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier | |
3818 | * chain. 0 is returned on success. A negative errno code is returned | |
3819 | * on a failure to set up the device, or if the name is a duplicate. | |
3820 | * | |
3821 | * Callers must hold the rtnl semaphore. You may want | |
3822 | * register_netdev() instead of this. | |
3823 | * | |
3824 | * BUGS: | |
3825 | * The locking appears insufficient to guarantee two parallel registers | |
3826 | * will not get the same name. | |
3827 | */ | |
3828 | ||
3829 | int register_netdevice(struct net_device *dev) | |
3830 | { | |
3831 | struct hlist_head *head; | |
3832 | struct hlist_node *p; | |
3833 | int ret; | |
881d966b | 3834 | struct net *net; |
1da177e4 LT |
3835 | |
3836 | BUG_ON(dev_boot_phase); | |
3837 | ASSERT_RTNL(); | |
3838 | ||
b17a7c17 SH |
3839 | might_sleep(); |
3840 | ||
1da177e4 LT |
3841 | /* When net_device's are persistent, this will be fatal. */ |
3842 | BUG_ON(dev->reg_state != NETREG_UNINITIALIZED); | |
c346dca1 YH |
3843 | BUG_ON(!dev_net(dev)); |
3844 | net = dev_net(dev); | |
1da177e4 | 3845 | |
f1f28aa3 | 3846 | spin_lock_init(&dev->addr_list_lock); |
c773e847 | 3847 | netdev_init_queue_locks(dev); |
1da177e4 | 3848 | |
1da177e4 LT |
3849 | dev->iflink = -1; |
3850 | ||
3851 | /* Init, if this function is available */ | |
3852 | if (dev->init) { | |
3853 | ret = dev->init(dev); | |
3854 | if (ret) { | |
3855 | if (ret > 0) | |
3856 | ret = -EIO; | |
90833aa4 | 3857 | goto out; |
1da177e4 LT |
3858 | } |
3859 | } | |
4ec93edb | 3860 | |
1da177e4 LT |
3861 | if (!dev_valid_name(dev->name)) { |
3862 | ret = -EINVAL; | |
7ce1b0ed | 3863 | goto err_uninit; |
1da177e4 LT |
3864 | } |
3865 | ||
881d966b | 3866 | dev->ifindex = dev_new_index(net); |
1da177e4 LT |
3867 | if (dev->iflink == -1) |
3868 | dev->iflink = dev->ifindex; | |
3869 | ||
3870 | /* Check for existence of name */ | |
881d966b | 3871 | head = dev_name_hash(net, dev->name); |
1da177e4 LT |
3872 | hlist_for_each(p, head) { |
3873 | struct net_device *d | |
3874 | = hlist_entry(p, struct net_device, name_hlist); | |
3875 | if (!strncmp(d->name, dev->name, IFNAMSIZ)) { | |
3876 | ret = -EEXIST; | |
7ce1b0ed | 3877 | goto err_uninit; |
1da177e4 | 3878 | } |
4ec93edb | 3879 | } |
1da177e4 | 3880 | |
d212f87b SH |
3881 | /* Fix illegal checksum combinations */ |
3882 | if ((dev->features & NETIF_F_HW_CSUM) && | |
3883 | (dev->features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { | |
3884 | printk(KERN_NOTICE "%s: mixed HW and IP checksum settings.\n", | |
3885 | dev->name); | |
3886 | dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); | |
3887 | } | |
3888 | ||
3889 | if ((dev->features & NETIF_F_NO_CSUM) && | |
3890 | (dev->features & (NETIF_F_HW_CSUM|NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { | |
3891 | printk(KERN_NOTICE "%s: mixed no checksumming and other settings.\n", | |
3892 | dev->name); | |
3893 | dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM|NETIF_F_HW_CSUM); | |
3894 | } | |
3895 | ||
3896 | ||
1da177e4 LT |
3897 | /* Fix illegal SG+CSUM combinations. */ |
3898 | if ((dev->features & NETIF_F_SG) && | |
8648b305 | 3899 | !(dev->features & NETIF_F_ALL_CSUM)) { |
5a8da02b | 3900 | printk(KERN_NOTICE "%s: Dropping NETIF_F_SG since no checksum feature.\n", |
1da177e4 LT |
3901 | dev->name); |
3902 | dev->features &= ~NETIF_F_SG; | |
3903 | } | |
3904 | ||
3905 | /* TSO requires that SG is present as well. */ | |
3906 | if ((dev->features & NETIF_F_TSO) && | |
3907 | !(dev->features & NETIF_F_SG)) { | |
5a8da02b | 3908 | printk(KERN_NOTICE "%s: Dropping NETIF_F_TSO since no SG feature.\n", |
1da177e4 LT |
3909 | dev->name); |
3910 | dev->features &= ~NETIF_F_TSO; | |
3911 | } | |
e89e9cf5 AR |
3912 | if (dev->features & NETIF_F_UFO) { |
3913 | if (!(dev->features & NETIF_F_HW_CSUM)) { | |
3914 | printk(KERN_ERR "%s: Dropping NETIF_F_UFO since no " | |
3915 | "NETIF_F_HW_CSUM feature.\n", | |
3916 | dev->name); | |
3917 | dev->features &= ~NETIF_F_UFO; | |
3918 | } | |
3919 | if (!(dev->features & NETIF_F_SG)) { | |
3920 | printk(KERN_ERR "%s: Dropping NETIF_F_UFO since no " | |
3921 | "NETIF_F_SG feature.\n", | |
3922 | dev->name); | |
3923 | dev->features &= ~NETIF_F_UFO; | |
3924 | } | |
3925 | } | |
1da177e4 | 3926 | |
aaf8cdc3 | 3927 | netdev_initialize_kobject(dev); |
8b41d188 | 3928 | ret = netdev_register_kobject(dev); |
b17a7c17 | 3929 | if (ret) |
7ce1b0ed | 3930 | goto err_uninit; |
b17a7c17 SH |
3931 | dev->reg_state = NETREG_REGISTERED; |
3932 | ||
1da177e4 LT |
3933 | /* |
3934 | * Default initial state at registry is that the | |
3935 | * device is present. | |
3936 | */ | |
3937 | ||
3938 | set_bit(__LINK_STATE_PRESENT, &dev->state); | |
3939 | ||
1da177e4 | 3940 | dev_init_scheduler(dev); |
1da177e4 | 3941 | dev_hold(dev); |
ce286d32 | 3942 | list_netdevice(dev); |
1da177e4 LT |
3943 | |
3944 | /* Notify protocols, that a new device appeared. */ | |
056925ab | 3945 | ret = call_netdevice_notifiers(NETDEV_REGISTER, dev); |
fcc5a03a | 3946 | ret = notifier_to_errno(ret); |
93ee31f1 DL |
3947 | if (ret) { |
3948 | rollback_registered(dev); | |
3949 | dev->reg_state = NETREG_UNREGISTERED; | |
3950 | } | |
1da177e4 LT |
3951 | |
3952 | out: | |
3953 | return ret; | |
7ce1b0ed HX |
3954 | |
3955 | err_uninit: | |
3956 | if (dev->uninit) | |
3957 | dev->uninit(dev); | |
3958 | goto out; | |
1da177e4 LT |
3959 | } |
3960 | ||
3961 | /** | |
3962 | * register_netdev - register a network device | |
3963 | * @dev: device to register | |
3964 | * | |
3965 | * Take a completed network device structure and add it to the kernel | |
3966 | * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier | |
3967 | * chain. 0 is returned on success. A negative errno code is returned | |
3968 | * on a failure to set up the device, or if the name is a duplicate. | |
3969 | * | |
38b4da38 | 3970 | * This is a wrapper around register_netdevice that takes the rtnl semaphore |
1da177e4 LT |
3971 | * and expands the device name if you passed a format string to |
3972 | * alloc_netdev. | |
3973 | */ | |
3974 | int register_netdev(struct net_device *dev) | |
3975 | { | |
3976 | int err; | |
3977 | ||
3978 | rtnl_lock(); | |
3979 | ||
3980 | /* | |
3981 | * If the name is a format string the caller wants us to do a | |
3982 | * name allocation. | |
3983 | */ | |
3984 | if (strchr(dev->name, '%')) { | |
3985 | err = dev_alloc_name(dev, dev->name); | |
3986 | if (err < 0) | |
3987 | goto out; | |
3988 | } | |
4ec93edb | 3989 | |
1da177e4 LT |
3990 | err = register_netdevice(dev); |
3991 | out: | |
3992 | rtnl_unlock(); | |
3993 | return err; | |
3994 | } | |
3995 | EXPORT_SYMBOL(register_netdev); | |
3996 | ||
3997 | /* | |
3998 | * netdev_wait_allrefs - wait until all references are gone. | |
3999 | * | |
4000 | * This is called when unregistering network devices. | |
4001 | * | |
4002 | * Any protocol or device that holds a reference should register | |
4003 | * for netdevice notification, and cleanup and put back the | |
4004 | * reference if they receive an UNREGISTER event. | |
4005 | * We can get stuck here if buggy protocols don't correctly | |
4ec93edb | 4006 | * call dev_put. |
1da177e4 LT |
4007 | */ |
4008 | static void netdev_wait_allrefs(struct net_device *dev) | |
4009 | { | |
4010 | unsigned long rebroadcast_time, warning_time; | |
4011 | ||
4012 | rebroadcast_time = warning_time = jiffies; | |
4013 | while (atomic_read(&dev->refcnt) != 0) { | |
4014 | if (time_after(jiffies, rebroadcast_time + 1 * HZ)) { | |
6756ae4b | 4015 | rtnl_lock(); |
1da177e4 LT |
4016 | |
4017 | /* Rebroadcast unregister notification */ | |
056925ab | 4018 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); |
1da177e4 LT |
4019 | |
4020 | if (test_bit(__LINK_STATE_LINKWATCH_PENDING, | |
4021 | &dev->state)) { | |
4022 | /* We must not have linkwatch events | |
4023 | * pending on unregister. If this | |
4024 | * happens, we simply run the queue | |
4025 | * unscheduled, resulting in a noop | |
4026 | * for this device. | |
4027 | */ | |
4028 | linkwatch_run_queue(); | |
4029 | } | |
4030 | ||
6756ae4b | 4031 | __rtnl_unlock(); |
1da177e4 LT |
4032 | |
4033 | rebroadcast_time = jiffies; | |
4034 | } | |
4035 | ||
4036 | msleep(250); | |
4037 | ||
4038 | if (time_after(jiffies, warning_time + 10 * HZ)) { | |
4039 | printk(KERN_EMERG "unregister_netdevice: " | |
4040 | "waiting for %s to become free. Usage " | |
4041 | "count = %d\n", | |
4042 | dev->name, atomic_read(&dev->refcnt)); | |
4043 | warning_time = jiffies; | |
4044 | } | |
4045 | } | |
4046 | } | |
4047 | ||
4048 | /* The sequence is: | |
4049 | * | |
4050 | * rtnl_lock(); | |
4051 | * ... | |
4052 | * register_netdevice(x1); | |
4053 | * register_netdevice(x2); | |
4054 | * ... | |
4055 | * unregister_netdevice(y1); | |
4056 | * unregister_netdevice(y2); | |
4057 | * ... | |
4058 | * rtnl_unlock(); | |
4059 | * free_netdev(y1); | |
4060 | * free_netdev(y2); | |
4061 | * | |
4062 | * We are invoked by rtnl_unlock() after it drops the semaphore. | |
4063 | * This allows us to deal with problems: | |
b17a7c17 | 4064 | * 1) We can delete sysfs objects which invoke hotplug |
1da177e4 LT |
4065 | * without deadlocking with linkwatch via keventd. |
4066 | * 2) Since we run with the RTNL semaphore not held, we can sleep | |
4067 | * safely in order to wait for the netdev refcnt to drop to zero. | |
4068 | */ | |
4a3e2f71 | 4069 | static DEFINE_MUTEX(net_todo_run_mutex); |
1da177e4 LT |
4070 | void netdev_run_todo(void) |
4071 | { | |
626ab0e6 | 4072 | struct list_head list; |
1da177e4 LT |
4073 | |
4074 | /* Need to guard against multiple cpu's getting out of order. */ | |
4a3e2f71 | 4075 | mutex_lock(&net_todo_run_mutex); |
1da177e4 LT |
4076 | |
4077 | /* Not safe to do outside the semaphore. We must not return | |
4078 | * until all unregister events invoked by the local processor | |
4079 | * have been completed (either by this todo run, or one on | |
4080 | * another cpu). | |
4081 | */ | |
4082 | if (list_empty(&net_todo_list)) | |
4083 | goto out; | |
4084 | ||
4085 | /* Snapshot list, allow later requests */ | |
4086 | spin_lock(&net_todo_list_lock); | |
626ab0e6 | 4087 | list_replace_init(&net_todo_list, &list); |
1da177e4 | 4088 | spin_unlock(&net_todo_list_lock); |
626ab0e6 | 4089 | |
1da177e4 LT |
4090 | while (!list_empty(&list)) { |
4091 | struct net_device *dev | |
4092 | = list_entry(list.next, struct net_device, todo_list); | |
4093 | list_del(&dev->todo_list); | |
4094 | ||
b17a7c17 SH |
4095 | if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) { |
4096 | printk(KERN_ERR "network todo '%s' but state %d\n", | |
4097 | dev->name, dev->reg_state); | |
4098 | dump_stack(); | |
4099 | continue; | |
4100 | } | |
1da177e4 | 4101 | |
b17a7c17 | 4102 | dev->reg_state = NETREG_UNREGISTERED; |
1da177e4 | 4103 | |
b17a7c17 | 4104 | netdev_wait_allrefs(dev); |
1da177e4 | 4105 | |
b17a7c17 SH |
4106 | /* paranoia */ |
4107 | BUG_ON(atomic_read(&dev->refcnt)); | |
4108 | BUG_TRAP(!dev->ip_ptr); | |
4109 | BUG_TRAP(!dev->ip6_ptr); | |
4110 | BUG_TRAP(!dev->dn_ptr); | |
1da177e4 | 4111 | |
b17a7c17 SH |
4112 | if (dev->destructor) |
4113 | dev->destructor(dev); | |
9093bbb2 SH |
4114 | |
4115 | /* Free network device */ | |
4116 | kobject_put(&dev->dev.kobj); | |
1da177e4 LT |
4117 | } |
4118 | ||
4119 | out: | |
4a3e2f71 | 4120 | mutex_unlock(&net_todo_run_mutex); |
1da177e4 LT |
4121 | } |
4122 | ||
5a1b5898 | 4123 | static struct net_device_stats *internal_stats(struct net_device *dev) |
c45d286e | 4124 | { |
5a1b5898 | 4125 | return &dev->stats; |
c45d286e RR |
4126 | } |
4127 | ||
dc2b4847 | 4128 | static void netdev_init_one_queue(struct net_device *dev, |
e8a0464c DM |
4129 | struct netdev_queue *queue, |
4130 | void *_unused) | |
dc2b4847 DM |
4131 | { |
4132 | spin_lock_init(&queue->lock); | |
4133 | queue->dev = dev; | |
4134 | } | |
4135 | ||
bb949fbd DM |
4136 | static void netdev_init_queues(struct net_device *dev) |
4137 | { | |
e8a0464c DM |
4138 | netdev_init_one_queue(dev, &dev->rx_queue, NULL); |
4139 | netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL); | |
bb949fbd DM |
4140 | } |
4141 | ||
1da177e4 | 4142 | /** |
f25f4e44 | 4143 | * alloc_netdev_mq - allocate network device |
1da177e4 LT |
4144 | * @sizeof_priv: size of private data to allocate space for |
4145 | * @name: device name format string | |
4146 | * @setup: callback to initialize device | |
f25f4e44 | 4147 | * @queue_count: the number of subqueues to allocate |
1da177e4 LT |
4148 | * |
4149 | * Allocates a struct net_device with private data area for driver use | |
f25f4e44 PWJ |
4150 | * and performs basic initialization. Also allocates subquue structs |
4151 | * for each queue on the device at the end of the netdevice. | |
1da177e4 | 4152 | */ |
f25f4e44 PWJ |
4153 | struct net_device *alloc_netdev_mq(int sizeof_priv, const char *name, |
4154 | void (*setup)(struct net_device *), unsigned int queue_count) | |
1da177e4 | 4155 | { |
e8a0464c | 4156 | struct netdev_queue *tx; |
1da177e4 LT |
4157 | struct net_device *dev; |
4158 | int alloc_size; | |
e8a0464c | 4159 | void *p; |
1da177e4 | 4160 | |
b6fe17d6 SH |
4161 | BUG_ON(strlen(name) >= sizeof(dev->name)); |
4162 | ||
d1643d24 AD |
4163 | alloc_size = sizeof(struct net_device) + |
4164 | sizeof(struct net_device_subqueue) * (queue_count - 1); | |
4165 | if (sizeof_priv) { | |
4166 | /* ensure 32-byte alignment of private area */ | |
4167 | alloc_size = (alloc_size + NETDEV_ALIGN_CONST) & ~NETDEV_ALIGN_CONST; | |
4168 | alloc_size += sizeof_priv; | |
4169 | } | |
4170 | /* ensure 32-byte alignment of whole construct */ | |
4171 | alloc_size += NETDEV_ALIGN_CONST; | |
1da177e4 | 4172 | |
31380de9 | 4173 | p = kzalloc(alloc_size, GFP_KERNEL); |
1da177e4 | 4174 | if (!p) { |
b6fe17d6 | 4175 | printk(KERN_ERR "alloc_netdev: Unable to allocate device.\n"); |
1da177e4 LT |
4176 | return NULL; |
4177 | } | |
1da177e4 | 4178 | |
e8a0464c DM |
4179 | tx = kzalloc(sizeof(struct netdev_queue) * queue_count, GFP_KERNEL); |
4180 | if (!tx) { | |
4181 | printk(KERN_ERR "alloc_netdev: Unable to allocate " | |
4182 | "tx qdiscs.\n"); | |
4183 | kfree(p); | |
4184 | return NULL; | |
4185 | } | |
4186 | ||
1da177e4 LT |
4187 | dev = (struct net_device *) |
4188 | (((long)p + NETDEV_ALIGN_CONST) & ~NETDEV_ALIGN_CONST); | |
4189 | dev->padded = (char *)dev - (char *)p; | |
c346dca1 | 4190 | dev_net_set(dev, &init_net); |
1da177e4 | 4191 | |
e8a0464c DM |
4192 | dev->_tx = tx; |
4193 | dev->num_tx_queues = queue_count; | |
4194 | ||
f25f4e44 PWJ |
4195 | if (sizeof_priv) { |
4196 | dev->priv = ((char *)dev + | |
4197 | ((sizeof(struct net_device) + | |
4198 | (sizeof(struct net_device_subqueue) * | |
31ce72a6 | 4199 | (queue_count - 1)) + NETDEV_ALIGN_CONST) |
f25f4e44 PWJ |
4200 | & ~NETDEV_ALIGN_CONST)); |
4201 | } | |
4202 | ||
4203 | dev->egress_subqueue_count = queue_count; | |
82cc1a7a | 4204 | dev->gso_max_size = GSO_MAX_SIZE; |
1da177e4 | 4205 | |
bb949fbd DM |
4206 | netdev_init_queues(dev); |
4207 | ||
5a1b5898 | 4208 | dev->get_stats = internal_stats; |
bea3348e | 4209 | netpoll_netdev_init(dev); |
1da177e4 LT |
4210 | setup(dev); |
4211 | strcpy(dev->name, name); | |
4212 | return dev; | |
4213 | } | |
f25f4e44 | 4214 | EXPORT_SYMBOL(alloc_netdev_mq); |
1da177e4 LT |
4215 | |
4216 | /** | |
4217 | * free_netdev - free network device | |
4218 | * @dev: device | |
4219 | * | |
4ec93edb YH |
4220 | * This function does the last stage of destroying an allocated device |
4221 | * interface. The reference to the device object is released. | |
1da177e4 LT |
4222 | * If this is the last reference then it will be freed. |
4223 | */ | |
4224 | void free_netdev(struct net_device *dev) | |
4225 | { | |
f3005d7f DL |
4226 | release_net(dev_net(dev)); |
4227 | ||
e8a0464c DM |
4228 | kfree(dev->_tx); |
4229 | ||
3041a069 | 4230 | /* Compatibility with error handling in drivers */ |
1da177e4 LT |
4231 | if (dev->reg_state == NETREG_UNINITIALIZED) { |
4232 | kfree((char *)dev - dev->padded); | |
4233 | return; | |
4234 | } | |
4235 | ||
4236 | BUG_ON(dev->reg_state != NETREG_UNREGISTERED); | |
4237 | dev->reg_state = NETREG_RELEASED; | |
4238 | ||
43cb76d9 GKH |
4239 | /* will free via device release */ |
4240 | put_device(&dev->dev); | |
1da177e4 | 4241 | } |
4ec93edb | 4242 | |
1da177e4 | 4243 | /* Synchronize with packet receive processing. */ |
4ec93edb | 4244 | void synchronize_net(void) |
1da177e4 LT |
4245 | { |
4246 | might_sleep(); | |
fbd568a3 | 4247 | synchronize_rcu(); |
1da177e4 LT |
4248 | } |
4249 | ||
4250 | /** | |
4251 | * unregister_netdevice - remove device from the kernel | |
4252 | * @dev: device | |
4253 | * | |
4254 | * This function shuts down a device interface and removes it | |
d59b54b1 | 4255 | * from the kernel tables. |
1da177e4 LT |
4256 | * |
4257 | * Callers must hold the rtnl semaphore. You may want | |
4258 | * unregister_netdev() instead of this. | |
4259 | */ | |
4260 | ||
22f8cde5 | 4261 | void unregister_netdevice(struct net_device *dev) |
1da177e4 | 4262 | { |
a6620712 HX |
4263 | ASSERT_RTNL(); |
4264 | ||
93ee31f1 | 4265 | rollback_registered(dev); |
1da177e4 LT |
4266 | /* Finish processing unregister after unlock */ |
4267 | net_set_todo(dev); | |
1da177e4 LT |
4268 | } |
4269 | ||
4270 | /** | |
4271 | * unregister_netdev - remove device from the kernel | |
4272 | * @dev: device | |
4273 | * | |
4274 | * This function shuts down a device interface and removes it | |
d59b54b1 | 4275 | * from the kernel tables. |
1da177e4 LT |
4276 | * |
4277 | * This is just a wrapper for unregister_netdevice that takes | |
4278 | * the rtnl semaphore. In general you want to use this and not | |
4279 | * unregister_netdevice. | |
4280 | */ | |
4281 | void unregister_netdev(struct net_device *dev) | |
4282 | { | |
4283 | rtnl_lock(); | |
4284 | unregister_netdevice(dev); | |
4285 | rtnl_unlock(); | |
4286 | } | |
4287 | ||
4288 | EXPORT_SYMBOL(unregister_netdev); | |
4289 | ||
ce286d32 EB |
4290 | /** |
4291 | * dev_change_net_namespace - move device to different nethost namespace | |
4292 | * @dev: device | |
4293 | * @net: network namespace | |
4294 | * @pat: If not NULL name pattern to try if the current device name | |
4295 | * is already taken in the destination network namespace. | |
4296 | * | |
4297 | * This function shuts down a device interface and moves it | |
4298 | * to a new network namespace. On success 0 is returned, on | |
4299 | * a failure a netagive errno code is returned. | |
4300 | * | |
4301 | * Callers must hold the rtnl semaphore. | |
4302 | */ | |
4303 | ||
4304 | int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat) | |
4305 | { | |
4306 | char buf[IFNAMSIZ]; | |
4307 | const char *destname; | |
4308 | int err; | |
4309 | ||
4310 | ASSERT_RTNL(); | |
4311 | ||
4312 | /* Don't allow namespace local devices to be moved. */ | |
4313 | err = -EINVAL; | |
4314 | if (dev->features & NETIF_F_NETNS_LOCAL) | |
4315 | goto out; | |
4316 | ||
4317 | /* Ensure the device has been registrered */ | |
4318 | err = -EINVAL; | |
4319 | if (dev->reg_state != NETREG_REGISTERED) | |
4320 | goto out; | |
4321 | ||
4322 | /* Get out if there is nothing todo */ | |
4323 | err = 0; | |
878628fb | 4324 | if (net_eq(dev_net(dev), net)) |
ce286d32 EB |
4325 | goto out; |
4326 | ||
4327 | /* Pick the destination device name, and ensure | |
4328 | * we can use it in the destination network namespace. | |
4329 | */ | |
4330 | err = -EEXIST; | |
4331 | destname = dev->name; | |
4332 | if (__dev_get_by_name(net, destname)) { | |
4333 | /* We get here if we can't use the current device name */ | |
4334 | if (!pat) | |
4335 | goto out; | |
4336 | if (!dev_valid_name(pat)) | |
4337 | goto out; | |
4338 | if (strchr(pat, '%')) { | |
4339 | if (__dev_alloc_name(net, pat, buf) < 0) | |
4340 | goto out; | |
4341 | destname = buf; | |
4342 | } else | |
4343 | destname = pat; | |
4344 | if (__dev_get_by_name(net, destname)) | |
4345 | goto out; | |
4346 | } | |
4347 | ||
4348 | /* | |
4349 | * And now a mini version of register_netdevice unregister_netdevice. | |
4350 | */ | |
4351 | ||
4352 | /* If device is running close it first. */ | |
9b772652 | 4353 | dev_close(dev); |
ce286d32 EB |
4354 | |
4355 | /* And unlink it from device chain */ | |
4356 | err = -ENODEV; | |
4357 | unlist_netdevice(dev); | |
4358 | ||
4359 | synchronize_net(); | |
4360 | ||
4361 | /* Shutdown queueing discipline. */ | |
4362 | dev_shutdown(dev); | |
4363 | ||
4364 | /* Notify protocols, that we are about to destroy | |
4365 | this device. They should clean all the things. | |
4366 | */ | |
4367 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); | |
4368 | ||
4369 | /* | |
4370 | * Flush the unicast and multicast chains | |
4371 | */ | |
4372 | dev_addr_discard(dev); | |
4373 | ||
4374 | /* Actually switch the network namespace */ | |
c346dca1 | 4375 | dev_net_set(dev, net); |
ce286d32 EB |
4376 | |
4377 | /* Assign the new device name */ | |
4378 | if (destname != dev->name) | |
4379 | strcpy(dev->name, destname); | |
4380 | ||
4381 | /* If there is an ifindex conflict assign a new one */ | |
4382 | if (__dev_get_by_index(net, dev->ifindex)) { | |
4383 | int iflink = (dev->iflink == dev->ifindex); | |
4384 | dev->ifindex = dev_new_index(net); | |
4385 | if (iflink) | |
4386 | dev->iflink = dev->ifindex; | |
4387 | } | |
4388 | ||
8b41d188 | 4389 | /* Fixup kobjects */ |
aaf8cdc3 DL |
4390 | netdev_unregister_kobject(dev); |
4391 | err = netdev_register_kobject(dev); | |
8b41d188 | 4392 | WARN_ON(err); |
ce286d32 EB |
4393 | |
4394 | /* Add the device back in the hashes */ | |
4395 | list_netdevice(dev); | |
4396 | ||
4397 | /* Notify protocols, that a new device appeared. */ | |
4398 | call_netdevice_notifiers(NETDEV_REGISTER, dev); | |
4399 | ||
4400 | synchronize_net(); | |
4401 | err = 0; | |
4402 | out: | |
4403 | return err; | |
4404 | } | |
4405 | ||
1da177e4 LT |
4406 | static int dev_cpu_callback(struct notifier_block *nfb, |
4407 | unsigned long action, | |
4408 | void *ocpu) | |
4409 | { | |
4410 | struct sk_buff **list_skb; | |
ee609cb3 | 4411 | struct netdev_queue **list_net; |
1da177e4 LT |
4412 | struct sk_buff *skb; |
4413 | unsigned int cpu, oldcpu = (unsigned long)ocpu; | |
4414 | struct softnet_data *sd, *oldsd; | |
4415 | ||
8bb78442 | 4416 | if (action != CPU_DEAD && action != CPU_DEAD_FROZEN) |
1da177e4 LT |
4417 | return NOTIFY_OK; |
4418 | ||
4419 | local_irq_disable(); | |
4420 | cpu = smp_processor_id(); | |
4421 | sd = &per_cpu(softnet_data, cpu); | |
4422 | oldsd = &per_cpu(softnet_data, oldcpu); | |
4423 | ||
4424 | /* Find end of our completion_queue. */ | |
4425 | list_skb = &sd->completion_queue; | |
4426 | while (*list_skb) | |
4427 | list_skb = &(*list_skb)->next; | |
4428 | /* Append completion queue from offline CPU. */ | |
4429 | *list_skb = oldsd->completion_queue; | |
4430 | oldsd->completion_queue = NULL; | |
4431 | ||
4432 | /* Find end of our output_queue. */ | |
4433 | list_net = &sd->output_queue; | |
4434 | while (*list_net) | |
4435 | list_net = &(*list_net)->next_sched; | |
4436 | /* Append output queue from offline CPU. */ | |
4437 | *list_net = oldsd->output_queue; | |
4438 | oldsd->output_queue = NULL; | |
4439 | ||
4440 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
4441 | local_irq_enable(); | |
4442 | ||
4443 | /* Process offline CPU's input_pkt_queue */ | |
4444 | while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) | |
4445 | netif_rx(skb); | |
4446 | ||
4447 | return NOTIFY_OK; | |
4448 | } | |
1da177e4 | 4449 | |
db217334 CL |
4450 | #ifdef CONFIG_NET_DMA |
4451 | /** | |
0ed72ec4 RD |
4452 | * net_dma_rebalance - try to maintain one DMA channel per CPU |
4453 | * @net_dma: DMA client and associated data (lock, channels, channel_mask) | |
4454 | * | |
4455 | * This is called when the number of channels allocated to the net_dma client | |
4456 | * changes. The net_dma client tries to have one DMA channel per CPU. | |
db217334 | 4457 | */ |
d379b01e DW |
4458 | |
4459 | static void net_dma_rebalance(struct net_dma *net_dma) | |
db217334 | 4460 | { |
d379b01e | 4461 | unsigned int cpu, i, n, chan_idx; |
db217334 CL |
4462 | struct dma_chan *chan; |
4463 | ||
d379b01e | 4464 | if (cpus_empty(net_dma->channel_mask)) { |
db217334 | 4465 | for_each_online_cpu(cpu) |
29bbd72d | 4466 | rcu_assign_pointer(per_cpu(softnet_data, cpu).net_dma, NULL); |
db217334 CL |
4467 | return; |
4468 | } | |
4469 | ||
4470 | i = 0; | |
4471 | cpu = first_cpu(cpu_online_map); | |
4472 | ||
d379b01e DW |
4473 | for_each_cpu_mask(chan_idx, net_dma->channel_mask) { |
4474 | chan = net_dma->channels[chan_idx]; | |
4475 | ||
4476 | n = ((num_online_cpus() / cpus_weight(net_dma->channel_mask)) | |
4477 | + (i < (num_online_cpus() % | |
4478 | cpus_weight(net_dma->channel_mask)) ? 1 : 0)); | |
db217334 CL |
4479 | |
4480 | while(n) { | |
29bbd72d | 4481 | per_cpu(softnet_data, cpu).net_dma = chan; |
db217334 CL |
4482 | cpu = next_cpu(cpu, cpu_online_map); |
4483 | n--; | |
4484 | } | |
4485 | i++; | |
4486 | } | |
db217334 CL |
4487 | } |
4488 | ||
4489 | /** | |
4490 | * netdev_dma_event - event callback for the net_dma_client | |
4491 | * @client: should always be net_dma_client | |
f4b8ea78 | 4492 | * @chan: DMA channel for the event |
0ed72ec4 | 4493 | * @state: DMA state to be handled |
db217334 | 4494 | */ |
d379b01e DW |
4495 | static enum dma_state_client |
4496 | netdev_dma_event(struct dma_client *client, struct dma_chan *chan, | |
4497 | enum dma_state state) | |
4498 | { | |
4499 | int i, found = 0, pos = -1; | |
4500 | struct net_dma *net_dma = | |
4501 | container_of(client, struct net_dma, client); | |
4502 | enum dma_state_client ack = DMA_DUP; /* default: take no action */ | |
4503 | ||
4504 | spin_lock(&net_dma->lock); | |
4505 | switch (state) { | |
4506 | case DMA_RESOURCE_AVAILABLE: | |
0c0b0aca | 4507 | for (i = 0; i < nr_cpu_ids; i++) |
d379b01e DW |
4508 | if (net_dma->channels[i] == chan) { |
4509 | found = 1; | |
4510 | break; | |
4511 | } else if (net_dma->channels[i] == NULL && pos < 0) | |
4512 | pos = i; | |
4513 | ||
4514 | if (!found && pos >= 0) { | |
4515 | ack = DMA_ACK; | |
4516 | net_dma->channels[pos] = chan; | |
4517 | cpu_set(pos, net_dma->channel_mask); | |
4518 | net_dma_rebalance(net_dma); | |
4519 | } | |
db217334 CL |
4520 | break; |
4521 | case DMA_RESOURCE_REMOVED: | |
0c0b0aca | 4522 | for (i = 0; i < nr_cpu_ids; i++) |
d379b01e DW |
4523 | if (net_dma->channels[i] == chan) { |
4524 | found = 1; | |
4525 | pos = i; | |
4526 | break; | |
4527 | } | |
4528 | ||
4529 | if (found) { | |
4530 | ack = DMA_ACK; | |
4531 | cpu_clear(pos, net_dma->channel_mask); | |
4532 | net_dma->channels[i] = NULL; | |
4533 | net_dma_rebalance(net_dma); | |
4534 | } | |
db217334 CL |
4535 | break; |
4536 | default: | |
4537 | break; | |
4538 | } | |
d379b01e DW |
4539 | spin_unlock(&net_dma->lock); |
4540 | ||
4541 | return ack; | |
db217334 CL |
4542 | } |
4543 | ||
4544 | /** | |
4545 | * netdev_dma_regiser - register the networking subsystem as a DMA client | |
4546 | */ | |
4547 | static int __init netdev_dma_register(void) | |
4548 | { | |
0c0b0aca MT |
4549 | net_dma.channels = kzalloc(nr_cpu_ids * sizeof(struct net_dma), |
4550 | GFP_KERNEL); | |
4551 | if (unlikely(!net_dma.channels)) { | |
4552 | printk(KERN_NOTICE | |
4553 | "netdev_dma: no memory for net_dma.channels\n"); | |
4554 | return -ENOMEM; | |
4555 | } | |
d379b01e DW |
4556 | spin_lock_init(&net_dma.lock); |
4557 | dma_cap_set(DMA_MEMCPY, net_dma.client.cap_mask); | |
4558 | dma_async_client_register(&net_dma.client); | |
4559 | dma_async_client_chan_request(&net_dma.client); | |
db217334 CL |
4560 | return 0; |
4561 | } | |
4562 | ||
4563 | #else | |
4564 | static int __init netdev_dma_register(void) { return -ENODEV; } | |
4565 | #endif /* CONFIG_NET_DMA */ | |
1da177e4 | 4566 | |
7f353bf2 HX |
4567 | /** |
4568 | * netdev_compute_feature - compute conjunction of two feature sets | |
4569 | * @all: first feature set | |
4570 | * @one: second feature set | |
4571 | * | |
4572 | * Computes a new feature set after adding a device with feature set | |
4573 | * @one to the master device with current feature set @all. Returns | |
4574 | * the new feature set. | |
4575 | */ | |
4576 | int netdev_compute_features(unsigned long all, unsigned long one) | |
4577 | { | |
4578 | /* if device needs checksumming, downgrade to hw checksumming */ | |
4579 | if (all & NETIF_F_NO_CSUM && !(one & NETIF_F_NO_CSUM)) | |
4580 | all ^= NETIF_F_NO_CSUM | NETIF_F_HW_CSUM; | |
4581 | ||
4582 | /* if device can't do all checksum, downgrade to ipv4/ipv6 */ | |
4583 | if (all & NETIF_F_HW_CSUM && !(one & NETIF_F_HW_CSUM)) | |
4584 | all ^= NETIF_F_HW_CSUM | |
4585 | | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM; | |
4586 | ||
4587 | if (one & NETIF_F_GSO) | |
4588 | one |= NETIF_F_GSO_SOFTWARE; | |
4589 | one |= NETIF_F_GSO; | |
4590 | ||
4591 | /* If even one device supports robust GSO, enable it for all. */ | |
4592 | if (one & NETIF_F_GSO_ROBUST) | |
4593 | all |= NETIF_F_GSO_ROBUST; | |
4594 | ||
4595 | all &= one | NETIF_F_LLTX; | |
4596 | ||
4597 | if (!(all & NETIF_F_ALL_CSUM)) | |
4598 | all &= ~NETIF_F_SG; | |
4599 | if (!(all & NETIF_F_SG)) | |
4600 | all &= ~NETIF_F_GSO_MASK; | |
4601 | ||
4602 | return all; | |
4603 | } | |
4604 | EXPORT_SYMBOL(netdev_compute_features); | |
4605 | ||
30d97d35 PE |
4606 | static struct hlist_head *netdev_create_hash(void) |
4607 | { | |
4608 | int i; | |
4609 | struct hlist_head *hash; | |
4610 | ||
4611 | hash = kmalloc(sizeof(*hash) * NETDEV_HASHENTRIES, GFP_KERNEL); | |
4612 | if (hash != NULL) | |
4613 | for (i = 0; i < NETDEV_HASHENTRIES; i++) | |
4614 | INIT_HLIST_HEAD(&hash[i]); | |
4615 | ||
4616 | return hash; | |
4617 | } | |
4618 | ||
881d966b | 4619 | /* Initialize per network namespace state */ |
4665079c | 4620 | static int __net_init netdev_init(struct net *net) |
881d966b | 4621 | { |
881d966b | 4622 | INIT_LIST_HEAD(&net->dev_base_head); |
881d966b | 4623 | |
30d97d35 PE |
4624 | net->dev_name_head = netdev_create_hash(); |
4625 | if (net->dev_name_head == NULL) | |
4626 | goto err_name; | |
881d966b | 4627 | |
30d97d35 PE |
4628 | net->dev_index_head = netdev_create_hash(); |
4629 | if (net->dev_index_head == NULL) | |
4630 | goto err_idx; | |
881d966b EB |
4631 | |
4632 | return 0; | |
30d97d35 PE |
4633 | |
4634 | err_idx: | |
4635 | kfree(net->dev_name_head); | |
4636 | err_name: | |
4637 | return -ENOMEM; | |
881d966b EB |
4638 | } |
4639 | ||
4665079c | 4640 | static void __net_exit netdev_exit(struct net *net) |
881d966b EB |
4641 | { |
4642 | kfree(net->dev_name_head); | |
4643 | kfree(net->dev_index_head); | |
4644 | } | |
4645 | ||
022cbae6 | 4646 | static struct pernet_operations __net_initdata netdev_net_ops = { |
881d966b EB |
4647 | .init = netdev_init, |
4648 | .exit = netdev_exit, | |
4649 | }; | |
4650 | ||
4665079c | 4651 | static void __net_exit default_device_exit(struct net *net) |
ce286d32 EB |
4652 | { |
4653 | struct net_device *dev, *next; | |
4654 | /* | |
4655 | * Push all migratable of the network devices back to the | |
4656 | * initial network namespace | |
4657 | */ | |
4658 | rtnl_lock(); | |
4659 | for_each_netdev_safe(net, dev, next) { | |
4660 | int err; | |
aca51397 | 4661 | char fb_name[IFNAMSIZ]; |
ce286d32 EB |
4662 | |
4663 | /* Ignore unmoveable devices (i.e. loopback) */ | |
4664 | if (dev->features & NETIF_F_NETNS_LOCAL) | |
4665 | continue; | |
4666 | ||
4667 | /* Push remaing network devices to init_net */ | |
aca51397 PE |
4668 | snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex); |
4669 | err = dev_change_net_namespace(dev, &init_net, fb_name); | |
ce286d32 | 4670 | if (err) { |
aca51397 | 4671 | printk(KERN_EMERG "%s: failed to move %s to init_net: %d\n", |
ce286d32 | 4672 | __func__, dev->name, err); |
aca51397 | 4673 | BUG(); |
ce286d32 EB |
4674 | } |
4675 | } | |
4676 | rtnl_unlock(); | |
4677 | } | |
4678 | ||
022cbae6 | 4679 | static struct pernet_operations __net_initdata default_device_ops = { |
ce286d32 EB |
4680 | .exit = default_device_exit, |
4681 | }; | |
4682 | ||
1da177e4 LT |
4683 | /* |
4684 | * Initialize the DEV module. At boot time this walks the device list and | |
4685 | * unhooks any devices that fail to initialise (normally hardware not | |
4686 | * present) and leaves us with a valid list of present and active devices. | |
4687 | * | |
4688 | */ | |
4689 | ||
4690 | /* | |
4691 | * This is called single threaded during boot, so no need | |
4692 | * to take the rtnl semaphore. | |
4693 | */ | |
4694 | static int __init net_dev_init(void) | |
4695 | { | |
4696 | int i, rc = -ENOMEM; | |
4697 | ||
4698 | BUG_ON(!dev_boot_phase); | |
4699 | ||
1da177e4 LT |
4700 | if (dev_proc_init()) |
4701 | goto out; | |
4702 | ||
8b41d188 | 4703 | if (netdev_kobject_init()) |
1da177e4 LT |
4704 | goto out; |
4705 | ||
4706 | INIT_LIST_HEAD(&ptype_all); | |
82d8a867 | 4707 | for (i = 0; i < PTYPE_HASH_SIZE; i++) |
1da177e4 LT |
4708 | INIT_LIST_HEAD(&ptype_base[i]); |
4709 | ||
881d966b EB |
4710 | if (register_pernet_subsys(&netdev_net_ops)) |
4711 | goto out; | |
1da177e4 | 4712 | |
ce286d32 EB |
4713 | if (register_pernet_device(&default_device_ops)) |
4714 | goto out; | |
4715 | ||
1da177e4 LT |
4716 | /* |
4717 | * Initialise the packet receive queues. | |
4718 | */ | |
4719 | ||
6f912042 | 4720 | for_each_possible_cpu(i) { |
1da177e4 LT |
4721 | struct softnet_data *queue; |
4722 | ||
4723 | queue = &per_cpu(softnet_data, i); | |
4724 | skb_queue_head_init(&queue->input_pkt_queue); | |
1da177e4 LT |
4725 | queue->completion_queue = NULL; |
4726 | INIT_LIST_HEAD(&queue->poll_list); | |
bea3348e SH |
4727 | |
4728 | queue->backlog.poll = process_backlog; | |
4729 | queue->backlog.weight = weight_p; | |
1da177e4 LT |
4730 | } |
4731 | ||
db217334 CL |
4732 | netdev_dma_register(); |
4733 | ||
1da177e4 LT |
4734 | dev_boot_phase = 0; |
4735 | ||
4736 | open_softirq(NET_TX_SOFTIRQ, net_tx_action, NULL); | |
4737 | open_softirq(NET_RX_SOFTIRQ, net_rx_action, NULL); | |
4738 | ||
4739 | hotcpu_notifier(dev_cpu_callback, 0); | |
4740 | dst_init(); | |
4741 | dev_mcast_init(); | |
4742 | rc = 0; | |
4743 | out: | |
4744 | return rc; | |
4745 | } | |
4746 | ||
4747 | subsys_initcall(net_dev_init); | |
4748 | ||
4749 | EXPORT_SYMBOL(__dev_get_by_index); | |
4750 | EXPORT_SYMBOL(__dev_get_by_name); | |
4751 | EXPORT_SYMBOL(__dev_remove_pack); | |
c2373ee9 | 4752 | EXPORT_SYMBOL(dev_valid_name); |
1da177e4 LT |
4753 | EXPORT_SYMBOL(dev_add_pack); |
4754 | EXPORT_SYMBOL(dev_alloc_name); | |
4755 | EXPORT_SYMBOL(dev_close); | |
4756 | EXPORT_SYMBOL(dev_get_by_flags); | |
4757 | EXPORT_SYMBOL(dev_get_by_index); | |
4758 | EXPORT_SYMBOL(dev_get_by_name); | |
1da177e4 LT |
4759 | EXPORT_SYMBOL(dev_open); |
4760 | EXPORT_SYMBOL(dev_queue_xmit); | |
4761 | EXPORT_SYMBOL(dev_remove_pack); | |
4762 | EXPORT_SYMBOL(dev_set_allmulti); | |
4763 | EXPORT_SYMBOL(dev_set_promiscuity); | |
4764 | EXPORT_SYMBOL(dev_change_flags); | |
4765 | EXPORT_SYMBOL(dev_set_mtu); | |
4766 | EXPORT_SYMBOL(dev_set_mac_address); | |
4767 | EXPORT_SYMBOL(free_netdev); | |
4768 | EXPORT_SYMBOL(netdev_boot_setup_check); | |
4769 | EXPORT_SYMBOL(netdev_set_master); | |
4770 | EXPORT_SYMBOL(netdev_state_change); | |
4771 | EXPORT_SYMBOL(netif_receive_skb); | |
4772 | EXPORT_SYMBOL(netif_rx); | |
4773 | EXPORT_SYMBOL(register_gifconf); | |
4774 | EXPORT_SYMBOL(register_netdevice); | |
4775 | EXPORT_SYMBOL(register_netdevice_notifier); | |
4776 | EXPORT_SYMBOL(skb_checksum_help); | |
4777 | EXPORT_SYMBOL(synchronize_net); | |
4778 | EXPORT_SYMBOL(unregister_netdevice); | |
4779 | EXPORT_SYMBOL(unregister_netdevice_notifier); | |
4780 | EXPORT_SYMBOL(net_enable_timestamp); | |
4781 | EXPORT_SYMBOL(net_disable_timestamp); | |
4782 | EXPORT_SYMBOL(dev_get_flags); | |
4783 | ||
4784 | #if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE) | |
4785 | EXPORT_SYMBOL(br_handle_frame_hook); | |
4786 | EXPORT_SYMBOL(br_fdb_get_hook); | |
4787 | EXPORT_SYMBOL(br_fdb_put_hook); | |
4788 | #endif | |
4789 | ||
4790 | #ifdef CONFIG_KMOD | |
4791 | EXPORT_SYMBOL(dev_load); | |
4792 | #endif | |
4793 | ||
4794 | EXPORT_PER_CPU_SYMBOL(softnet_data); |