]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NET3 Protocol independent device support routines. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public License | |
6 | * as published by the Free Software Foundation; either version | |
7 | * 2 of the License, or (at your option) any later version. | |
8 | * | |
9 | * Derived from the non IP parts of dev.c 1.0.19 | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
12 | * Mark Evans, <evansmp@uhura.aston.ac.uk> | |
13 | * | |
14 | * Additional Authors: | |
15 | * Florian la Roche <rzsfl@rz.uni-sb.de> | |
16 | * Alan Cox <gw4pts@gw4pts.ampr.org> | |
17 | * David Hinds <dahinds@users.sourceforge.net> | |
18 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> | |
19 | * Adam Sulmicki <adam@cfar.umd.edu> | |
20 | * Pekka Riikonen <priikone@poesidon.pspt.fi> | |
21 | * | |
22 | * Changes: | |
23 | * D.J. Barrow : Fixed bug where dev->refcnt gets set | |
24 | * to 2 if register_netdev gets called | |
25 | * before net_dev_init & also removed a | |
26 | * few lines of code in the process. | |
27 | * Alan Cox : device private ioctl copies fields back. | |
28 | * Alan Cox : Transmit queue code does relevant | |
29 | * stunts to keep the queue safe. | |
30 | * Alan Cox : Fixed double lock. | |
31 | * Alan Cox : Fixed promisc NULL pointer trap | |
32 | * ???????? : Support the full private ioctl range | |
33 | * Alan Cox : Moved ioctl permission check into | |
34 | * drivers | |
35 | * Tim Kordas : SIOCADDMULTI/SIOCDELMULTI | |
36 | * Alan Cox : 100 backlog just doesn't cut it when | |
37 | * you start doing multicast video 8) | |
38 | * Alan Cox : Rewrote net_bh and list manager. | |
39 | * Alan Cox : Fix ETH_P_ALL echoback lengths. | |
40 | * Alan Cox : Took out transmit every packet pass | |
41 | * Saved a few bytes in the ioctl handler | |
42 | * Alan Cox : Network driver sets packet type before | |
43 | * calling netif_rx. Saves a function | |
44 | * call a packet. | |
45 | * Alan Cox : Hashed net_bh() | |
46 | * Richard Kooijman: Timestamp fixes. | |
47 | * Alan Cox : Wrong field in SIOCGIFDSTADDR | |
48 | * Alan Cox : Device lock protection. | |
49 | * Alan Cox : Fixed nasty side effect of device close | |
50 | * changes. | |
51 | * Rudi Cilibrasi : Pass the right thing to | |
52 | * set_mac_address() | |
53 | * Dave Miller : 32bit quantity for the device lock to | |
54 | * make it work out on a Sparc. | |
55 | * Bjorn Ekwall : Added KERNELD hack. | |
56 | * Alan Cox : Cleaned up the backlog initialise. | |
57 | * Craig Metz : SIOCGIFCONF fix if space for under | |
58 | * 1 device. | |
59 | * Thomas Bogendoerfer : Return ENODEV for dev_open, if there | |
60 | * is no device open function. | |
61 | * Andi Kleen : Fix error reporting for SIOCGIFCONF | |
62 | * Michael Chastain : Fix signed/unsigned for SIOCGIFCONF | |
63 | * Cyrus Durgin : Cleaned for KMOD | |
64 | * Adam Sulmicki : Bug Fix : Network Device Unload | |
65 | * A network device unload needs to purge | |
66 | * the backlog queue. | |
67 | * Paul Rusty Russell : SIOCSIFNAME | |
68 | * Pekka Riikonen : Netdev boot-time settings code | |
69 | * Andrew Morton : Make unregister_netdevice wait | |
70 | * indefinitely on dev->refcnt | |
71 | * J Hadi Salim : - Backlog queue sampling | |
72 | * - netif_rx() feedback | |
73 | */ | |
74 | ||
75 | #include <asm/uaccess.h> | |
76 | #include <asm/system.h> | |
77 | #include <linux/bitops.h> | |
4fc268d2 | 78 | #include <linux/capability.h> |
1da177e4 LT |
79 | #include <linux/cpu.h> |
80 | #include <linux/types.h> | |
81 | #include <linux/kernel.h> | |
82 | #include <linux/sched.h> | |
4a3e2f71 | 83 | #include <linux/mutex.h> |
1da177e4 LT |
84 | #include <linux/string.h> |
85 | #include <linux/mm.h> | |
86 | #include <linux/socket.h> | |
87 | #include <linux/sockios.h> | |
88 | #include <linux/errno.h> | |
89 | #include <linux/interrupt.h> | |
90 | #include <linux/if_ether.h> | |
91 | #include <linux/netdevice.h> | |
92 | #include <linux/etherdevice.h> | |
0187bdfb | 93 | #include <linux/ethtool.h> |
1da177e4 LT |
94 | #include <linux/notifier.h> |
95 | #include <linux/skbuff.h> | |
457c4cbc | 96 | #include <net/net_namespace.h> |
1da177e4 LT |
97 | #include <net/sock.h> |
98 | #include <linux/rtnetlink.h> | |
99 | #include <linux/proc_fs.h> | |
100 | #include <linux/seq_file.h> | |
101 | #include <linux/stat.h> | |
102 | #include <linux/if_bridge.h> | |
b863ceb7 | 103 | #include <linux/if_macvlan.h> |
1da177e4 LT |
104 | #include <net/dst.h> |
105 | #include <net/pkt_sched.h> | |
106 | #include <net/checksum.h> | |
107 | #include <linux/highmem.h> | |
108 | #include <linux/init.h> | |
109 | #include <linux/kmod.h> | |
110 | #include <linux/module.h> | |
1da177e4 LT |
111 | #include <linux/netpoll.h> |
112 | #include <linux/rcupdate.h> | |
113 | #include <linux/delay.h> | |
295f4a1f | 114 | #include <net/wext.h> |
1da177e4 | 115 | #include <net/iw_handler.h> |
1da177e4 | 116 | #include <asm/current.h> |
5bdb9886 | 117 | #include <linux/audit.h> |
db217334 | 118 | #include <linux/dmaengine.h> |
f6a78bfc | 119 | #include <linux/err.h> |
c7fa9d18 | 120 | #include <linux/ctype.h> |
723e98b7 | 121 | #include <linux/if_arp.h> |
6de329e2 | 122 | #include <linux/if_vlan.h> |
8f0f2223 | 123 | #include <linux/ip.h> |
ad55dcaf | 124 | #include <net/ip.h> |
8f0f2223 DM |
125 | #include <linux/ipv6.h> |
126 | #include <linux/in.h> | |
b6b2fed1 DM |
127 | #include <linux/jhash.h> |
128 | #include <linux/random.h> | |
1da177e4 | 129 | |
342709ef PE |
130 | #include "net-sysfs.h" |
131 | ||
d565b0a1 HX |
132 | /* Instead of increasing this, you should create a hash table. */ |
133 | #define MAX_GRO_SKBS 8 | |
134 | ||
5d38a079 HX |
135 | /* This should be increased if a protocol with a bigger head is added. */ |
136 | #define GRO_MAX_HEAD (MAX_HEADER + 128) | |
137 | ||
1da177e4 LT |
138 | /* |
139 | * The list of packet types we will receive (as opposed to discard) | |
140 | * and the routines to invoke. | |
141 | * | |
142 | * Why 16. Because with 16 the only overlap we get on a hash of the | |
143 | * low nibble of the protocol value is RARP/SNAP/X.25. | |
144 | * | |
145 | * NOTE: That is no longer true with the addition of VLAN tags. Not | |
146 | * sure which should go first, but I bet it won't make much | |
147 | * difference if we are running VLANs. The good news is that | |
148 | * this protocol won't be in the list unless compiled in, so | |
3041a069 | 149 | * the average user (w/out VLANs) will not be adversely affected. |
1da177e4 LT |
150 | * --BLG |
151 | * | |
152 | * 0800 IP | |
153 | * 8100 802.1Q VLAN | |
154 | * 0001 802.3 | |
155 | * 0002 AX.25 | |
156 | * 0004 802.2 | |
157 | * 8035 RARP | |
158 | * 0005 SNAP | |
159 | * 0805 X.25 | |
160 | * 0806 ARP | |
161 | * 8137 IPX | |
162 | * 0009 Localtalk | |
163 | * 86DD IPv6 | |
164 | */ | |
165 | ||
82d8a867 PE |
166 | #define PTYPE_HASH_SIZE (16) |
167 | #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) | |
168 | ||
1da177e4 | 169 | static DEFINE_SPINLOCK(ptype_lock); |
82d8a867 | 170 | static struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; |
6b2bedc3 | 171 | static struct list_head ptype_all __read_mostly; /* Taps */ |
1da177e4 | 172 | |
1da177e4 | 173 | /* |
7562f876 | 174 | * The @dev_base_head list is protected by @dev_base_lock and the rtnl |
1da177e4 LT |
175 | * semaphore. |
176 | * | |
177 | * Pure readers hold dev_base_lock for reading. | |
178 | * | |
179 | * Writers must hold the rtnl semaphore while they loop through the | |
7562f876 | 180 | * dev_base_head list, and hold dev_base_lock for writing when they do the |
1da177e4 LT |
181 | * actual updates. This allows pure readers to access the list even |
182 | * while a writer is preparing to update it. | |
183 | * | |
184 | * To put it another way, dev_base_lock is held for writing only to | |
185 | * protect against pure readers; the rtnl semaphore provides the | |
186 | * protection against other writers. | |
187 | * | |
188 | * See, for example usages, register_netdevice() and | |
189 | * unregister_netdevice(), which must be called with the rtnl | |
190 | * semaphore held. | |
191 | */ | |
1da177e4 LT |
192 | DEFINE_RWLOCK(dev_base_lock); |
193 | ||
1da177e4 LT |
194 | EXPORT_SYMBOL(dev_base_lock); |
195 | ||
196 | #define NETDEV_HASHBITS 8 | |
881d966b | 197 | #define NETDEV_HASHENTRIES (1 << NETDEV_HASHBITS) |
1da177e4 | 198 | |
881d966b | 199 | static inline struct hlist_head *dev_name_hash(struct net *net, const char *name) |
1da177e4 LT |
200 | { |
201 | unsigned hash = full_name_hash(name, strnlen(name, IFNAMSIZ)); | |
881d966b | 202 | return &net->dev_name_head[hash & ((1 << NETDEV_HASHBITS) - 1)]; |
1da177e4 LT |
203 | } |
204 | ||
881d966b | 205 | static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex) |
1da177e4 | 206 | { |
881d966b | 207 | return &net->dev_index_head[ifindex & ((1 << NETDEV_HASHBITS) - 1)]; |
1da177e4 LT |
208 | } |
209 | ||
ce286d32 EB |
210 | /* Device list insertion */ |
211 | static int list_netdevice(struct net_device *dev) | |
212 | { | |
c346dca1 | 213 | struct net *net = dev_net(dev); |
ce286d32 EB |
214 | |
215 | ASSERT_RTNL(); | |
216 | ||
217 | write_lock_bh(&dev_base_lock); | |
218 | list_add_tail(&dev->dev_list, &net->dev_base_head); | |
219 | hlist_add_head(&dev->name_hlist, dev_name_hash(net, dev->name)); | |
220 | hlist_add_head(&dev->index_hlist, dev_index_hash(net, dev->ifindex)); | |
221 | write_unlock_bh(&dev_base_lock); | |
222 | return 0; | |
223 | } | |
224 | ||
225 | /* Device list removal */ | |
226 | static void unlist_netdevice(struct net_device *dev) | |
227 | { | |
228 | ASSERT_RTNL(); | |
229 | ||
230 | /* Unlink dev from the device chain */ | |
231 | write_lock_bh(&dev_base_lock); | |
232 | list_del(&dev->dev_list); | |
233 | hlist_del(&dev->name_hlist); | |
234 | hlist_del(&dev->index_hlist); | |
235 | write_unlock_bh(&dev_base_lock); | |
236 | } | |
237 | ||
1da177e4 LT |
238 | /* |
239 | * Our notifier list | |
240 | */ | |
241 | ||
f07d5b94 | 242 | static RAW_NOTIFIER_HEAD(netdev_chain); |
1da177e4 LT |
243 | |
244 | /* | |
245 | * Device drivers call our routines to queue packets here. We empty the | |
246 | * queue in the local softnet handler. | |
247 | */ | |
bea3348e SH |
248 | |
249 | DEFINE_PER_CPU(struct softnet_data, softnet_data); | |
1da177e4 | 250 | |
cf508b12 | 251 | #ifdef CONFIG_LOCKDEP |
723e98b7 | 252 | /* |
c773e847 | 253 | * register_netdevice() inits txq->_xmit_lock and sets lockdep class |
723e98b7 JP |
254 | * according to dev->type |
255 | */ | |
256 | static const unsigned short netdev_lock_type[] = | |
257 | {ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25, | |
258 | ARPHRD_PRONET, ARPHRD_CHAOS, ARPHRD_IEEE802, ARPHRD_ARCNET, | |
259 | ARPHRD_APPLETLK, ARPHRD_DLCI, ARPHRD_ATM, ARPHRD_METRICOM, | |
260 | ARPHRD_IEEE1394, ARPHRD_EUI64, ARPHRD_INFINIBAND, ARPHRD_SLIP, | |
261 | ARPHRD_CSLIP, ARPHRD_SLIP6, ARPHRD_CSLIP6, ARPHRD_RSRVD, | |
262 | ARPHRD_ADAPT, ARPHRD_ROSE, ARPHRD_X25, ARPHRD_HWX25, | |
263 | ARPHRD_PPP, ARPHRD_CISCO, ARPHRD_LAPB, ARPHRD_DDCMP, | |
264 | ARPHRD_RAWHDLC, ARPHRD_TUNNEL, ARPHRD_TUNNEL6, ARPHRD_FRAD, | |
265 | ARPHRD_SKIP, ARPHRD_LOOPBACK, ARPHRD_LOCALTLK, ARPHRD_FDDI, | |
266 | ARPHRD_BIF, ARPHRD_SIT, ARPHRD_IPDDP, ARPHRD_IPGRE, | |
267 | ARPHRD_PIMREG, ARPHRD_HIPPI, ARPHRD_ASH, ARPHRD_ECONET, | |
268 | ARPHRD_IRDA, ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL, | |
269 | ARPHRD_FCFABRIC, ARPHRD_IEEE802_TR, ARPHRD_IEEE80211, | |
2d91d78b | 270 | ARPHRD_IEEE80211_PRISM, ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, |
57c81fff | 271 | ARPHRD_PHONET_PIPE, ARPHRD_VOID, ARPHRD_NONE}; |
723e98b7 JP |
272 | |
273 | static const char *netdev_lock_name[] = | |
274 | {"_xmit_NETROM", "_xmit_ETHER", "_xmit_EETHER", "_xmit_AX25", | |
275 | "_xmit_PRONET", "_xmit_CHAOS", "_xmit_IEEE802", "_xmit_ARCNET", | |
276 | "_xmit_APPLETLK", "_xmit_DLCI", "_xmit_ATM", "_xmit_METRICOM", | |
277 | "_xmit_IEEE1394", "_xmit_EUI64", "_xmit_INFINIBAND", "_xmit_SLIP", | |
278 | "_xmit_CSLIP", "_xmit_SLIP6", "_xmit_CSLIP6", "_xmit_RSRVD", | |
279 | "_xmit_ADAPT", "_xmit_ROSE", "_xmit_X25", "_xmit_HWX25", | |
280 | "_xmit_PPP", "_xmit_CISCO", "_xmit_LAPB", "_xmit_DDCMP", | |
281 | "_xmit_RAWHDLC", "_xmit_TUNNEL", "_xmit_TUNNEL6", "_xmit_FRAD", | |
282 | "_xmit_SKIP", "_xmit_LOOPBACK", "_xmit_LOCALTLK", "_xmit_FDDI", | |
283 | "_xmit_BIF", "_xmit_SIT", "_xmit_IPDDP", "_xmit_IPGRE", | |
284 | "_xmit_PIMREG", "_xmit_HIPPI", "_xmit_ASH", "_xmit_ECONET", | |
285 | "_xmit_IRDA", "_xmit_FCPP", "_xmit_FCAL", "_xmit_FCPL", | |
286 | "_xmit_FCFABRIC", "_xmit_IEEE802_TR", "_xmit_IEEE80211", | |
2d91d78b | 287 | "_xmit_IEEE80211_PRISM", "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", |
57c81fff | 288 | "_xmit_PHONET_PIPE", "_xmit_VOID", "_xmit_NONE"}; |
723e98b7 JP |
289 | |
290 | static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)]; | |
cf508b12 | 291 | static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)]; |
723e98b7 JP |
292 | |
293 | static inline unsigned short netdev_lock_pos(unsigned short dev_type) | |
294 | { | |
295 | int i; | |
296 | ||
297 | for (i = 0; i < ARRAY_SIZE(netdev_lock_type); i++) | |
298 | if (netdev_lock_type[i] == dev_type) | |
299 | return i; | |
300 | /* the last key is used by default */ | |
301 | return ARRAY_SIZE(netdev_lock_type) - 1; | |
302 | } | |
303 | ||
cf508b12 DM |
304 | static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock, |
305 | unsigned short dev_type) | |
723e98b7 JP |
306 | { |
307 | int i; | |
308 | ||
309 | i = netdev_lock_pos(dev_type); | |
310 | lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i], | |
311 | netdev_lock_name[i]); | |
312 | } | |
cf508b12 DM |
313 | |
314 | static inline void netdev_set_addr_lockdep_class(struct net_device *dev) | |
315 | { | |
316 | int i; | |
317 | ||
318 | i = netdev_lock_pos(dev->type); | |
319 | lockdep_set_class_and_name(&dev->addr_list_lock, | |
320 | &netdev_addr_lock_key[i], | |
321 | netdev_lock_name[i]); | |
322 | } | |
723e98b7 | 323 | #else |
cf508b12 DM |
324 | static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock, |
325 | unsigned short dev_type) | |
326 | { | |
327 | } | |
328 | static inline void netdev_set_addr_lockdep_class(struct net_device *dev) | |
723e98b7 JP |
329 | { |
330 | } | |
331 | #endif | |
1da177e4 LT |
332 | |
333 | /******************************************************************************* | |
334 | ||
335 | Protocol management and registration routines | |
336 | ||
337 | *******************************************************************************/ | |
338 | ||
1da177e4 LT |
339 | /* |
340 | * Add a protocol ID to the list. Now that the input handler is | |
341 | * smarter we can dispense with all the messy stuff that used to be | |
342 | * here. | |
343 | * | |
344 | * BEWARE!!! Protocol handlers, mangling input packets, | |
345 | * MUST BE last in hash buckets and checking protocol handlers | |
346 | * MUST start from promiscuous ptype_all chain in net_bh. | |
347 | * It is true now, do not change it. | |
348 | * Explanation follows: if protocol handler, mangling packet, will | |
349 | * be the first on list, it is not able to sense, that packet | |
350 | * is cloned and should be copied-on-write, so that it will | |
351 | * change it and subsequent readers will get broken packet. | |
352 | * --ANK (980803) | |
353 | */ | |
354 | ||
355 | /** | |
356 | * dev_add_pack - add packet handler | |
357 | * @pt: packet type declaration | |
358 | * | |
359 | * Add a protocol handler to the networking stack. The passed &packet_type | |
360 | * is linked into kernel lists and may not be freed until it has been | |
361 | * removed from the kernel lists. | |
362 | * | |
4ec93edb | 363 | * This call does not sleep therefore it can not |
1da177e4 LT |
364 | * guarantee all CPU's that are in middle of receiving packets |
365 | * will see the new packet type (until the next received packet). | |
366 | */ | |
367 | ||
368 | void dev_add_pack(struct packet_type *pt) | |
369 | { | |
370 | int hash; | |
371 | ||
372 | spin_lock_bh(&ptype_lock); | |
9be9a6b9 | 373 | if (pt->type == htons(ETH_P_ALL)) |
1da177e4 | 374 | list_add_rcu(&pt->list, &ptype_all); |
9be9a6b9 | 375 | else { |
82d8a867 | 376 | hash = ntohs(pt->type) & PTYPE_HASH_MASK; |
1da177e4 LT |
377 | list_add_rcu(&pt->list, &ptype_base[hash]); |
378 | } | |
379 | spin_unlock_bh(&ptype_lock); | |
380 | } | |
381 | ||
1da177e4 LT |
382 | /** |
383 | * __dev_remove_pack - remove packet handler | |
384 | * @pt: packet type declaration | |
385 | * | |
386 | * Remove a protocol handler that was previously added to the kernel | |
387 | * protocol handlers by dev_add_pack(). The passed &packet_type is removed | |
388 | * from the kernel lists and can be freed or reused once this function | |
4ec93edb | 389 | * returns. |
1da177e4 LT |
390 | * |
391 | * The packet type might still be in use by receivers | |
392 | * and must not be freed until after all the CPU's have gone | |
393 | * through a quiescent state. | |
394 | */ | |
395 | void __dev_remove_pack(struct packet_type *pt) | |
396 | { | |
397 | struct list_head *head; | |
398 | struct packet_type *pt1; | |
399 | ||
400 | spin_lock_bh(&ptype_lock); | |
401 | ||
9be9a6b9 | 402 | if (pt->type == htons(ETH_P_ALL)) |
1da177e4 | 403 | head = &ptype_all; |
9be9a6b9 | 404 | else |
82d8a867 | 405 | head = &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK]; |
1da177e4 LT |
406 | |
407 | list_for_each_entry(pt1, head, list) { | |
408 | if (pt == pt1) { | |
409 | list_del_rcu(&pt->list); | |
410 | goto out; | |
411 | } | |
412 | } | |
413 | ||
414 | printk(KERN_WARNING "dev_remove_pack: %p not found.\n", pt); | |
415 | out: | |
416 | spin_unlock_bh(&ptype_lock); | |
417 | } | |
418 | /** | |
419 | * dev_remove_pack - remove packet handler | |
420 | * @pt: packet type declaration | |
421 | * | |
422 | * Remove a protocol handler that was previously added to the kernel | |
423 | * protocol handlers by dev_add_pack(). The passed &packet_type is removed | |
424 | * from the kernel lists and can be freed or reused once this function | |
425 | * returns. | |
426 | * | |
427 | * This call sleeps to guarantee that no CPU is looking at the packet | |
428 | * type after return. | |
429 | */ | |
430 | void dev_remove_pack(struct packet_type *pt) | |
431 | { | |
432 | __dev_remove_pack(pt); | |
4ec93edb | 433 | |
1da177e4 LT |
434 | synchronize_net(); |
435 | } | |
436 | ||
437 | /****************************************************************************** | |
438 | ||
439 | Device Boot-time Settings Routines | |
440 | ||
441 | *******************************************************************************/ | |
442 | ||
443 | /* Boot time configuration table */ | |
444 | static struct netdev_boot_setup dev_boot_setup[NETDEV_BOOT_SETUP_MAX]; | |
445 | ||
446 | /** | |
447 | * netdev_boot_setup_add - add new setup entry | |
448 | * @name: name of the device | |
449 | * @map: configured settings for the device | |
450 | * | |
451 | * Adds new setup entry to the dev_boot_setup list. The function | |
452 | * returns 0 on error and 1 on success. This is a generic routine to | |
453 | * all netdevices. | |
454 | */ | |
455 | static int netdev_boot_setup_add(char *name, struct ifmap *map) | |
456 | { | |
457 | struct netdev_boot_setup *s; | |
458 | int i; | |
459 | ||
460 | s = dev_boot_setup; | |
461 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { | |
462 | if (s[i].name[0] == '\0' || s[i].name[0] == ' ') { | |
463 | memset(s[i].name, 0, sizeof(s[i].name)); | |
93b3cff9 | 464 | strlcpy(s[i].name, name, IFNAMSIZ); |
1da177e4 LT |
465 | memcpy(&s[i].map, map, sizeof(s[i].map)); |
466 | break; | |
467 | } | |
468 | } | |
469 | ||
470 | return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1; | |
471 | } | |
472 | ||
473 | /** | |
474 | * netdev_boot_setup_check - check boot time settings | |
475 | * @dev: the netdevice | |
476 | * | |
477 | * Check boot time settings for the device. | |
478 | * The found settings are set for the device to be used | |
479 | * later in the device probing. | |
480 | * Returns 0 if no settings found, 1 if they are. | |
481 | */ | |
482 | int netdev_boot_setup_check(struct net_device *dev) | |
483 | { | |
484 | struct netdev_boot_setup *s = dev_boot_setup; | |
485 | int i; | |
486 | ||
487 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { | |
488 | if (s[i].name[0] != '\0' && s[i].name[0] != ' ' && | |
93b3cff9 | 489 | !strcmp(dev->name, s[i].name)) { |
1da177e4 LT |
490 | dev->irq = s[i].map.irq; |
491 | dev->base_addr = s[i].map.base_addr; | |
492 | dev->mem_start = s[i].map.mem_start; | |
493 | dev->mem_end = s[i].map.mem_end; | |
494 | return 1; | |
495 | } | |
496 | } | |
497 | return 0; | |
498 | } | |
499 | ||
500 | ||
501 | /** | |
502 | * netdev_boot_base - get address from boot time settings | |
503 | * @prefix: prefix for network device | |
504 | * @unit: id for network device | |
505 | * | |
506 | * Check boot time settings for the base address of device. | |
507 | * The found settings are set for the device to be used | |
508 | * later in the device probing. | |
509 | * Returns 0 if no settings found. | |
510 | */ | |
511 | unsigned long netdev_boot_base(const char *prefix, int unit) | |
512 | { | |
513 | const struct netdev_boot_setup *s = dev_boot_setup; | |
514 | char name[IFNAMSIZ]; | |
515 | int i; | |
516 | ||
517 | sprintf(name, "%s%d", prefix, unit); | |
518 | ||
519 | /* | |
520 | * If device already registered then return base of 1 | |
521 | * to indicate not to probe for this interface | |
522 | */ | |
881d966b | 523 | if (__dev_get_by_name(&init_net, name)) |
1da177e4 LT |
524 | return 1; |
525 | ||
526 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) | |
527 | if (!strcmp(name, s[i].name)) | |
528 | return s[i].map.base_addr; | |
529 | return 0; | |
530 | } | |
531 | ||
532 | /* | |
533 | * Saves at boot time configured settings for any netdevice. | |
534 | */ | |
535 | int __init netdev_boot_setup(char *str) | |
536 | { | |
537 | int ints[5]; | |
538 | struct ifmap map; | |
539 | ||
540 | str = get_options(str, ARRAY_SIZE(ints), ints); | |
541 | if (!str || !*str) | |
542 | return 0; | |
543 | ||
544 | /* Save settings */ | |
545 | memset(&map, 0, sizeof(map)); | |
546 | if (ints[0] > 0) | |
547 | map.irq = ints[1]; | |
548 | if (ints[0] > 1) | |
549 | map.base_addr = ints[2]; | |
550 | if (ints[0] > 2) | |
551 | map.mem_start = ints[3]; | |
552 | if (ints[0] > 3) | |
553 | map.mem_end = ints[4]; | |
554 | ||
555 | /* Add new entry to the list */ | |
556 | return netdev_boot_setup_add(str, &map); | |
557 | } | |
558 | ||
559 | __setup("netdev=", netdev_boot_setup); | |
560 | ||
561 | /******************************************************************************* | |
562 | ||
563 | Device Interface Subroutines | |
564 | ||
565 | *******************************************************************************/ | |
566 | ||
567 | /** | |
568 | * __dev_get_by_name - find a device by its name | |
c4ea43c5 | 569 | * @net: the applicable net namespace |
1da177e4 LT |
570 | * @name: name to find |
571 | * | |
572 | * Find an interface by name. Must be called under RTNL semaphore | |
573 | * or @dev_base_lock. If the name is found a pointer to the device | |
574 | * is returned. If the name is not found then %NULL is returned. The | |
575 | * reference counters are not incremented so the caller must be | |
576 | * careful with locks. | |
577 | */ | |
578 | ||
881d966b | 579 | struct net_device *__dev_get_by_name(struct net *net, const char *name) |
1da177e4 LT |
580 | { |
581 | struct hlist_node *p; | |
582 | ||
881d966b | 583 | hlist_for_each(p, dev_name_hash(net, name)) { |
1da177e4 LT |
584 | struct net_device *dev |
585 | = hlist_entry(p, struct net_device, name_hlist); | |
586 | if (!strncmp(dev->name, name, IFNAMSIZ)) | |
587 | return dev; | |
588 | } | |
589 | return NULL; | |
590 | } | |
591 | ||
592 | /** | |
593 | * dev_get_by_name - find a device by its name | |
c4ea43c5 | 594 | * @net: the applicable net namespace |
1da177e4 LT |
595 | * @name: name to find |
596 | * | |
597 | * Find an interface by name. This can be called from any | |
598 | * context and does its own locking. The returned handle has | |
599 | * the usage count incremented and the caller must use dev_put() to | |
600 | * release it when it is no longer needed. %NULL is returned if no | |
601 | * matching device is found. | |
602 | */ | |
603 | ||
881d966b | 604 | struct net_device *dev_get_by_name(struct net *net, const char *name) |
1da177e4 LT |
605 | { |
606 | struct net_device *dev; | |
607 | ||
608 | read_lock(&dev_base_lock); | |
881d966b | 609 | dev = __dev_get_by_name(net, name); |
1da177e4 LT |
610 | if (dev) |
611 | dev_hold(dev); | |
612 | read_unlock(&dev_base_lock); | |
613 | return dev; | |
614 | } | |
615 | ||
616 | /** | |
617 | * __dev_get_by_index - find a device by its ifindex | |
c4ea43c5 | 618 | * @net: the applicable net namespace |
1da177e4 LT |
619 | * @ifindex: index of device |
620 | * | |
621 | * Search for an interface by index. Returns %NULL if the device | |
622 | * is not found or a pointer to the device. The device has not | |
623 | * had its reference counter increased so the caller must be careful | |
624 | * about locking. The caller must hold either the RTNL semaphore | |
625 | * or @dev_base_lock. | |
626 | */ | |
627 | ||
881d966b | 628 | struct net_device *__dev_get_by_index(struct net *net, int ifindex) |
1da177e4 LT |
629 | { |
630 | struct hlist_node *p; | |
631 | ||
881d966b | 632 | hlist_for_each(p, dev_index_hash(net, ifindex)) { |
1da177e4 LT |
633 | struct net_device *dev |
634 | = hlist_entry(p, struct net_device, index_hlist); | |
635 | if (dev->ifindex == ifindex) | |
636 | return dev; | |
637 | } | |
638 | return NULL; | |
639 | } | |
640 | ||
641 | ||
642 | /** | |
643 | * dev_get_by_index - find a device by its ifindex | |
c4ea43c5 | 644 | * @net: the applicable net namespace |
1da177e4 LT |
645 | * @ifindex: index of device |
646 | * | |
647 | * Search for an interface by index. Returns NULL if the device | |
648 | * is not found or a pointer to the device. The device returned has | |
649 | * had a reference added and the pointer is safe until the user calls | |
650 | * dev_put to indicate they have finished with it. | |
651 | */ | |
652 | ||
881d966b | 653 | struct net_device *dev_get_by_index(struct net *net, int ifindex) |
1da177e4 LT |
654 | { |
655 | struct net_device *dev; | |
656 | ||
657 | read_lock(&dev_base_lock); | |
881d966b | 658 | dev = __dev_get_by_index(net, ifindex); |
1da177e4 LT |
659 | if (dev) |
660 | dev_hold(dev); | |
661 | read_unlock(&dev_base_lock); | |
662 | return dev; | |
663 | } | |
664 | ||
665 | /** | |
666 | * dev_getbyhwaddr - find a device by its hardware address | |
c4ea43c5 | 667 | * @net: the applicable net namespace |
1da177e4 LT |
668 | * @type: media type of device |
669 | * @ha: hardware address | |
670 | * | |
671 | * Search for an interface by MAC address. Returns NULL if the device | |
672 | * is not found or a pointer to the device. The caller must hold the | |
673 | * rtnl semaphore. The returned device has not had its ref count increased | |
674 | * and the caller must therefore be careful about locking | |
675 | * | |
676 | * BUGS: | |
677 | * If the API was consistent this would be __dev_get_by_hwaddr | |
678 | */ | |
679 | ||
881d966b | 680 | struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, char *ha) |
1da177e4 LT |
681 | { |
682 | struct net_device *dev; | |
683 | ||
684 | ASSERT_RTNL(); | |
685 | ||
81103a52 | 686 | for_each_netdev(net, dev) |
1da177e4 LT |
687 | if (dev->type == type && |
688 | !memcmp(dev->dev_addr, ha, dev->addr_len)) | |
7562f876 PE |
689 | return dev; |
690 | ||
691 | return NULL; | |
1da177e4 LT |
692 | } |
693 | ||
cf309e3f JF |
694 | EXPORT_SYMBOL(dev_getbyhwaddr); |
695 | ||
881d966b | 696 | struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type) |
1da177e4 LT |
697 | { |
698 | struct net_device *dev; | |
699 | ||
4e9cac2b | 700 | ASSERT_RTNL(); |
881d966b | 701 | for_each_netdev(net, dev) |
4e9cac2b | 702 | if (dev->type == type) |
7562f876 PE |
703 | return dev; |
704 | ||
705 | return NULL; | |
4e9cac2b PM |
706 | } |
707 | ||
708 | EXPORT_SYMBOL(__dev_getfirstbyhwtype); | |
709 | ||
881d966b | 710 | struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type) |
4e9cac2b PM |
711 | { |
712 | struct net_device *dev; | |
713 | ||
714 | rtnl_lock(); | |
881d966b | 715 | dev = __dev_getfirstbyhwtype(net, type); |
4e9cac2b PM |
716 | if (dev) |
717 | dev_hold(dev); | |
1da177e4 LT |
718 | rtnl_unlock(); |
719 | return dev; | |
720 | } | |
721 | ||
722 | EXPORT_SYMBOL(dev_getfirstbyhwtype); | |
723 | ||
724 | /** | |
725 | * dev_get_by_flags - find any device with given flags | |
c4ea43c5 | 726 | * @net: the applicable net namespace |
1da177e4 LT |
727 | * @if_flags: IFF_* values |
728 | * @mask: bitmask of bits in if_flags to check | |
729 | * | |
730 | * Search for any interface with the given flags. Returns NULL if a device | |
4ec93edb | 731 | * is not found or a pointer to the device. The device returned has |
1da177e4 LT |
732 | * had a reference added and the pointer is safe until the user calls |
733 | * dev_put to indicate they have finished with it. | |
734 | */ | |
735 | ||
881d966b | 736 | struct net_device * dev_get_by_flags(struct net *net, unsigned short if_flags, unsigned short mask) |
1da177e4 | 737 | { |
7562f876 | 738 | struct net_device *dev, *ret; |
1da177e4 | 739 | |
7562f876 | 740 | ret = NULL; |
1da177e4 | 741 | read_lock(&dev_base_lock); |
881d966b | 742 | for_each_netdev(net, dev) { |
1da177e4 LT |
743 | if (((dev->flags ^ if_flags) & mask) == 0) { |
744 | dev_hold(dev); | |
7562f876 | 745 | ret = dev; |
1da177e4 LT |
746 | break; |
747 | } | |
748 | } | |
749 | read_unlock(&dev_base_lock); | |
7562f876 | 750 | return ret; |
1da177e4 LT |
751 | } |
752 | ||
753 | /** | |
754 | * dev_valid_name - check if name is okay for network device | |
755 | * @name: name string | |
756 | * | |
757 | * Network device names need to be valid file names to | |
c7fa9d18 DM |
758 | * to allow sysfs to work. We also disallow any kind of |
759 | * whitespace. | |
1da177e4 | 760 | */ |
c2373ee9 | 761 | int dev_valid_name(const char *name) |
1da177e4 | 762 | { |
c7fa9d18 DM |
763 | if (*name == '\0') |
764 | return 0; | |
b6fe17d6 SH |
765 | if (strlen(name) >= IFNAMSIZ) |
766 | return 0; | |
c7fa9d18 DM |
767 | if (!strcmp(name, ".") || !strcmp(name, "..")) |
768 | return 0; | |
769 | ||
770 | while (*name) { | |
771 | if (*name == '/' || isspace(*name)) | |
772 | return 0; | |
773 | name++; | |
774 | } | |
775 | return 1; | |
1da177e4 LT |
776 | } |
777 | ||
778 | /** | |
b267b179 EB |
779 | * __dev_alloc_name - allocate a name for a device |
780 | * @net: network namespace to allocate the device name in | |
1da177e4 | 781 | * @name: name format string |
b267b179 | 782 | * @buf: scratch buffer and result name string |
1da177e4 LT |
783 | * |
784 | * Passed a format string - eg "lt%d" it will try and find a suitable | |
3041a069 SH |
785 | * id. It scans list of devices to build up a free map, then chooses |
786 | * the first empty slot. The caller must hold the dev_base or rtnl lock | |
787 | * while allocating the name and adding the device in order to avoid | |
788 | * duplicates. | |
789 | * Limited to bits_per_byte * page size devices (ie 32K on most platforms). | |
790 | * Returns the number of the unit assigned or a negative errno code. | |
1da177e4 LT |
791 | */ |
792 | ||
b267b179 | 793 | static int __dev_alloc_name(struct net *net, const char *name, char *buf) |
1da177e4 LT |
794 | { |
795 | int i = 0; | |
1da177e4 LT |
796 | const char *p; |
797 | const int max_netdevices = 8*PAGE_SIZE; | |
cfcabdcc | 798 | unsigned long *inuse; |
1da177e4 LT |
799 | struct net_device *d; |
800 | ||
801 | p = strnchr(name, IFNAMSIZ-1, '%'); | |
802 | if (p) { | |
803 | /* | |
804 | * Verify the string as this thing may have come from | |
805 | * the user. There must be either one "%d" and no other "%" | |
806 | * characters. | |
807 | */ | |
808 | if (p[1] != 'd' || strchr(p + 2, '%')) | |
809 | return -EINVAL; | |
810 | ||
811 | /* Use one page as a bit array of possible slots */ | |
cfcabdcc | 812 | inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC); |
1da177e4 LT |
813 | if (!inuse) |
814 | return -ENOMEM; | |
815 | ||
881d966b | 816 | for_each_netdev(net, d) { |
1da177e4 LT |
817 | if (!sscanf(d->name, name, &i)) |
818 | continue; | |
819 | if (i < 0 || i >= max_netdevices) | |
820 | continue; | |
821 | ||
822 | /* avoid cases where sscanf is not exact inverse of printf */ | |
b267b179 | 823 | snprintf(buf, IFNAMSIZ, name, i); |
1da177e4 LT |
824 | if (!strncmp(buf, d->name, IFNAMSIZ)) |
825 | set_bit(i, inuse); | |
826 | } | |
827 | ||
828 | i = find_first_zero_bit(inuse, max_netdevices); | |
829 | free_page((unsigned long) inuse); | |
830 | } | |
831 | ||
b267b179 EB |
832 | snprintf(buf, IFNAMSIZ, name, i); |
833 | if (!__dev_get_by_name(net, buf)) | |
1da177e4 | 834 | return i; |
1da177e4 LT |
835 | |
836 | /* It is possible to run out of possible slots | |
837 | * when the name is long and there isn't enough space left | |
838 | * for the digits, or if all bits are used. | |
839 | */ | |
840 | return -ENFILE; | |
841 | } | |
842 | ||
b267b179 EB |
843 | /** |
844 | * dev_alloc_name - allocate a name for a device | |
845 | * @dev: device | |
846 | * @name: name format string | |
847 | * | |
848 | * Passed a format string - eg "lt%d" it will try and find a suitable | |
849 | * id. It scans list of devices to build up a free map, then chooses | |
850 | * the first empty slot. The caller must hold the dev_base or rtnl lock | |
851 | * while allocating the name and adding the device in order to avoid | |
852 | * duplicates. | |
853 | * Limited to bits_per_byte * page size devices (ie 32K on most platforms). | |
854 | * Returns the number of the unit assigned or a negative errno code. | |
855 | */ | |
856 | ||
857 | int dev_alloc_name(struct net_device *dev, const char *name) | |
858 | { | |
859 | char buf[IFNAMSIZ]; | |
860 | struct net *net; | |
861 | int ret; | |
862 | ||
c346dca1 YH |
863 | BUG_ON(!dev_net(dev)); |
864 | net = dev_net(dev); | |
b267b179 EB |
865 | ret = __dev_alloc_name(net, name, buf); |
866 | if (ret >= 0) | |
867 | strlcpy(dev->name, buf, IFNAMSIZ); | |
868 | return ret; | |
869 | } | |
870 | ||
1da177e4 LT |
871 | |
872 | /** | |
873 | * dev_change_name - change name of a device | |
874 | * @dev: device | |
875 | * @newname: name (or format string) must be at least IFNAMSIZ | |
876 | * | |
877 | * Change name of a device, can pass format strings "eth%d". | |
878 | * for wildcarding. | |
879 | */ | |
cf04a4c7 | 880 | int dev_change_name(struct net_device *dev, const char *newname) |
1da177e4 | 881 | { |
fcc5a03a | 882 | char oldname[IFNAMSIZ]; |
1da177e4 | 883 | int err = 0; |
fcc5a03a | 884 | int ret; |
881d966b | 885 | struct net *net; |
1da177e4 LT |
886 | |
887 | ASSERT_RTNL(); | |
c346dca1 | 888 | BUG_ON(!dev_net(dev)); |
1da177e4 | 889 | |
c346dca1 | 890 | net = dev_net(dev); |
1da177e4 LT |
891 | if (dev->flags & IFF_UP) |
892 | return -EBUSY; | |
893 | ||
894 | if (!dev_valid_name(newname)) | |
895 | return -EINVAL; | |
896 | ||
c8d90dca SH |
897 | if (strncmp(newname, dev->name, IFNAMSIZ) == 0) |
898 | return 0; | |
899 | ||
fcc5a03a HX |
900 | memcpy(oldname, dev->name, IFNAMSIZ); |
901 | ||
1da177e4 LT |
902 | if (strchr(newname, '%')) { |
903 | err = dev_alloc_name(dev, newname); | |
904 | if (err < 0) | |
905 | return err; | |
1da177e4 | 906 | } |
881d966b | 907 | else if (__dev_get_by_name(net, newname)) |
1da177e4 LT |
908 | return -EEXIST; |
909 | else | |
910 | strlcpy(dev->name, newname, IFNAMSIZ); | |
911 | ||
fcc5a03a | 912 | rollback: |
3891845e EB |
913 | /* For now only devices in the initial network namespace |
914 | * are in sysfs. | |
915 | */ | |
916 | if (net == &init_net) { | |
917 | ret = device_rename(&dev->dev, dev->name); | |
918 | if (ret) { | |
919 | memcpy(dev->name, oldname, IFNAMSIZ); | |
920 | return ret; | |
921 | } | |
dcc99773 | 922 | } |
7f988eab HX |
923 | |
924 | write_lock_bh(&dev_base_lock); | |
92749821 | 925 | hlist_del(&dev->name_hlist); |
881d966b | 926 | hlist_add_head(&dev->name_hlist, dev_name_hash(net, dev->name)); |
7f988eab HX |
927 | write_unlock_bh(&dev_base_lock); |
928 | ||
056925ab | 929 | ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev); |
fcc5a03a HX |
930 | ret = notifier_to_errno(ret); |
931 | ||
932 | if (ret) { | |
933 | if (err) { | |
934 | printk(KERN_ERR | |
935 | "%s: name change rollback failed: %d.\n", | |
936 | dev->name, ret); | |
937 | } else { | |
938 | err = ret; | |
939 | memcpy(dev->name, oldname, IFNAMSIZ); | |
940 | goto rollback; | |
941 | } | |
942 | } | |
1da177e4 LT |
943 | |
944 | return err; | |
945 | } | |
946 | ||
0b815a1a SH |
947 | /** |
948 | * dev_set_alias - change ifalias of a device | |
949 | * @dev: device | |
950 | * @alias: name up to IFALIASZ | |
f0db275a | 951 | * @len: limit of bytes to copy from info |
0b815a1a SH |
952 | * |
953 | * Set ifalias for a device, | |
954 | */ | |
955 | int dev_set_alias(struct net_device *dev, const char *alias, size_t len) | |
956 | { | |
957 | ASSERT_RTNL(); | |
958 | ||
959 | if (len >= IFALIASZ) | |
960 | return -EINVAL; | |
961 | ||
96ca4a2c OH |
962 | if (!len) { |
963 | if (dev->ifalias) { | |
964 | kfree(dev->ifalias); | |
965 | dev->ifalias = NULL; | |
966 | } | |
967 | return 0; | |
968 | } | |
969 | ||
0b815a1a SH |
970 | dev->ifalias = krealloc(dev->ifalias, len+1, GFP_KERNEL); |
971 | if (!dev->ifalias) | |
972 | return -ENOMEM; | |
973 | ||
974 | strlcpy(dev->ifalias, alias, len+1); | |
975 | return len; | |
976 | } | |
977 | ||
978 | ||
d8a33ac4 | 979 | /** |
3041a069 | 980 | * netdev_features_change - device changes features |
d8a33ac4 SH |
981 | * @dev: device to cause notification |
982 | * | |
983 | * Called to indicate a device has changed features. | |
984 | */ | |
985 | void netdev_features_change(struct net_device *dev) | |
986 | { | |
056925ab | 987 | call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev); |
d8a33ac4 SH |
988 | } |
989 | EXPORT_SYMBOL(netdev_features_change); | |
990 | ||
1da177e4 LT |
991 | /** |
992 | * netdev_state_change - device changes state | |
993 | * @dev: device to cause notification | |
994 | * | |
995 | * Called to indicate a device has changed state. This function calls | |
996 | * the notifier chains for netdev_chain and sends a NEWLINK message | |
997 | * to the routing socket. | |
998 | */ | |
999 | void netdev_state_change(struct net_device *dev) | |
1000 | { | |
1001 | if (dev->flags & IFF_UP) { | |
056925ab | 1002 | call_netdevice_notifiers(NETDEV_CHANGE, dev); |
1da177e4 LT |
1003 | rtmsg_ifinfo(RTM_NEWLINK, dev, 0); |
1004 | } | |
1005 | } | |
1006 | ||
c1da4ac7 OG |
1007 | void netdev_bonding_change(struct net_device *dev) |
1008 | { | |
1009 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, dev); | |
1010 | } | |
1011 | EXPORT_SYMBOL(netdev_bonding_change); | |
1012 | ||
1da177e4 LT |
1013 | /** |
1014 | * dev_load - load a network module | |
c4ea43c5 | 1015 | * @net: the applicable net namespace |
1da177e4 LT |
1016 | * @name: name of interface |
1017 | * | |
1018 | * If a network interface is not present and the process has suitable | |
1019 | * privileges this function loads the module. If module loading is not | |
1020 | * available in this kernel then it becomes a nop. | |
1021 | */ | |
1022 | ||
881d966b | 1023 | void dev_load(struct net *net, const char *name) |
1da177e4 | 1024 | { |
4ec93edb | 1025 | struct net_device *dev; |
1da177e4 LT |
1026 | |
1027 | read_lock(&dev_base_lock); | |
881d966b | 1028 | dev = __dev_get_by_name(net, name); |
1da177e4 LT |
1029 | read_unlock(&dev_base_lock); |
1030 | ||
1031 | if (!dev && capable(CAP_SYS_MODULE)) | |
1032 | request_module("%s", name); | |
1033 | } | |
1034 | ||
1da177e4 LT |
1035 | /** |
1036 | * dev_open - prepare an interface for use. | |
1037 | * @dev: device to open | |
1038 | * | |
1039 | * Takes a device from down to up state. The device's private open | |
1040 | * function is invoked and then the multicast lists are loaded. Finally | |
1041 | * the device is moved into the up state and a %NETDEV_UP message is | |
1042 | * sent to the netdev notifier chain. | |
1043 | * | |
1044 | * Calling this function on an active interface is a nop. On a failure | |
1045 | * a negative errno code is returned. | |
1046 | */ | |
1047 | int dev_open(struct net_device *dev) | |
1048 | { | |
d314774c | 1049 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
1050 | int ret = 0; |
1051 | ||
e46b66bc BH |
1052 | ASSERT_RTNL(); |
1053 | ||
1da177e4 LT |
1054 | /* |
1055 | * Is it already up? | |
1056 | */ | |
1057 | ||
1058 | if (dev->flags & IFF_UP) | |
1059 | return 0; | |
1060 | ||
1061 | /* | |
1062 | * Is it even present? | |
1063 | */ | |
1064 | if (!netif_device_present(dev)) | |
1065 | return -ENODEV; | |
1066 | ||
1067 | /* | |
1068 | * Call device private open method | |
1069 | */ | |
1070 | set_bit(__LINK_STATE_START, &dev->state); | |
bada339b | 1071 | |
d314774c SH |
1072 | if (ops->ndo_validate_addr) |
1073 | ret = ops->ndo_validate_addr(dev); | |
bada339b | 1074 | |
d314774c SH |
1075 | if (!ret && ops->ndo_open) |
1076 | ret = ops->ndo_open(dev); | |
1da177e4 | 1077 | |
4ec93edb | 1078 | /* |
1da177e4 LT |
1079 | * If it went open OK then: |
1080 | */ | |
1081 | ||
bada339b JG |
1082 | if (ret) |
1083 | clear_bit(__LINK_STATE_START, &dev->state); | |
1084 | else { | |
1da177e4 LT |
1085 | /* |
1086 | * Set the flags. | |
1087 | */ | |
1088 | dev->flags |= IFF_UP; | |
1089 | ||
649274d9 DW |
1090 | /* |
1091 | * Enable NET_DMA | |
1092 | */ | |
b4bd07c2 | 1093 | net_dmaengine_get(); |
649274d9 | 1094 | |
1da177e4 LT |
1095 | /* |
1096 | * Initialize multicasting status | |
1097 | */ | |
4417da66 | 1098 | dev_set_rx_mode(dev); |
1da177e4 LT |
1099 | |
1100 | /* | |
1101 | * Wakeup transmit queue engine | |
1102 | */ | |
1103 | dev_activate(dev); | |
1104 | ||
1105 | /* | |
1106 | * ... and announce new interface. | |
1107 | */ | |
056925ab | 1108 | call_netdevice_notifiers(NETDEV_UP, dev); |
1da177e4 | 1109 | } |
bada339b | 1110 | |
1da177e4 LT |
1111 | return ret; |
1112 | } | |
1113 | ||
1114 | /** | |
1115 | * dev_close - shutdown an interface. | |
1116 | * @dev: device to shutdown | |
1117 | * | |
1118 | * This function moves an active device into down state. A | |
1119 | * %NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device | |
1120 | * is then deactivated and finally a %NETDEV_DOWN is sent to the notifier | |
1121 | * chain. | |
1122 | */ | |
1123 | int dev_close(struct net_device *dev) | |
1124 | { | |
d314774c | 1125 | const struct net_device_ops *ops = dev->netdev_ops; |
e46b66bc BH |
1126 | ASSERT_RTNL(); |
1127 | ||
9d5010db DM |
1128 | might_sleep(); |
1129 | ||
1da177e4 LT |
1130 | if (!(dev->flags & IFF_UP)) |
1131 | return 0; | |
1132 | ||
1133 | /* | |
1134 | * Tell people we are going down, so that they can | |
1135 | * prepare to death, when device is still operating. | |
1136 | */ | |
056925ab | 1137 | call_netdevice_notifiers(NETDEV_GOING_DOWN, dev); |
1da177e4 | 1138 | |
1da177e4 LT |
1139 | clear_bit(__LINK_STATE_START, &dev->state); |
1140 | ||
1141 | /* Synchronize to scheduled poll. We cannot touch poll list, | |
bea3348e SH |
1142 | * it can be even on different cpu. So just clear netif_running(). |
1143 | * | |
1144 | * dev->stop() will invoke napi_disable() on all of it's | |
1145 | * napi_struct instances on this device. | |
1146 | */ | |
1da177e4 | 1147 | smp_mb__after_clear_bit(); /* Commit netif_running(). */ |
1da177e4 | 1148 | |
d8b2a4d2 ML |
1149 | dev_deactivate(dev); |
1150 | ||
1da177e4 LT |
1151 | /* |
1152 | * Call the device specific close. This cannot fail. | |
1153 | * Only if device is UP | |
1154 | * | |
1155 | * We allow it to be called even after a DETACH hot-plug | |
1156 | * event. | |
1157 | */ | |
d314774c SH |
1158 | if (ops->ndo_stop) |
1159 | ops->ndo_stop(dev); | |
1da177e4 LT |
1160 | |
1161 | /* | |
1162 | * Device is now down. | |
1163 | */ | |
1164 | ||
1165 | dev->flags &= ~IFF_UP; | |
1166 | ||
1167 | /* | |
1168 | * Tell people we are down | |
1169 | */ | |
056925ab | 1170 | call_netdevice_notifiers(NETDEV_DOWN, dev); |
1da177e4 | 1171 | |
649274d9 DW |
1172 | /* |
1173 | * Shutdown NET_DMA | |
1174 | */ | |
b4bd07c2 | 1175 | net_dmaengine_put(); |
649274d9 | 1176 | |
1da177e4 LT |
1177 | return 0; |
1178 | } | |
1179 | ||
1180 | ||
0187bdfb BH |
1181 | /** |
1182 | * dev_disable_lro - disable Large Receive Offload on a device | |
1183 | * @dev: device | |
1184 | * | |
1185 | * Disable Large Receive Offload (LRO) on a net device. Must be | |
1186 | * called under RTNL. This is needed if received packets may be | |
1187 | * forwarded to another interface. | |
1188 | */ | |
1189 | void dev_disable_lro(struct net_device *dev) | |
1190 | { | |
1191 | if (dev->ethtool_ops && dev->ethtool_ops->get_flags && | |
1192 | dev->ethtool_ops->set_flags) { | |
1193 | u32 flags = dev->ethtool_ops->get_flags(dev); | |
1194 | if (flags & ETH_FLAG_LRO) { | |
1195 | flags &= ~ETH_FLAG_LRO; | |
1196 | dev->ethtool_ops->set_flags(dev, flags); | |
1197 | } | |
1198 | } | |
1199 | WARN_ON(dev->features & NETIF_F_LRO); | |
1200 | } | |
1201 | EXPORT_SYMBOL(dev_disable_lro); | |
1202 | ||
1203 | ||
881d966b EB |
1204 | static int dev_boot_phase = 1; |
1205 | ||
1da177e4 LT |
1206 | /* |
1207 | * Device change register/unregister. These are not inline or static | |
1208 | * as we export them to the world. | |
1209 | */ | |
1210 | ||
1211 | /** | |
1212 | * register_netdevice_notifier - register a network notifier block | |
1213 | * @nb: notifier | |
1214 | * | |
1215 | * Register a notifier to be called when network device events occur. | |
1216 | * The notifier passed is linked into the kernel structures and must | |
1217 | * not be reused until it has been unregistered. A negative errno code | |
1218 | * is returned on a failure. | |
1219 | * | |
1220 | * When registered all registration and up events are replayed | |
4ec93edb | 1221 | * to the new notifier to allow device to have a race free |
1da177e4 LT |
1222 | * view of the network device list. |
1223 | */ | |
1224 | ||
1225 | int register_netdevice_notifier(struct notifier_block *nb) | |
1226 | { | |
1227 | struct net_device *dev; | |
fcc5a03a | 1228 | struct net_device *last; |
881d966b | 1229 | struct net *net; |
1da177e4 LT |
1230 | int err; |
1231 | ||
1232 | rtnl_lock(); | |
f07d5b94 | 1233 | err = raw_notifier_chain_register(&netdev_chain, nb); |
fcc5a03a HX |
1234 | if (err) |
1235 | goto unlock; | |
881d966b EB |
1236 | if (dev_boot_phase) |
1237 | goto unlock; | |
1238 | for_each_net(net) { | |
1239 | for_each_netdev(net, dev) { | |
1240 | err = nb->notifier_call(nb, NETDEV_REGISTER, dev); | |
1241 | err = notifier_to_errno(err); | |
1242 | if (err) | |
1243 | goto rollback; | |
1244 | ||
1245 | if (!(dev->flags & IFF_UP)) | |
1246 | continue; | |
1da177e4 | 1247 | |
881d966b EB |
1248 | nb->notifier_call(nb, NETDEV_UP, dev); |
1249 | } | |
1da177e4 | 1250 | } |
fcc5a03a HX |
1251 | |
1252 | unlock: | |
1da177e4 LT |
1253 | rtnl_unlock(); |
1254 | return err; | |
fcc5a03a HX |
1255 | |
1256 | rollback: | |
1257 | last = dev; | |
881d966b EB |
1258 | for_each_net(net) { |
1259 | for_each_netdev(net, dev) { | |
1260 | if (dev == last) | |
1261 | break; | |
fcc5a03a | 1262 | |
881d966b EB |
1263 | if (dev->flags & IFF_UP) { |
1264 | nb->notifier_call(nb, NETDEV_GOING_DOWN, dev); | |
1265 | nb->notifier_call(nb, NETDEV_DOWN, dev); | |
1266 | } | |
1267 | nb->notifier_call(nb, NETDEV_UNREGISTER, dev); | |
fcc5a03a | 1268 | } |
fcc5a03a | 1269 | } |
c67625a1 PE |
1270 | |
1271 | raw_notifier_chain_unregister(&netdev_chain, nb); | |
fcc5a03a | 1272 | goto unlock; |
1da177e4 LT |
1273 | } |
1274 | ||
1275 | /** | |
1276 | * unregister_netdevice_notifier - unregister a network notifier block | |
1277 | * @nb: notifier | |
1278 | * | |
1279 | * Unregister a notifier previously registered by | |
1280 | * register_netdevice_notifier(). The notifier is unlinked into the | |
1281 | * kernel structures and may then be reused. A negative errno code | |
1282 | * is returned on a failure. | |
1283 | */ | |
1284 | ||
1285 | int unregister_netdevice_notifier(struct notifier_block *nb) | |
1286 | { | |
9f514950 HX |
1287 | int err; |
1288 | ||
1289 | rtnl_lock(); | |
f07d5b94 | 1290 | err = raw_notifier_chain_unregister(&netdev_chain, nb); |
9f514950 HX |
1291 | rtnl_unlock(); |
1292 | return err; | |
1da177e4 LT |
1293 | } |
1294 | ||
1295 | /** | |
1296 | * call_netdevice_notifiers - call all network notifier blocks | |
1297 | * @val: value passed unmodified to notifier function | |
c4ea43c5 | 1298 | * @dev: net_device pointer passed unmodified to notifier function |
1da177e4 LT |
1299 | * |
1300 | * Call all network notifier blocks. Parameters and return value | |
f07d5b94 | 1301 | * are as for raw_notifier_call_chain(). |
1da177e4 LT |
1302 | */ |
1303 | ||
ad7379d4 | 1304 | int call_netdevice_notifiers(unsigned long val, struct net_device *dev) |
1da177e4 | 1305 | { |
ad7379d4 | 1306 | return raw_notifier_call_chain(&netdev_chain, val, dev); |
1da177e4 LT |
1307 | } |
1308 | ||
1309 | /* When > 0 there are consumers of rx skb time stamps */ | |
1310 | static atomic_t netstamp_needed = ATOMIC_INIT(0); | |
1311 | ||
1312 | void net_enable_timestamp(void) | |
1313 | { | |
1314 | atomic_inc(&netstamp_needed); | |
1315 | } | |
1316 | ||
1317 | void net_disable_timestamp(void) | |
1318 | { | |
1319 | atomic_dec(&netstamp_needed); | |
1320 | } | |
1321 | ||
a61bbcf2 | 1322 | static inline void net_timestamp(struct sk_buff *skb) |
1da177e4 LT |
1323 | { |
1324 | if (atomic_read(&netstamp_needed)) | |
a61bbcf2 | 1325 | __net_timestamp(skb); |
b7aa0bf7 ED |
1326 | else |
1327 | skb->tstamp.tv64 = 0; | |
1da177e4 LT |
1328 | } |
1329 | ||
1330 | /* | |
1331 | * Support routine. Sends outgoing frames to any network | |
1332 | * taps currently in use. | |
1333 | */ | |
1334 | ||
f6a78bfc | 1335 | static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
1336 | { |
1337 | struct packet_type *ptype; | |
a61bbcf2 PM |
1338 | |
1339 | net_timestamp(skb); | |
1da177e4 LT |
1340 | |
1341 | rcu_read_lock(); | |
1342 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
1343 | /* Never send packets back to the socket | |
1344 | * they originated from - MvS (miquels@drinkel.ow.org) | |
1345 | */ | |
1346 | if ((ptype->dev == dev || !ptype->dev) && | |
1347 | (ptype->af_packet_priv == NULL || | |
1348 | (struct sock *)ptype->af_packet_priv != skb->sk)) { | |
1349 | struct sk_buff *skb2= skb_clone(skb, GFP_ATOMIC); | |
1350 | if (!skb2) | |
1351 | break; | |
1352 | ||
1353 | /* skb->nh should be correctly | |
1354 | set by sender, so that the second statement is | |
1355 | just protection against buggy protocols. | |
1356 | */ | |
459a98ed | 1357 | skb_reset_mac_header(skb2); |
1da177e4 | 1358 | |
d56f90a7 | 1359 | if (skb_network_header(skb2) < skb2->data || |
27a884dc | 1360 | skb2->network_header > skb2->tail) { |
1da177e4 LT |
1361 | if (net_ratelimit()) |
1362 | printk(KERN_CRIT "protocol %04x is " | |
1363 | "buggy, dev %s\n", | |
1364 | skb2->protocol, dev->name); | |
c1d2bbe1 | 1365 | skb_reset_network_header(skb2); |
1da177e4 LT |
1366 | } |
1367 | ||
b0e380b1 | 1368 | skb2->transport_header = skb2->network_header; |
1da177e4 | 1369 | skb2->pkt_type = PACKET_OUTGOING; |
f2ccd8fa | 1370 | ptype->func(skb2, skb->dev, ptype, skb->dev); |
1da177e4 LT |
1371 | } |
1372 | } | |
1373 | rcu_read_unlock(); | |
1374 | } | |
1375 | ||
56079431 | 1376 | |
def82a1d | 1377 | static inline void __netif_reschedule(struct Qdisc *q) |
56079431 | 1378 | { |
def82a1d JP |
1379 | struct softnet_data *sd; |
1380 | unsigned long flags; | |
56079431 | 1381 | |
def82a1d JP |
1382 | local_irq_save(flags); |
1383 | sd = &__get_cpu_var(softnet_data); | |
1384 | q->next_sched = sd->output_queue; | |
1385 | sd->output_queue = q; | |
1386 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
1387 | local_irq_restore(flags); | |
1388 | } | |
1389 | ||
1390 | void __netif_schedule(struct Qdisc *q) | |
1391 | { | |
1392 | if (!test_and_set_bit(__QDISC_STATE_SCHED, &q->state)) | |
1393 | __netif_reschedule(q); | |
56079431 DV |
1394 | } |
1395 | EXPORT_SYMBOL(__netif_schedule); | |
1396 | ||
bea3348e | 1397 | void dev_kfree_skb_irq(struct sk_buff *skb) |
56079431 | 1398 | { |
bea3348e SH |
1399 | if (atomic_dec_and_test(&skb->users)) { |
1400 | struct softnet_data *sd; | |
1401 | unsigned long flags; | |
56079431 | 1402 | |
bea3348e SH |
1403 | local_irq_save(flags); |
1404 | sd = &__get_cpu_var(softnet_data); | |
1405 | skb->next = sd->completion_queue; | |
1406 | sd->completion_queue = skb; | |
1407 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
1408 | local_irq_restore(flags); | |
1409 | } | |
56079431 | 1410 | } |
bea3348e | 1411 | EXPORT_SYMBOL(dev_kfree_skb_irq); |
56079431 DV |
1412 | |
1413 | void dev_kfree_skb_any(struct sk_buff *skb) | |
1414 | { | |
1415 | if (in_irq() || irqs_disabled()) | |
1416 | dev_kfree_skb_irq(skb); | |
1417 | else | |
1418 | dev_kfree_skb(skb); | |
1419 | } | |
1420 | EXPORT_SYMBOL(dev_kfree_skb_any); | |
1421 | ||
1422 | ||
bea3348e SH |
1423 | /** |
1424 | * netif_device_detach - mark device as removed | |
1425 | * @dev: network device | |
1426 | * | |
1427 | * Mark device as removed from system and therefore no longer available. | |
1428 | */ | |
56079431 DV |
1429 | void netif_device_detach(struct net_device *dev) |
1430 | { | |
1431 | if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) && | |
1432 | netif_running(dev)) { | |
1433 | netif_stop_queue(dev); | |
1434 | } | |
1435 | } | |
1436 | EXPORT_SYMBOL(netif_device_detach); | |
1437 | ||
bea3348e SH |
1438 | /** |
1439 | * netif_device_attach - mark device as attached | |
1440 | * @dev: network device | |
1441 | * | |
1442 | * Mark device as attached from system and restart if needed. | |
1443 | */ | |
56079431 DV |
1444 | void netif_device_attach(struct net_device *dev) |
1445 | { | |
1446 | if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) && | |
1447 | netif_running(dev)) { | |
1448 | netif_wake_queue(dev); | |
4ec93edb | 1449 | __netdev_watchdog_up(dev); |
56079431 DV |
1450 | } |
1451 | } | |
1452 | EXPORT_SYMBOL(netif_device_attach); | |
1453 | ||
6de329e2 BH |
1454 | static bool can_checksum_protocol(unsigned long features, __be16 protocol) |
1455 | { | |
1456 | return ((features & NETIF_F_GEN_CSUM) || | |
1457 | ((features & NETIF_F_IP_CSUM) && | |
1458 | protocol == htons(ETH_P_IP)) || | |
1459 | ((features & NETIF_F_IPV6_CSUM) && | |
1460 | protocol == htons(ETH_P_IPV6))); | |
1461 | } | |
1462 | ||
1463 | static bool dev_can_checksum(struct net_device *dev, struct sk_buff *skb) | |
1464 | { | |
1465 | if (can_checksum_protocol(dev->features, skb->protocol)) | |
1466 | return true; | |
1467 | ||
1468 | if (skb->protocol == htons(ETH_P_8021Q)) { | |
1469 | struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data; | |
1470 | if (can_checksum_protocol(dev->features & dev->vlan_features, | |
1471 | veh->h_vlan_encapsulated_proto)) | |
1472 | return true; | |
1473 | } | |
1474 | ||
1475 | return false; | |
1476 | } | |
56079431 | 1477 | |
1da177e4 LT |
1478 | /* |
1479 | * Invalidate hardware checksum when packet is to be mangled, and | |
1480 | * complete checksum manually on outgoing path. | |
1481 | */ | |
84fa7933 | 1482 | int skb_checksum_help(struct sk_buff *skb) |
1da177e4 | 1483 | { |
d3bc23e7 | 1484 | __wsum csum; |
663ead3b | 1485 | int ret = 0, offset; |
1da177e4 | 1486 | |
84fa7933 | 1487 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
a430a43d HX |
1488 | goto out_set_summed; |
1489 | ||
1490 | if (unlikely(skb_shinfo(skb)->gso_size)) { | |
a430a43d HX |
1491 | /* Let GSO fix up the checksum. */ |
1492 | goto out_set_summed; | |
1da177e4 LT |
1493 | } |
1494 | ||
a030847e HX |
1495 | offset = skb->csum_start - skb_headroom(skb); |
1496 | BUG_ON(offset >= skb_headlen(skb)); | |
1497 | csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
1498 | ||
1499 | offset += skb->csum_offset; | |
1500 | BUG_ON(offset + sizeof(__sum16) > skb_headlen(skb)); | |
1501 | ||
1502 | if (skb_cloned(skb) && | |
1503 | !skb_clone_writable(skb, offset + sizeof(__sum16))) { | |
1da177e4 LT |
1504 | ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC); |
1505 | if (ret) | |
1506 | goto out; | |
1507 | } | |
1508 | ||
a030847e | 1509 | *(__sum16 *)(skb->data + offset) = csum_fold(csum); |
a430a43d | 1510 | out_set_summed: |
1da177e4 | 1511 | skb->ip_summed = CHECKSUM_NONE; |
4ec93edb | 1512 | out: |
1da177e4 LT |
1513 | return ret; |
1514 | } | |
1515 | ||
f6a78bfc HX |
1516 | /** |
1517 | * skb_gso_segment - Perform segmentation on skb. | |
1518 | * @skb: buffer to segment | |
576a30eb | 1519 | * @features: features for the output path (see dev->features) |
f6a78bfc HX |
1520 | * |
1521 | * This function segments the given skb and returns a list of segments. | |
576a30eb HX |
1522 | * |
1523 | * It may return NULL if the skb requires no segmentation. This is | |
1524 | * only possible when GSO is used for verifying header integrity. | |
f6a78bfc | 1525 | */ |
576a30eb | 1526 | struct sk_buff *skb_gso_segment(struct sk_buff *skb, int features) |
f6a78bfc HX |
1527 | { |
1528 | struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT); | |
1529 | struct packet_type *ptype; | |
252e3346 | 1530 | __be16 type = skb->protocol; |
a430a43d | 1531 | int err; |
f6a78bfc | 1532 | |
459a98ed | 1533 | skb_reset_mac_header(skb); |
b0e380b1 | 1534 | skb->mac_len = skb->network_header - skb->mac_header; |
f6a78bfc HX |
1535 | __skb_pull(skb, skb->mac_len); |
1536 | ||
67fd1a73 HX |
1537 | if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { |
1538 | struct net_device *dev = skb->dev; | |
1539 | struct ethtool_drvinfo info = {}; | |
1540 | ||
1541 | if (dev && dev->ethtool_ops && dev->ethtool_ops->get_drvinfo) | |
1542 | dev->ethtool_ops->get_drvinfo(dev, &info); | |
1543 | ||
1544 | WARN(1, "%s: caps=(0x%lx, 0x%lx) len=%d data_len=%d " | |
1545 | "ip_summed=%d", | |
1546 | info.driver, dev ? dev->features : 0L, | |
1547 | skb->sk ? skb->sk->sk_route_caps : 0L, | |
1548 | skb->len, skb->data_len, skb->ip_summed); | |
1549 | ||
a430a43d HX |
1550 | if (skb_header_cloned(skb) && |
1551 | (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) | |
1552 | return ERR_PTR(err); | |
1553 | } | |
1554 | ||
f6a78bfc | 1555 | rcu_read_lock(); |
82d8a867 PE |
1556 | list_for_each_entry_rcu(ptype, |
1557 | &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { | |
f6a78bfc | 1558 | if (ptype->type == type && !ptype->dev && ptype->gso_segment) { |
84fa7933 | 1559 | if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { |
a430a43d HX |
1560 | err = ptype->gso_send_check(skb); |
1561 | segs = ERR_PTR(err); | |
1562 | if (err || skb_gso_ok(skb, features)) | |
1563 | break; | |
d56f90a7 ACM |
1564 | __skb_push(skb, (skb->data - |
1565 | skb_network_header(skb))); | |
a430a43d | 1566 | } |
576a30eb | 1567 | segs = ptype->gso_segment(skb, features); |
f6a78bfc HX |
1568 | break; |
1569 | } | |
1570 | } | |
1571 | rcu_read_unlock(); | |
1572 | ||
98e399f8 | 1573 | __skb_push(skb, skb->data - skb_mac_header(skb)); |
576a30eb | 1574 | |
f6a78bfc HX |
1575 | return segs; |
1576 | } | |
1577 | ||
1578 | EXPORT_SYMBOL(skb_gso_segment); | |
1579 | ||
fb286bb2 HX |
1580 | /* Take action when hardware reception checksum errors are detected. */ |
1581 | #ifdef CONFIG_BUG | |
1582 | void netdev_rx_csum_fault(struct net_device *dev) | |
1583 | { | |
1584 | if (net_ratelimit()) { | |
4ec93edb | 1585 | printk(KERN_ERR "%s: hw csum failure.\n", |
246a4212 | 1586 | dev ? dev->name : "<unknown>"); |
fb286bb2 HX |
1587 | dump_stack(); |
1588 | } | |
1589 | } | |
1590 | EXPORT_SYMBOL(netdev_rx_csum_fault); | |
1591 | #endif | |
1592 | ||
1da177e4 LT |
1593 | /* Actually, we should eliminate this check as soon as we know, that: |
1594 | * 1. IOMMU is present and allows to map all the memory. | |
1595 | * 2. No high memory really exists on this machine. | |
1596 | */ | |
1597 | ||
1598 | static inline int illegal_highdma(struct net_device *dev, struct sk_buff *skb) | |
1599 | { | |
3d3a8533 | 1600 | #ifdef CONFIG_HIGHMEM |
1da177e4 LT |
1601 | int i; |
1602 | ||
1603 | if (dev->features & NETIF_F_HIGHDMA) | |
1604 | return 0; | |
1605 | ||
1606 | for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) | |
1607 | if (PageHighMem(skb_shinfo(skb)->frags[i].page)) | |
1608 | return 1; | |
1609 | ||
3d3a8533 | 1610 | #endif |
1da177e4 LT |
1611 | return 0; |
1612 | } | |
1da177e4 | 1613 | |
f6a78bfc HX |
1614 | struct dev_gso_cb { |
1615 | void (*destructor)(struct sk_buff *skb); | |
1616 | }; | |
1617 | ||
1618 | #define DEV_GSO_CB(skb) ((struct dev_gso_cb *)(skb)->cb) | |
1619 | ||
1620 | static void dev_gso_skb_destructor(struct sk_buff *skb) | |
1621 | { | |
1622 | struct dev_gso_cb *cb; | |
1623 | ||
1624 | do { | |
1625 | struct sk_buff *nskb = skb->next; | |
1626 | ||
1627 | skb->next = nskb->next; | |
1628 | nskb->next = NULL; | |
1629 | kfree_skb(nskb); | |
1630 | } while (skb->next); | |
1631 | ||
1632 | cb = DEV_GSO_CB(skb); | |
1633 | if (cb->destructor) | |
1634 | cb->destructor(skb); | |
1635 | } | |
1636 | ||
1637 | /** | |
1638 | * dev_gso_segment - Perform emulated hardware segmentation on skb. | |
1639 | * @skb: buffer to segment | |
1640 | * | |
1641 | * This function segments the given skb and stores the list of segments | |
1642 | * in skb->next. | |
1643 | */ | |
1644 | static int dev_gso_segment(struct sk_buff *skb) | |
1645 | { | |
1646 | struct net_device *dev = skb->dev; | |
1647 | struct sk_buff *segs; | |
576a30eb HX |
1648 | int features = dev->features & ~(illegal_highdma(dev, skb) ? |
1649 | NETIF_F_SG : 0); | |
1650 | ||
1651 | segs = skb_gso_segment(skb, features); | |
1652 | ||
1653 | /* Verifying header integrity only. */ | |
1654 | if (!segs) | |
1655 | return 0; | |
f6a78bfc | 1656 | |
801678c5 | 1657 | if (IS_ERR(segs)) |
f6a78bfc HX |
1658 | return PTR_ERR(segs); |
1659 | ||
1660 | skb->next = segs; | |
1661 | DEV_GSO_CB(skb)->destructor = skb->destructor; | |
1662 | skb->destructor = dev_gso_skb_destructor; | |
1663 | ||
1664 | return 0; | |
1665 | } | |
1666 | ||
fd2ea0a7 DM |
1667 | int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, |
1668 | struct netdev_queue *txq) | |
f6a78bfc | 1669 | { |
00829823 | 1670 | const struct net_device_ops *ops = dev->netdev_ops; |
ac45f602 | 1671 | int rc; |
00829823 | 1672 | |
f6a78bfc | 1673 | if (likely(!skb->next)) { |
9be9a6b9 | 1674 | if (!list_empty(&ptype_all)) |
f6a78bfc HX |
1675 | dev_queue_xmit_nit(skb, dev); |
1676 | ||
576a30eb HX |
1677 | if (netif_needs_gso(dev, skb)) { |
1678 | if (unlikely(dev_gso_segment(skb))) | |
1679 | goto out_kfree_skb; | |
1680 | if (skb->next) | |
1681 | goto gso; | |
1682 | } | |
f6a78bfc | 1683 | |
ac45f602 PO |
1684 | rc = ops->ndo_start_xmit(skb, dev); |
1685 | /* | |
1686 | * TODO: if skb_orphan() was called by | |
1687 | * dev->hard_start_xmit() (for example, the unmodified | |
1688 | * igb driver does that; bnx2 doesn't), then | |
1689 | * skb_tx_software_timestamp() will be unable to send | |
1690 | * back the time stamp. | |
1691 | * | |
1692 | * How can this be prevented? Always create another | |
1693 | * reference to the socket before calling | |
1694 | * dev->hard_start_xmit()? Prevent that skb_orphan() | |
1695 | * does anything in dev->hard_start_xmit() by clearing | |
1696 | * the skb destructor before the call and restoring it | |
1697 | * afterwards, then doing the skb_orphan() ourselves? | |
1698 | */ | |
ac45f602 | 1699 | return rc; |
f6a78bfc HX |
1700 | } |
1701 | ||
576a30eb | 1702 | gso: |
f6a78bfc HX |
1703 | do { |
1704 | struct sk_buff *nskb = skb->next; | |
f6a78bfc HX |
1705 | |
1706 | skb->next = nskb->next; | |
1707 | nskb->next = NULL; | |
00829823 | 1708 | rc = ops->ndo_start_xmit(nskb, dev); |
f6a78bfc | 1709 | if (unlikely(rc)) { |
f54d9e8d | 1710 | nskb->next = skb->next; |
f6a78bfc HX |
1711 | skb->next = nskb; |
1712 | return rc; | |
1713 | } | |
fd2ea0a7 | 1714 | if (unlikely(netif_tx_queue_stopped(txq) && skb->next)) |
f54d9e8d | 1715 | return NETDEV_TX_BUSY; |
f6a78bfc | 1716 | } while (skb->next); |
4ec93edb | 1717 | |
f6a78bfc HX |
1718 | skb->destructor = DEV_GSO_CB(skb)->destructor; |
1719 | ||
1720 | out_kfree_skb: | |
1721 | kfree_skb(skb); | |
1722 | return 0; | |
1723 | } | |
1724 | ||
7019298a | 1725 | static u32 skb_tx_hashrnd; |
b6b2fed1 | 1726 | |
9247744e | 1727 | u16 skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb) |
8f0f2223 | 1728 | { |
7019298a | 1729 | u32 hash; |
b6b2fed1 | 1730 | |
d5a9e24a | 1731 | if (skb_rx_queue_recorded(skb)) { |
7019298a DM |
1732 | hash = skb_get_rx_queue(skb); |
1733 | } else if (skb->sk && skb->sk->sk_hash) { | |
1734 | hash = skb->sk->sk_hash; | |
1735 | } else | |
1736 | hash = skb->protocol; | |
d5a9e24a | 1737 | |
7019298a | 1738 | hash = jhash_1word(hash, skb_tx_hashrnd); |
b6b2fed1 DM |
1739 | |
1740 | return (u16) (((u64) hash * dev->real_num_tx_queues) >> 32); | |
8f0f2223 | 1741 | } |
9247744e | 1742 | EXPORT_SYMBOL(skb_tx_hash); |
8f0f2223 | 1743 | |
e8a0464c DM |
1744 | static struct netdev_queue *dev_pick_tx(struct net_device *dev, |
1745 | struct sk_buff *skb) | |
1746 | { | |
00829823 | 1747 | const struct net_device_ops *ops = dev->netdev_ops; |
fd2ea0a7 DM |
1748 | u16 queue_index = 0; |
1749 | ||
00829823 SH |
1750 | if (ops->ndo_select_queue) |
1751 | queue_index = ops->ndo_select_queue(dev, skb); | |
8f0f2223 | 1752 | else if (dev->real_num_tx_queues > 1) |
7019298a | 1753 | queue_index = skb_tx_hash(dev, skb); |
eae792b7 | 1754 | |
fd2ea0a7 DM |
1755 | skb_set_queue_mapping(skb, queue_index); |
1756 | return netdev_get_tx_queue(dev, queue_index); | |
e8a0464c DM |
1757 | } |
1758 | ||
d29f749e DJ |
1759 | /** |
1760 | * dev_queue_xmit - transmit a buffer | |
1761 | * @skb: buffer to transmit | |
1762 | * | |
1763 | * Queue a buffer for transmission to a network device. The caller must | |
1764 | * have set the device and priority and built the buffer before calling | |
1765 | * this function. The function can be called from an interrupt. | |
1766 | * | |
1767 | * A negative errno code is returned on a failure. A success does not | |
1768 | * guarantee the frame will be transmitted as it may be dropped due | |
1769 | * to congestion or traffic shaping. | |
1770 | * | |
1771 | * ----------------------------------------------------------------------------------- | |
1772 | * I notice this method can also return errors from the queue disciplines, | |
1773 | * including NET_XMIT_DROP, which is a positive value. So, errors can also | |
1774 | * be positive. | |
1775 | * | |
1776 | * Regardless of the return value, the skb is consumed, so it is currently | |
1777 | * difficult to retry a send to this method. (You can bump the ref count | |
1778 | * before sending to hold a reference for retry if you are careful.) | |
1779 | * | |
1780 | * When calling this method, interrupts MUST be enabled. This is because | |
1781 | * the BH enable code must have IRQs enabled so that it will not deadlock. | |
1782 | * --BLG | |
1783 | */ | |
1da177e4 LT |
1784 | int dev_queue_xmit(struct sk_buff *skb) |
1785 | { | |
1786 | struct net_device *dev = skb->dev; | |
dc2b4847 | 1787 | struct netdev_queue *txq; |
1da177e4 LT |
1788 | struct Qdisc *q; |
1789 | int rc = -ENOMEM; | |
1790 | ||
f6a78bfc HX |
1791 | /* GSO will handle the following emulations directly. */ |
1792 | if (netif_needs_gso(dev, skb)) | |
1793 | goto gso; | |
1794 | ||
1da177e4 LT |
1795 | if (skb_shinfo(skb)->frag_list && |
1796 | !(dev->features & NETIF_F_FRAGLIST) && | |
364c6bad | 1797 | __skb_linearize(skb)) |
1da177e4 LT |
1798 | goto out_kfree_skb; |
1799 | ||
1800 | /* Fragmented skb is linearized if device does not support SG, | |
1801 | * or if at least one of fragments is in highmem and device | |
1802 | * does not support DMA from it. | |
1803 | */ | |
1804 | if (skb_shinfo(skb)->nr_frags && | |
1805 | (!(dev->features & NETIF_F_SG) || illegal_highdma(dev, skb)) && | |
364c6bad | 1806 | __skb_linearize(skb)) |
1da177e4 LT |
1807 | goto out_kfree_skb; |
1808 | ||
1809 | /* If packet is not checksummed and device does not support | |
1810 | * checksumming for this protocol, complete checksumming here. | |
1811 | */ | |
663ead3b HX |
1812 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
1813 | skb_set_transport_header(skb, skb->csum_start - | |
1814 | skb_headroom(skb)); | |
6de329e2 BH |
1815 | if (!dev_can_checksum(dev, skb) && skb_checksum_help(skb)) |
1816 | goto out_kfree_skb; | |
663ead3b | 1817 | } |
1da177e4 | 1818 | |
f6a78bfc | 1819 | gso: |
4ec93edb YH |
1820 | /* Disable soft irqs for various locks below. Also |
1821 | * stops preemption for RCU. | |
1da177e4 | 1822 | */ |
4ec93edb | 1823 | rcu_read_lock_bh(); |
1da177e4 | 1824 | |
eae792b7 | 1825 | txq = dev_pick_tx(dev, skb); |
b0e1e646 | 1826 | q = rcu_dereference(txq->qdisc); |
37437bb2 | 1827 | |
1da177e4 LT |
1828 | #ifdef CONFIG_NET_CLS_ACT |
1829 | skb->tc_verd = SET_TC_AT(skb->tc_verd,AT_EGRESS); | |
1830 | #endif | |
1831 | if (q->enqueue) { | |
5fb66229 | 1832 | spinlock_t *root_lock = qdisc_lock(q); |
37437bb2 DM |
1833 | |
1834 | spin_lock(root_lock); | |
1835 | ||
a9312ae8 | 1836 | if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) { |
96d20316 | 1837 | kfree_skb(skb); |
a9312ae8 | 1838 | rc = NET_XMIT_DROP; |
96d20316 DM |
1839 | } else { |
1840 | rc = qdisc_enqueue_root(skb, q); | |
1841 | qdisc_run(q); | |
a9312ae8 | 1842 | } |
37437bb2 DM |
1843 | spin_unlock(root_lock); |
1844 | ||
37437bb2 | 1845 | goto out; |
1da177e4 LT |
1846 | } |
1847 | ||
1848 | /* The device has no queue. Common case for software devices: | |
1849 | loopback, all the sorts of tunnels... | |
1850 | ||
932ff279 HX |
1851 | Really, it is unlikely that netif_tx_lock protection is necessary |
1852 | here. (f.e. loopback and IP tunnels are clean ignoring statistics | |
1da177e4 LT |
1853 | counters.) |
1854 | However, it is possible, that they rely on protection | |
1855 | made by us here. | |
1856 | ||
1857 | Check this and shot the lock. It is not prone from deadlocks. | |
1858 | Either shot noqueue qdisc, it is even simpler 8) | |
1859 | */ | |
1860 | if (dev->flags & IFF_UP) { | |
1861 | int cpu = smp_processor_id(); /* ok because BHs are off */ | |
1862 | ||
c773e847 | 1863 | if (txq->xmit_lock_owner != cpu) { |
1da177e4 | 1864 | |
c773e847 | 1865 | HARD_TX_LOCK(dev, txq, cpu); |
1da177e4 | 1866 | |
fd2ea0a7 | 1867 | if (!netif_tx_queue_stopped(txq)) { |
1da177e4 | 1868 | rc = 0; |
fd2ea0a7 | 1869 | if (!dev_hard_start_xmit(skb, dev, txq)) { |
c773e847 | 1870 | HARD_TX_UNLOCK(dev, txq); |
1da177e4 LT |
1871 | goto out; |
1872 | } | |
1873 | } | |
c773e847 | 1874 | HARD_TX_UNLOCK(dev, txq); |
1da177e4 LT |
1875 | if (net_ratelimit()) |
1876 | printk(KERN_CRIT "Virtual device %s asks to " | |
1877 | "queue packet!\n", dev->name); | |
1878 | } else { | |
1879 | /* Recursion is detected! It is possible, | |
1880 | * unfortunately */ | |
1881 | if (net_ratelimit()) | |
1882 | printk(KERN_CRIT "Dead loop on virtual device " | |
1883 | "%s, fix it urgently!\n", dev->name); | |
1884 | } | |
1885 | } | |
1886 | ||
1887 | rc = -ENETDOWN; | |
d4828d85 | 1888 | rcu_read_unlock_bh(); |
1da177e4 LT |
1889 | |
1890 | out_kfree_skb: | |
1891 | kfree_skb(skb); | |
1892 | return rc; | |
1893 | out: | |
d4828d85 | 1894 | rcu_read_unlock_bh(); |
1da177e4 LT |
1895 | return rc; |
1896 | } | |
1897 | ||
1898 | ||
1899 | /*======================================================================= | |
1900 | Receiver routines | |
1901 | =======================================================================*/ | |
1902 | ||
6b2bedc3 SH |
1903 | int netdev_max_backlog __read_mostly = 1000; |
1904 | int netdev_budget __read_mostly = 300; | |
1905 | int weight_p __read_mostly = 64; /* old backlog weight */ | |
1da177e4 LT |
1906 | |
1907 | DEFINE_PER_CPU(struct netif_rx_stats, netdev_rx_stat) = { 0, }; | |
1908 | ||
1909 | ||
1da177e4 LT |
1910 | /** |
1911 | * netif_rx - post buffer to the network code | |
1912 | * @skb: buffer to post | |
1913 | * | |
1914 | * This function receives a packet from a device driver and queues it for | |
1915 | * the upper (protocol) levels to process. It always succeeds. The buffer | |
1916 | * may be dropped during processing for congestion control or by the | |
1917 | * protocol layers. | |
1918 | * | |
1919 | * return values: | |
1920 | * NET_RX_SUCCESS (no congestion) | |
1da177e4 LT |
1921 | * NET_RX_DROP (packet was dropped) |
1922 | * | |
1923 | */ | |
1924 | ||
1925 | int netif_rx(struct sk_buff *skb) | |
1926 | { | |
1da177e4 LT |
1927 | struct softnet_data *queue; |
1928 | unsigned long flags; | |
1929 | ||
1930 | /* if netpoll wants it, pretend we never saw it */ | |
1931 | if (netpoll_rx(skb)) | |
1932 | return NET_RX_DROP; | |
1933 | ||
b7aa0bf7 | 1934 | if (!skb->tstamp.tv64) |
a61bbcf2 | 1935 | net_timestamp(skb); |
1da177e4 LT |
1936 | |
1937 | /* | |
1938 | * The code is rearranged so that the path is the most | |
1939 | * short when CPU is congested, but is still operating. | |
1940 | */ | |
1941 | local_irq_save(flags); | |
1da177e4 LT |
1942 | queue = &__get_cpu_var(softnet_data); |
1943 | ||
1944 | __get_cpu_var(netdev_rx_stat).total++; | |
1945 | if (queue->input_pkt_queue.qlen <= netdev_max_backlog) { | |
1946 | if (queue->input_pkt_queue.qlen) { | |
1da177e4 | 1947 | enqueue: |
1da177e4 | 1948 | __skb_queue_tail(&queue->input_pkt_queue, skb); |
1da177e4 | 1949 | local_irq_restore(flags); |
34008d8c | 1950 | return NET_RX_SUCCESS; |
1da177e4 LT |
1951 | } |
1952 | ||
bea3348e | 1953 | napi_schedule(&queue->backlog); |
1da177e4 LT |
1954 | goto enqueue; |
1955 | } | |
1956 | ||
1da177e4 LT |
1957 | __get_cpu_var(netdev_rx_stat).dropped++; |
1958 | local_irq_restore(flags); | |
1959 | ||
1960 | kfree_skb(skb); | |
1961 | return NET_RX_DROP; | |
1962 | } | |
1963 | ||
1964 | int netif_rx_ni(struct sk_buff *skb) | |
1965 | { | |
1966 | int err; | |
1967 | ||
1968 | preempt_disable(); | |
1969 | err = netif_rx(skb); | |
1970 | if (local_softirq_pending()) | |
1971 | do_softirq(); | |
1972 | preempt_enable(); | |
1973 | ||
1974 | return err; | |
1975 | } | |
1976 | ||
1977 | EXPORT_SYMBOL(netif_rx_ni); | |
1978 | ||
1da177e4 LT |
1979 | static void net_tx_action(struct softirq_action *h) |
1980 | { | |
1981 | struct softnet_data *sd = &__get_cpu_var(softnet_data); | |
1982 | ||
1983 | if (sd->completion_queue) { | |
1984 | struct sk_buff *clist; | |
1985 | ||
1986 | local_irq_disable(); | |
1987 | clist = sd->completion_queue; | |
1988 | sd->completion_queue = NULL; | |
1989 | local_irq_enable(); | |
1990 | ||
1991 | while (clist) { | |
1992 | struct sk_buff *skb = clist; | |
1993 | clist = clist->next; | |
1994 | ||
547b792c | 1995 | WARN_ON(atomic_read(&skb->users)); |
1da177e4 LT |
1996 | __kfree_skb(skb); |
1997 | } | |
1998 | } | |
1999 | ||
2000 | if (sd->output_queue) { | |
37437bb2 | 2001 | struct Qdisc *head; |
1da177e4 LT |
2002 | |
2003 | local_irq_disable(); | |
2004 | head = sd->output_queue; | |
2005 | sd->output_queue = NULL; | |
2006 | local_irq_enable(); | |
2007 | ||
2008 | while (head) { | |
37437bb2 DM |
2009 | struct Qdisc *q = head; |
2010 | spinlock_t *root_lock; | |
2011 | ||
1da177e4 LT |
2012 | head = head->next_sched; |
2013 | ||
5fb66229 | 2014 | root_lock = qdisc_lock(q); |
37437bb2 | 2015 | if (spin_trylock(root_lock)) { |
def82a1d JP |
2016 | smp_mb__before_clear_bit(); |
2017 | clear_bit(__QDISC_STATE_SCHED, | |
2018 | &q->state); | |
37437bb2 DM |
2019 | qdisc_run(q); |
2020 | spin_unlock(root_lock); | |
1da177e4 | 2021 | } else { |
195648bb | 2022 | if (!test_bit(__QDISC_STATE_DEACTIVATED, |
e8a83e10 | 2023 | &q->state)) { |
195648bb | 2024 | __netif_reschedule(q); |
e8a83e10 JP |
2025 | } else { |
2026 | smp_mb__before_clear_bit(); | |
2027 | clear_bit(__QDISC_STATE_SCHED, | |
2028 | &q->state); | |
2029 | } | |
1da177e4 LT |
2030 | } |
2031 | } | |
2032 | } | |
2033 | } | |
2034 | ||
6f05f629 SH |
2035 | static inline int deliver_skb(struct sk_buff *skb, |
2036 | struct packet_type *pt_prev, | |
2037 | struct net_device *orig_dev) | |
1da177e4 LT |
2038 | { |
2039 | atomic_inc(&skb->users); | |
f2ccd8fa | 2040 | return pt_prev->func(skb, skb->dev, pt_prev, orig_dev); |
1da177e4 LT |
2041 | } |
2042 | ||
2043 | #if defined(CONFIG_BRIDGE) || defined (CONFIG_BRIDGE_MODULE) | |
6229e362 | 2044 | /* These hooks defined here for ATM */ |
1da177e4 LT |
2045 | struct net_bridge; |
2046 | struct net_bridge_fdb_entry *(*br_fdb_get_hook)(struct net_bridge *br, | |
2047 | unsigned char *addr); | |
6229e362 | 2048 | void (*br_fdb_put_hook)(struct net_bridge_fdb_entry *ent) __read_mostly; |
1da177e4 | 2049 | |
6229e362 SH |
2050 | /* |
2051 | * If bridge module is loaded call bridging hook. | |
2052 | * returns NULL if packet was consumed. | |
2053 | */ | |
2054 | struct sk_buff *(*br_handle_frame_hook)(struct net_bridge_port *p, | |
2055 | struct sk_buff *skb) __read_mostly; | |
2056 | static inline struct sk_buff *handle_bridge(struct sk_buff *skb, | |
2057 | struct packet_type **pt_prev, int *ret, | |
2058 | struct net_device *orig_dev) | |
1da177e4 LT |
2059 | { |
2060 | struct net_bridge_port *port; | |
2061 | ||
6229e362 SH |
2062 | if (skb->pkt_type == PACKET_LOOPBACK || |
2063 | (port = rcu_dereference(skb->dev->br_port)) == NULL) | |
2064 | return skb; | |
1da177e4 LT |
2065 | |
2066 | if (*pt_prev) { | |
6229e362 | 2067 | *ret = deliver_skb(skb, *pt_prev, orig_dev); |
1da177e4 | 2068 | *pt_prev = NULL; |
4ec93edb YH |
2069 | } |
2070 | ||
6229e362 | 2071 | return br_handle_frame_hook(port, skb); |
1da177e4 LT |
2072 | } |
2073 | #else | |
6229e362 | 2074 | #define handle_bridge(skb, pt_prev, ret, orig_dev) (skb) |
1da177e4 LT |
2075 | #endif |
2076 | ||
b863ceb7 PM |
2077 | #if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) |
2078 | struct sk_buff *(*macvlan_handle_frame_hook)(struct sk_buff *skb) __read_mostly; | |
2079 | EXPORT_SYMBOL_GPL(macvlan_handle_frame_hook); | |
2080 | ||
2081 | static inline struct sk_buff *handle_macvlan(struct sk_buff *skb, | |
2082 | struct packet_type **pt_prev, | |
2083 | int *ret, | |
2084 | struct net_device *orig_dev) | |
2085 | { | |
2086 | if (skb->dev->macvlan_port == NULL) | |
2087 | return skb; | |
2088 | ||
2089 | if (*pt_prev) { | |
2090 | *ret = deliver_skb(skb, *pt_prev, orig_dev); | |
2091 | *pt_prev = NULL; | |
2092 | } | |
2093 | return macvlan_handle_frame_hook(skb); | |
2094 | } | |
2095 | #else | |
2096 | #define handle_macvlan(skb, pt_prev, ret, orig_dev) (skb) | |
2097 | #endif | |
2098 | ||
1da177e4 LT |
2099 | #ifdef CONFIG_NET_CLS_ACT |
2100 | /* TODO: Maybe we should just force sch_ingress to be compiled in | |
2101 | * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions | |
2102 | * a compare and 2 stores extra right now if we dont have it on | |
2103 | * but have CONFIG_NET_CLS_ACT | |
4ec93edb | 2104 | * NOTE: This doesnt stop any functionality; if you dont have |
1da177e4 LT |
2105 | * the ingress scheduler, you just cant add policies on ingress. |
2106 | * | |
2107 | */ | |
4ec93edb | 2108 | static int ing_filter(struct sk_buff *skb) |
1da177e4 | 2109 | { |
1da177e4 | 2110 | struct net_device *dev = skb->dev; |
f697c3e8 | 2111 | u32 ttl = G_TC_RTTL(skb->tc_verd); |
555353cf DM |
2112 | struct netdev_queue *rxq; |
2113 | int result = TC_ACT_OK; | |
2114 | struct Qdisc *q; | |
4ec93edb | 2115 | |
f697c3e8 HX |
2116 | if (MAX_RED_LOOP < ttl++) { |
2117 | printk(KERN_WARNING | |
2118 | "Redir loop detected Dropping packet (%d->%d)\n", | |
2119 | skb->iif, dev->ifindex); | |
2120 | return TC_ACT_SHOT; | |
2121 | } | |
1da177e4 | 2122 | |
f697c3e8 HX |
2123 | skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl); |
2124 | skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS); | |
1da177e4 | 2125 | |
555353cf DM |
2126 | rxq = &dev->rx_queue; |
2127 | ||
83874000 | 2128 | q = rxq->qdisc; |
8d50b53d | 2129 | if (q != &noop_qdisc) { |
83874000 | 2130 | spin_lock(qdisc_lock(q)); |
a9312ae8 DM |
2131 | if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) |
2132 | result = qdisc_enqueue_root(skb, q); | |
83874000 DM |
2133 | spin_unlock(qdisc_lock(q)); |
2134 | } | |
f697c3e8 HX |
2135 | |
2136 | return result; | |
2137 | } | |
86e65da9 | 2138 | |
f697c3e8 HX |
2139 | static inline struct sk_buff *handle_ing(struct sk_buff *skb, |
2140 | struct packet_type **pt_prev, | |
2141 | int *ret, struct net_device *orig_dev) | |
2142 | { | |
8d50b53d | 2143 | if (skb->dev->rx_queue.qdisc == &noop_qdisc) |
f697c3e8 | 2144 | goto out; |
1da177e4 | 2145 | |
f697c3e8 HX |
2146 | if (*pt_prev) { |
2147 | *ret = deliver_skb(skb, *pt_prev, orig_dev); | |
2148 | *pt_prev = NULL; | |
2149 | } else { | |
2150 | /* Huh? Why does turning on AF_PACKET affect this? */ | |
2151 | skb->tc_verd = SET_TC_OK2MUNGE(skb->tc_verd); | |
1da177e4 LT |
2152 | } |
2153 | ||
f697c3e8 HX |
2154 | switch (ing_filter(skb)) { |
2155 | case TC_ACT_SHOT: | |
2156 | case TC_ACT_STOLEN: | |
2157 | kfree_skb(skb); | |
2158 | return NULL; | |
2159 | } | |
2160 | ||
2161 | out: | |
2162 | skb->tc_verd = 0; | |
2163 | return skb; | |
1da177e4 LT |
2164 | } |
2165 | #endif | |
2166 | ||
bc1d0411 PM |
2167 | /* |
2168 | * netif_nit_deliver - deliver received packets to network taps | |
2169 | * @skb: buffer | |
2170 | * | |
2171 | * This function is used to deliver incoming packets to network | |
2172 | * taps. It should be used when the normal netif_receive_skb path | |
2173 | * is bypassed, for example because of VLAN acceleration. | |
2174 | */ | |
2175 | void netif_nit_deliver(struct sk_buff *skb) | |
2176 | { | |
2177 | struct packet_type *ptype; | |
2178 | ||
2179 | if (list_empty(&ptype_all)) | |
2180 | return; | |
2181 | ||
2182 | skb_reset_network_header(skb); | |
2183 | skb_reset_transport_header(skb); | |
2184 | skb->mac_len = skb->network_header - skb->mac_header; | |
2185 | ||
2186 | rcu_read_lock(); | |
2187 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
2188 | if (!ptype->dev || ptype->dev == skb->dev) | |
2189 | deliver_skb(skb, ptype, skb->dev); | |
2190 | } | |
2191 | rcu_read_unlock(); | |
2192 | } | |
2193 | ||
3b582cc1 SH |
2194 | /** |
2195 | * netif_receive_skb - process receive buffer from network | |
2196 | * @skb: buffer to process | |
2197 | * | |
2198 | * netif_receive_skb() is the main receive data processing function. | |
2199 | * It always succeeds. The buffer may be dropped during processing | |
2200 | * for congestion control or by the protocol layers. | |
2201 | * | |
2202 | * This function may only be called from softirq context and interrupts | |
2203 | * should be enabled. | |
2204 | * | |
2205 | * Return values (usually ignored): | |
2206 | * NET_RX_SUCCESS: no congestion | |
2207 | * NET_RX_DROP: packet was dropped | |
2208 | */ | |
1da177e4 LT |
2209 | int netif_receive_skb(struct sk_buff *skb) |
2210 | { | |
2211 | struct packet_type *ptype, *pt_prev; | |
f2ccd8fa | 2212 | struct net_device *orig_dev; |
0d7a3681 | 2213 | struct net_device *null_or_orig; |
1da177e4 | 2214 | int ret = NET_RX_DROP; |
252e3346 | 2215 | __be16 type; |
1da177e4 | 2216 | |
9b22ea56 PM |
2217 | if (skb->vlan_tci && vlan_hwaccel_do_receive(skb)) |
2218 | return NET_RX_SUCCESS; | |
2219 | ||
1da177e4 | 2220 | /* if we've gotten here through NAPI, check netpoll */ |
bea3348e | 2221 | if (netpoll_receive_skb(skb)) |
1da177e4 LT |
2222 | return NET_RX_DROP; |
2223 | ||
b7aa0bf7 | 2224 | if (!skb->tstamp.tv64) |
a61bbcf2 | 2225 | net_timestamp(skb); |
1da177e4 | 2226 | |
c01003c2 PM |
2227 | if (!skb->iif) |
2228 | skb->iif = skb->dev->ifindex; | |
86e65da9 | 2229 | |
0d7a3681 | 2230 | null_or_orig = NULL; |
cc9bd5ce JE |
2231 | orig_dev = skb->dev; |
2232 | if (orig_dev->master) { | |
0d7a3681 JE |
2233 | if (skb_bond_should_drop(skb)) |
2234 | null_or_orig = orig_dev; /* deliver only exact match */ | |
2235 | else | |
2236 | skb->dev = orig_dev->master; | |
cc9bd5ce | 2237 | } |
8f903c70 | 2238 | |
1da177e4 LT |
2239 | __get_cpu_var(netdev_rx_stat).total++; |
2240 | ||
c1d2bbe1 | 2241 | skb_reset_network_header(skb); |
badff6d0 | 2242 | skb_reset_transport_header(skb); |
b0e380b1 | 2243 | skb->mac_len = skb->network_header - skb->mac_header; |
1da177e4 LT |
2244 | |
2245 | pt_prev = NULL; | |
2246 | ||
2247 | rcu_read_lock(); | |
2248 | ||
2249 | #ifdef CONFIG_NET_CLS_ACT | |
2250 | if (skb->tc_verd & TC_NCLS) { | |
2251 | skb->tc_verd = CLR_TC_NCLS(skb->tc_verd); | |
2252 | goto ncls; | |
2253 | } | |
2254 | #endif | |
2255 | ||
2256 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
f982307f JE |
2257 | if (ptype->dev == null_or_orig || ptype->dev == skb->dev || |
2258 | ptype->dev == orig_dev) { | |
4ec93edb | 2259 | if (pt_prev) |
f2ccd8fa | 2260 | ret = deliver_skb(skb, pt_prev, orig_dev); |
1da177e4 LT |
2261 | pt_prev = ptype; |
2262 | } | |
2263 | } | |
2264 | ||
2265 | #ifdef CONFIG_NET_CLS_ACT | |
f697c3e8 HX |
2266 | skb = handle_ing(skb, &pt_prev, &ret, orig_dev); |
2267 | if (!skb) | |
1da177e4 | 2268 | goto out; |
1da177e4 LT |
2269 | ncls: |
2270 | #endif | |
2271 | ||
6229e362 | 2272 | skb = handle_bridge(skb, &pt_prev, &ret, orig_dev); |
b863ceb7 PM |
2273 | if (!skb) |
2274 | goto out; | |
2275 | skb = handle_macvlan(skb, &pt_prev, &ret, orig_dev); | |
6229e362 | 2276 | if (!skb) |
1da177e4 LT |
2277 | goto out; |
2278 | ||
9a279bcb HX |
2279 | skb_orphan(skb); |
2280 | ||
1da177e4 | 2281 | type = skb->protocol; |
82d8a867 PE |
2282 | list_for_each_entry_rcu(ptype, |
2283 | &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { | |
1da177e4 | 2284 | if (ptype->type == type && |
f982307f JE |
2285 | (ptype->dev == null_or_orig || ptype->dev == skb->dev || |
2286 | ptype->dev == orig_dev)) { | |
4ec93edb | 2287 | if (pt_prev) |
f2ccd8fa | 2288 | ret = deliver_skb(skb, pt_prev, orig_dev); |
1da177e4 LT |
2289 | pt_prev = ptype; |
2290 | } | |
2291 | } | |
2292 | ||
2293 | if (pt_prev) { | |
f2ccd8fa | 2294 | ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev); |
1da177e4 LT |
2295 | } else { |
2296 | kfree_skb(skb); | |
2297 | /* Jamal, now you will not able to escape explaining | |
2298 | * me how you were going to use this. :-) | |
2299 | */ | |
2300 | ret = NET_RX_DROP; | |
2301 | } | |
2302 | ||
2303 | out: | |
2304 | rcu_read_unlock(); | |
2305 | return ret; | |
2306 | } | |
2307 | ||
6e583ce5 SH |
2308 | /* Network device is going away, flush any packets still pending */ |
2309 | static void flush_backlog(void *arg) | |
2310 | { | |
2311 | struct net_device *dev = arg; | |
2312 | struct softnet_data *queue = &__get_cpu_var(softnet_data); | |
2313 | struct sk_buff *skb, *tmp; | |
2314 | ||
2315 | skb_queue_walk_safe(&queue->input_pkt_queue, skb, tmp) | |
2316 | if (skb->dev == dev) { | |
2317 | __skb_unlink(skb, &queue->input_pkt_queue); | |
2318 | kfree_skb(skb); | |
2319 | } | |
2320 | } | |
2321 | ||
d565b0a1 HX |
2322 | static int napi_gro_complete(struct sk_buff *skb) |
2323 | { | |
2324 | struct packet_type *ptype; | |
2325 | __be16 type = skb->protocol; | |
2326 | struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK]; | |
2327 | int err = -ENOENT; | |
2328 | ||
5d38a079 | 2329 | if (NAPI_GRO_CB(skb)->count == 1) |
d565b0a1 HX |
2330 | goto out; |
2331 | ||
2332 | rcu_read_lock(); | |
2333 | list_for_each_entry_rcu(ptype, head, list) { | |
2334 | if (ptype->type != type || ptype->dev || !ptype->gro_complete) | |
2335 | continue; | |
2336 | ||
2337 | err = ptype->gro_complete(skb); | |
2338 | break; | |
2339 | } | |
2340 | rcu_read_unlock(); | |
2341 | ||
2342 | if (err) { | |
2343 | WARN_ON(&ptype->list == head); | |
2344 | kfree_skb(skb); | |
2345 | return NET_RX_SUCCESS; | |
2346 | } | |
2347 | ||
2348 | out: | |
b530256d | 2349 | skb_shinfo(skb)->gso_size = 0; |
d565b0a1 HX |
2350 | return netif_receive_skb(skb); |
2351 | } | |
2352 | ||
2353 | void napi_gro_flush(struct napi_struct *napi) | |
2354 | { | |
2355 | struct sk_buff *skb, *next; | |
2356 | ||
2357 | for (skb = napi->gro_list; skb; skb = next) { | |
2358 | next = skb->next; | |
2359 | skb->next = NULL; | |
2360 | napi_gro_complete(skb); | |
2361 | } | |
2362 | ||
4ae5544f | 2363 | napi->gro_count = 0; |
d565b0a1 HX |
2364 | napi->gro_list = NULL; |
2365 | } | |
2366 | EXPORT_SYMBOL(napi_gro_flush); | |
2367 | ||
86911732 HX |
2368 | void *skb_gro_header(struct sk_buff *skb, unsigned int hlen) |
2369 | { | |
2370 | unsigned int offset = skb_gro_offset(skb); | |
2371 | ||
2372 | hlen += offset; | |
2373 | if (hlen <= skb_headlen(skb)) | |
2374 | return skb->data + offset; | |
2375 | ||
2376 | if (unlikely(!skb_shinfo(skb)->nr_frags || | |
2377 | skb_shinfo(skb)->frags[0].size <= | |
2378 | hlen - skb_headlen(skb) || | |
2379 | PageHighMem(skb_shinfo(skb)->frags[0].page))) | |
2380 | return pskb_may_pull(skb, hlen) ? skb->data + offset : NULL; | |
2381 | ||
2382 | return page_address(skb_shinfo(skb)->frags[0].page) + | |
56035022 HX |
2383 | skb_shinfo(skb)->frags[0].page_offset + |
2384 | offset - skb_headlen(skb); | |
86911732 HX |
2385 | } |
2386 | EXPORT_SYMBOL(skb_gro_header); | |
2387 | ||
96e93eab | 2388 | int dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb) |
d565b0a1 HX |
2389 | { |
2390 | struct sk_buff **pp = NULL; | |
2391 | struct packet_type *ptype; | |
2392 | __be16 type = skb->protocol; | |
2393 | struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK]; | |
0da2afd5 | 2394 | int same_flow; |
d565b0a1 | 2395 | int mac_len; |
5d0d9be8 | 2396 | int ret; |
d565b0a1 HX |
2397 | |
2398 | if (!(skb->dev->features & NETIF_F_GRO)) | |
2399 | goto normal; | |
2400 | ||
f17f5c91 HX |
2401 | if (skb_is_gso(skb) || skb_shinfo(skb)->frag_list) |
2402 | goto normal; | |
2403 | ||
d565b0a1 HX |
2404 | rcu_read_lock(); |
2405 | list_for_each_entry_rcu(ptype, head, list) { | |
d565b0a1 HX |
2406 | if (ptype->type != type || ptype->dev || !ptype->gro_receive) |
2407 | continue; | |
2408 | ||
86911732 | 2409 | skb_set_network_header(skb, skb_gro_offset(skb)); |
d565b0a1 HX |
2410 | mac_len = skb->network_header - skb->mac_header; |
2411 | skb->mac_len = mac_len; | |
2412 | NAPI_GRO_CB(skb)->same_flow = 0; | |
2413 | NAPI_GRO_CB(skb)->flush = 0; | |
5d38a079 | 2414 | NAPI_GRO_CB(skb)->free = 0; |
d565b0a1 | 2415 | |
d565b0a1 HX |
2416 | pp = ptype->gro_receive(&napi->gro_list, skb); |
2417 | break; | |
2418 | } | |
2419 | rcu_read_unlock(); | |
2420 | ||
2421 | if (&ptype->list == head) | |
2422 | goto normal; | |
2423 | ||
0da2afd5 | 2424 | same_flow = NAPI_GRO_CB(skb)->same_flow; |
5d0d9be8 | 2425 | ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED; |
0da2afd5 | 2426 | |
d565b0a1 HX |
2427 | if (pp) { |
2428 | struct sk_buff *nskb = *pp; | |
2429 | ||
2430 | *pp = nskb->next; | |
2431 | nskb->next = NULL; | |
2432 | napi_gro_complete(nskb); | |
4ae5544f | 2433 | napi->gro_count--; |
d565b0a1 HX |
2434 | } |
2435 | ||
0da2afd5 | 2436 | if (same_flow) |
d565b0a1 HX |
2437 | goto ok; |
2438 | ||
4ae5544f | 2439 | if (NAPI_GRO_CB(skb)->flush || napi->gro_count >= MAX_GRO_SKBS) |
d565b0a1 | 2440 | goto normal; |
d565b0a1 | 2441 | |
4ae5544f | 2442 | napi->gro_count++; |
d565b0a1 | 2443 | NAPI_GRO_CB(skb)->count = 1; |
86911732 | 2444 | skb_shinfo(skb)->gso_size = skb_gro_len(skb); |
d565b0a1 HX |
2445 | skb->next = napi->gro_list; |
2446 | napi->gro_list = skb; | |
5d0d9be8 | 2447 | ret = GRO_HELD; |
d565b0a1 | 2448 | |
ad0f9904 HX |
2449 | pull: |
2450 | if (unlikely(!pskb_may_pull(skb, skb_gro_offset(skb)))) { | |
2451 | if (napi->gro_list == skb) | |
2452 | napi->gro_list = skb->next; | |
2453 | ret = GRO_DROP; | |
2454 | } | |
2455 | ||
d565b0a1 | 2456 | ok: |
5d0d9be8 | 2457 | return ret; |
d565b0a1 HX |
2458 | |
2459 | normal: | |
ad0f9904 HX |
2460 | ret = GRO_NORMAL; |
2461 | goto pull; | |
5d38a079 | 2462 | } |
96e93eab HX |
2463 | EXPORT_SYMBOL(dev_gro_receive); |
2464 | ||
2465 | static int __napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb) | |
2466 | { | |
2467 | struct sk_buff *p; | |
2468 | ||
d1c76af9 HX |
2469 | if (netpoll_rx_on(skb)) |
2470 | return GRO_NORMAL; | |
2471 | ||
96e93eab | 2472 | for (p = napi->gro_list; p; p = p->next) { |
aa4b9f53 HX |
2473 | NAPI_GRO_CB(p)->same_flow = !compare_ether_header( |
2474 | skb_mac_header(p), skb_gro_mac_header(skb)); | |
96e93eab HX |
2475 | NAPI_GRO_CB(p)->flush = 0; |
2476 | } | |
2477 | ||
2478 | return dev_gro_receive(napi, skb); | |
2479 | } | |
5d38a079 | 2480 | |
5d0d9be8 | 2481 | int napi_skb_finish(int ret, struct sk_buff *skb) |
5d38a079 | 2482 | { |
5d0d9be8 HX |
2483 | int err = NET_RX_SUCCESS; |
2484 | ||
2485 | switch (ret) { | |
2486 | case GRO_NORMAL: | |
5d38a079 HX |
2487 | return netif_receive_skb(skb); |
2488 | ||
5d0d9be8 HX |
2489 | case GRO_DROP: |
2490 | err = NET_RX_DROP; | |
2491 | /* fall through */ | |
2492 | ||
2493 | case GRO_MERGED_FREE: | |
5d38a079 HX |
2494 | kfree_skb(skb); |
2495 | break; | |
2496 | } | |
2497 | ||
5d0d9be8 HX |
2498 | return err; |
2499 | } | |
2500 | EXPORT_SYMBOL(napi_skb_finish); | |
2501 | ||
2502 | int napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb) | |
2503 | { | |
86911732 HX |
2504 | skb_gro_reset_offset(skb); |
2505 | ||
5d0d9be8 | 2506 | return napi_skb_finish(__napi_gro_receive(napi, skb), skb); |
d565b0a1 HX |
2507 | } |
2508 | EXPORT_SYMBOL(napi_gro_receive); | |
2509 | ||
96e93eab HX |
2510 | void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb) |
2511 | { | |
96e93eab HX |
2512 | __skb_pull(skb, skb_headlen(skb)); |
2513 | skb_reserve(skb, NET_IP_ALIGN - skb_headroom(skb)); | |
2514 | ||
2515 | napi->skb = skb; | |
2516 | } | |
2517 | EXPORT_SYMBOL(napi_reuse_skb); | |
2518 | ||
2519 | struct sk_buff *napi_fraginfo_skb(struct napi_struct *napi, | |
2520 | struct napi_gro_fraginfo *info) | |
5d38a079 HX |
2521 | { |
2522 | struct net_device *dev = napi->dev; | |
2523 | struct sk_buff *skb = napi->skb; | |
86911732 | 2524 | struct ethhdr *eth; |
80595d59 HX |
2525 | skb_frag_t *frag; |
2526 | int i; | |
5d38a079 HX |
2527 | |
2528 | napi->skb = NULL; | |
2529 | ||
2530 | if (!skb) { | |
2531 | skb = netdev_alloc_skb(dev, GRO_MAX_HEAD + NET_IP_ALIGN); | |
2532 | if (!skb) | |
2533 | goto out; | |
2534 | ||
2535 | skb_reserve(skb, NET_IP_ALIGN); | |
2536 | } | |
2537 | ||
2538 | BUG_ON(info->nr_frags > MAX_SKB_FRAGS); | |
80595d59 HX |
2539 | frag = &info->frags[info->nr_frags - 1]; |
2540 | ||
2541 | for (i = skb_shinfo(skb)->nr_frags; i < info->nr_frags; i++) { | |
2542 | skb_fill_page_desc(skb, i, frag->page, frag->page_offset, | |
2543 | frag->size); | |
2544 | frag++; | |
2545 | } | |
5d38a079 | 2546 | skb_shinfo(skb)->nr_frags = info->nr_frags; |
5d38a079 HX |
2547 | |
2548 | skb->data_len = info->len; | |
2549 | skb->len += info->len; | |
2550 | skb->truesize += info->len; | |
2551 | ||
86911732 HX |
2552 | skb_reset_mac_header(skb); |
2553 | skb_gro_reset_offset(skb); | |
2554 | ||
2555 | eth = skb_gro_header(skb, sizeof(*eth)); | |
2556 | if (!eth) { | |
96e93eab | 2557 | napi_reuse_skb(napi, skb); |
9a8e47ff | 2558 | skb = NULL; |
96e93eab HX |
2559 | goto out; |
2560 | } | |
5d38a079 | 2561 | |
86911732 HX |
2562 | skb_gro_pull(skb, sizeof(*eth)); |
2563 | ||
2564 | /* | |
2565 | * This works because the only protocols we care about don't require | |
2566 | * special handling. We'll fix it up properly at the end. | |
2567 | */ | |
2568 | skb->protocol = eth->h_proto; | |
5d38a079 HX |
2569 | |
2570 | skb->ip_summed = info->ip_summed; | |
2571 | skb->csum = info->csum; | |
2572 | ||
96e93eab HX |
2573 | out: |
2574 | return skb; | |
2575 | } | |
2576 | EXPORT_SYMBOL(napi_fraginfo_skb); | |
2577 | ||
5d0d9be8 | 2578 | int napi_frags_finish(struct napi_struct *napi, struct sk_buff *skb, int ret) |
96e93eab | 2579 | { |
5d0d9be8 | 2580 | int err = NET_RX_SUCCESS; |
96e93eab | 2581 | |
5d0d9be8 HX |
2582 | switch (ret) { |
2583 | case GRO_NORMAL: | |
86911732 | 2584 | case GRO_HELD: |
86911732 HX |
2585 | skb->protocol = eth_type_trans(skb, napi->dev); |
2586 | ||
2587 | if (ret == GRO_NORMAL) | |
2588 | return netif_receive_skb(skb); | |
2589 | ||
2590 | skb_gro_pull(skb, -ETH_HLEN); | |
2591 | break; | |
5d38a079 | 2592 | |
5d0d9be8 HX |
2593 | case GRO_DROP: |
2594 | err = NET_RX_DROP; | |
2595 | /* fall through */ | |
5d38a079 | 2596 | |
5d0d9be8 HX |
2597 | case GRO_MERGED_FREE: |
2598 | napi_reuse_skb(napi, skb); | |
2599 | break; | |
2600 | } | |
5d38a079 | 2601 | |
5d38a079 HX |
2602 | return err; |
2603 | } | |
5d0d9be8 HX |
2604 | EXPORT_SYMBOL(napi_frags_finish); |
2605 | ||
2606 | int napi_gro_frags(struct napi_struct *napi, struct napi_gro_fraginfo *info) | |
2607 | { | |
2608 | struct sk_buff *skb = napi_fraginfo_skb(napi, info); | |
2609 | ||
2610 | if (!skb) | |
2611 | return NET_RX_DROP; | |
2612 | ||
2613 | return napi_frags_finish(napi, skb, __napi_gro_receive(napi, skb)); | |
2614 | } | |
5d38a079 HX |
2615 | EXPORT_SYMBOL(napi_gro_frags); |
2616 | ||
bea3348e | 2617 | static int process_backlog(struct napi_struct *napi, int quota) |
1da177e4 LT |
2618 | { |
2619 | int work = 0; | |
1da177e4 LT |
2620 | struct softnet_data *queue = &__get_cpu_var(softnet_data); |
2621 | unsigned long start_time = jiffies; | |
2622 | ||
bea3348e SH |
2623 | napi->weight = weight_p; |
2624 | do { | |
1da177e4 | 2625 | struct sk_buff *skb; |
1da177e4 LT |
2626 | |
2627 | local_irq_disable(); | |
2628 | skb = __skb_dequeue(&queue->input_pkt_queue); | |
bea3348e | 2629 | if (!skb) { |
bea3348e | 2630 | local_irq_enable(); |
303c6a02 HX |
2631 | napi_complete(napi); |
2632 | goto out; | |
bea3348e | 2633 | } |
1da177e4 LT |
2634 | local_irq_enable(); |
2635 | ||
d565b0a1 | 2636 | napi_gro_receive(napi, skb); |
bea3348e | 2637 | } while (++work < quota && jiffies == start_time); |
1da177e4 | 2638 | |
d565b0a1 HX |
2639 | napi_gro_flush(napi); |
2640 | ||
303c6a02 | 2641 | out: |
bea3348e SH |
2642 | return work; |
2643 | } | |
1da177e4 | 2644 | |
bea3348e SH |
2645 | /** |
2646 | * __napi_schedule - schedule for receive | |
c4ea43c5 | 2647 | * @n: entry to schedule |
bea3348e SH |
2648 | * |
2649 | * The entry's receive function will be scheduled to run | |
2650 | */ | |
b5606c2d | 2651 | void __napi_schedule(struct napi_struct *n) |
bea3348e SH |
2652 | { |
2653 | unsigned long flags; | |
1da177e4 | 2654 | |
bea3348e SH |
2655 | local_irq_save(flags); |
2656 | list_add_tail(&n->poll_list, &__get_cpu_var(softnet_data).poll_list); | |
2657 | __raise_softirq_irqoff(NET_RX_SOFTIRQ); | |
2658 | local_irq_restore(flags); | |
1da177e4 | 2659 | } |
bea3348e SH |
2660 | EXPORT_SYMBOL(__napi_schedule); |
2661 | ||
d565b0a1 HX |
2662 | void __napi_complete(struct napi_struct *n) |
2663 | { | |
2664 | BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); | |
2665 | BUG_ON(n->gro_list); | |
2666 | ||
2667 | list_del(&n->poll_list); | |
2668 | smp_mb__before_clear_bit(); | |
2669 | clear_bit(NAPI_STATE_SCHED, &n->state); | |
2670 | } | |
2671 | EXPORT_SYMBOL(__napi_complete); | |
2672 | ||
2673 | void napi_complete(struct napi_struct *n) | |
2674 | { | |
2675 | unsigned long flags; | |
2676 | ||
2677 | /* | |
2678 | * don't let napi dequeue from the cpu poll list | |
2679 | * just in case its running on a different cpu | |
2680 | */ | |
2681 | if (unlikely(test_bit(NAPI_STATE_NPSVC, &n->state))) | |
2682 | return; | |
2683 | ||
2684 | napi_gro_flush(n); | |
2685 | local_irq_save(flags); | |
2686 | __napi_complete(n); | |
2687 | local_irq_restore(flags); | |
2688 | } | |
2689 | EXPORT_SYMBOL(napi_complete); | |
2690 | ||
2691 | void netif_napi_add(struct net_device *dev, struct napi_struct *napi, | |
2692 | int (*poll)(struct napi_struct *, int), int weight) | |
2693 | { | |
2694 | INIT_LIST_HEAD(&napi->poll_list); | |
4ae5544f | 2695 | napi->gro_count = 0; |
d565b0a1 | 2696 | napi->gro_list = NULL; |
5d38a079 | 2697 | napi->skb = NULL; |
d565b0a1 HX |
2698 | napi->poll = poll; |
2699 | napi->weight = weight; | |
2700 | list_add(&napi->dev_list, &dev->napi_list); | |
d565b0a1 | 2701 | napi->dev = dev; |
5d38a079 | 2702 | #ifdef CONFIG_NETPOLL |
d565b0a1 HX |
2703 | spin_lock_init(&napi->poll_lock); |
2704 | napi->poll_owner = -1; | |
2705 | #endif | |
2706 | set_bit(NAPI_STATE_SCHED, &napi->state); | |
2707 | } | |
2708 | EXPORT_SYMBOL(netif_napi_add); | |
2709 | ||
2710 | void netif_napi_del(struct napi_struct *napi) | |
2711 | { | |
2712 | struct sk_buff *skb, *next; | |
2713 | ||
d7b06636 | 2714 | list_del_init(&napi->dev_list); |
e4a389a9 | 2715 | kfree_skb(napi->skb); |
d565b0a1 HX |
2716 | |
2717 | for (skb = napi->gro_list; skb; skb = next) { | |
2718 | next = skb->next; | |
2719 | skb->next = NULL; | |
2720 | kfree_skb(skb); | |
2721 | } | |
2722 | ||
2723 | napi->gro_list = NULL; | |
4ae5544f | 2724 | napi->gro_count = 0; |
d565b0a1 HX |
2725 | } |
2726 | EXPORT_SYMBOL(netif_napi_del); | |
2727 | ||
1da177e4 LT |
2728 | |
2729 | static void net_rx_action(struct softirq_action *h) | |
2730 | { | |
bea3348e | 2731 | struct list_head *list = &__get_cpu_var(softnet_data).poll_list; |
24f8b238 | 2732 | unsigned long time_limit = jiffies + 2; |
51b0bded | 2733 | int budget = netdev_budget; |
53fb95d3 MM |
2734 | void *have; |
2735 | ||
1da177e4 LT |
2736 | local_irq_disable(); |
2737 | ||
bea3348e SH |
2738 | while (!list_empty(list)) { |
2739 | struct napi_struct *n; | |
2740 | int work, weight; | |
1da177e4 | 2741 | |
bea3348e | 2742 | /* If softirq window is exhuasted then punt. |
24f8b238 SH |
2743 | * Allow this to run for 2 jiffies since which will allow |
2744 | * an average latency of 1.5/HZ. | |
bea3348e | 2745 | */ |
24f8b238 | 2746 | if (unlikely(budget <= 0 || time_after(jiffies, time_limit))) |
1da177e4 LT |
2747 | goto softnet_break; |
2748 | ||
2749 | local_irq_enable(); | |
2750 | ||
bea3348e SH |
2751 | /* Even though interrupts have been re-enabled, this |
2752 | * access is safe because interrupts can only add new | |
2753 | * entries to the tail of this list, and only ->poll() | |
2754 | * calls can remove this head entry from the list. | |
2755 | */ | |
2756 | n = list_entry(list->next, struct napi_struct, poll_list); | |
1da177e4 | 2757 | |
bea3348e SH |
2758 | have = netpoll_poll_lock(n); |
2759 | ||
2760 | weight = n->weight; | |
2761 | ||
0a7606c1 DM |
2762 | /* This NAPI_STATE_SCHED test is for avoiding a race |
2763 | * with netpoll's poll_napi(). Only the entity which | |
2764 | * obtains the lock and sees NAPI_STATE_SCHED set will | |
2765 | * actually make the ->poll() call. Therefore we avoid | |
2766 | * accidently calling ->poll() when NAPI is not scheduled. | |
2767 | */ | |
2768 | work = 0; | |
2769 | if (test_bit(NAPI_STATE_SCHED, &n->state)) | |
2770 | work = n->poll(n, weight); | |
bea3348e SH |
2771 | |
2772 | WARN_ON_ONCE(work > weight); | |
2773 | ||
2774 | budget -= work; | |
2775 | ||
2776 | local_irq_disable(); | |
2777 | ||
2778 | /* Drivers must not modify the NAPI state if they | |
2779 | * consume the entire weight. In such cases this code | |
2780 | * still "owns" the NAPI instance and therefore can | |
2781 | * move the instance around on the list at-will. | |
2782 | */ | |
fed17f30 DM |
2783 | if (unlikely(work == weight)) { |
2784 | if (unlikely(napi_disable_pending(n))) | |
2785 | __napi_complete(n); | |
2786 | else | |
2787 | list_move_tail(&n->poll_list, list); | |
2788 | } | |
bea3348e SH |
2789 | |
2790 | netpoll_poll_unlock(have); | |
1da177e4 LT |
2791 | } |
2792 | out: | |
515e06c4 | 2793 | local_irq_enable(); |
bea3348e | 2794 | |
db217334 CL |
2795 | #ifdef CONFIG_NET_DMA |
2796 | /* | |
2797 | * There may not be any more sk_buffs coming right now, so push | |
2798 | * any pending DMA copies to hardware | |
2799 | */ | |
2ba05622 | 2800 | dma_issue_pending_all(); |
db217334 | 2801 | #endif |
bea3348e | 2802 | |
1da177e4 LT |
2803 | return; |
2804 | ||
2805 | softnet_break: | |
2806 | __get_cpu_var(netdev_rx_stat).time_squeeze++; | |
2807 | __raise_softirq_irqoff(NET_RX_SOFTIRQ); | |
2808 | goto out; | |
2809 | } | |
2810 | ||
2811 | static gifconf_func_t * gifconf_list [NPROTO]; | |
2812 | ||
2813 | /** | |
2814 | * register_gifconf - register a SIOCGIF handler | |
2815 | * @family: Address family | |
2816 | * @gifconf: Function handler | |
2817 | * | |
2818 | * Register protocol dependent address dumping routines. The handler | |
2819 | * that is passed must not be freed or reused until it has been replaced | |
2820 | * by another handler. | |
2821 | */ | |
2822 | int register_gifconf(unsigned int family, gifconf_func_t * gifconf) | |
2823 | { | |
2824 | if (family >= NPROTO) | |
2825 | return -EINVAL; | |
2826 | gifconf_list[family] = gifconf; | |
2827 | return 0; | |
2828 | } | |
2829 | ||
2830 | ||
2831 | /* | |
2832 | * Map an interface index to its name (SIOCGIFNAME) | |
2833 | */ | |
2834 | ||
2835 | /* | |
2836 | * We need this ioctl for efficient implementation of the | |
2837 | * if_indextoname() function required by the IPv6 API. Without | |
2838 | * it, we would have to search all the interfaces to find a | |
2839 | * match. --pb | |
2840 | */ | |
2841 | ||
881d966b | 2842 | static int dev_ifname(struct net *net, struct ifreq __user *arg) |
1da177e4 LT |
2843 | { |
2844 | struct net_device *dev; | |
2845 | struct ifreq ifr; | |
2846 | ||
2847 | /* | |
2848 | * Fetch the caller's info block. | |
2849 | */ | |
2850 | ||
2851 | if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) | |
2852 | return -EFAULT; | |
2853 | ||
2854 | read_lock(&dev_base_lock); | |
881d966b | 2855 | dev = __dev_get_by_index(net, ifr.ifr_ifindex); |
1da177e4 LT |
2856 | if (!dev) { |
2857 | read_unlock(&dev_base_lock); | |
2858 | return -ENODEV; | |
2859 | } | |
2860 | ||
2861 | strcpy(ifr.ifr_name, dev->name); | |
2862 | read_unlock(&dev_base_lock); | |
2863 | ||
2864 | if (copy_to_user(arg, &ifr, sizeof(struct ifreq))) | |
2865 | return -EFAULT; | |
2866 | return 0; | |
2867 | } | |
2868 | ||
2869 | /* | |
2870 | * Perform a SIOCGIFCONF call. This structure will change | |
2871 | * size eventually, and there is nothing I can do about it. | |
2872 | * Thus we will need a 'compatibility mode'. | |
2873 | */ | |
2874 | ||
881d966b | 2875 | static int dev_ifconf(struct net *net, char __user *arg) |
1da177e4 LT |
2876 | { |
2877 | struct ifconf ifc; | |
2878 | struct net_device *dev; | |
2879 | char __user *pos; | |
2880 | int len; | |
2881 | int total; | |
2882 | int i; | |
2883 | ||
2884 | /* | |
2885 | * Fetch the caller's info block. | |
2886 | */ | |
2887 | ||
2888 | if (copy_from_user(&ifc, arg, sizeof(struct ifconf))) | |
2889 | return -EFAULT; | |
2890 | ||
2891 | pos = ifc.ifc_buf; | |
2892 | len = ifc.ifc_len; | |
2893 | ||
2894 | /* | |
2895 | * Loop over the interfaces, and write an info block for each. | |
2896 | */ | |
2897 | ||
2898 | total = 0; | |
881d966b | 2899 | for_each_netdev(net, dev) { |
1da177e4 LT |
2900 | for (i = 0; i < NPROTO; i++) { |
2901 | if (gifconf_list[i]) { | |
2902 | int done; | |
2903 | if (!pos) | |
2904 | done = gifconf_list[i](dev, NULL, 0); | |
2905 | else | |
2906 | done = gifconf_list[i](dev, pos + total, | |
2907 | len - total); | |
2908 | if (done < 0) | |
2909 | return -EFAULT; | |
2910 | total += done; | |
2911 | } | |
2912 | } | |
4ec93edb | 2913 | } |
1da177e4 LT |
2914 | |
2915 | /* | |
2916 | * All done. Write the updated control block back to the caller. | |
2917 | */ | |
2918 | ifc.ifc_len = total; | |
2919 | ||
2920 | /* | |
2921 | * Both BSD and Solaris return 0 here, so we do too. | |
2922 | */ | |
2923 | return copy_to_user(arg, &ifc, sizeof(struct ifconf)) ? -EFAULT : 0; | |
2924 | } | |
2925 | ||
2926 | #ifdef CONFIG_PROC_FS | |
2927 | /* | |
2928 | * This is invoked by the /proc filesystem handler to display a device | |
2929 | * in detail. | |
2930 | */ | |
7562f876 | 2931 | void *dev_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2932 | __acquires(dev_base_lock) |
1da177e4 | 2933 | { |
e372c414 | 2934 | struct net *net = seq_file_net(seq); |
7562f876 | 2935 | loff_t off; |
1da177e4 | 2936 | struct net_device *dev; |
1da177e4 | 2937 | |
7562f876 PE |
2938 | read_lock(&dev_base_lock); |
2939 | if (!*pos) | |
2940 | return SEQ_START_TOKEN; | |
1da177e4 | 2941 | |
7562f876 | 2942 | off = 1; |
881d966b | 2943 | for_each_netdev(net, dev) |
7562f876 PE |
2944 | if (off++ == *pos) |
2945 | return dev; | |
1da177e4 | 2946 | |
7562f876 | 2947 | return NULL; |
1da177e4 LT |
2948 | } |
2949 | ||
2950 | void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2951 | { | |
e372c414 | 2952 | struct net *net = seq_file_net(seq); |
1da177e4 | 2953 | ++*pos; |
7562f876 | 2954 | return v == SEQ_START_TOKEN ? |
881d966b | 2955 | first_net_device(net) : next_net_device((struct net_device *)v); |
1da177e4 LT |
2956 | } |
2957 | ||
2958 | void dev_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2959 | __releases(dev_base_lock) |
1da177e4 LT |
2960 | { |
2961 | read_unlock(&dev_base_lock); | |
2962 | } | |
2963 | ||
2964 | static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev) | |
2965 | { | |
eeda3fd6 | 2966 | const struct net_device_stats *stats = dev_get_stats(dev); |
1da177e4 | 2967 | |
5a1b5898 RR |
2968 | seq_printf(seq, "%6s:%8lu %7lu %4lu %4lu %4lu %5lu %10lu %9lu " |
2969 | "%8lu %7lu %4lu %4lu %4lu %5lu %7lu %10lu\n", | |
2970 | dev->name, stats->rx_bytes, stats->rx_packets, | |
2971 | stats->rx_errors, | |
2972 | stats->rx_dropped + stats->rx_missed_errors, | |
2973 | stats->rx_fifo_errors, | |
2974 | stats->rx_length_errors + stats->rx_over_errors + | |
2975 | stats->rx_crc_errors + stats->rx_frame_errors, | |
2976 | stats->rx_compressed, stats->multicast, | |
2977 | stats->tx_bytes, stats->tx_packets, | |
2978 | stats->tx_errors, stats->tx_dropped, | |
2979 | stats->tx_fifo_errors, stats->collisions, | |
2980 | stats->tx_carrier_errors + | |
2981 | stats->tx_aborted_errors + | |
2982 | stats->tx_window_errors + | |
2983 | stats->tx_heartbeat_errors, | |
2984 | stats->tx_compressed); | |
1da177e4 LT |
2985 | } |
2986 | ||
2987 | /* | |
2988 | * Called from the PROCfs module. This now uses the new arbitrary sized | |
2989 | * /proc/net interface to create /proc/net/dev | |
2990 | */ | |
2991 | static int dev_seq_show(struct seq_file *seq, void *v) | |
2992 | { | |
2993 | if (v == SEQ_START_TOKEN) | |
2994 | seq_puts(seq, "Inter-| Receive " | |
2995 | " | Transmit\n" | |
2996 | " face |bytes packets errs drop fifo frame " | |
2997 | "compressed multicast|bytes packets errs " | |
2998 | "drop fifo colls carrier compressed\n"); | |
2999 | else | |
3000 | dev_seq_printf_stats(seq, v); | |
3001 | return 0; | |
3002 | } | |
3003 | ||
3004 | static struct netif_rx_stats *softnet_get_online(loff_t *pos) | |
3005 | { | |
3006 | struct netif_rx_stats *rc = NULL; | |
3007 | ||
0c0b0aca | 3008 | while (*pos < nr_cpu_ids) |
4ec93edb | 3009 | if (cpu_online(*pos)) { |
1da177e4 LT |
3010 | rc = &per_cpu(netdev_rx_stat, *pos); |
3011 | break; | |
3012 | } else | |
3013 | ++*pos; | |
3014 | return rc; | |
3015 | } | |
3016 | ||
3017 | static void *softnet_seq_start(struct seq_file *seq, loff_t *pos) | |
3018 | { | |
3019 | return softnet_get_online(pos); | |
3020 | } | |
3021 | ||
3022 | static void *softnet_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3023 | { | |
3024 | ++*pos; | |
3025 | return softnet_get_online(pos); | |
3026 | } | |
3027 | ||
3028 | static void softnet_seq_stop(struct seq_file *seq, void *v) | |
3029 | { | |
3030 | } | |
3031 | ||
3032 | static int softnet_seq_show(struct seq_file *seq, void *v) | |
3033 | { | |
3034 | struct netif_rx_stats *s = v; | |
3035 | ||
3036 | seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n", | |
31aa02c5 | 3037 | s->total, s->dropped, s->time_squeeze, 0, |
c1ebcdb8 SH |
3038 | 0, 0, 0, 0, /* was fastroute */ |
3039 | s->cpu_collision ); | |
1da177e4 LT |
3040 | return 0; |
3041 | } | |
3042 | ||
f690808e | 3043 | static const struct seq_operations dev_seq_ops = { |
1da177e4 LT |
3044 | .start = dev_seq_start, |
3045 | .next = dev_seq_next, | |
3046 | .stop = dev_seq_stop, | |
3047 | .show = dev_seq_show, | |
3048 | }; | |
3049 | ||
3050 | static int dev_seq_open(struct inode *inode, struct file *file) | |
3051 | { | |
e372c414 DL |
3052 | return seq_open_net(inode, file, &dev_seq_ops, |
3053 | sizeof(struct seq_net_private)); | |
1da177e4 LT |
3054 | } |
3055 | ||
9a32144e | 3056 | static const struct file_operations dev_seq_fops = { |
1da177e4 LT |
3057 | .owner = THIS_MODULE, |
3058 | .open = dev_seq_open, | |
3059 | .read = seq_read, | |
3060 | .llseek = seq_lseek, | |
e372c414 | 3061 | .release = seq_release_net, |
1da177e4 LT |
3062 | }; |
3063 | ||
f690808e | 3064 | static const struct seq_operations softnet_seq_ops = { |
1da177e4 LT |
3065 | .start = softnet_seq_start, |
3066 | .next = softnet_seq_next, | |
3067 | .stop = softnet_seq_stop, | |
3068 | .show = softnet_seq_show, | |
3069 | }; | |
3070 | ||
3071 | static int softnet_seq_open(struct inode *inode, struct file *file) | |
3072 | { | |
3073 | return seq_open(file, &softnet_seq_ops); | |
3074 | } | |
3075 | ||
9a32144e | 3076 | static const struct file_operations softnet_seq_fops = { |
1da177e4 LT |
3077 | .owner = THIS_MODULE, |
3078 | .open = softnet_seq_open, | |
3079 | .read = seq_read, | |
3080 | .llseek = seq_lseek, | |
3081 | .release = seq_release, | |
3082 | }; | |
3083 | ||
0e1256ff SH |
3084 | static void *ptype_get_idx(loff_t pos) |
3085 | { | |
3086 | struct packet_type *pt = NULL; | |
3087 | loff_t i = 0; | |
3088 | int t; | |
3089 | ||
3090 | list_for_each_entry_rcu(pt, &ptype_all, list) { | |
3091 | if (i == pos) | |
3092 | return pt; | |
3093 | ++i; | |
3094 | } | |
3095 | ||
82d8a867 | 3096 | for (t = 0; t < PTYPE_HASH_SIZE; t++) { |
0e1256ff SH |
3097 | list_for_each_entry_rcu(pt, &ptype_base[t], list) { |
3098 | if (i == pos) | |
3099 | return pt; | |
3100 | ++i; | |
3101 | } | |
3102 | } | |
3103 | return NULL; | |
3104 | } | |
3105 | ||
3106 | static void *ptype_seq_start(struct seq_file *seq, loff_t *pos) | |
72348a42 | 3107 | __acquires(RCU) |
0e1256ff SH |
3108 | { |
3109 | rcu_read_lock(); | |
3110 | return *pos ? ptype_get_idx(*pos - 1) : SEQ_START_TOKEN; | |
3111 | } | |
3112 | ||
3113 | static void *ptype_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3114 | { | |
3115 | struct packet_type *pt; | |
3116 | struct list_head *nxt; | |
3117 | int hash; | |
3118 | ||
3119 | ++*pos; | |
3120 | if (v == SEQ_START_TOKEN) | |
3121 | return ptype_get_idx(0); | |
3122 | ||
3123 | pt = v; | |
3124 | nxt = pt->list.next; | |
3125 | if (pt->type == htons(ETH_P_ALL)) { | |
3126 | if (nxt != &ptype_all) | |
3127 | goto found; | |
3128 | hash = 0; | |
3129 | nxt = ptype_base[0].next; | |
3130 | } else | |
82d8a867 | 3131 | hash = ntohs(pt->type) & PTYPE_HASH_MASK; |
0e1256ff SH |
3132 | |
3133 | while (nxt == &ptype_base[hash]) { | |
82d8a867 | 3134 | if (++hash >= PTYPE_HASH_SIZE) |
0e1256ff SH |
3135 | return NULL; |
3136 | nxt = ptype_base[hash].next; | |
3137 | } | |
3138 | found: | |
3139 | return list_entry(nxt, struct packet_type, list); | |
3140 | } | |
3141 | ||
3142 | static void ptype_seq_stop(struct seq_file *seq, void *v) | |
72348a42 | 3143 | __releases(RCU) |
0e1256ff SH |
3144 | { |
3145 | rcu_read_unlock(); | |
3146 | } | |
3147 | ||
0e1256ff SH |
3148 | static int ptype_seq_show(struct seq_file *seq, void *v) |
3149 | { | |
3150 | struct packet_type *pt = v; | |
3151 | ||
3152 | if (v == SEQ_START_TOKEN) | |
3153 | seq_puts(seq, "Type Device Function\n"); | |
c346dca1 | 3154 | else if (pt->dev == NULL || dev_net(pt->dev) == seq_file_net(seq)) { |
0e1256ff SH |
3155 | if (pt->type == htons(ETH_P_ALL)) |
3156 | seq_puts(seq, "ALL "); | |
3157 | else | |
3158 | seq_printf(seq, "%04x", ntohs(pt->type)); | |
3159 | ||
908cd2da AD |
3160 | seq_printf(seq, " %-8s %pF\n", |
3161 | pt->dev ? pt->dev->name : "", pt->func); | |
0e1256ff SH |
3162 | } |
3163 | ||
3164 | return 0; | |
3165 | } | |
3166 | ||
3167 | static const struct seq_operations ptype_seq_ops = { | |
3168 | .start = ptype_seq_start, | |
3169 | .next = ptype_seq_next, | |
3170 | .stop = ptype_seq_stop, | |
3171 | .show = ptype_seq_show, | |
3172 | }; | |
3173 | ||
3174 | static int ptype_seq_open(struct inode *inode, struct file *file) | |
3175 | { | |
2feb27db PE |
3176 | return seq_open_net(inode, file, &ptype_seq_ops, |
3177 | sizeof(struct seq_net_private)); | |
0e1256ff SH |
3178 | } |
3179 | ||
3180 | static const struct file_operations ptype_seq_fops = { | |
3181 | .owner = THIS_MODULE, | |
3182 | .open = ptype_seq_open, | |
3183 | .read = seq_read, | |
3184 | .llseek = seq_lseek, | |
2feb27db | 3185 | .release = seq_release_net, |
0e1256ff SH |
3186 | }; |
3187 | ||
3188 | ||
4665079c | 3189 | static int __net_init dev_proc_net_init(struct net *net) |
1da177e4 LT |
3190 | { |
3191 | int rc = -ENOMEM; | |
3192 | ||
881d966b | 3193 | if (!proc_net_fops_create(net, "dev", S_IRUGO, &dev_seq_fops)) |
1da177e4 | 3194 | goto out; |
881d966b | 3195 | if (!proc_net_fops_create(net, "softnet_stat", S_IRUGO, &softnet_seq_fops)) |
1da177e4 | 3196 | goto out_dev; |
881d966b | 3197 | if (!proc_net_fops_create(net, "ptype", S_IRUGO, &ptype_seq_fops)) |
457c4cbc | 3198 | goto out_softnet; |
0e1256ff | 3199 | |
881d966b | 3200 | if (wext_proc_init(net)) |
457c4cbc | 3201 | goto out_ptype; |
1da177e4 LT |
3202 | rc = 0; |
3203 | out: | |
3204 | return rc; | |
457c4cbc | 3205 | out_ptype: |
881d966b | 3206 | proc_net_remove(net, "ptype"); |
1da177e4 | 3207 | out_softnet: |
881d966b | 3208 | proc_net_remove(net, "softnet_stat"); |
1da177e4 | 3209 | out_dev: |
881d966b | 3210 | proc_net_remove(net, "dev"); |
1da177e4 LT |
3211 | goto out; |
3212 | } | |
881d966b | 3213 | |
4665079c | 3214 | static void __net_exit dev_proc_net_exit(struct net *net) |
881d966b EB |
3215 | { |
3216 | wext_proc_exit(net); | |
3217 | ||
3218 | proc_net_remove(net, "ptype"); | |
3219 | proc_net_remove(net, "softnet_stat"); | |
3220 | proc_net_remove(net, "dev"); | |
3221 | } | |
3222 | ||
022cbae6 | 3223 | static struct pernet_operations __net_initdata dev_proc_ops = { |
881d966b EB |
3224 | .init = dev_proc_net_init, |
3225 | .exit = dev_proc_net_exit, | |
3226 | }; | |
3227 | ||
3228 | static int __init dev_proc_init(void) | |
3229 | { | |
3230 | return register_pernet_subsys(&dev_proc_ops); | |
3231 | } | |
1da177e4 LT |
3232 | #else |
3233 | #define dev_proc_init() 0 | |
3234 | #endif /* CONFIG_PROC_FS */ | |
3235 | ||
3236 | ||
3237 | /** | |
3238 | * netdev_set_master - set up master/slave pair | |
3239 | * @slave: slave device | |
3240 | * @master: new master device | |
3241 | * | |
3242 | * Changes the master device of the slave. Pass %NULL to break the | |
3243 | * bonding. The caller must hold the RTNL semaphore. On a failure | |
3244 | * a negative errno code is returned. On success the reference counts | |
3245 | * are adjusted, %RTM_NEWLINK is sent to the routing socket and the | |
3246 | * function returns zero. | |
3247 | */ | |
3248 | int netdev_set_master(struct net_device *slave, struct net_device *master) | |
3249 | { | |
3250 | struct net_device *old = slave->master; | |
3251 | ||
3252 | ASSERT_RTNL(); | |
3253 | ||
3254 | if (master) { | |
3255 | if (old) | |
3256 | return -EBUSY; | |
3257 | dev_hold(master); | |
3258 | } | |
3259 | ||
3260 | slave->master = master; | |
4ec93edb | 3261 | |
1da177e4 LT |
3262 | synchronize_net(); |
3263 | ||
3264 | if (old) | |
3265 | dev_put(old); | |
3266 | ||
3267 | if (master) | |
3268 | slave->flags |= IFF_SLAVE; | |
3269 | else | |
3270 | slave->flags &= ~IFF_SLAVE; | |
3271 | ||
3272 | rtmsg_ifinfo(RTM_NEWLINK, slave, IFF_SLAVE); | |
3273 | return 0; | |
3274 | } | |
3275 | ||
b6c40d68 PM |
3276 | static void dev_change_rx_flags(struct net_device *dev, int flags) |
3277 | { | |
d314774c SH |
3278 | const struct net_device_ops *ops = dev->netdev_ops; |
3279 | ||
3280 | if ((dev->flags & IFF_UP) && ops->ndo_change_rx_flags) | |
3281 | ops->ndo_change_rx_flags(dev, flags); | |
b6c40d68 PM |
3282 | } |
3283 | ||
dad9b335 | 3284 | static int __dev_set_promiscuity(struct net_device *dev, int inc) |
1da177e4 LT |
3285 | { |
3286 | unsigned short old_flags = dev->flags; | |
8192b0c4 DH |
3287 | uid_t uid; |
3288 | gid_t gid; | |
1da177e4 | 3289 | |
24023451 PM |
3290 | ASSERT_RTNL(); |
3291 | ||
dad9b335 WC |
3292 | dev->flags |= IFF_PROMISC; |
3293 | dev->promiscuity += inc; | |
3294 | if (dev->promiscuity == 0) { | |
3295 | /* | |
3296 | * Avoid overflow. | |
3297 | * If inc causes overflow, untouch promisc and return error. | |
3298 | */ | |
3299 | if (inc < 0) | |
3300 | dev->flags &= ~IFF_PROMISC; | |
3301 | else { | |
3302 | dev->promiscuity -= inc; | |
3303 | printk(KERN_WARNING "%s: promiscuity touches roof, " | |
3304 | "set promiscuity failed, promiscuity feature " | |
3305 | "of device might be broken.\n", dev->name); | |
3306 | return -EOVERFLOW; | |
3307 | } | |
3308 | } | |
52609c0b | 3309 | if (dev->flags != old_flags) { |
1da177e4 LT |
3310 | printk(KERN_INFO "device %s %s promiscuous mode\n", |
3311 | dev->name, (dev->flags & IFF_PROMISC) ? "entered" : | |
4ec93edb | 3312 | "left"); |
8192b0c4 DH |
3313 | if (audit_enabled) { |
3314 | current_uid_gid(&uid, &gid); | |
7759db82 KHK |
3315 | audit_log(current->audit_context, GFP_ATOMIC, |
3316 | AUDIT_ANOM_PROMISCUOUS, | |
3317 | "dev=%s prom=%d old_prom=%d auid=%u uid=%u gid=%u ses=%u", | |
3318 | dev->name, (dev->flags & IFF_PROMISC), | |
3319 | (old_flags & IFF_PROMISC), | |
3320 | audit_get_loginuid(current), | |
8192b0c4 | 3321 | uid, gid, |
7759db82 | 3322 | audit_get_sessionid(current)); |
8192b0c4 | 3323 | } |
24023451 | 3324 | |
b6c40d68 | 3325 | dev_change_rx_flags(dev, IFF_PROMISC); |
1da177e4 | 3326 | } |
dad9b335 | 3327 | return 0; |
1da177e4 LT |
3328 | } |
3329 | ||
4417da66 PM |
3330 | /** |
3331 | * dev_set_promiscuity - update promiscuity count on a device | |
3332 | * @dev: device | |
3333 | * @inc: modifier | |
3334 | * | |
3335 | * Add or remove promiscuity from a device. While the count in the device | |
3336 | * remains above zero the interface remains promiscuous. Once it hits zero | |
3337 | * the device reverts back to normal filtering operation. A negative inc | |
3338 | * value is used to drop promiscuity on the device. | |
dad9b335 | 3339 | * Return 0 if successful or a negative errno code on error. |
4417da66 | 3340 | */ |
dad9b335 | 3341 | int dev_set_promiscuity(struct net_device *dev, int inc) |
4417da66 PM |
3342 | { |
3343 | unsigned short old_flags = dev->flags; | |
dad9b335 | 3344 | int err; |
4417da66 | 3345 | |
dad9b335 | 3346 | err = __dev_set_promiscuity(dev, inc); |
4b5a698e | 3347 | if (err < 0) |
dad9b335 | 3348 | return err; |
4417da66 PM |
3349 | if (dev->flags != old_flags) |
3350 | dev_set_rx_mode(dev); | |
dad9b335 | 3351 | return err; |
4417da66 PM |
3352 | } |
3353 | ||
1da177e4 LT |
3354 | /** |
3355 | * dev_set_allmulti - update allmulti count on a device | |
3356 | * @dev: device | |
3357 | * @inc: modifier | |
3358 | * | |
3359 | * Add or remove reception of all multicast frames to a device. While the | |
3360 | * count in the device remains above zero the interface remains listening | |
3361 | * to all interfaces. Once it hits zero the device reverts back to normal | |
3362 | * filtering operation. A negative @inc value is used to drop the counter | |
3363 | * when releasing a resource needing all multicasts. | |
dad9b335 | 3364 | * Return 0 if successful or a negative errno code on error. |
1da177e4 LT |
3365 | */ |
3366 | ||
dad9b335 | 3367 | int dev_set_allmulti(struct net_device *dev, int inc) |
1da177e4 LT |
3368 | { |
3369 | unsigned short old_flags = dev->flags; | |
3370 | ||
24023451 PM |
3371 | ASSERT_RTNL(); |
3372 | ||
1da177e4 | 3373 | dev->flags |= IFF_ALLMULTI; |
dad9b335 WC |
3374 | dev->allmulti += inc; |
3375 | if (dev->allmulti == 0) { | |
3376 | /* | |
3377 | * Avoid overflow. | |
3378 | * If inc causes overflow, untouch allmulti and return error. | |
3379 | */ | |
3380 | if (inc < 0) | |
3381 | dev->flags &= ~IFF_ALLMULTI; | |
3382 | else { | |
3383 | dev->allmulti -= inc; | |
3384 | printk(KERN_WARNING "%s: allmulti touches roof, " | |
3385 | "set allmulti failed, allmulti feature of " | |
3386 | "device might be broken.\n", dev->name); | |
3387 | return -EOVERFLOW; | |
3388 | } | |
3389 | } | |
24023451 | 3390 | if (dev->flags ^ old_flags) { |
b6c40d68 | 3391 | dev_change_rx_flags(dev, IFF_ALLMULTI); |
4417da66 | 3392 | dev_set_rx_mode(dev); |
24023451 | 3393 | } |
dad9b335 | 3394 | return 0; |
4417da66 PM |
3395 | } |
3396 | ||
3397 | /* | |
3398 | * Upload unicast and multicast address lists to device and | |
3399 | * configure RX filtering. When the device doesn't support unicast | |
53ccaae1 | 3400 | * filtering it is put in promiscuous mode while unicast addresses |
4417da66 PM |
3401 | * are present. |
3402 | */ | |
3403 | void __dev_set_rx_mode(struct net_device *dev) | |
3404 | { | |
d314774c SH |
3405 | const struct net_device_ops *ops = dev->netdev_ops; |
3406 | ||
4417da66 PM |
3407 | /* dev_open will call this function so the list will stay sane. */ |
3408 | if (!(dev->flags&IFF_UP)) | |
3409 | return; | |
3410 | ||
3411 | if (!netif_device_present(dev)) | |
40b77c94 | 3412 | return; |
4417da66 | 3413 | |
d314774c SH |
3414 | if (ops->ndo_set_rx_mode) |
3415 | ops->ndo_set_rx_mode(dev); | |
4417da66 PM |
3416 | else { |
3417 | /* Unicast addresses changes may only happen under the rtnl, | |
3418 | * therefore calling __dev_set_promiscuity here is safe. | |
3419 | */ | |
3420 | if (dev->uc_count > 0 && !dev->uc_promisc) { | |
3421 | __dev_set_promiscuity(dev, 1); | |
3422 | dev->uc_promisc = 1; | |
3423 | } else if (dev->uc_count == 0 && dev->uc_promisc) { | |
3424 | __dev_set_promiscuity(dev, -1); | |
3425 | dev->uc_promisc = 0; | |
3426 | } | |
3427 | ||
d314774c SH |
3428 | if (ops->ndo_set_multicast_list) |
3429 | ops->ndo_set_multicast_list(dev); | |
4417da66 PM |
3430 | } |
3431 | } | |
3432 | ||
3433 | void dev_set_rx_mode(struct net_device *dev) | |
3434 | { | |
b9e40857 | 3435 | netif_addr_lock_bh(dev); |
4417da66 | 3436 | __dev_set_rx_mode(dev); |
b9e40857 | 3437 | netif_addr_unlock_bh(dev); |
1da177e4 LT |
3438 | } |
3439 | ||
61cbc2fc PM |
3440 | int __dev_addr_delete(struct dev_addr_list **list, int *count, |
3441 | void *addr, int alen, int glbl) | |
bf742482 PM |
3442 | { |
3443 | struct dev_addr_list *da; | |
3444 | ||
3445 | for (; (da = *list) != NULL; list = &da->next) { | |
3446 | if (memcmp(da->da_addr, addr, da->da_addrlen) == 0 && | |
3447 | alen == da->da_addrlen) { | |
3448 | if (glbl) { | |
3449 | int old_glbl = da->da_gusers; | |
3450 | da->da_gusers = 0; | |
3451 | if (old_glbl == 0) | |
3452 | break; | |
3453 | } | |
3454 | if (--da->da_users) | |
3455 | return 0; | |
3456 | ||
3457 | *list = da->next; | |
3458 | kfree(da); | |
61cbc2fc | 3459 | (*count)--; |
bf742482 PM |
3460 | return 0; |
3461 | } | |
3462 | } | |
3463 | return -ENOENT; | |
3464 | } | |
3465 | ||
61cbc2fc PM |
3466 | int __dev_addr_add(struct dev_addr_list **list, int *count, |
3467 | void *addr, int alen, int glbl) | |
bf742482 PM |
3468 | { |
3469 | struct dev_addr_list *da; | |
3470 | ||
3471 | for (da = *list; da != NULL; da = da->next) { | |
3472 | if (memcmp(da->da_addr, addr, da->da_addrlen) == 0 && | |
3473 | da->da_addrlen == alen) { | |
3474 | if (glbl) { | |
3475 | int old_glbl = da->da_gusers; | |
3476 | da->da_gusers = 1; | |
3477 | if (old_glbl) | |
3478 | return 0; | |
3479 | } | |
3480 | da->da_users++; | |
3481 | return 0; | |
3482 | } | |
3483 | } | |
3484 | ||
12aa343a | 3485 | da = kzalloc(sizeof(*da), GFP_ATOMIC); |
bf742482 PM |
3486 | if (da == NULL) |
3487 | return -ENOMEM; | |
3488 | memcpy(da->da_addr, addr, alen); | |
3489 | da->da_addrlen = alen; | |
3490 | da->da_users = 1; | |
3491 | da->da_gusers = glbl ? 1 : 0; | |
3492 | da->next = *list; | |
3493 | *list = da; | |
61cbc2fc | 3494 | (*count)++; |
bf742482 PM |
3495 | return 0; |
3496 | } | |
3497 | ||
4417da66 PM |
3498 | /** |
3499 | * dev_unicast_delete - Release secondary unicast address. | |
3500 | * @dev: device | |
0ed72ec4 RD |
3501 | * @addr: address to delete |
3502 | * @alen: length of @addr | |
4417da66 PM |
3503 | * |
3504 | * Release reference to a secondary unicast address and remove it | |
0ed72ec4 | 3505 | * from the device if the reference count drops to zero. |
4417da66 PM |
3506 | * |
3507 | * The caller must hold the rtnl_mutex. | |
3508 | */ | |
3509 | int dev_unicast_delete(struct net_device *dev, void *addr, int alen) | |
3510 | { | |
3511 | int err; | |
3512 | ||
3513 | ASSERT_RTNL(); | |
3514 | ||
b9e40857 | 3515 | netif_addr_lock_bh(dev); |
61cbc2fc PM |
3516 | err = __dev_addr_delete(&dev->uc_list, &dev->uc_count, addr, alen, 0); |
3517 | if (!err) | |
4417da66 | 3518 | __dev_set_rx_mode(dev); |
b9e40857 | 3519 | netif_addr_unlock_bh(dev); |
4417da66 PM |
3520 | return err; |
3521 | } | |
3522 | EXPORT_SYMBOL(dev_unicast_delete); | |
3523 | ||
3524 | /** | |
3525 | * dev_unicast_add - add a secondary unicast address | |
3526 | * @dev: device | |
5dbaec5d | 3527 | * @addr: address to add |
0ed72ec4 | 3528 | * @alen: length of @addr |
4417da66 PM |
3529 | * |
3530 | * Add a secondary unicast address to the device or increase | |
3531 | * the reference count if it already exists. | |
3532 | * | |
3533 | * The caller must hold the rtnl_mutex. | |
3534 | */ | |
3535 | int dev_unicast_add(struct net_device *dev, void *addr, int alen) | |
3536 | { | |
3537 | int err; | |
3538 | ||
3539 | ASSERT_RTNL(); | |
3540 | ||
b9e40857 | 3541 | netif_addr_lock_bh(dev); |
61cbc2fc PM |
3542 | err = __dev_addr_add(&dev->uc_list, &dev->uc_count, addr, alen, 0); |
3543 | if (!err) | |
4417da66 | 3544 | __dev_set_rx_mode(dev); |
b9e40857 | 3545 | netif_addr_unlock_bh(dev); |
4417da66 PM |
3546 | return err; |
3547 | } | |
3548 | EXPORT_SYMBOL(dev_unicast_add); | |
3549 | ||
e83a2ea8 CL |
3550 | int __dev_addr_sync(struct dev_addr_list **to, int *to_count, |
3551 | struct dev_addr_list **from, int *from_count) | |
3552 | { | |
3553 | struct dev_addr_list *da, *next; | |
3554 | int err = 0; | |
3555 | ||
3556 | da = *from; | |
3557 | while (da != NULL) { | |
3558 | next = da->next; | |
3559 | if (!da->da_synced) { | |
3560 | err = __dev_addr_add(to, to_count, | |
3561 | da->da_addr, da->da_addrlen, 0); | |
3562 | if (err < 0) | |
3563 | break; | |
3564 | da->da_synced = 1; | |
3565 | da->da_users++; | |
3566 | } else if (da->da_users == 1) { | |
3567 | __dev_addr_delete(to, to_count, | |
3568 | da->da_addr, da->da_addrlen, 0); | |
3569 | __dev_addr_delete(from, from_count, | |
3570 | da->da_addr, da->da_addrlen, 0); | |
3571 | } | |
3572 | da = next; | |
3573 | } | |
3574 | return err; | |
3575 | } | |
3576 | ||
3577 | void __dev_addr_unsync(struct dev_addr_list **to, int *to_count, | |
3578 | struct dev_addr_list **from, int *from_count) | |
3579 | { | |
3580 | struct dev_addr_list *da, *next; | |
3581 | ||
3582 | da = *from; | |
3583 | while (da != NULL) { | |
3584 | next = da->next; | |
3585 | if (da->da_synced) { | |
3586 | __dev_addr_delete(to, to_count, | |
3587 | da->da_addr, da->da_addrlen, 0); | |
3588 | da->da_synced = 0; | |
3589 | __dev_addr_delete(from, from_count, | |
3590 | da->da_addr, da->da_addrlen, 0); | |
3591 | } | |
3592 | da = next; | |
3593 | } | |
3594 | } | |
3595 | ||
3596 | /** | |
3597 | * dev_unicast_sync - Synchronize device's unicast list to another device | |
3598 | * @to: destination device | |
3599 | * @from: source device | |
3600 | * | |
3601 | * Add newly added addresses to the destination device and release | |
3602 | * addresses that have no users left. The source device must be | |
3603 | * locked by netif_tx_lock_bh. | |
3604 | * | |
3605 | * This function is intended to be called from the dev->set_rx_mode | |
3606 | * function of layered software devices. | |
3607 | */ | |
3608 | int dev_unicast_sync(struct net_device *to, struct net_device *from) | |
3609 | { | |
3610 | int err = 0; | |
3611 | ||
b9e40857 | 3612 | netif_addr_lock_bh(to); |
e83a2ea8 CL |
3613 | err = __dev_addr_sync(&to->uc_list, &to->uc_count, |
3614 | &from->uc_list, &from->uc_count); | |
3615 | if (!err) | |
3616 | __dev_set_rx_mode(to); | |
b9e40857 | 3617 | netif_addr_unlock_bh(to); |
e83a2ea8 CL |
3618 | return err; |
3619 | } | |
3620 | EXPORT_SYMBOL(dev_unicast_sync); | |
3621 | ||
3622 | /** | |
bc2cda1e | 3623 | * dev_unicast_unsync - Remove synchronized addresses from the destination device |
e83a2ea8 CL |
3624 | * @to: destination device |
3625 | * @from: source device | |
3626 | * | |
3627 | * Remove all addresses that were added to the destination device by | |
3628 | * dev_unicast_sync(). This function is intended to be called from the | |
3629 | * dev->stop function of layered software devices. | |
3630 | */ | |
3631 | void dev_unicast_unsync(struct net_device *to, struct net_device *from) | |
3632 | { | |
b9e40857 | 3633 | netif_addr_lock_bh(from); |
e308a5d8 | 3634 | netif_addr_lock(to); |
e83a2ea8 CL |
3635 | |
3636 | __dev_addr_unsync(&to->uc_list, &to->uc_count, | |
3637 | &from->uc_list, &from->uc_count); | |
3638 | __dev_set_rx_mode(to); | |
3639 | ||
e308a5d8 | 3640 | netif_addr_unlock(to); |
b9e40857 | 3641 | netif_addr_unlock_bh(from); |
e83a2ea8 CL |
3642 | } |
3643 | EXPORT_SYMBOL(dev_unicast_unsync); | |
3644 | ||
12972621 DC |
3645 | static void __dev_addr_discard(struct dev_addr_list **list) |
3646 | { | |
3647 | struct dev_addr_list *tmp; | |
3648 | ||
3649 | while (*list != NULL) { | |
3650 | tmp = *list; | |
3651 | *list = tmp->next; | |
3652 | if (tmp->da_users > tmp->da_gusers) | |
3653 | printk("__dev_addr_discard: address leakage! " | |
3654 | "da_users=%d\n", tmp->da_users); | |
3655 | kfree(tmp); | |
3656 | } | |
3657 | } | |
3658 | ||
26cc2522 | 3659 | static void dev_addr_discard(struct net_device *dev) |
4417da66 | 3660 | { |
b9e40857 | 3661 | netif_addr_lock_bh(dev); |
26cc2522 | 3662 | |
4417da66 PM |
3663 | __dev_addr_discard(&dev->uc_list); |
3664 | dev->uc_count = 0; | |
4417da66 | 3665 | |
456ad75c DC |
3666 | __dev_addr_discard(&dev->mc_list); |
3667 | dev->mc_count = 0; | |
26cc2522 | 3668 | |
b9e40857 | 3669 | netif_addr_unlock_bh(dev); |
456ad75c DC |
3670 | } |
3671 | ||
f0db275a SH |
3672 | /** |
3673 | * dev_get_flags - get flags reported to userspace | |
3674 | * @dev: device | |
3675 | * | |
3676 | * Get the combination of flag bits exported through APIs to userspace. | |
3677 | */ | |
1da177e4 LT |
3678 | unsigned dev_get_flags(const struct net_device *dev) |
3679 | { | |
3680 | unsigned flags; | |
3681 | ||
3682 | flags = (dev->flags & ~(IFF_PROMISC | | |
3683 | IFF_ALLMULTI | | |
b00055aa SR |
3684 | IFF_RUNNING | |
3685 | IFF_LOWER_UP | | |
3686 | IFF_DORMANT)) | | |
1da177e4 LT |
3687 | (dev->gflags & (IFF_PROMISC | |
3688 | IFF_ALLMULTI)); | |
3689 | ||
b00055aa SR |
3690 | if (netif_running(dev)) { |
3691 | if (netif_oper_up(dev)) | |
3692 | flags |= IFF_RUNNING; | |
3693 | if (netif_carrier_ok(dev)) | |
3694 | flags |= IFF_LOWER_UP; | |
3695 | if (netif_dormant(dev)) | |
3696 | flags |= IFF_DORMANT; | |
3697 | } | |
1da177e4 LT |
3698 | |
3699 | return flags; | |
3700 | } | |
3701 | ||
f0db275a SH |
3702 | /** |
3703 | * dev_change_flags - change device settings | |
3704 | * @dev: device | |
3705 | * @flags: device state flags | |
3706 | * | |
3707 | * Change settings on device based state flags. The flags are | |
3708 | * in the userspace exported format. | |
3709 | */ | |
1da177e4 LT |
3710 | int dev_change_flags(struct net_device *dev, unsigned flags) |
3711 | { | |
7c355f53 | 3712 | int ret, changes; |
1da177e4 LT |
3713 | int old_flags = dev->flags; |
3714 | ||
24023451 PM |
3715 | ASSERT_RTNL(); |
3716 | ||
1da177e4 LT |
3717 | /* |
3718 | * Set the flags on our device. | |
3719 | */ | |
3720 | ||
3721 | dev->flags = (flags & (IFF_DEBUG | IFF_NOTRAILERS | IFF_NOARP | | |
3722 | IFF_DYNAMIC | IFF_MULTICAST | IFF_PORTSEL | | |
3723 | IFF_AUTOMEDIA)) | | |
3724 | (dev->flags & (IFF_UP | IFF_VOLATILE | IFF_PROMISC | | |
3725 | IFF_ALLMULTI)); | |
3726 | ||
3727 | /* | |
3728 | * Load in the correct multicast list now the flags have changed. | |
3729 | */ | |
3730 | ||
b6c40d68 PM |
3731 | if ((old_flags ^ flags) & IFF_MULTICAST) |
3732 | dev_change_rx_flags(dev, IFF_MULTICAST); | |
24023451 | 3733 | |
4417da66 | 3734 | dev_set_rx_mode(dev); |
1da177e4 LT |
3735 | |
3736 | /* | |
3737 | * Have we downed the interface. We handle IFF_UP ourselves | |
3738 | * according to user attempts to set it, rather than blindly | |
3739 | * setting it. | |
3740 | */ | |
3741 | ||
3742 | ret = 0; | |
3743 | if ((old_flags ^ flags) & IFF_UP) { /* Bit is different ? */ | |
3744 | ret = ((old_flags & IFF_UP) ? dev_close : dev_open)(dev); | |
3745 | ||
3746 | if (!ret) | |
4417da66 | 3747 | dev_set_rx_mode(dev); |
1da177e4 LT |
3748 | } |
3749 | ||
3750 | if (dev->flags & IFF_UP && | |
3751 | ((old_flags ^ dev->flags) &~ (IFF_UP | IFF_PROMISC | IFF_ALLMULTI | | |
3752 | IFF_VOLATILE))) | |
056925ab | 3753 | call_netdevice_notifiers(NETDEV_CHANGE, dev); |
1da177e4 LT |
3754 | |
3755 | if ((flags ^ dev->gflags) & IFF_PROMISC) { | |
3756 | int inc = (flags & IFF_PROMISC) ? +1 : -1; | |
3757 | dev->gflags ^= IFF_PROMISC; | |
3758 | dev_set_promiscuity(dev, inc); | |
3759 | } | |
3760 | ||
3761 | /* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI | |
3762 | is important. Some (broken) drivers set IFF_PROMISC, when | |
3763 | IFF_ALLMULTI is requested not asking us and not reporting. | |
3764 | */ | |
3765 | if ((flags ^ dev->gflags) & IFF_ALLMULTI) { | |
3766 | int inc = (flags & IFF_ALLMULTI) ? +1 : -1; | |
3767 | dev->gflags ^= IFF_ALLMULTI; | |
3768 | dev_set_allmulti(dev, inc); | |
3769 | } | |
3770 | ||
7c355f53 TG |
3771 | /* Exclude state transition flags, already notified */ |
3772 | changes = (old_flags ^ dev->flags) & ~(IFF_UP | IFF_RUNNING); | |
3773 | if (changes) | |
3774 | rtmsg_ifinfo(RTM_NEWLINK, dev, changes); | |
1da177e4 LT |
3775 | |
3776 | return ret; | |
3777 | } | |
3778 | ||
f0db275a SH |
3779 | /** |
3780 | * dev_set_mtu - Change maximum transfer unit | |
3781 | * @dev: device | |
3782 | * @new_mtu: new transfer unit | |
3783 | * | |
3784 | * Change the maximum transfer size of the network device. | |
3785 | */ | |
1da177e4 LT |
3786 | int dev_set_mtu(struct net_device *dev, int new_mtu) |
3787 | { | |
d314774c | 3788 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
3789 | int err; |
3790 | ||
3791 | if (new_mtu == dev->mtu) | |
3792 | return 0; | |
3793 | ||
3794 | /* MTU must be positive. */ | |
3795 | if (new_mtu < 0) | |
3796 | return -EINVAL; | |
3797 | ||
3798 | if (!netif_device_present(dev)) | |
3799 | return -ENODEV; | |
3800 | ||
3801 | err = 0; | |
d314774c SH |
3802 | if (ops->ndo_change_mtu) |
3803 | err = ops->ndo_change_mtu(dev, new_mtu); | |
1da177e4 LT |
3804 | else |
3805 | dev->mtu = new_mtu; | |
d314774c | 3806 | |
1da177e4 | 3807 | if (!err && dev->flags & IFF_UP) |
056925ab | 3808 | call_netdevice_notifiers(NETDEV_CHANGEMTU, dev); |
1da177e4 LT |
3809 | return err; |
3810 | } | |
3811 | ||
f0db275a SH |
3812 | /** |
3813 | * dev_set_mac_address - Change Media Access Control Address | |
3814 | * @dev: device | |
3815 | * @sa: new address | |
3816 | * | |
3817 | * Change the hardware (MAC) address of the device | |
3818 | */ | |
1da177e4 LT |
3819 | int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa) |
3820 | { | |
d314774c | 3821 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
3822 | int err; |
3823 | ||
d314774c | 3824 | if (!ops->ndo_set_mac_address) |
1da177e4 LT |
3825 | return -EOPNOTSUPP; |
3826 | if (sa->sa_family != dev->type) | |
3827 | return -EINVAL; | |
3828 | if (!netif_device_present(dev)) | |
3829 | return -ENODEV; | |
d314774c | 3830 | err = ops->ndo_set_mac_address(dev, sa); |
1da177e4 | 3831 | if (!err) |
056925ab | 3832 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); |
1da177e4 LT |
3833 | return err; |
3834 | } | |
3835 | ||
3836 | /* | |
14e3e079 | 3837 | * Perform the SIOCxIFxxx calls, inside read_lock(dev_base_lock) |
1da177e4 | 3838 | */ |
14e3e079 | 3839 | static int dev_ifsioc_locked(struct net *net, struct ifreq *ifr, unsigned int cmd) |
1da177e4 LT |
3840 | { |
3841 | int err; | |
881d966b | 3842 | struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name); |
1da177e4 LT |
3843 | |
3844 | if (!dev) | |
3845 | return -ENODEV; | |
3846 | ||
3847 | switch (cmd) { | |
3848 | case SIOCGIFFLAGS: /* Get interface flags */ | |
3849 | ifr->ifr_flags = dev_get_flags(dev); | |
3850 | return 0; | |
3851 | ||
1da177e4 LT |
3852 | case SIOCGIFMETRIC: /* Get the metric on the interface |
3853 | (currently unused) */ | |
3854 | ifr->ifr_metric = 0; | |
3855 | return 0; | |
3856 | ||
1da177e4 LT |
3857 | case SIOCGIFMTU: /* Get the MTU of a device */ |
3858 | ifr->ifr_mtu = dev->mtu; | |
3859 | return 0; | |
3860 | ||
1da177e4 LT |
3861 | case SIOCGIFHWADDR: |
3862 | if (!dev->addr_len) | |
3863 | memset(ifr->ifr_hwaddr.sa_data, 0, sizeof ifr->ifr_hwaddr.sa_data); | |
3864 | else | |
3865 | memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr, | |
3866 | min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); | |
3867 | ifr->ifr_hwaddr.sa_family = dev->type; | |
3868 | return 0; | |
3869 | ||
14e3e079 JG |
3870 | case SIOCGIFSLAVE: |
3871 | err = -EINVAL; | |
3872 | break; | |
3873 | ||
3874 | case SIOCGIFMAP: | |
3875 | ifr->ifr_map.mem_start = dev->mem_start; | |
3876 | ifr->ifr_map.mem_end = dev->mem_end; | |
3877 | ifr->ifr_map.base_addr = dev->base_addr; | |
3878 | ifr->ifr_map.irq = dev->irq; | |
3879 | ifr->ifr_map.dma = dev->dma; | |
3880 | ifr->ifr_map.port = dev->if_port; | |
3881 | return 0; | |
3882 | ||
3883 | case SIOCGIFINDEX: | |
3884 | ifr->ifr_ifindex = dev->ifindex; | |
3885 | return 0; | |
3886 | ||
3887 | case SIOCGIFTXQLEN: | |
3888 | ifr->ifr_qlen = dev->tx_queue_len; | |
3889 | return 0; | |
3890 | ||
3891 | default: | |
3892 | /* dev_ioctl() should ensure this case | |
3893 | * is never reached | |
3894 | */ | |
3895 | WARN_ON(1); | |
3896 | err = -EINVAL; | |
3897 | break; | |
3898 | ||
3899 | } | |
3900 | return err; | |
3901 | } | |
3902 | ||
3903 | /* | |
3904 | * Perform the SIOCxIFxxx calls, inside rtnl_lock() | |
3905 | */ | |
3906 | static int dev_ifsioc(struct net *net, struct ifreq *ifr, unsigned int cmd) | |
3907 | { | |
3908 | int err; | |
3909 | struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name); | |
5f2f6da7 | 3910 | const struct net_device_ops *ops; |
14e3e079 JG |
3911 | |
3912 | if (!dev) | |
3913 | return -ENODEV; | |
3914 | ||
5f2f6da7 JP |
3915 | ops = dev->netdev_ops; |
3916 | ||
14e3e079 JG |
3917 | switch (cmd) { |
3918 | case SIOCSIFFLAGS: /* Set interface flags */ | |
3919 | return dev_change_flags(dev, ifr->ifr_flags); | |
3920 | ||
3921 | case SIOCSIFMETRIC: /* Set the metric on the interface | |
3922 | (currently unused) */ | |
3923 | return -EOPNOTSUPP; | |
3924 | ||
3925 | case SIOCSIFMTU: /* Set the MTU of a device */ | |
3926 | return dev_set_mtu(dev, ifr->ifr_mtu); | |
3927 | ||
1da177e4 LT |
3928 | case SIOCSIFHWADDR: |
3929 | return dev_set_mac_address(dev, &ifr->ifr_hwaddr); | |
3930 | ||
3931 | case SIOCSIFHWBROADCAST: | |
3932 | if (ifr->ifr_hwaddr.sa_family != dev->type) | |
3933 | return -EINVAL; | |
3934 | memcpy(dev->broadcast, ifr->ifr_hwaddr.sa_data, | |
3935 | min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); | |
056925ab | 3936 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); |
1da177e4 LT |
3937 | return 0; |
3938 | ||
1da177e4 | 3939 | case SIOCSIFMAP: |
d314774c | 3940 | if (ops->ndo_set_config) { |
1da177e4 LT |
3941 | if (!netif_device_present(dev)) |
3942 | return -ENODEV; | |
d314774c | 3943 | return ops->ndo_set_config(dev, &ifr->ifr_map); |
1da177e4 LT |
3944 | } |
3945 | return -EOPNOTSUPP; | |
3946 | ||
3947 | case SIOCADDMULTI: | |
d314774c | 3948 | if ((!ops->ndo_set_multicast_list && !ops->ndo_set_rx_mode) || |
1da177e4 LT |
3949 | ifr->ifr_hwaddr.sa_family != AF_UNSPEC) |
3950 | return -EINVAL; | |
3951 | if (!netif_device_present(dev)) | |
3952 | return -ENODEV; | |
3953 | return dev_mc_add(dev, ifr->ifr_hwaddr.sa_data, | |
3954 | dev->addr_len, 1); | |
3955 | ||
3956 | case SIOCDELMULTI: | |
d314774c | 3957 | if ((!ops->ndo_set_multicast_list && !ops->ndo_set_rx_mode) || |
1da177e4 LT |
3958 | ifr->ifr_hwaddr.sa_family != AF_UNSPEC) |
3959 | return -EINVAL; | |
3960 | if (!netif_device_present(dev)) | |
3961 | return -ENODEV; | |
3962 | return dev_mc_delete(dev, ifr->ifr_hwaddr.sa_data, | |
3963 | dev->addr_len, 1); | |
3964 | ||
1da177e4 LT |
3965 | case SIOCSIFTXQLEN: |
3966 | if (ifr->ifr_qlen < 0) | |
3967 | return -EINVAL; | |
3968 | dev->tx_queue_len = ifr->ifr_qlen; | |
3969 | return 0; | |
3970 | ||
3971 | case SIOCSIFNAME: | |
3972 | ifr->ifr_newname[IFNAMSIZ-1] = '\0'; | |
3973 | return dev_change_name(dev, ifr->ifr_newname); | |
3974 | ||
3975 | /* | |
3976 | * Unknown or private ioctl | |
3977 | */ | |
3978 | ||
3979 | default: | |
3980 | if ((cmd >= SIOCDEVPRIVATE && | |
3981 | cmd <= SIOCDEVPRIVATE + 15) || | |
3982 | cmd == SIOCBONDENSLAVE || | |
3983 | cmd == SIOCBONDRELEASE || | |
3984 | cmd == SIOCBONDSETHWADDR || | |
3985 | cmd == SIOCBONDSLAVEINFOQUERY || | |
3986 | cmd == SIOCBONDINFOQUERY || | |
3987 | cmd == SIOCBONDCHANGEACTIVE || | |
3988 | cmd == SIOCGMIIPHY || | |
3989 | cmd == SIOCGMIIREG || | |
3990 | cmd == SIOCSMIIREG || | |
3991 | cmd == SIOCBRADDIF || | |
3992 | cmd == SIOCBRDELIF || | |
d24fff22 | 3993 | cmd == SIOCSHWTSTAMP || |
1da177e4 LT |
3994 | cmd == SIOCWANDEV) { |
3995 | err = -EOPNOTSUPP; | |
d314774c | 3996 | if (ops->ndo_do_ioctl) { |
1da177e4 | 3997 | if (netif_device_present(dev)) |
d314774c | 3998 | err = ops->ndo_do_ioctl(dev, ifr, cmd); |
1da177e4 LT |
3999 | else |
4000 | err = -ENODEV; | |
4001 | } | |
4002 | } else | |
4003 | err = -EINVAL; | |
4004 | ||
4005 | } | |
4006 | return err; | |
4007 | } | |
4008 | ||
4009 | /* | |
4010 | * This function handles all "interface"-type I/O control requests. The actual | |
4011 | * 'doing' part of this is dev_ifsioc above. | |
4012 | */ | |
4013 | ||
4014 | /** | |
4015 | * dev_ioctl - network device ioctl | |
c4ea43c5 | 4016 | * @net: the applicable net namespace |
1da177e4 LT |
4017 | * @cmd: command to issue |
4018 | * @arg: pointer to a struct ifreq in user space | |
4019 | * | |
4020 | * Issue ioctl functions to devices. This is normally called by the | |
4021 | * user space syscall interfaces but can sometimes be useful for | |
4022 | * other purposes. The return value is the return from the syscall if | |
4023 | * positive or a negative errno code on error. | |
4024 | */ | |
4025 | ||
881d966b | 4026 | int dev_ioctl(struct net *net, unsigned int cmd, void __user *arg) |
1da177e4 LT |
4027 | { |
4028 | struct ifreq ifr; | |
4029 | int ret; | |
4030 | char *colon; | |
4031 | ||
4032 | /* One special case: SIOCGIFCONF takes ifconf argument | |
4033 | and requires shared lock, because it sleeps writing | |
4034 | to user space. | |
4035 | */ | |
4036 | ||
4037 | if (cmd == SIOCGIFCONF) { | |
6756ae4b | 4038 | rtnl_lock(); |
881d966b | 4039 | ret = dev_ifconf(net, (char __user *) arg); |
6756ae4b | 4040 | rtnl_unlock(); |
1da177e4 LT |
4041 | return ret; |
4042 | } | |
4043 | if (cmd == SIOCGIFNAME) | |
881d966b | 4044 | return dev_ifname(net, (struct ifreq __user *)arg); |
1da177e4 LT |
4045 | |
4046 | if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) | |
4047 | return -EFAULT; | |
4048 | ||
4049 | ifr.ifr_name[IFNAMSIZ-1] = 0; | |
4050 | ||
4051 | colon = strchr(ifr.ifr_name, ':'); | |
4052 | if (colon) | |
4053 | *colon = 0; | |
4054 | ||
4055 | /* | |
4056 | * See which interface the caller is talking about. | |
4057 | */ | |
4058 | ||
4059 | switch (cmd) { | |
4060 | /* | |
4061 | * These ioctl calls: | |
4062 | * - can be done by all. | |
4063 | * - atomic and do not require locking. | |
4064 | * - return a value | |
4065 | */ | |
4066 | case SIOCGIFFLAGS: | |
4067 | case SIOCGIFMETRIC: | |
4068 | case SIOCGIFMTU: | |
4069 | case SIOCGIFHWADDR: | |
4070 | case SIOCGIFSLAVE: | |
4071 | case SIOCGIFMAP: | |
4072 | case SIOCGIFINDEX: | |
4073 | case SIOCGIFTXQLEN: | |
881d966b | 4074 | dev_load(net, ifr.ifr_name); |
1da177e4 | 4075 | read_lock(&dev_base_lock); |
14e3e079 | 4076 | ret = dev_ifsioc_locked(net, &ifr, cmd); |
1da177e4 LT |
4077 | read_unlock(&dev_base_lock); |
4078 | if (!ret) { | |
4079 | if (colon) | |
4080 | *colon = ':'; | |
4081 | if (copy_to_user(arg, &ifr, | |
4082 | sizeof(struct ifreq))) | |
4083 | ret = -EFAULT; | |
4084 | } | |
4085 | return ret; | |
4086 | ||
4087 | case SIOCETHTOOL: | |
881d966b | 4088 | dev_load(net, ifr.ifr_name); |
1da177e4 | 4089 | rtnl_lock(); |
881d966b | 4090 | ret = dev_ethtool(net, &ifr); |
1da177e4 LT |
4091 | rtnl_unlock(); |
4092 | if (!ret) { | |
4093 | if (colon) | |
4094 | *colon = ':'; | |
4095 | if (copy_to_user(arg, &ifr, | |
4096 | sizeof(struct ifreq))) | |
4097 | ret = -EFAULT; | |
4098 | } | |
4099 | return ret; | |
4100 | ||
4101 | /* | |
4102 | * These ioctl calls: | |
4103 | * - require superuser power. | |
4104 | * - require strict serialization. | |
4105 | * - return a value | |
4106 | */ | |
4107 | case SIOCGMIIPHY: | |
4108 | case SIOCGMIIREG: | |
4109 | case SIOCSIFNAME: | |
4110 | if (!capable(CAP_NET_ADMIN)) | |
4111 | return -EPERM; | |
881d966b | 4112 | dev_load(net, ifr.ifr_name); |
1da177e4 | 4113 | rtnl_lock(); |
881d966b | 4114 | ret = dev_ifsioc(net, &ifr, cmd); |
1da177e4 LT |
4115 | rtnl_unlock(); |
4116 | if (!ret) { | |
4117 | if (colon) | |
4118 | *colon = ':'; | |
4119 | if (copy_to_user(arg, &ifr, | |
4120 | sizeof(struct ifreq))) | |
4121 | ret = -EFAULT; | |
4122 | } | |
4123 | return ret; | |
4124 | ||
4125 | /* | |
4126 | * These ioctl calls: | |
4127 | * - require superuser power. | |
4128 | * - require strict serialization. | |
4129 | * - do not return a value | |
4130 | */ | |
4131 | case SIOCSIFFLAGS: | |
4132 | case SIOCSIFMETRIC: | |
4133 | case SIOCSIFMTU: | |
4134 | case SIOCSIFMAP: | |
4135 | case SIOCSIFHWADDR: | |
4136 | case SIOCSIFSLAVE: | |
4137 | case SIOCADDMULTI: | |
4138 | case SIOCDELMULTI: | |
4139 | case SIOCSIFHWBROADCAST: | |
4140 | case SIOCSIFTXQLEN: | |
4141 | case SIOCSMIIREG: | |
4142 | case SIOCBONDENSLAVE: | |
4143 | case SIOCBONDRELEASE: | |
4144 | case SIOCBONDSETHWADDR: | |
1da177e4 LT |
4145 | case SIOCBONDCHANGEACTIVE: |
4146 | case SIOCBRADDIF: | |
4147 | case SIOCBRDELIF: | |
d24fff22 | 4148 | case SIOCSHWTSTAMP: |
1da177e4 LT |
4149 | if (!capable(CAP_NET_ADMIN)) |
4150 | return -EPERM; | |
cabcac0b TG |
4151 | /* fall through */ |
4152 | case SIOCBONDSLAVEINFOQUERY: | |
4153 | case SIOCBONDINFOQUERY: | |
881d966b | 4154 | dev_load(net, ifr.ifr_name); |
1da177e4 | 4155 | rtnl_lock(); |
881d966b | 4156 | ret = dev_ifsioc(net, &ifr, cmd); |
1da177e4 LT |
4157 | rtnl_unlock(); |
4158 | return ret; | |
4159 | ||
4160 | case SIOCGIFMEM: | |
4161 | /* Get the per device memory space. We can add this but | |
4162 | * currently do not support it */ | |
4163 | case SIOCSIFMEM: | |
4164 | /* Set the per device memory buffer space. | |
4165 | * Not applicable in our case */ | |
4166 | case SIOCSIFLINK: | |
4167 | return -EINVAL; | |
4168 | ||
4169 | /* | |
4170 | * Unknown or private ioctl. | |
4171 | */ | |
4172 | default: | |
4173 | if (cmd == SIOCWANDEV || | |
4174 | (cmd >= SIOCDEVPRIVATE && | |
4175 | cmd <= SIOCDEVPRIVATE + 15)) { | |
881d966b | 4176 | dev_load(net, ifr.ifr_name); |
1da177e4 | 4177 | rtnl_lock(); |
881d966b | 4178 | ret = dev_ifsioc(net, &ifr, cmd); |
1da177e4 LT |
4179 | rtnl_unlock(); |
4180 | if (!ret && copy_to_user(arg, &ifr, | |
4181 | sizeof(struct ifreq))) | |
4182 | ret = -EFAULT; | |
4183 | return ret; | |
4184 | } | |
1da177e4 | 4185 | /* Take care of Wireless Extensions */ |
295f4a1f | 4186 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) |
881d966b | 4187 | return wext_handle_ioctl(net, &ifr, cmd, arg); |
1da177e4 LT |
4188 | return -EINVAL; |
4189 | } | |
4190 | } | |
4191 | ||
4192 | ||
4193 | /** | |
4194 | * dev_new_index - allocate an ifindex | |
c4ea43c5 | 4195 | * @net: the applicable net namespace |
1da177e4 LT |
4196 | * |
4197 | * Returns a suitable unique value for a new device interface | |
4198 | * number. The caller must hold the rtnl semaphore or the | |
4199 | * dev_base_lock to be sure it remains unique. | |
4200 | */ | |
881d966b | 4201 | static int dev_new_index(struct net *net) |
1da177e4 LT |
4202 | { |
4203 | static int ifindex; | |
4204 | for (;;) { | |
4205 | if (++ifindex <= 0) | |
4206 | ifindex = 1; | |
881d966b | 4207 | if (!__dev_get_by_index(net, ifindex)) |
1da177e4 LT |
4208 | return ifindex; |
4209 | } | |
4210 | } | |
4211 | ||
1da177e4 | 4212 | /* Delayed registration/unregisteration */ |
3b5b34fd | 4213 | static LIST_HEAD(net_todo_list); |
1da177e4 | 4214 | |
6f05f629 | 4215 | static void net_set_todo(struct net_device *dev) |
1da177e4 | 4216 | { |
1da177e4 | 4217 | list_add_tail(&dev->todo_list, &net_todo_list); |
1da177e4 LT |
4218 | } |
4219 | ||
93ee31f1 DL |
4220 | static void rollback_registered(struct net_device *dev) |
4221 | { | |
4222 | BUG_ON(dev_boot_phase); | |
4223 | ASSERT_RTNL(); | |
4224 | ||
4225 | /* Some devices call without registering for initialization unwind. */ | |
4226 | if (dev->reg_state == NETREG_UNINITIALIZED) { | |
4227 | printk(KERN_DEBUG "unregister_netdevice: device %s/%p never " | |
4228 | "was registered\n", dev->name, dev); | |
4229 | ||
4230 | WARN_ON(1); | |
4231 | return; | |
4232 | } | |
4233 | ||
4234 | BUG_ON(dev->reg_state != NETREG_REGISTERED); | |
4235 | ||
4236 | /* If device is running, close it first. */ | |
4237 | dev_close(dev); | |
4238 | ||
4239 | /* And unlink it from device chain. */ | |
4240 | unlist_netdevice(dev); | |
4241 | ||
4242 | dev->reg_state = NETREG_UNREGISTERING; | |
4243 | ||
4244 | synchronize_net(); | |
4245 | ||
4246 | /* Shutdown queueing discipline. */ | |
4247 | dev_shutdown(dev); | |
4248 | ||
4249 | ||
4250 | /* Notify protocols, that we are about to destroy | |
4251 | this device. They should clean all the things. | |
4252 | */ | |
4253 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); | |
4254 | ||
4255 | /* | |
4256 | * Flush the unicast and multicast chains | |
4257 | */ | |
4258 | dev_addr_discard(dev); | |
4259 | ||
d314774c SH |
4260 | if (dev->netdev_ops->ndo_uninit) |
4261 | dev->netdev_ops->ndo_uninit(dev); | |
93ee31f1 DL |
4262 | |
4263 | /* Notifier chain MUST detach us from master device. */ | |
547b792c | 4264 | WARN_ON(dev->master); |
93ee31f1 DL |
4265 | |
4266 | /* Remove entries from kobject tree */ | |
4267 | netdev_unregister_kobject(dev); | |
4268 | ||
4269 | synchronize_net(); | |
4270 | ||
4271 | dev_put(dev); | |
4272 | } | |
4273 | ||
e8a0464c DM |
4274 | static void __netdev_init_queue_locks_one(struct net_device *dev, |
4275 | struct netdev_queue *dev_queue, | |
4276 | void *_unused) | |
c773e847 DM |
4277 | { |
4278 | spin_lock_init(&dev_queue->_xmit_lock); | |
cf508b12 | 4279 | netdev_set_xmit_lockdep_class(&dev_queue->_xmit_lock, dev->type); |
c773e847 DM |
4280 | dev_queue->xmit_lock_owner = -1; |
4281 | } | |
4282 | ||
4283 | static void netdev_init_queue_locks(struct net_device *dev) | |
4284 | { | |
e8a0464c DM |
4285 | netdev_for_each_tx_queue(dev, __netdev_init_queue_locks_one, NULL); |
4286 | __netdev_init_queue_locks_one(dev, &dev->rx_queue, NULL); | |
c773e847 DM |
4287 | } |
4288 | ||
b63365a2 HX |
4289 | unsigned long netdev_fix_features(unsigned long features, const char *name) |
4290 | { | |
4291 | /* Fix illegal SG+CSUM combinations. */ | |
4292 | if ((features & NETIF_F_SG) && | |
4293 | !(features & NETIF_F_ALL_CSUM)) { | |
4294 | if (name) | |
4295 | printk(KERN_NOTICE "%s: Dropping NETIF_F_SG since no " | |
4296 | "checksum feature.\n", name); | |
4297 | features &= ~NETIF_F_SG; | |
4298 | } | |
4299 | ||
4300 | /* TSO requires that SG is present as well. */ | |
4301 | if ((features & NETIF_F_TSO) && !(features & NETIF_F_SG)) { | |
4302 | if (name) | |
4303 | printk(KERN_NOTICE "%s: Dropping NETIF_F_TSO since no " | |
4304 | "SG feature.\n", name); | |
4305 | features &= ~NETIF_F_TSO; | |
4306 | } | |
4307 | ||
4308 | if (features & NETIF_F_UFO) { | |
4309 | if (!(features & NETIF_F_GEN_CSUM)) { | |
4310 | if (name) | |
4311 | printk(KERN_ERR "%s: Dropping NETIF_F_UFO " | |
4312 | "since no NETIF_F_HW_CSUM feature.\n", | |
4313 | name); | |
4314 | features &= ~NETIF_F_UFO; | |
4315 | } | |
4316 | ||
4317 | if (!(features & NETIF_F_SG)) { | |
4318 | if (name) | |
4319 | printk(KERN_ERR "%s: Dropping NETIF_F_UFO " | |
4320 | "since no NETIF_F_SG feature.\n", name); | |
4321 | features &= ~NETIF_F_UFO; | |
4322 | } | |
4323 | } | |
4324 | ||
4325 | return features; | |
4326 | } | |
4327 | EXPORT_SYMBOL(netdev_fix_features); | |
4328 | ||
9d40bbda DM |
4329 | /* Some devices need to (re-)set their netdev_ops inside |
4330 | * ->init() or similar. If that happens, we have to setup | |
4331 | * the compat pointers again. | |
4332 | */ | |
4333 | void netdev_resync_ops(struct net_device *dev) | |
4334 | { | |
4335 | #ifdef CONFIG_COMPAT_NET_DEV_OPS | |
4336 | const struct net_device_ops *ops = dev->netdev_ops; | |
4337 | ||
4338 | dev->init = ops->ndo_init; | |
4339 | dev->uninit = ops->ndo_uninit; | |
4340 | dev->open = ops->ndo_open; | |
4341 | dev->change_rx_flags = ops->ndo_change_rx_flags; | |
4342 | dev->set_rx_mode = ops->ndo_set_rx_mode; | |
4343 | dev->set_multicast_list = ops->ndo_set_multicast_list; | |
4344 | dev->set_mac_address = ops->ndo_set_mac_address; | |
4345 | dev->validate_addr = ops->ndo_validate_addr; | |
4346 | dev->do_ioctl = ops->ndo_do_ioctl; | |
4347 | dev->set_config = ops->ndo_set_config; | |
4348 | dev->change_mtu = ops->ndo_change_mtu; | |
4349 | dev->neigh_setup = ops->ndo_neigh_setup; | |
4350 | dev->tx_timeout = ops->ndo_tx_timeout; | |
4351 | dev->get_stats = ops->ndo_get_stats; | |
4352 | dev->vlan_rx_register = ops->ndo_vlan_rx_register; | |
4353 | dev->vlan_rx_add_vid = ops->ndo_vlan_rx_add_vid; | |
4354 | dev->vlan_rx_kill_vid = ops->ndo_vlan_rx_kill_vid; | |
4355 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
4356 | dev->poll_controller = ops->ndo_poll_controller; | |
4357 | #endif | |
4358 | #endif | |
4359 | } | |
4360 | EXPORT_SYMBOL(netdev_resync_ops); | |
4361 | ||
1da177e4 LT |
4362 | /** |
4363 | * register_netdevice - register a network device | |
4364 | * @dev: device to register | |
4365 | * | |
4366 | * Take a completed network device structure and add it to the kernel | |
4367 | * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier | |
4368 | * chain. 0 is returned on success. A negative errno code is returned | |
4369 | * on a failure to set up the device, or if the name is a duplicate. | |
4370 | * | |
4371 | * Callers must hold the rtnl semaphore. You may want | |
4372 | * register_netdev() instead of this. | |
4373 | * | |
4374 | * BUGS: | |
4375 | * The locking appears insufficient to guarantee two parallel registers | |
4376 | * will not get the same name. | |
4377 | */ | |
4378 | ||
4379 | int register_netdevice(struct net_device *dev) | |
4380 | { | |
4381 | struct hlist_head *head; | |
4382 | struct hlist_node *p; | |
4383 | int ret; | |
d314774c | 4384 | struct net *net = dev_net(dev); |
1da177e4 LT |
4385 | |
4386 | BUG_ON(dev_boot_phase); | |
4387 | ASSERT_RTNL(); | |
4388 | ||
b17a7c17 SH |
4389 | might_sleep(); |
4390 | ||
1da177e4 LT |
4391 | /* When net_device's are persistent, this will be fatal. */ |
4392 | BUG_ON(dev->reg_state != NETREG_UNINITIALIZED); | |
d314774c | 4393 | BUG_ON(!net); |
1da177e4 | 4394 | |
f1f28aa3 | 4395 | spin_lock_init(&dev->addr_list_lock); |
cf508b12 | 4396 | netdev_set_addr_lockdep_class(dev); |
c773e847 | 4397 | netdev_init_queue_locks(dev); |
1da177e4 | 4398 | |
1da177e4 LT |
4399 | dev->iflink = -1; |
4400 | ||
d314774c SH |
4401 | #ifdef CONFIG_COMPAT_NET_DEV_OPS |
4402 | /* Netdevice_ops API compatiability support. | |
4403 | * This is temporary until all network devices are converted. | |
4404 | */ | |
4405 | if (dev->netdev_ops) { | |
9d40bbda | 4406 | netdev_resync_ops(dev); |
d314774c SH |
4407 | } else { |
4408 | char drivername[64]; | |
4409 | pr_info("%s (%s): not using net_device_ops yet\n", | |
4410 | dev->name, netdev_drivername(dev, drivername, 64)); | |
4411 | ||
4412 | /* This works only because net_device_ops and the | |
4413 | compatiablity structure are the same. */ | |
4414 | dev->netdev_ops = (void *) &(dev->init); | |
4415 | } | |
4416 | #endif | |
4417 | ||
1da177e4 | 4418 | /* Init, if this function is available */ |
d314774c SH |
4419 | if (dev->netdev_ops->ndo_init) { |
4420 | ret = dev->netdev_ops->ndo_init(dev); | |
1da177e4 LT |
4421 | if (ret) { |
4422 | if (ret > 0) | |
4423 | ret = -EIO; | |
90833aa4 | 4424 | goto out; |
1da177e4 LT |
4425 | } |
4426 | } | |
4ec93edb | 4427 | |
1da177e4 LT |
4428 | if (!dev_valid_name(dev->name)) { |
4429 | ret = -EINVAL; | |
7ce1b0ed | 4430 | goto err_uninit; |
1da177e4 LT |
4431 | } |
4432 | ||
881d966b | 4433 | dev->ifindex = dev_new_index(net); |
1da177e4 LT |
4434 | if (dev->iflink == -1) |
4435 | dev->iflink = dev->ifindex; | |
4436 | ||
4437 | /* Check for existence of name */ | |
881d966b | 4438 | head = dev_name_hash(net, dev->name); |
1da177e4 LT |
4439 | hlist_for_each(p, head) { |
4440 | struct net_device *d | |
4441 | = hlist_entry(p, struct net_device, name_hlist); | |
4442 | if (!strncmp(d->name, dev->name, IFNAMSIZ)) { | |
4443 | ret = -EEXIST; | |
7ce1b0ed | 4444 | goto err_uninit; |
1da177e4 | 4445 | } |
4ec93edb | 4446 | } |
1da177e4 | 4447 | |
d212f87b SH |
4448 | /* Fix illegal checksum combinations */ |
4449 | if ((dev->features & NETIF_F_HW_CSUM) && | |
4450 | (dev->features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { | |
4451 | printk(KERN_NOTICE "%s: mixed HW and IP checksum settings.\n", | |
4452 | dev->name); | |
4453 | dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); | |
4454 | } | |
4455 | ||
4456 | if ((dev->features & NETIF_F_NO_CSUM) && | |
4457 | (dev->features & (NETIF_F_HW_CSUM|NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { | |
4458 | printk(KERN_NOTICE "%s: mixed no checksumming and other settings.\n", | |
4459 | dev->name); | |
4460 | dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM|NETIF_F_HW_CSUM); | |
4461 | } | |
4462 | ||
b63365a2 | 4463 | dev->features = netdev_fix_features(dev->features, dev->name); |
1da177e4 | 4464 | |
e5a4a72d LB |
4465 | /* Enable software GSO if SG is supported. */ |
4466 | if (dev->features & NETIF_F_SG) | |
4467 | dev->features |= NETIF_F_GSO; | |
4468 | ||
aaf8cdc3 | 4469 | netdev_initialize_kobject(dev); |
8b41d188 | 4470 | ret = netdev_register_kobject(dev); |
b17a7c17 | 4471 | if (ret) |
7ce1b0ed | 4472 | goto err_uninit; |
b17a7c17 SH |
4473 | dev->reg_state = NETREG_REGISTERED; |
4474 | ||
1da177e4 LT |
4475 | /* |
4476 | * Default initial state at registry is that the | |
4477 | * device is present. | |
4478 | */ | |
4479 | ||
4480 | set_bit(__LINK_STATE_PRESENT, &dev->state); | |
4481 | ||
1da177e4 | 4482 | dev_init_scheduler(dev); |
1da177e4 | 4483 | dev_hold(dev); |
ce286d32 | 4484 | list_netdevice(dev); |
1da177e4 LT |
4485 | |
4486 | /* Notify protocols, that a new device appeared. */ | |
056925ab | 4487 | ret = call_netdevice_notifiers(NETDEV_REGISTER, dev); |
fcc5a03a | 4488 | ret = notifier_to_errno(ret); |
93ee31f1 DL |
4489 | if (ret) { |
4490 | rollback_registered(dev); | |
4491 | dev->reg_state = NETREG_UNREGISTERED; | |
4492 | } | |
1da177e4 LT |
4493 | |
4494 | out: | |
4495 | return ret; | |
7ce1b0ed HX |
4496 | |
4497 | err_uninit: | |
d314774c SH |
4498 | if (dev->netdev_ops->ndo_uninit) |
4499 | dev->netdev_ops->ndo_uninit(dev); | |
7ce1b0ed | 4500 | goto out; |
1da177e4 LT |
4501 | } |
4502 | ||
937f1ba5 BH |
4503 | /** |
4504 | * init_dummy_netdev - init a dummy network device for NAPI | |
4505 | * @dev: device to init | |
4506 | * | |
4507 | * This takes a network device structure and initialize the minimum | |
4508 | * amount of fields so it can be used to schedule NAPI polls without | |
4509 | * registering a full blown interface. This is to be used by drivers | |
4510 | * that need to tie several hardware interfaces to a single NAPI | |
4511 | * poll scheduler due to HW limitations. | |
4512 | */ | |
4513 | int init_dummy_netdev(struct net_device *dev) | |
4514 | { | |
4515 | /* Clear everything. Note we don't initialize spinlocks | |
4516 | * are they aren't supposed to be taken by any of the | |
4517 | * NAPI code and this dummy netdev is supposed to be | |
4518 | * only ever used for NAPI polls | |
4519 | */ | |
4520 | memset(dev, 0, sizeof(struct net_device)); | |
4521 | ||
4522 | /* make sure we BUG if trying to hit standard | |
4523 | * register/unregister code path | |
4524 | */ | |
4525 | dev->reg_state = NETREG_DUMMY; | |
4526 | ||
4527 | /* initialize the ref count */ | |
4528 | atomic_set(&dev->refcnt, 1); | |
4529 | ||
4530 | /* NAPI wants this */ | |
4531 | INIT_LIST_HEAD(&dev->napi_list); | |
4532 | ||
4533 | /* a dummy interface is started by default */ | |
4534 | set_bit(__LINK_STATE_PRESENT, &dev->state); | |
4535 | set_bit(__LINK_STATE_START, &dev->state); | |
4536 | ||
4537 | return 0; | |
4538 | } | |
4539 | EXPORT_SYMBOL_GPL(init_dummy_netdev); | |
4540 | ||
4541 | ||
1da177e4 LT |
4542 | /** |
4543 | * register_netdev - register a network device | |
4544 | * @dev: device to register | |
4545 | * | |
4546 | * Take a completed network device structure and add it to the kernel | |
4547 | * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier | |
4548 | * chain. 0 is returned on success. A negative errno code is returned | |
4549 | * on a failure to set up the device, or if the name is a duplicate. | |
4550 | * | |
38b4da38 | 4551 | * This is a wrapper around register_netdevice that takes the rtnl semaphore |
1da177e4 LT |
4552 | * and expands the device name if you passed a format string to |
4553 | * alloc_netdev. | |
4554 | */ | |
4555 | int register_netdev(struct net_device *dev) | |
4556 | { | |
4557 | int err; | |
4558 | ||
4559 | rtnl_lock(); | |
4560 | ||
4561 | /* | |
4562 | * If the name is a format string the caller wants us to do a | |
4563 | * name allocation. | |
4564 | */ | |
4565 | if (strchr(dev->name, '%')) { | |
4566 | err = dev_alloc_name(dev, dev->name); | |
4567 | if (err < 0) | |
4568 | goto out; | |
4569 | } | |
4ec93edb | 4570 | |
1da177e4 LT |
4571 | err = register_netdevice(dev); |
4572 | out: | |
4573 | rtnl_unlock(); | |
4574 | return err; | |
4575 | } | |
4576 | EXPORT_SYMBOL(register_netdev); | |
4577 | ||
4578 | /* | |
4579 | * netdev_wait_allrefs - wait until all references are gone. | |
4580 | * | |
4581 | * This is called when unregistering network devices. | |
4582 | * | |
4583 | * Any protocol or device that holds a reference should register | |
4584 | * for netdevice notification, and cleanup and put back the | |
4585 | * reference if they receive an UNREGISTER event. | |
4586 | * We can get stuck here if buggy protocols don't correctly | |
4ec93edb | 4587 | * call dev_put. |
1da177e4 LT |
4588 | */ |
4589 | static void netdev_wait_allrefs(struct net_device *dev) | |
4590 | { | |
4591 | unsigned long rebroadcast_time, warning_time; | |
4592 | ||
4593 | rebroadcast_time = warning_time = jiffies; | |
4594 | while (atomic_read(&dev->refcnt) != 0) { | |
4595 | if (time_after(jiffies, rebroadcast_time + 1 * HZ)) { | |
6756ae4b | 4596 | rtnl_lock(); |
1da177e4 LT |
4597 | |
4598 | /* Rebroadcast unregister notification */ | |
056925ab | 4599 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); |
1da177e4 LT |
4600 | |
4601 | if (test_bit(__LINK_STATE_LINKWATCH_PENDING, | |
4602 | &dev->state)) { | |
4603 | /* We must not have linkwatch events | |
4604 | * pending on unregister. If this | |
4605 | * happens, we simply run the queue | |
4606 | * unscheduled, resulting in a noop | |
4607 | * for this device. | |
4608 | */ | |
4609 | linkwatch_run_queue(); | |
4610 | } | |
4611 | ||
6756ae4b | 4612 | __rtnl_unlock(); |
1da177e4 LT |
4613 | |
4614 | rebroadcast_time = jiffies; | |
4615 | } | |
4616 | ||
4617 | msleep(250); | |
4618 | ||
4619 | if (time_after(jiffies, warning_time + 10 * HZ)) { | |
4620 | printk(KERN_EMERG "unregister_netdevice: " | |
4621 | "waiting for %s to become free. Usage " | |
4622 | "count = %d\n", | |
4623 | dev->name, atomic_read(&dev->refcnt)); | |
4624 | warning_time = jiffies; | |
4625 | } | |
4626 | } | |
4627 | } | |
4628 | ||
4629 | /* The sequence is: | |
4630 | * | |
4631 | * rtnl_lock(); | |
4632 | * ... | |
4633 | * register_netdevice(x1); | |
4634 | * register_netdevice(x2); | |
4635 | * ... | |
4636 | * unregister_netdevice(y1); | |
4637 | * unregister_netdevice(y2); | |
4638 | * ... | |
4639 | * rtnl_unlock(); | |
4640 | * free_netdev(y1); | |
4641 | * free_netdev(y2); | |
4642 | * | |
58ec3b4d | 4643 | * We are invoked by rtnl_unlock(). |
1da177e4 | 4644 | * This allows us to deal with problems: |
b17a7c17 | 4645 | * 1) We can delete sysfs objects which invoke hotplug |
1da177e4 LT |
4646 | * without deadlocking with linkwatch via keventd. |
4647 | * 2) Since we run with the RTNL semaphore not held, we can sleep | |
4648 | * safely in order to wait for the netdev refcnt to drop to zero. | |
58ec3b4d HX |
4649 | * |
4650 | * We must not return until all unregister events added during | |
4651 | * the interval the lock was held have been completed. | |
1da177e4 | 4652 | */ |
1da177e4 LT |
4653 | void netdev_run_todo(void) |
4654 | { | |
626ab0e6 | 4655 | struct list_head list; |
1da177e4 | 4656 | |
1da177e4 | 4657 | /* Snapshot list, allow later requests */ |
626ab0e6 | 4658 | list_replace_init(&net_todo_list, &list); |
58ec3b4d HX |
4659 | |
4660 | __rtnl_unlock(); | |
626ab0e6 | 4661 | |
1da177e4 LT |
4662 | while (!list_empty(&list)) { |
4663 | struct net_device *dev | |
4664 | = list_entry(list.next, struct net_device, todo_list); | |
4665 | list_del(&dev->todo_list); | |
4666 | ||
b17a7c17 SH |
4667 | if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) { |
4668 | printk(KERN_ERR "network todo '%s' but state %d\n", | |
4669 | dev->name, dev->reg_state); | |
4670 | dump_stack(); | |
4671 | continue; | |
4672 | } | |
1da177e4 | 4673 | |
b17a7c17 | 4674 | dev->reg_state = NETREG_UNREGISTERED; |
1da177e4 | 4675 | |
6e583ce5 SH |
4676 | on_each_cpu(flush_backlog, dev, 1); |
4677 | ||
b17a7c17 | 4678 | netdev_wait_allrefs(dev); |
1da177e4 | 4679 | |
b17a7c17 SH |
4680 | /* paranoia */ |
4681 | BUG_ON(atomic_read(&dev->refcnt)); | |
547b792c IJ |
4682 | WARN_ON(dev->ip_ptr); |
4683 | WARN_ON(dev->ip6_ptr); | |
4684 | WARN_ON(dev->dn_ptr); | |
1da177e4 | 4685 | |
b17a7c17 SH |
4686 | if (dev->destructor) |
4687 | dev->destructor(dev); | |
9093bbb2 SH |
4688 | |
4689 | /* Free network device */ | |
4690 | kobject_put(&dev->dev.kobj); | |
1da177e4 | 4691 | } |
1da177e4 LT |
4692 | } |
4693 | ||
eeda3fd6 SH |
4694 | /** |
4695 | * dev_get_stats - get network device statistics | |
4696 | * @dev: device to get statistics from | |
4697 | * | |
4698 | * Get network statistics from device. The device driver may provide | |
4699 | * its own method by setting dev->netdev_ops->get_stats; otherwise | |
4700 | * the internal statistics structure is used. | |
4701 | */ | |
4702 | const struct net_device_stats *dev_get_stats(struct net_device *dev) | |
4703 | { | |
4704 | const struct net_device_ops *ops = dev->netdev_ops; | |
4705 | ||
4706 | if (ops->ndo_get_stats) | |
4707 | return ops->ndo_get_stats(dev); | |
4708 | else | |
4709 | return &dev->stats; | |
c45d286e | 4710 | } |
eeda3fd6 | 4711 | EXPORT_SYMBOL(dev_get_stats); |
c45d286e | 4712 | |
dc2b4847 | 4713 | static void netdev_init_one_queue(struct net_device *dev, |
e8a0464c DM |
4714 | struct netdev_queue *queue, |
4715 | void *_unused) | |
dc2b4847 | 4716 | { |
dc2b4847 DM |
4717 | queue->dev = dev; |
4718 | } | |
4719 | ||
bb949fbd DM |
4720 | static void netdev_init_queues(struct net_device *dev) |
4721 | { | |
e8a0464c DM |
4722 | netdev_init_one_queue(dev, &dev->rx_queue, NULL); |
4723 | netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL); | |
c3f26a26 | 4724 | spin_lock_init(&dev->tx_global_lock); |
bb949fbd DM |
4725 | } |
4726 | ||
1da177e4 | 4727 | /** |
f25f4e44 | 4728 | * alloc_netdev_mq - allocate network device |
1da177e4 LT |
4729 | * @sizeof_priv: size of private data to allocate space for |
4730 | * @name: device name format string | |
4731 | * @setup: callback to initialize device | |
f25f4e44 | 4732 | * @queue_count: the number of subqueues to allocate |
1da177e4 LT |
4733 | * |
4734 | * Allocates a struct net_device with private data area for driver use | |
f25f4e44 PWJ |
4735 | * and performs basic initialization. Also allocates subquue structs |
4736 | * for each queue on the device at the end of the netdevice. | |
1da177e4 | 4737 | */ |
f25f4e44 PWJ |
4738 | struct net_device *alloc_netdev_mq(int sizeof_priv, const char *name, |
4739 | void (*setup)(struct net_device *), unsigned int queue_count) | |
1da177e4 | 4740 | { |
e8a0464c | 4741 | struct netdev_queue *tx; |
1da177e4 | 4742 | struct net_device *dev; |
7943986c | 4743 | size_t alloc_size; |
e8a0464c | 4744 | void *p; |
1da177e4 | 4745 | |
b6fe17d6 SH |
4746 | BUG_ON(strlen(name) >= sizeof(dev->name)); |
4747 | ||
fd2ea0a7 | 4748 | alloc_size = sizeof(struct net_device); |
d1643d24 AD |
4749 | if (sizeof_priv) { |
4750 | /* ensure 32-byte alignment of private area */ | |
4751 | alloc_size = (alloc_size + NETDEV_ALIGN_CONST) & ~NETDEV_ALIGN_CONST; | |
4752 | alloc_size += sizeof_priv; | |
4753 | } | |
4754 | /* ensure 32-byte alignment of whole construct */ | |
4755 | alloc_size += NETDEV_ALIGN_CONST; | |
1da177e4 | 4756 | |
31380de9 | 4757 | p = kzalloc(alloc_size, GFP_KERNEL); |
1da177e4 | 4758 | if (!p) { |
b6fe17d6 | 4759 | printk(KERN_ERR "alloc_netdev: Unable to allocate device.\n"); |
1da177e4 LT |
4760 | return NULL; |
4761 | } | |
1da177e4 | 4762 | |
7943986c | 4763 | tx = kcalloc(queue_count, sizeof(struct netdev_queue), GFP_KERNEL); |
e8a0464c DM |
4764 | if (!tx) { |
4765 | printk(KERN_ERR "alloc_netdev: Unable to allocate " | |
4766 | "tx qdiscs.\n"); | |
4767 | kfree(p); | |
4768 | return NULL; | |
4769 | } | |
4770 | ||
1da177e4 LT |
4771 | dev = (struct net_device *) |
4772 | (((long)p + NETDEV_ALIGN_CONST) & ~NETDEV_ALIGN_CONST); | |
4773 | dev->padded = (char *)dev - (char *)p; | |
c346dca1 | 4774 | dev_net_set(dev, &init_net); |
1da177e4 | 4775 | |
e8a0464c DM |
4776 | dev->_tx = tx; |
4777 | dev->num_tx_queues = queue_count; | |
fd2ea0a7 | 4778 | dev->real_num_tx_queues = queue_count; |
e8a0464c | 4779 | |
82cc1a7a | 4780 | dev->gso_max_size = GSO_MAX_SIZE; |
1da177e4 | 4781 | |
bb949fbd DM |
4782 | netdev_init_queues(dev); |
4783 | ||
d565b0a1 | 4784 | INIT_LIST_HEAD(&dev->napi_list); |
1da177e4 LT |
4785 | setup(dev); |
4786 | strcpy(dev->name, name); | |
4787 | return dev; | |
4788 | } | |
f25f4e44 | 4789 | EXPORT_SYMBOL(alloc_netdev_mq); |
1da177e4 LT |
4790 | |
4791 | /** | |
4792 | * free_netdev - free network device | |
4793 | * @dev: device | |
4794 | * | |
4ec93edb YH |
4795 | * This function does the last stage of destroying an allocated device |
4796 | * interface. The reference to the device object is released. | |
1da177e4 LT |
4797 | * If this is the last reference then it will be freed. |
4798 | */ | |
4799 | void free_netdev(struct net_device *dev) | |
4800 | { | |
d565b0a1 HX |
4801 | struct napi_struct *p, *n; |
4802 | ||
f3005d7f DL |
4803 | release_net(dev_net(dev)); |
4804 | ||
e8a0464c DM |
4805 | kfree(dev->_tx); |
4806 | ||
d565b0a1 HX |
4807 | list_for_each_entry_safe(p, n, &dev->napi_list, dev_list) |
4808 | netif_napi_del(p); | |
4809 | ||
3041a069 | 4810 | /* Compatibility with error handling in drivers */ |
1da177e4 LT |
4811 | if (dev->reg_state == NETREG_UNINITIALIZED) { |
4812 | kfree((char *)dev - dev->padded); | |
4813 | return; | |
4814 | } | |
4815 | ||
4816 | BUG_ON(dev->reg_state != NETREG_UNREGISTERED); | |
4817 | dev->reg_state = NETREG_RELEASED; | |
4818 | ||
43cb76d9 GKH |
4819 | /* will free via device release */ |
4820 | put_device(&dev->dev); | |
1da177e4 | 4821 | } |
4ec93edb | 4822 | |
f0db275a SH |
4823 | /** |
4824 | * synchronize_net - Synchronize with packet receive processing | |
4825 | * | |
4826 | * Wait for packets currently being received to be done. | |
4827 | * Does not block later packets from starting. | |
4828 | */ | |
4ec93edb | 4829 | void synchronize_net(void) |
1da177e4 LT |
4830 | { |
4831 | might_sleep(); | |
fbd568a3 | 4832 | synchronize_rcu(); |
1da177e4 LT |
4833 | } |
4834 | ||
4835 | /** | |
4836 | * unregister_netdevice - remove device from the kernel | |
4837 | * @dev: device | |
4838 | * | |
4839 | * This function shuts down a device interface and removes it | |
d59b54b1 | 4840 | * from the kernel tables. |
1da177e4 LT |
4841 | * |
4842 | * Callers must hold the rtnl semaphore. You may want | |
4843 | * unregister_netdev() instead of this. | |
4844 | */ | |
4845 | ||
22f8cde5 | 4846 | void unregister_netdevice(struct net_device *dev) |
1da177e4 | 4847 | { |
a6620712 HX |
4848 | ASSERT_RTNL(); |
4849 | ||
93ee31f1 | 4850 | rollback_registered(dev); |
1da177e4 LT |
4851 | /* Finish processing unregister after unlock */ |
4852 | net_set_todo(dev); | |
1da177e4 LT |
4853 | } |
4854 | ||
4855 | /** | |
4856 | * unregister_netdev - remove device from the kernel | |
4857 | * @dev: device | |
4858 | * | |
4859 | * This function shuts down a device interface and removes it | |
d59b54b1 | 4860 | * from the kernel tables. |
1da177e4 LT |
4861 | * |
4862 | * This is just a wrapper for unregister_netdevice that takes | |
4863 | * the rtnl semaphore. In general you want to use this and not | |
4864 | * unregister_netdevice. | |
4865 | */ | |
4866 | void unregister_netdev(struct net_device *dev) | |
4867 | { | |
4868 | rtnl_lock(); | |
4869 | unregister_netdevice(dev); | |
4870 | rtnl_unlock(); | |
4871 | } | |
4872 | ||
4873 | EXPORT_SYMBOL(unregister_netdev); | |
4874 | ||
ce286d32 EB |
4875 | /** |
4876 | * dev_change_net_namespace - move device to different nethost namespace | |
4877 | * @dev: device | |
4878 | * @net: network namespace | |
4879 | * @pat: If not NULL name pattern to try if the current device name | |
4880 | * is already taken in the destination network namespace. | |
4881 | * | |
4882 | * This function shuts down a device interface and moves it | |
4883 | * to a new network namespace. On success 0 is returned, on | |
4884 | * a failure a netagive errno code is returned. | |
4885 | * | |
4886 | * Callers must hold the rtnl semaphore. | |
4887 | */ | |
4888 | ||
4889 | int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat) | |
4890 | { | |
4891 | char buf[IFNAMSIZ]; | |
4892 | const char *destname; | |
4893 | int err; | |
4894 | ||
4895 | ASSERT_RTNL(); | |
4896 | ||
4897 | /* Don't allow namespace local devices to be moved. */ | |
4898 | err = -EINVAL; | |
4899 | if (dev->features & NETIF_F_NETNS_LOCAL) | |
4900 | goto out; | |
4901 | ||
3891845e EB |
4902 | #ifdef CONFIG_SYSFS |
4903 | /* Don't allow real devices to be moved when sysfs | |
4904 | * is enabled. | |
4905 | */ | |
4906 | err = -EINVAL; | |
4907 | if (dev->dev.parent) | |
4908 | goto out; | |
4909 | #endif | |
4910 | ||
ce286d32 EB |
4911 | /* Ensure the device has been registrered */ |
4912 | err = -EINVAL; | |
4913 | if (dev->reg_state != NETREG_REGISTERED) | |
4914 | goto out; | |
4915 | ||
4916 | /* Get out if there is nothing todo */ | |
4917 | err = 0; | |
878628fb | 4918 | if (net_eq(dev_net(dev), net)) |
ce286d32 EB |
4919 | goto out; |
4920 | ||
4921 | /* Pick the destination device name, and ensure | |
4922 | * we can use it in the destination network namespace. | |
4923 | */ | |
4924 | err = -EEXIST; | |
4925 | destname = dev->name; | |
4926 | if (__dev_get_by_name(net, destname)) { | |
4927 | /* We get here if we can't use the current device name */ | |
4928 | if (!pat) | |
4929 | goto out; | |
4930 | if (!dev_valid_name(pat)) | |
4931 | goto out; | |
4932 | if (strchr(pat, '%')) { | |
4933 | if (__dev_alloc_name(net, pat, buf) < 0) | |
4934 | goto out; | |
4935 | destname = buf; | |
4936 | } else | |
4937 | destname = pat; | |
4938 | if (__dev_get_by_name(net, destname)) | |
4939 | goto out; | |
4940 | } | |
4941 | ||
4942 | /* | |
4943 | * And now a mini version of register_netdevice unregister_netdevice. | |
4944 | */ | |
4945 | ||
4946 | /* If device is running close it first. */ | |
9b772652 | 4947 | dev_close(dev); |
ce286d32 EB |
4948 | |
4949 | /* And unlink it from device chain */ | |
4950 | err = -ENODEV; | |
4951 | unlist_netdevice(dev); | |
4952 | ||
4953 | synchronize_net(); | |
4954 | ||
4955 | /* Shutdown queueing discipline. */ | |
4956 | dev_shutdown(dev); | |
4957 | ||
4958 | /* Notify protocols, that we are about to destroy | |
4959 | this device. They should clean all the things. | |
4960 | */ | |
4961 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); | |
4962 | ||
4963 | /* | |
4964 | * Flush the unicast and multicast chains | |
4965 | */ | |
4966 | dev_addr_discard(dev); | |
4967 | ||
3891845e EB |
4968 | netdev_unregister_kobject(dev); |
4969 | ||
ce286d32 | 4970 | /* Actually switch the network namespace */ |
c346dca1 | 4971 | dev_net_set(dev, net); |
ce286d32 EB |
4972 | |
4973 | /* Assign the new device name */ | |
4974 | if (destname != dev->name) | |
4975 | strcpy(dev->name, destname); | |
4976 | ||
4977 | /* If there is an ifindex conflict assign a new one */ | |
4978 | if (__dev_get_by_index(net, dev->ifindex)) { | |
4979 | int iflink = (dev->iflink == dev->ifindex); | |
4980 | dev->ifindex = dev_new_index(net); | |
4981 | if (iflink) | |
4982 | dev->iflink = dev->ifindex; | |
4983 | } | |
4984 | ||
8b41d188 | 4985 | /* Fixup kobjects */ |
aaf8cdc3 | 4986 | err = netdev_register_kobject(dev); |
8b41d188 | 4987 | WARN_ON(err); |
ce286d32 EB |
4988 | |
4989 | /* Add the device back in the hashes */ | |
4990 | list_netdevice(dev); | |
4991 | ||
4992 | /* Notify protocols, that a new device appeared. */ | |
4993 | call_netdevice_notifiers(NETDEV_REGISTER, dev); | |
4994 | ||
4995 | synchronize_net(); | |
4996 | err = 0; | |
4997 | out: | |
4998 | return err; | |
4999 | } | |
5000 | ||
1da177e4 LT |
5001 | static int dev_cpu_callback(struct notifier_block *nfb, |
5002 | unsigned long action, | |
5003 | void *ocpu) | |
5004 | { | |
5005 | struct sk_buff **list_skb; | |
37437bb2 | 5006 | struct Qdisc **list_net; |
1da177e4 LT |
5007 | struct sk_buff *skb; |
5008 | unsigned int cpu, oldcpu = (unsigned long)ocpu; | |
5009 | struct softnet_data *sd, *oldsd; | |
5010 | ||
8bb78442 | 5011 | if (action != CPU_DEAD && action != CPU_DEAD_FROZEN) |
1da177e4 LT |
5012 | return NOTIFY_OK; |
5013 | ||
5014 | local_irq_disable(); | |
5015 | cpu = smp_processor_id(); | |
5016 | sd = &per_cpu(softnet_data, cpu); | |
5017 | oldsd = &per_cpu(softnet_data, oldcpu); | |
5018 | ||
5019 | /* Find end of our completion_queue. */ | |
5020 | list_skb = &sd->completion_queue; | |
5021 | while (*list_skb) | |
5022 | list_skb = &(*list_skb)->next; | |
5023 | /* Append completion queue from offline CPU. */ | |
5024 | *list_skb = oldsd->completion_queue; | |
5025 | oldsd->completion_queue = NULL; | |
5026 | ||
5027 | /* Find end of our output_queue. */ | |
5028 | list_net = &sd->output_queue; | |
5029 | while (*list_net) | |
5030 | list_net = &(*list_net)->next_sched; | |
5031 | /* Append output queue from offline CPU. */ | |
5032 | *list_net = oldsd->output_queue; | |
5033 | oldsd->output_queue = NULL; | |
5034 | ||
5035 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
5036 | local_irq_enable(); | |
5037 | ||
5038 | /* Process offline CPU's input_pkt_queue */ | |
5039 | while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) | |
5040 | netif_rx(skb); | |
5041 | ||
5042 | return NOTIFY_OK; | |
5043 | } | |
1da177e4 LT |
5044 | |
5045 | ||
7f353bf2 | 5046 | /** |
b63365a2 HX |
5047 | * netdev_increment_features - increment feature set by one |
5048 | * @all: current feature set | |
5049 | * @one: new feature set | |
5050 | * @mask: mask feature set | |
7f353bf2 HX |
5051 | * |
5052 | * Computes a new feature set after adding a device with feature set | |
b63365a2 HX |
5053 | * @one to the master device with current feature set @all. Will not |
5054 | * enable anything that is off in @mask. Returns the new feature set. | |
7f353bf2 | 5055 | */ |
b63365a2 HX |
5056 | unsigned long netdev_increment_features(unsigned long all, unsigned long one, |
5057 | unsigned long mask) | |
5058 | { | |
5059 | /* If device needs checksumming, downgrade to it. */ | |
5060 | if (all & NETIF_F_NO_CSUM && !(one & NETIF_F_NO_CSUM)) | |
5061 | all ^= NETIF_F_NO_CSUM | (one & NETIF_F_ALL_CSUM); | |
5062 | else if (mask & NETIF_F_ALL_CSUM) { | |
5063 | /* If one device supports v4/v6 checksumming, set for all. */ | |
5064 | if (one & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM) && | |
5065 | !(all & NETIF_F_GEN_CSUM)) { | |
5066 | all &= ~NETIF_F_ALL_CSUM; | |
5067 | all |= one & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM); | |
5068 | } | |
e2a6b852 | 5069 | |
b63365a2 HX |
5070 | /* If one device supports hw checksumming, set for all. */ |
5071 | if (one & NETIF_F_GEN_CSUM && !(all & NETIF_F_GEN_CSUM)) { | |
5072 | all &= ~NETIF_F_ALL_CSUM; | |
5073 | all |= NETIF_F_HW_CSUM; | |
5074 | } | |
5075 | } | |
7f353bf2 | 5076 | |
b63365a2 | 5077 | one |= NETIF_F_ALL_CSUM; |
7f353bf2 | 5078 | |
b63365a2 HX |
5079 | one |= all & NETIF_F_ONE_FOR_ALL; |
5080 | all &= one | NETIF_F_LLTX | NETIF_F_GSO; | |
5081 | all |= one & mask & NETIF_F_ONE_FOR_ALL; | |
7f353bf2 HX |
5082 | |
5083 | return all; | |
5084 | } | |
b63365a2 | 5085 | EXPORT_SYMBOL(netdev_increment_features); |
7f353bf2 | 5086 | |
30d97d35 PE |
5087 | static struct hlist_head *netdev_create_hash(void) |
5088 | { | |
5089 | int i; | |
5090 | struct hlist_head *hash; | |
5091 | ||
5092 | hash = kmalloc(sizeof(*hash) * NETDEV_HASHENTRIES, GFP_KERNEL); | |
5093 | if (hash != NULL) | |
5094 | for (i = 0; i < NETDEV_HASHENTRIES; i++) | |
5095 | INIT_HLIST_HEAD(&hash[i]); | |
5096 | ||
5097 | return hash; | |
5098 | } | |
5099 | ||
881d966b | 5100 | /* Initialize per network namespace state */ |
4665079c | 5101 | static int __net_init netdev_init(struct net *net) |
881d966b | 5102 | { |
881d966b | 5103 | INIT_LIST_HEAD(&net->dev_base_head); |
881d966b | 5104 | |
30d97d35 PE |
5105 | net->dev_name_head = netdev_create_hash(); |
5106 | if (net->dev_name_head == NULL) | |
5107 | goto err_name; | |
881d966b | 5108 | |
30d97d35 PE |
5109 | net->dev_index_head = netdev_create_hash(); |
5110 | if (net->dev_index_head == NULL) | |
5111 | goto err_idx; | |
881d966b EB |
5112 | |
5113 | return 0; | |
30d97d35 PE |
5114 | |
5115 | err_idx: | |
5116 | kfree(net->dev_name_head); | |
5117 | err_name: | |
5118 | return -ENOMEM; | |
881d966b EB |
5119 | } |
5120 | ||
f0db275a SH |
5121 | /** |
5122 | * netdev_drivername - network driver for the device | |
5123 | * @dev: network device | |
5124 | * @buffer: buffer for resulting name | |
5125 | * @len: size of buffer | |
5126 | * | |
5127 | * Determine network driver for device. | |
5128 | */ | |
cf04a4c7 | 5129 | char *netdev_drivername(const struct net_device *dev, char *buffer, int len) |
6579e57b | 5130 | { |
cf04a4c7 SH |
5131 | const struct device_driver *driver; |
5132 | const struct device *parent; | |
6579e57b AV |
5133 | |
5134 | if (len <= 0 || !buffer) | |
5135 | return buffer; | |
5136 | buffer[0] = 0; | |
5137 | ||
5138 | parent = dev->dev.parent; | |
5139 | ||
5140 | if (!parent) | |
5141 | return buffer; | |
5142 | ||
5143 | driver = parent->driver; | |
5144 | if (driver && driver->name) | |
5145 | strlcpy(buffer, driver->name, len); | |
5146 | return buffer; | |
5147 | } | |
5148 | ||
4665079c | 5149 | static void __net_exit netdev_exit(struct net *net) |
881d966b EB |
5150 | { |
5151 | kfree(net->dev_name_head); | |
5152 | kfree(net->dev_index_head); | |
5153 | } | |
5154 | ||
022cbae6 | 5155 | static struct pernet_operations __net_initdata netdev_net_ops = { |
881d966b EB |
5156 | .init = netdev_init, |
5157 | .exit = netdev_exit, | |
5158 | }; | |
5159 | ||
4665079c | 5160 | static void __net_exit default_device_exit(struct net *net) |
ce286d32 | 5161 | { |
8eb79863 | 5162 | struct net_device *dev; |
ce286d32 EB |
5163 | /* |
5164 | * Push all migratable of the network devices back to the | |
5165 | * initial network namespace | |
5166 | */ | |
5167 | rtnl_lock(); | |
8eb79863 EB |
5168 | restart: |
5169 | for_each_netdev(net, dev) { | |
ce286d32 | 5170 | int err; |
aca51397 | 5171 | char fb_name[IFNAMSIZ]; |
ce286d32 EB |
5172 | |
5173 | /* Ignore unmoveable devices (i.e. loopback) */ | |
5174 | if (dev->features & NETIF_F_NETNS_LOCAL) | |
5175 | continue; | |
5176 | ||
d0c082ce EB |
5177 | /* Delete virtual devices */ |
5178 | if (dev->rtnl_link_ops && dev->rtnl_link_ops->dellink) { | |
5179 | dev->rtnl_link_ops->dellink(dev); | |
8eb79863 | 5180 | goto restart; |
d0c082ce EB |
5181 | } |
5182 | ||
ce286d32 | 5183 | /* Push remaing network devices to init_net */ |
aca51397 PE |
5184 | snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex); |
5185 | err = dev_change_net_namespace(dev, &init_net, fb_name); | |
ce286d32 | 5186 | if (err) { |
aca51397 | 5187 | printk(KERN_EMERG "%s: failed to move %s to init_net: %d\n", |
ce286d32 | 5188 | __func__, dev->name, err); |
aca51397 | 5189 | BUG(); |
ce286d32 | 5190 | } |
8eb79863 | 5191 | goto restart; |
ce286d32 EB |
5192 | } |
5193 | rtnl_unlock(); | |
5194 | } | |
5195 | ||
022cbae6 | 5196 | static struct pernet_operations __net_initdata default_device_ops = { |
ce286d32 EB |
5197 | .exit = default_device_exit, |
5198 | }; | |
5199 | ||
1da177e4 LT |
5200 | /* |
5201 | * Initialize the DEV module. At boot time this walks the device list and | |
5202 | * unhooks any devices that fail to initialise (normally hardware not | |
5203 | * present) and leaves us with a valid list of present and active devices. | |
5204 | * | |
5205 | */ | |
5206 | ||
5207 | /* | |
5208 | * This is called single threaded during boot, so no need | |
5209 | * to take the rtnl semaphore. | |
5210 | */ | |
5211 | static int __init net_dev_init(void) | |
5212 | { | |
5213 | int i, rc = -ENOMEM; | |
5214 | ||
5215 | BUG_ON(!dev_boot_phase); | |
5216 | ||
1da177e4 LT |
5217 | if (dev_proc_init()) |
5218 | goto out; | |
5219 | ||
8b41d188 | 5220 | if (netdev_kobject_init()) |
1da177e4 LT |
5221 | goto out; |
5222 | ||
5223 | INIT_LIST_HEAD(&ptype_all); | |
82d8a867 | 5224 | for (i = 0; i < PTYPE_HASH_SIZE; i++) |
1da177e4 LT |
5225 | INIT_LIST_HEAD(&ptype_base[i]); |
5226 | ||
881d966b EB |
5227 | if (register_pernet_subsys(&netdev_net_ops)) |
5228 | goto out; | |
1da177e4 LT |
5229 | |
5230 | /* | |
5231 | * Initialise the packet receive queues. | |
5232 | */ | |
5233 | ||
6f912042 | 5234 | for_each_possible_cpu(i) { |
1da177e4 LT |
5235 | struct softnet_data *queue; |
5236 | ||
5237 | queue = &per_cpu(softnet_data, i); | |
5238 | skb_queue_head_init(&queue->input_pkt_queue); | |
1da177e4 LT |
5239 | queue->completion_queue = NULL; |
5240 | INIT_LIST_HEAD(&queue->poll_list); | |
bea3348e SH |
5241 | |
5242 | queue->backlog.poll = process_backlog; | |
5243 | queue->backlog.weight = weight_p; | |
d565b0a1 | 5244 | queue->backlog.gro_list = NULL; |
4ae5544f | 5245 | queue->backlog.gro_count = 0; |
1da177e4 LT |
5246 | } |
5247 | ||
1da177e4 LT |
5248 | dev_boot_phase = 0; |
5249 | ||
505d4f73 EB |
5250 | /* The loopback device is special if any other network devices |
5251 | * is present in a network namespace the loopback device must | |
5252 | * be present. Since we now dynamically allocate and free the | |
5253 | * loopback device ensure this invariant is maintained by | |
5254 | * keeping the loopback device as the first device on the | |
5255 | * list of network devices. Ensuring the loopback devices | |
5256 | * is the first device that appears and the last network device | |
5257 | * that disappears. | |
5258 | */ | |
5259 | if (register_pernet_device(&loopback_net_ops)) | |
5260 | goto out; | |
5261 | ||
5262 | if (register_pernet_device(&default_device_ops)) | |
5263 | goto out; | |
5264 | ||
962cf36c CM |
5265 | open_softirq(NET_TX_SOFTIRQ, net_tx_action); |
5266 | open_softirq(NET_RX_SOFTIRQ, net_rx_action); | |
1da177e4 LT |
5267 | |
5268 | hotcpu_notifier(dev_cpu_callback, 0); | |
5269 | dst_init(); | |
5270 | dev_mcast_init(); | |
5271 | rc = 0; | |
5272 | out: | |
5273 | return rc; | |
5274 | } | |
5275 | ||
5276 | subsys_initcall(net_dev_init); | |
5277 | ||
e88721f8 KK |
5278 | static int __init initialize_hashrnd(void) |
5279 | { | |
5280 | get_random_bytes(&skb_tx_hashrnd, sizeof(skb_tx_hashrnd)); | |
5281 | return 0; | |
5282 | } | |
5283 | ||
5284 | late_initcall_sync(initialize_hashrnd); | |
5285 | ||
1da177e4 LT |
5286 | EXPORT_SYMBOL(__dev_get_by_index); |
5287 | EXPORT_SYMBOL(__dev_get_by_name); | |
5288 | EXPORT_SYMBOL(__dev_remove_pack); | |
c2373ee9 | 5289 | EXPORT_SYMBOL(dev_valid_name); |
1da177e4 LT |
5290 | EXPORT_SYMBOL(dev_add_pack); |
5291 | EXPORT_SYMBOL(dev_alloc_name); | |
5292 | EXPORT_SYMBOL(dev_close); | |
5293 | EXPORT_SYMBOL(dev_get_by_flags); | |
5294 | EXPORT_SYMBOL(dev_get_by_index); | |
5295 | EXPORT_SYMBOL(dev_get_by_name); | |
1da177e4 LT |
5296 | EXPORT_SYMBOL(dev_open); |
5297 | EXPORT_SYMBOL(dev_queue_xmit); | |
5298 | EXPORT_SYMBOL(dev_remove_pack); | |
5299 | EXPORT_SYMBOL(dev_set_allmulti); | |
5300 | EXPORT_SYMBOL(dev_set_promiscuity); | |
5301 | EXPORT_SYMBOL(dev_change_flags); | |
5302 | EXPORT_SYMBOL(dev_set_mtu); | |
5303 | EXPORT_SYMBOL(dev_set_mac_address); | |
5304 | EXPORT_SYMBOL(free_netdev); | |
5305 | EXPORT_SYMBOL(netdev_boot_setup_check); | |
5306 | EXPORT_SYMBOL(netdev_set_master); | |
5307 | EXPORT_SYMBOL(netdev_state_change); | |
5308 | EXPORT_SYMBOL(netif_receive_skb); | |
5309 | EXPORT_SYMBOL(netif_rx); | |
5310 | EXPORT_SYMBOL(register_gifconf); | |
5311 | EXPORT_SYMBOL(register_netdevice); | |
5312 | EXPORT_SYMBOL(register_netdevice_notifier); | |
5313 | EXPORT_SYMBOL(skb_checksum_help); | |
5314 | EXPORT_SYMBOL(synchronize_net); | |
5315 | EXPORT_SYMBOL(unregister_netdevice); | |
5316 | EXPORT_SYMBOL(unregister_netdevice_notifier); | |
5317 | EXPORT_SYMBOL(net_enable_timestamp); | |
5318 | EXPORT_SYMBOL(net_disable_timestamp); | |
5319 | EXPORT_SYMBOL(dev_get_flags); | |
5320 | ||
5321 | #if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE) | |
5322 | EXPORT_SYMBOL(br_handle_frame_hook); | |
5323 | EXPORT_SYMBOL(br_fdb_get_hook); | |
5324 | EXPORT_SYMBOL(br_fdb_put_hook); | |
5325 | #endif | |
5326 | ||
1da177e4 | 5327 | EXPORT_SYMBOL(dev_load); |
1da177e4 LT |
5328 | |
5329 | EXPORT_PER_CPU_SYMBOL(softnet_data); |