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> | |
08e9897d | 82 | #include <linux/hash.h> |
5a0e3ad6 | 83 | #include <linux/slab.h> |
1da177e4 | 84 | #include <linux/sched.h> |
4a3e2f71 | 85 | #include <linux/mutex.h> |
1da177e4 LT |
86 | #include <linux/string.h> |
87 | #include <linux/mm.h> | |
88 | #include <linux/socket.h> | |
89 | #include <linux/sockios.h> | |
90 | #include <linux/errno.h> | |
91 | #include <linux/interrupt.h> | |
92 | #include <linux/if_ether.h> | |
93 | #include <linux/netdevice.h> | |
94 | #include <linux/etherdevice.h> | |
0187bdfb | 95 | #include <linux/ethtool.h> |
1da177e4 LT |
96 | #include <linux/notifier.h> |
97 | #include <linux/skbuff.h> | |
457c4cbc | 98 | #include <net/net_namespace.h> |
1da177e4 LT |
99 | #include <net/sock.h> |
100 | #include <linux/rtnetlink.h> | |
101 | #include <linux/proc_fs.h> | |
102 | #include <linux/seq_file.h> | |
103 | #include <linux/stat.h> | |
104 | #include <linux/if_bridge.h> | |
b863ceb7 | 105 | #include <linux/if_macvlan.h> |
1da177e4 LT |
106 | #include <net/dst.h> |
107 | #include <net/pkt_sched.h> | |
108 | #include <net/checksum.h> | |
44540960 | 109 | #include <net/xfrm.h> |
1da177e4 LT |
110 | #include <linux/highmem.h> |
111 | #include <linux/init.h> | |
112 | #include <linux/kmod.h> | |
113 | #include <linux/module.h> | |
1da177e4 LT |
114 | #include <linux/netpoll.h> |
115 | #include <linux/rcupdate.h> | |
116 | #include <linux/delay.h> | |
295f4a1f | 117 | #include <net/wext.h> |
1da177e4 | 118 | #include <net/iw_handler.h> |
1da177e4 | 119 | #include <asm/current.h> |
5bdb9886 | 120 | #include <linux/audit.h> |
db217334 | 121 | #include <linux/dmaengine.h> |
f6a78bfc | 122 | #include <linux/err.h> |
c7fa9d18 | 123 | #include <linux/ctype.h> |
723e98b7 | 124 | #include <linux/if_arp.h> |
6de329e2 | 125 | #include <linux/if_vlan.h> |
8f0f2223 | 126 | #include <linux/ip.h> |
ad55dcaf | 127 | #include <net/ip.h> |
8f0f2223 DM |
128 | #include <linux/ipv6.h> |
129 | #include <linux/in.h> | |
b6b2fed1 DM |
130 | #include <linux/jhash.h> |
131 | #include <linux/random.h> | |
9cbc1cb8 | 132 | #include <trace/events/napi.h> |
1da177e4 | 133 | |
342709ef PE |
134 | #include "net-sysfs.h" |
135 | ||
d565b0a1 HX |
136 | /* Instead of increasing this, you should create a hash table. */ |
137 | #define MAX_GRO_SKBS 8 | |
138 | ||
5d38a079 HX |
139 | /* This should be increased if a protocol with a bigger head is added. */ |
140 | #define GRO_MAX_HEAD (MAX_HEADER + 128) | |
141 | ||
1da177e4 LT |
142 | /* |
143 | * The list of packet types we will receive (as opposed to discard) | |
144 | * and the routines to invoke. | |
145 | * | |
146 | * Why 16. Because with 16 the only overlap we get on a hash of the | |
147 | * low nibble of the protocol value is RARP/SNAP/X.25. | |
148 | * | |
149 | * NOTE: That is no longer true with the addition of VLAN tags. Not | |
150 | * sure which should go first, but I bet it won't make much | |
151 | * difference if we are running VLANs. The good news is that | |
152 | * this protocol won't be in the list unless compiled in, so | |
3041a069 | 153 | * the average user (w/out VLANs) will not be adversely affected. |
1da177e4 LT |
154 | * --BLG |
155 | * | |
156 | * 0800 IP | |
157 | * 8100 802.1Q VLAN | |
158 | * 0001 802.3 | |
159 | * 0002 AX.25 | |
160 | * 0004 802.2 | |
161 | * 8035 RARP | |
162 | * 0005 SNAP | |
163 | * 0805 X.25 | |
164 | * 0806 ARP | |
165 | * 8137 IPX | |
166 | * 0009 Localtalk | |
167 | * 86DD IPv6 | |
168 | */ | |
169 | ||
82d8a867 PE |
170 | #define PTYPE_HASH_SIZE (16) |
171 | #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) | |
172 | ||
1da177e4 | 173 | static DEFINE_SPINLOCK(ptype_lock); |
82d8a867 | 174 | static struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; |
6b2bedc3 | 175 | static struct list_head ptype_all __read_mostly; /* Taps */ |
1da177e4 | 176 | |
1da177e4 | 177 | /* |
7562f876 | 178 | * The @dev_base_head list is protected by @dev_base_lock and the rtnl |
1da177e4 LT |
179 | * semaphore. |
180 | * | |
c6d14c84 | 181 | * Pure readers hold dev_base_lock for reading, or rcu_read_lock() |
1da177e4 LT |
182 | * |
183 | * Writers must hold the rtnl semaphore while they loop through the | |
7562f876 | 184 | * dev_base_head list, and hold dev_base_lock for writing when they do the |
1da177e4 LT |
185 | * actual updates. This allows pure readers to access the list even |
186 | * while a writer is preparing to update it. | |
187 | * | |
188 | * To put it another way, dev_base_lock is held for writing only to | |
189 | * protect against pure readers; the rtnl semaphore provides the | |
190 | * protection against other writers. | |
191 | * | |
192 | * See, for example usages, register_netdevice() and | |
193 | * unregister_netdevice(), which must be called with the rtnl | |
194 | * semaphore held. | |
195 | */ | |
1da177e4 | 196 | DEFINE_RWLOCK(dev_base_lock); |
1da177e4 LT |
197 | EXPORT_SYMBOL(dev_base_lock); |
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)); | |
08e9897d | 202 | return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)]; |
1da177e4 LT |
203 | } |
204 | ||
881d966b | 205 | static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex) |
1da177e4 | 206 | { |
7c28bd0b | 207 | return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 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); | |
c6d14c84 | 218 | list_add_tail_rcu(&dev->dev_list, &net->dev_base_head); |
72c9528b | 219 | hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name)); |
fb699dfd ED |
220 | hlist_add_head_rcu(&dev->index_hlist, |
221 | dev_index_hash(net, dev->ifindex)); | |
ce286d32 EB |
222 | write_unlock_bh(&dev_base_lock); |
223 | return 0; | |
224 | } | |
225 | ||
fb699dfd ED |
226 | /* Device list removal |
227 | * caller must respect a RCU grace period before freeing/reusing dev | |
228 | */ | |
ce286d32 EB |
229 | static void unlist_netdevice(struct net_device *dev) |
230 | { | |
231 | ASSERT_RTNL(); | |
232 | ||
233 | /* Unlink dev from the device chain */ | |
234 | write_lock_bh(&dev_base_lock); | |
c6d14c84 | 235 | list_del_rcu(&dev->dev_list); |
72c9528b | 236 | hlist_del_rcu(&dev->name_hlist); |
fb699dfd | 237 | hlist_del_rcu(&dev->index_hlist); |
ce286d32 EB |
238 | write_unlock_bh(&dev_base_lock); |
239 | } | |
240 | ||
1da177e4 LT |
241 | /* |
242 | * Our notifier list | |
243 | */ | |
244 | ||
f07d5b94 | 245 | static RAW_NOTIFIER_HEAD(netdev_chain); |
1da177e4 LT |
246 | |
247 | /* | |
248 | * Device drivers call our routines to queue packets here. We empty the | |
249 | * queue in the local softnet handler. | |
250 | */ | |
bea3348e SH |
251 | |
252 | DEFINE_PER_CPU(struct softnet_data, softnet_data); | |
d1b19dff | 253 | EXPORT_PER_CPU_SYMBOL(softnet_data); |
1da177e4 | 254 | |
cf508b12 | 255 | #ifdef CONFIG_LOCKDEP |
723e98b7 | 256 | /* |
c773e847 | 257 | * register_netdevice() inits txq->_xmit_lock and sets lockdep class |
723e98b7 JP |
258 | * according to dev->type |
259 | */ | |
260 | static const unsigned short netdev_lock_type[] = | |
261 | {ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25, | |
262 | ARPHRD_PRONET, ARPHRD_CHAOS, ARPHRD_IEEE802, ARPHRD_ARCNET, | |
263 | ARPHRD_APPLETLK, ARPHRD_DLCI, ARPHRD_ATM, ARPHRD_METRICOM, | |
264 | ARPHRD_IEEE1394, ARPHRD_EUI64, ARPHRD_INFINIBAND, ARPHRD_SLIP, | |
265 | ARPHRD_CSLIP, ARPHRD_SLIP6, ARPHRD_CSLIP6, ARPHRD_RSRVD, | |
266 | ARPHRD_ADAPT, ARPHRD_ROSE, ARPHRD_X25, ARPHRD_HWX25, | |
267 | ARPHRD_PPP, ARPHRD_CISCO, ARPHRD_LAPB, ARPHRD_DDCMP, | |
268 | ARPHRD_RAWHDLC, ARPHRD_TUNNEL, ARPHRD_TUNNEL6, ARPHRD_FRAD, | |
269 | ARPHRD_SKIP, ARPHRD_LOOPBACK, ARPHRD_LOCALTLK, ARPHRD_FDDI, | |
270 | ARPHRD_BIF, ARPHRD_SIT, ARPHRD_IPDDP, ARPHRD_IPGRE, | |
271 | ARPHRD_PIMREG, ARPHRD_HIPPI, ARPHRD_ASH, ARPHRD_ECONET, | |
272 | ARPHRD_IRDA, ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL, | |
273 | ARPHRD_FCFABRIC, ARPHRD_IEEE802_TR, ARPHRD_IEEE80211, | |
2d91d78b | 274 | ARPHRD_IEEE80211_PRISM, ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, |
929122cd | 275 | ARPHRD_PHONET_PIPE, ARPHRD_IEEE802154, |
fcb94e42 | 276 | ARPHRD_VOID, ARPHRD_NONE}; |
723e98b7 | 277 | |
36cbd3dc | 278 | static const char *const netdev_lock_name[] = |
723e98b7 JP |
279 | {"_xmit_NETROM", "_xmit_ETHER", "_xmit_EETHER", "_xmit_AX25", |
280 | "_xmit_PRONET", "_xmit_CHAOS", "_xmit_IEEE802", "_xmit_ARCNET", | |
281 | "_xmit_APPLETLK", "_xmit_DLCI", "_xmit_ATM", "_xmit_METRICOM", | |
282 | "_xmit_IEEE1394", "_xmit_EUI64", "_xmit_INFINIBAND", "_xmit_SLIP", | |
283 | "_xmit_CSLIP", "_xmit_SLIP6", "_xmit_CSLIP6", "_xmit_RSRVD", | |
284 | "_xmit_ADAPT", "_xmit_ROSE", "_xmit_X25", "_xmit_HWX25", | |
285 | "_xmit_PPP", "_xmit_CISCO", "_xmit_LAPB", "_xmit_DDCMP", | |
286 | "_xmit_RAWHDLC", "_xmit_TUNNEL", "_xmit_TUNNEL6", "_xmit_FRAD", | |
287 | "_xmit_SKIP", "_xmit_LOOPBACK", "_xmit_LOCALTLK", "_xmit_FDDI", | |
288 | "_xmit_BIF", "_xmit_SIT", "_xmit_IPDDP", "_xmit_IPGRE", | |
289 | "_xmit_PIMREG", "_xmit_HIPPI", "_xmit_ASH", "_xmit_ECONET", | |
290 | "_xmit_IRDA", "_xmit_FCPP", "_xmit_FCAL", "_xmit_FCPL", | |
291 | "_xmit_FCFABRIC", "_xmit_IEEE802_TR", "_xmit_IEEE80211", | |
2d91d78b | 292 | "_xmit_IEEE80211_PRISM", "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", |
929122cd | 293 | "_xmit_PHONET_PIPE", "_xmit_IEEE802154", |
fcb94e42 | 294 | "_xmit_VOID", "_xmit_NONE"}; |
723e98b7 JP |
295 | |
296 | static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)]; | |
cf508b12 | 297 | static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)]; |
723e98b7 JP |
298 | |
299 | static inline unsigned short netdev_lock_pos(unsigned short dev_type) | |
300 | { | |
301 | int i; | |
302 | ||
303 | for (i = 0; i < ARRAY_SIZE(netdev_lock_type); i++) | |
304 | if (netdev_lock_type[i] == dev_type) | |
305 | return i; | |
306 | /* the last key is used by default */ | |
307 | return ARRAY_SIZE(netdev_lock_type) - 1; | |
308 | } | |
309 | ||
cf508b12 DM |
310 | static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock, |
311 | unsigned short dev_type) | |
723e98b7 JP |
312 | { |
313 | int i; | |
314 | ||
315 | i = netdev_lock_pos(dev_type); | |
316 | lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i], | |
317 | netdev_lock_name[i]); | |
318 | } | |
cf508b12 DM |
319 | |
320 | static inline void netdev_set_addr_lockdep_class(struct net_device *dev) | |
321 | { | |
322 | int i; | |
323 | ||
324 | i = netdev_lock_pos(dev->type); | |
325 | lockdep_set_class_and_name(&dev->addr_list_lock, | |
326 | &netdev_addr_lock_key[i], | |
327 | netdev_lock_name[i]); | |
328 | } | |
723e98b7 | 329 | #else |
cf508b12 DM |
330 | static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock, |
331 | unsigned short dev_type) | |
332 | { | |
333 | } | |
334 | static inline void netdev_set_addr_lockdep_class(struct net_device *dev) | |
723e98b7 JP |
335 | { |
336 | } | |
337 | #endif | |
1da177e4 LT |
338 | |
339 | /******************************************************************************* | |
340 | ||
341 | Protocol management and registration routines | |
342 | ||
343 | *******************************************************************************/ | |
344 | ||
1da177e4 LT |
345 | /* |
346 | * Add a protocol ID to the list. Now that the input handler is | |
347 | * smarter we can dispense with all the messy stuff that used to be | |
348 | * here. | |
349 | * | |
350 | * BEWARE!!! Protocol handlers, mangling input packets, | |
351 | * MUST BE last in hash buckets and checking protocol handlers | |
352 | * MUST start from promiscuous ptype_all chain in net_bh. | |
353 | * It is true now, do not change it. | |
354 | * Explanation follows: if protocol handler, mangling packet, will | |
355 | * be the first on list, it is not able to sense, that packet | |
356 | * is cloned and should be copied-on-write, so that it will | |
357 | * change it and subsequent readers will get broken packet. | |
358 | * --ANK (980803) | |
359 | */ | |
360 | ||
361 | /** | |
362 | * dev_add_pack - add packet handler | |
363 | * @pt: packet type declaration | |
364 | * | |
365 | * Add a protocol handler to the networking stack. The passed &packet_type | |
366 | * is linked into kernel lists and may not be freed until it has been | |
367 | * removed from the kernel lists. | |
368 | * | |
4ec93edb | 369 | * This call does not sleep therefore it can not |
1da177e4 LT |
370 | * guarantee all CPU's that are in middle of receiving packets |
371 | * will see the new packet type (until the next received packet). | |
372 | */ | |
373 | ||
374 | void dev_add_pack(struct packet_type *pt) | |
375 | { | |
376 | int hash; | |
377 | ||
378 | spin_lock_bh(&ptype_lock); | |
9be9a6b9 | 379 | if (pt->type == htons(ETH_P_ALL)) |
1da177e4 | 380 | list_add_rcu(&pt->list, &ptype_all); |
9be9a6b9 | 381 | else { |
82d8a867 | 382 | hash = ntohs(pt->type) & PTYPE_HASH_MASK; |
1da177e4 LT |
383 | list_add_rcu(&pt->list, &ptype_base[hash]); |
384 | } | |
385 | spin_unlock_bh(&ptype_lock); | |
386 | } | |
d1b19dff | 387 | EXPORT_SYMBOL(dev_add_pack); |
1da177e4 | 388 | |
1da177e4 LT |
389 | /** |
390 | * __dev_remove_pack - remove packet handler | |
391 | * @pt: packet type declaration | |
392 | * | |
393 | * Remove a protocol handler that was previously added to the kernel | |
394 | * protocol handlers by dev_add_pack(). The passed &packet_type is removed | |
395 | * from the kernel lists and can be freed or reused once this function | |
4ec93edb | 396 | * returns. |
1da177e4 LT |
397 | * |
398 | * The packet type might still be in use by receivers | |
399 | * and must not be freed until after all the CPU's have gone | |
400 | * through a quiescent state. | |
401 | */ | |
402 | void __dev_remove_pack(struct packet_type *pt) | |
403 | { | |
404 | struct list_head *head; | |
405 | struct packet_type *pt1; | |
406 | ||
407 | spin_lock_bh(&ptype_lock); | |
408 | ||
9be9a6b9 | 409 | if (pt->type == htons(ETH_P_ALL)) |
1da177e4 | 410 | head = &ptype_all; |
9be9a6b9 | 411 | else |
82d8a867 | 412 | head = &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK]; |
1da177e4 LT |
413 | |
414 | list_for_each_entry(pt1, head, list) { | |
415 | if (pt == pt1) { | |
416 | list_del_rcu(&pt->list); | |
417 | goto out; | |
418 | } | |
419 | } | |
420 | ||
421 | printk(KERN_WARNING "dev_remove_pack: %p not found.\n", pt); | |
422 | out: | |
423 | spin_unlock_bh(&ptype_lock); | |
424 | } | |
d1b19dff ED |
425 | EXPORT_SYMBOL(__dev_remove_pack); |
426 | ||
1da177e4 LT |
427 | /** |
428 | * dev_remove_pack - remove packet handler | |
429 | * @pt: packet type declaration | |
430 | * | |
431 | * Remove a protocol handler that was previously added to the kernel | |
432 | * protocol handlers by dev_add_pack(). The passed &packet_type is removed | |
433 | * from the kernel lists and can be freed or reused once this function | |
434 | * returns. | |
435 | * | |
436 | * This call sleeps to guarantee that no CPU is looking at the packet | |
437 | * type after return. | |
438 | */ | |
439 | void dev_remove_pack(struct packet_type *pt) | |
440 | { | |
441 | __dev_remove_pack(pt); | |
4ec93edb | 442 | |
1da177e4 LT |
443 | synchronize_net(); |
444 | } | |
d1b19dff | 445 | EXPORT_SYMBOL(dev_remove_pack); |
1da177e4 LT |
446 | |
447 | /****************************************************************************** | |
448 | ||
449 | Device Boot-time Settings Routines | |
450 | ||
451 | *******************************************************************************/ | |
452 | ||
453 | /* Boot time configuration table */ | |
454 | static struct netdev_boot_setup dev_boot_setup[NETDEV_BOOT_SETUP_MAX]; | |
455 | ||
456 | /** | |
457 | * netdev_boot_setup_add - add new setup entry | |
458 | * @name: name of the device | |
459 | * @map: configured settings for the device | |
460 | * | |
461 | * Adds new setup entry to the dev_boot_setup list. The function | |
462 | * returns 0 on error and 1 on success. This is a generic routine to | |
463 | * all netdevices. | |
464 | */ | |
465 | static int netdev_boot_setup_add(char *name, struct ifmap *map) | |
466 | { | |
467 | struct netdev_boot_setup *s; | |
468 | int i; | |
469 | ||
470 | s = dev_boot_setup; | |
471 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { | |
472 | if (s[i].name[0] == '\0' || s[i].name[0] == ' ') { | |
473 | memset(s[i].name, 0, sizeof(s[i].name)); | |
93b3cff9 | 474 | strlcpy(s[i].name, name, IFNAMSIZ); |
1da177e4 LT |
475 | memcpy(&s[i].map, map, sizeof(s[i].map)); |
476 | break; | |
477 | } | |
478 | } | |
479 | ||
480 | return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1; | |
481 | } | |
482 | ||
483 | /** | |
484 | * netdev_boot_setup_check - check boot time settings | |
485 | * @dev: the netdevice | |
486 | * | |
487 | * Check boot time settings for the device. | |
488 | * The found settings are set for the device to be used | |
489 | * later in the device probing. | |
490 | * Returns 0 if no settings found, 1 if they are. | |
491 | */ | |
492 | int netdev_boot_setup_check(struct net_device *dev) | |
493 | { | |
494 | struct netdev_boot_setup *s = dev_boot_setup; | |
495 | int i; | |
496 | ||
497 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { | |
498 | if (s[i].name[0] != '\0' && s[i].name[0] != ' ' && | |
93b3cff9 | 499 | !strcmp(dev->name, s[i].name)) { |
1da177e4 LT |
500 | dev->irq = s[i].map.irq; |
501 | dev->base_addr = s[i].map.base_addr; | |
502 | dev->mem_start = s[i].map.mem_start; | |
503 | dev->mem_end = s[i].map.mem_end; | |
504 | return 1; | |
505 | } | |
506 | } | |
507 | return 0; | |
508 | } | |
d1b19dff | 509 | EXPORT_SYMBOL(netdev_boot_setup_check); |
1da177e4 LT |
510 | |
511 | ||
512 | /** | |
513 | * netdev_boot_base - get address from boot time settings | |
514 | * @prefix: prefix for network device | |
515 | * @unit: id for network device | |
516 | * | |
517 | * Check boot time settings for the base address of device. | |
518 | * The found settings are set for the device to be used | |
519 | * later in the device probing. | |
520 | * Returns 0 if no settings found. | |
521 | */ | |
522 | unsigned long netdev_boot_base(const char *prefix, int unit) | |
523 | { | |
524 | const struct netdev_boot_setup *s = dev_boot_setup; | |
525 | char name[IFNAMSIZ]; | |
526 | int i; | |
527 | ||
528 | sprintf(name, "%s%d", prefix, unit); | |
529 | ||
530 | /* | |
531 | * If device already registered then return base of 1 | |
532 | * to indicate not to probe for this interface | |
533 | */ | |
881d966b | 534 | if (__dev_get_by_name(&init_net, name)) |
1da177e4 LT |
535 | return 1; |
536 | ||
537 | for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) | |
538 | if (!strcmp(name, s[i].name)) | |
539 | return s[i].map.base_addr; | |
540 | return 0; | |
541 | } | |
542 | ||
543 | /* | |
544 | * Saves at boot time configured settings for any netdevice. | |
545 | */ | |
546 | int __init netdev_boot_setup(char *str) | |
547 | { | |
548 | int ints[5]; | |
549 | struct ifmap map; | |
550 | ||
551 | str = get_options(str, ARRAY_SIZE(ints), ints); | |
552 | if (!str || !*str) | |
553 | return 0; | |
554 | ||
555 | /* Save settings */ | |
556 | memset(&map, 0, sizeof(map)); | |
557 | if (ints[0] > 0) | |
558 | map.irq = ints[1]; | |
559 | if (ints[0] > 1) | |
560 | map.base_addr = ints[2]; | |
561 | if (ints[0] > 2) | |
562 | map.mem_start = ints[3]; | |
563 | if (ints[0] > 3) | |
564 | map.mem_end = ints[4]; | |
565 | ||
566 | /* Add new entry to the list */ | |
567 | return netdev_boot_setup_add(str, &map); | |
568 | } | |
569 | ||
570 | __setup("netdev=", netdev_boot_setup); | |
571 | ||
572 | /******************************************************************************* | |
573 | ||
574 | Device Interface Subroutines | |
575 | ||
576 | *******************************************************************************/ | |
577 | ||
578 | /** | |
579 | * __dev_get_by_name - find a device by its name | |
c4ea43c5 | 580 | * @net: the applicable net namespace |
1da177e4 LT |
581 | * @name: name to find |
582 | * | |
583 | * Find an interface by name. Must be called under RTNL semaphore | |
584 | * or @dev_base_lock. If the name is found a pointer to the device | |
585 | * is returned. If the name is not found then %NULL is returned. The | |
586 | * reference counters are not incremented so the caller must be | |
587 | * careful with locks. | |
588 | */ | |
589 | ||
881d966b | 590 | struct net_device *__dev_get_by_name(struct net *net, const char *name) |
1da177e4 LT |
591 | { |
592 | struct hlist_node *p; | |
0bd8d536 ED |
593 | struct net_device *dev; |
594 | struct hlist_head *head = dev_name_hash(net, name); | |
1da177e4 | 595 | |
0bd8d536 | 596 | hlist_for_each_entry(dev, p, head, name_hlist) |
1da177e4 LT |
597 | if (!strncmp(dev->name, name, IFNAMSIZ)) |
598 | return dev; | |
0bd8d536 | 599 | |
1da177e4 LT |
600 | return NULL; |
601 | } | |
d1b19dff | 602 | EXPORT_SYMBOL(__dev_get_by_name); |
1da177e4 | 603 | |
72c9528b ED |
604 | /** |
605 | * dev_get_by_name_rcu - find a device by its name | |
606 | * @net: the applicable net namespace | |
607 | * @name: name to find | |
608 | * | |
609 | * Find an interface by name. | |
610 | * If the name is found a pointer to the device is returned. | |
611 | * If the name is not found then %NULL is returned. | |
612 | * The reference counters are not incremented so the caller must be | |
613 | * careful with locks. The caller must hold RCU lock. | |
614 | */ | |
615 | ||
616 | struct net_device *dev_get_by_name_rcu(struct net *net, const char *name) | |
617 | { | |
618 | struct hlist_node *p; | |
619 | struct net_device *dev; | |
620 | struct hlist_head *head = dev_name_hash(net, name); | |
621 | ||
622 | hlist_for_each_entry_rcu(dev, p, head, name_hlist) | |
623 | if (!strncmp(dev->name, name, IFNAMSIZ)) | |
624 | return dev; | |
625 | ||
626 | return NULL; | |
627 | } | |
628 | EXPORT_SYMBOL(dev_get_by_name_rcu); | |
629 | ||
1da177e4 LT |
630 | /** |
631 | * dev_get_by_name - find a device by its name | |
c4ea43c5 | 632 | * @net: the applicable net namespace |
1da177e4 LT |
633 | * @name: name to find |
634 | * | |
635 | * Find an interface by name. This can be called from any | |
636 | * context and does its own locking. The returned handle has | |
637 | * the usage count incremented and the caller must use dev_put() to | |
638 | * release it when it is no longer needed. %NULL is returned if no | |
639 | * matching device is found. | |
640 | */ | |
641 | ||
881d966b | 642 | struct net_device *dev_get_by_name(struct net *net, const char *name) |
1da177e4 LT |
643 | { |
644 | struct net_device *dev; | |
645 | ||
72c9528b ED |
646 | rcu_read_lock(); |
647 | dev = dev_get_by_name_rcu(net, name); | |
1da177e4 LT |
648 | if (dev) |
649 | dev_hold(dev); | |
72c9528b | 650 | rcu_read_unlock(); |
1da177e4 LT |
651 | return dev; |
652 | } | |
d1b19dff | 653 | EXPORT_SYMBOL(dev_get_by_name); |
1da177e4 LT |
654 | |
655 | /** | |
656 | * __dev_get_by_index - find a device by its ifindex | |
c4ea43c5 | 657 | * @net: the applicable net namespace |
1da177e4 LT |
658 | * @ifindex: index of device |
659 | * | |
660 | * Search for an interface by index. Returns %NULL if the device | |
661 | * is not found or a pointer to the device. The device has not | |
662 | * had its reference counter increased so the caller must be careful | |
663 | * about locking. The caller must hold either the RTNL semaphore | |
664 | * or @dev_base_lock. | |
665 | */ | |
666 | ||
881d966b | 667 | struct net_device *__dev_get_by_index(struct net *net, int ifindex) |
1da177e4 LT |
668 | { |
669 | struct hlist_node *p; | |
0bd8d536 ED |
670 | struct net_device *dev; |
671 | struct hlist_head *head = dev_index_hash(net, ifindex); | |
1da177e4 | 672 | |
0bd8d536 | 673 | hlist_for_each_entry(dev, p, head, index_hlist) |
1da177e4 LT |
674 | if (dev->ifindex == ifindex) |
675 | return dev; | |
0bd8d536 | 676 | |
1da177e4 LT |
677 | return NULL; |
678 | } | |
d1b19dff | 679 | EXPORT_SYMBOL(__dev_get_by_index); |
1da177e4 | 680 | |
fb699dfd ED |
681 | /** |
682 | * dev_get_by_index_rcu - find a device by its ifindex | |
683 | * @net: the applicable net namespace | |
684 | * @ifindex: index of device | |
685 | * | |
686 | * Search for an interface by index. Returns %NULL if the device | |
687 | * is not found or a pointer to the device. The device has not | |
688 | * had its reference counter increased so the caller must be careful | |
689 | * about locking. The caller must hold RCU lock. | |
690 | */ | |
691 | ||
692 | struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex) | |
693 | { | |
694 | struct hlist_node *p; | |
695 | struct net_device *dev; | |
696 | struct hlist_head *head = dev_index_hash(net, ifindex); | |
697 | ||
698 | hlist_for_each_entry_rcu(dev, p, head, index_hlist) | |
699 | if (dev->ifindex == ifindex) | |
700 | return dev; | |
701 | ||
702 | return NULL; | |
703 | } | |
704 | EXPORT_SYMBOL(dev_get_by_index_rcu); | |
705 | ||
1da177e4 LT |
706 | |
707 | /** | |
708 | * dev_get_by_index - find a device by its ifindex | |
c4ea43c5 | 709 | * @net: the applicable net namespace |
1da177e4 LT |
710 | * @ifindex: index of device |
711 | * | |
712 | * Search for an interface by index. Returns NULL if the device | |
713 | * is not found or a pointer to the device. The device returned has | |
714 | * had a reference added and the pointer is safe until the user calls | |
715 | * dev_put to indicate they have finished with it. | |
716 | */ | |
717 | ||
881d966b | 718 | struct net_device *dev_get_by_index(struct net *net, int ifindex) |
1da177e4 LT |
719 | { |
720 | struct net_device *dev; | |
721 | ||
fb699dfd ED |
722 | rcu_read_lock(); |
723 | dev = dev_get_by_index_rcu(net, ifindex); | |
1da177e4 LT |
724 | if (dev) |
725 | dev_hold(dev); | |
fb699dfd | 726 | rcu_read_unlock(); |
1da177e4 LT |
727 | return dev; |
728 | } | |
d1b19dff | 729 | EXPORT_SYMBOL(dev_get_by_index); |
1da177e4 LT |
730 | |
731 | /** | |
732 | * dev_getbyhwaddr - find a device by its hardware address | |
c4ea43c5 | 733 | * @net: the applicable net namespace |
1da177e4 LT |
734 | * @type: media type of device |
735 | * @ha: hardware address | |
736 | * | |
737 | * Search for an interface by MAC address. Returns NULL if the device | |
738 | * is not found or a pointer to the device. The caller must hold the | |
739 | * rtnl semaphore. The returned device has not had its ref count increased | |
740 | * and the caller must therefore be careful about locking | |
741 | * | |
742 | * BUGS: | |
743 | * If the API was consistent this would be __dev_get_by_hwaddr | |
744 | */ | |
745 | ||
881d966b | 746 | struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, char *ha) |
1da177e4 LT |
747 | { |
748 | struct net_device *dev; | |
749 | ||
750 | ASSERT_RTNL(); | |
751 | ||
81103a52 | 752 | for_each_netdev(net, dev) |
1da177e4 LT |
753 | if (dev->type == type && |
754 | !memcmp(dev->dev_addr, ha, dev->addr_len)) | |
7562f876 PE |
755 | return dev; |
756 | ||
757 | return NULL; | |
1da177e4 | 758 | } |
cf309e3f JF |
759 | EXPORT_SYMBOL(dev_getbyhwaddr); |
760 | ||
881d966b | 761 | struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type) |
1da177e4 LT |
762 | { |
763 | struct net_device *dev; | |
764 | ||
4e9cac2b | 765 | ASSERT_RTNL(); |
881d966b | 766 | for_each_netdev(net, dev) |
4e9cac2b | 767 | if (dev->type == type) |
7562f876 PE |
768 | return dev; |
769 | ||
770 | return NULL; | |
4e9cac2b | 771 | } |
4e9cac2b PM |
772 | EXPORT_SYMBOL(__dev_getfirstbyhwtype); |
773 | ||
881d966b | 774 | struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type) |
4e9cac2b PM |
775 | { |
776 | struct net_device *dev; | |
777 | ||
778 | rtnl_lock(); | |
881d966b | 779 | dev = __dev_getfirstbyhwtype(net, type); |
4e9cac2b PM |
780 | if (dev) |
781 | dev_hold(dev); | |
1da177e4 LT |
782 | rtnl_unlock(); |
783 | return dev; | |
784 | } | |
1da177e4 LT |
785 | EXPORT_SYMBOL(dev_getfirstbyhwtype); |
786 | ||
787 | /** | |
788 | * dev_get_by_flags - find any device with given flags | |
c4ea43c5 | 789 | * @net: the applicable net namespace |
1da177e4 LT |
790 | * @if_flags: IFF_* values |
791 | * @mask: bitmask of bits in if_flags to check | |
792 | * | |
793 | * Search for any interface with the given flags. Returns NULL if a device | |
4ec93edb | 794 | * is not found or a pointer to the device. The device returned has |
1da177e4 LT |
795 | * had a reference added and the pointer is safe until the user calls |
796 | * dev_put to indicate they have finished with it. | |
797 | */ | |
798 | ||
d1b19dff ED |
799 | struct net_device *dev_get_by_flags(struct net *net, unsigned short if_flags, |
800 | unsigned short mask) | |
1da177e4 | 801 | { |
7562f876 | 802 | struct net_device *dev, *ret; |
1da177e4 | 803 | |
7562f876 | 804 | ret = NULL; |
c6d14c84 ED |
805 | rcu_read_lock(); |
806 | for_each_netdev_rcu(net, dev) { | |
1da177e4 LT |
807 | if (((dev->flags ^ if_flags) & mask) == 0) { |
808 | dev_hold(dev); | |
7562f876 | 809 | ret = dev; |
1da177e4 LT |
810 | break; |
811 | } | |
812 | } | |
c6d14c84 | 813 | rcu_read_unlock(); |
7562f876 | 814 | return ret; |
1da177e4 | 815 | } |
d1b19dff | 816 | EXPORT_SYMBOL(dev_get_by_flags); |
1da177e4 LT |
817 | |
818 | /** | |
819 | * dev_valid_name - check if name is okay for network device | |
820 | * @name: name string | |
821 | * | |
822 | * Network device names need to be valid file names to | |
c7fa9d18 DM |
823 | * to allow sysfs to work. We also disallow any kind of |
824 | * whitespace. | |
1da177e4 | 825 | */ |
c2373ee9 | 826 | int dev_valid_name(const char *name) |
1da177e4 | 827 | { |
c7fa9d18 DM |
828 | if (*name == '\0') |
829 | return 0; | |
b6fe17d6 SH |
830 | if (strlen(name) >= IFNAMSIZ) |
831 | return 0; | |
c7fa9d18 DM |
832 | if (!strcmp(name, ".") || !strcmp(name, "..")) |
833 | return 0; | |
834 | ||
835 | while (*name) { | |
836 | if (*name == '/' || isspace(*name)) | |
837 | return 0; | |
838 | name++; | |
839 | } | |
840 | return 1; | |
1da177e4 | 841 | } |
d1b19dff | 842 | EXPORT_SYMBOL(dev_valid_name); |
1da177e4 LT |
843 | |
844 | /** | |
b267b179 EB |
845 | * __dev_alloc_name - allocate a name for a device |
846 | * @net: network namespace to allocate the device name in | |
1da177e4 | 847 | * @name: name format string |
b267b179 | 848 | * @buf: scratch buffer and result name string |
1da177e4 LT |
849 | * |
850 | * Passed a format string - eg "lt%d" it will try and find a suitable | |
3041a069 SH |
851 | * id. It scans list of devices to build up a free map, then chooses |
852 | * the first empty slot. The caller must hold the dev_base or rtnl lock | |
853 | * while allocating the name and adding the device in order to avoid | |
854 | * duplicates. | |
855 | * Limited to bits_per_byte * page size devices (ie 32K on most platforms). | |
856 | * Returns the number of the unit assigned or a negative errno code. | |
1da177e4 LT |
857 | */ |
858 | ||
b267b179 | 859 | static int __dev_alloc_name(struct net *net, const char *name, char *buf) |
1da177e4 LT |
860 | { |
861 | int i = 0; | |
1da177e4 LT |
862 | const char *p; |
863 | const int max_netdevices = 8*PAGE_SIZE; | |
cfcabdcc | 864 | unsigned long *inuse; |
1da177e4 LT |
865 | struct net_device *d; |
866 | ||
867 | p = strnchr(name, IFNAMSIZ-1, '%'); | |
868 | if (p) { | |
869 | /* | |
870 | * Verify the string as this thing may have come from | |
871 | * the user. There must be either one "%d" and no other "%" | |
872 | * characters. | |
873 | */ | |
874 | if (p[1] != 'd' || strchr(p + 2, '%')) | |
875 | return -EINVAL; | |
876 | ||
877 | /* Use one page as a bit array of possible slots */ | |
cfcabdcc | 878 | inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC); |
1da177e4 LT |
879 | if (!inuse) |
880 | return -ENOMEM; | |
881 | ||
881d966b | 882 | for_each_netdev(net, d) { |
1da177e4 LT |
883 | if (!sscanf(d->name, name, &i)) |
884 | continue; | |
885 | if (i < 0 || i >= max_netdevices) | |
886 | continue; | |
887 | ||
888 | /* avoid cases where sscanf is not exact inverse of printf */ | |
b267b179 | 889 | snprintf(buf, IFNAMSIZ, name, i); |
1da177e4 LT |
890 | if (!strncmp(buf, d->name, IFNAMSIZ)) |
891 | set_bit(i, inuse); | |
892 | } | |
893 | ||
894 | i = find_first_zero_bit(inuse, max_netdevices); | |
895 | free_page((unsigned long) inuse); | |
896 | } | |
897 | ||
d9031024 OP |
898 | if (buf != name) |
899 | snprintf(buf, IFNAMSIZ, name, i); | |
b267b179 | 900 | if (!__dev_get_by_name(net, buf)) |
1da177e4 | 901 | return i; |
1da177e4 LT |
902 | |
903 | /* It is possible to run out of possible slots | |
904 | * when the name is long and there isn't enough space left | |
905 | * for the digits, or if all bits are used. | |
906 | */ | |
907 | return -ENFILE; | |
908 | } | |
909 | ||
b267b179 EB |
910 | /** |
911 | * dev_alloc_name - allocate a name for a device | |
912 | * @dev: device | |
913 | * @name: name format string | |
914 | * | |
915 | * Passed a format string - eg "lt%d" it will try and find a suitable | |
916 | * id. It scans list of devices to build up a free map, then chooses | |
917 | * the first empty slot. The caller must hold the dev_base or rtnl lock | |
918 | * while allocating the name and adding the device in order to avoid | |
919 | * duplicates. | |
920 | * Limited to bits_per_byte * page size devices (ie 32K on most platforms). | |
921 | * Returns the number of the unit assigned or a negative errno code. | |
922 | */ | |
923 | ||
924 | int dev_alloc_name(struct net_device *dev, const char *name) | |
925 | { | |
926 | char buf[IFNAMSIZ]; | |
927 | struct net *net; | |
928 | int ret; | |
929 | ||
c346dca1 YH |
930 | BUG_ON(!dev_net(dev)); |
931 | net = dev_net(dev); | |
b267b179 EB |
932 | ret = __dev_alloc_name(net, name, buf); |
933 | if (ret >= 0) | |
934 | strlcpy(dev->name, buf, IFNAMSIZ); | |
935 | return ret; | |
936 | } | |
d1b19dff | 937 | EXPORT_SYMBOL(dev_alloc_name); |
b267b179 | 938 | |
d9031024 OP |
939 | static int dev_get_valid_name(struct net *net, const char *name, char *buf, |
940 | bool fmt) | |
941 | { | |
942 | if (!dev_valid_name(name)) | |
943 | return -EINVAL; | |
944 | ||
945 | if (fmt && strchr(name, '%')) | |
946 | return __dev_alloc_name(net, name, buf); | |
947 | else if (__dev_get_by_name(net, name)) | |
948 | return -EEXIST; | |
949 | else if (buf != name) | |
950 | strlcpy(buf, name, IFNAMSIZ); | |
951 | ||
952 | return 0; | |
953 | } | |
1da177e4 LT |
954 | |
955 | /** | |
956 | * dev_change_name - change name of a device | |
957 | * @dev: device | |
958 | * @newname: name (or format string) must be at least IFNAMSIZ | |
959 | * | |
960 | * Change name of a device, can pass format strings "eth%d". | |
961 | * for wildcarding. | |
962 | */ | |
cf04a4c7 | 963 | int dev_change_name(struct net_device *dev, const char *newname) |
1da177e4 | 964 | { |
fcc5a03a | 965 | char oldname[IFNAMSIZ]; |
1da177e4 | 966 | int err = 0; |
fcc5a03a | 967 | int ret; |
881d966b | 968 | struct net *net; |
1da177e4 LT |
969 | |
970 | ASSERT_RTNL(); | |
c346dca1 | 971 | BUG_ON(!dev_net(dev)); |
1da177e4 | 972 | |
c346dca1 | 973 | net = dev_net(dev); |
1da177e4 LT |
974 | if (dev->flags & IFF_UP) |
975 | return -EBUSY; | |
976 | ||
c8d90dca SH |
977 | if (strncmp(newname, dev->name, IFNAMSIZ) == 0) |
978 | return 0; | |
979 | ||
fcc5a03a HX |
980 | memcpy(oldname, dev->name, IFNAMSIZ); |
981 | ||
d9031024 OP |
982 | err = dev_get_valid_name(net, newname, dev->name, 1); |
983 | if (err < 0) | |
984 | return err; | |
1da177e4 | 985 | |
fcc5a03a | 986 | rollback: |
3891845e EB |
987 | /* For now only devices in the initial network namespace |
988 | * are in sysfs. | |
989 | */ | |
09ad9bc7 | 990 | if (net_eq(net, &init_net)) { |
3891845e EB |
991 | ret = device_rename(&dev->dev, dev->name); |
992 | if (ret) { | |
993 | memcpy(dev->name, oldname, IFNAMSIZ); | |
994 | return ret; | |
995 | } | |
dcc99773 | 996 | } |
7f988eab HX |
997 | |
998 | write_lock_bh(&dev_base_lock); | |
92749821 | 999 | hlist_del(&dev->name_hlist); |
72c9528b ED |
1000 | write_unlock_bh(&dev_base_lock); |
1001 | ||
1002 | synchronize_rcu(); | |
1003 | ||
1004 | write_lock_bh(&dev_base_lock); | |
1005 | hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name)); | |
7f988eab HX |
1006 | write_unlock_bh(&dev_base_lock); |
1007 | ||
056925ab | 1008 | ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev); |
fcc5a03a HX |
1009 | ret = notifier_to_errno(ret); |
1010 | ||
1011 | if (ret) { | |
91e9c07b ED |
1012 | /* err >= 0 after dev_alloc_name() or stores the first errno */ |
1013 | if (err >= 0) { | |
fcc5a03a HX |
1014 | err = ret; |
1015 | memcpy(dev->name, oldname, IFNAMSIZ); | |
1016 | goto rollback; | |
91e9c07b ED |
1017 | } else { |
1018 | printk(KERN_ERR | |
1019 | "%s: name change rollback failed: %d.\n", | |
1020 | dev->name, ret); | |
fcc5a03a HX |
1021 | } |
1022 | } | |
1da177e4 LT |
1023 | |
1024 | return err; | |
1025 | } | |
1026 | ||
0b815a1a SH |
1027 | /** |
1028 | * dev_set_alias - change ifalias of a device | |
1029 | * @dev: device | |
1030 | * @alias: name up to IFALIASZ | |
f0db275a | 1031 | * @len: limit of bytes to copy from info |
0b815a1a SH |
1032 | * |
1033 | * Set ifalias for a device, | |
1034 | */ | |
1035 | int dev_set_alias(struct net_device *dev, const char *alias, size_t len) | |
1036 | { | |
1037 | ASSERT_RTNL(); | |
1038 | ||
1039 | if (len >= IFALIASZ) | |
1040 | return -EINVAL; | |
1041 | ||
96ca4a2c OH |
1042 | if (!len) { |
1043 | if (dev->ifalias) { | |
1044 | kfree(dev->ifalias); | |
1045 | dev->ifalias = NULL; | |
1046 | } | |
1047 | return 0; | |
1048 | } | |
1049 | ||
d1b19dff | 1050 | dev->ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL); |
0b815a1a SH |
1051 | if (!dev->ifalias) |
1052 | return -ENOMEM; | |
1053 | ||
1054 | strlcpy(dev->ifalias, alias, len+1); | |
1055 | return len; | |
1056 | } | |
1057 | ||
1058 | ||
d8a33ac4 | 1059 | /** |
3041a069 | 1060 | * netdev_features_change - device changes features |
d8a33ac4 SH |
1061 | * @dev: device to cause notification |
1062 | * | |
1063 | * Called to indicate a device has changed features. | |
1064 | */ | |
1065 | void netdev_features_change(struct net_device *dev) | |
1066 | { | |
056925ab | 1067 | call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev); |
d8a33ac4 SH |
1068 | } |
1069 | EXPORT_SYMBOL(netdev_features_change); | |
1070 | ||
1da177e4 LT |
1071 | /** |
1072 | * netdev_state_change - device changes state | |
1073 | * @dev: device to cause notification | |
1074 | * | |
1075 | * Called to indicate a device has changed state. This function calls | |
1076 | * the notifier chains for netdev_chain and sends a NEWLINK message | |
1077 | * to the routing socket. | |
1078 | */ | |
1079 | void netdev_state_change(struct net_device *dev) | |
1080 | { | |
1081 | if (dev->flags & IFF_UP) { | |
056925ab | 1082 | call_netdevice_notifiers(NETDEV_CHANGE, dev); |
1da177e4 LT |
1083 | rtmsg_ifinfo(RTM_NEWLINK, dev, 0); |
1084 | } | |
1085 | } | |
d1b19dff | 1086 | EXPORT_SYMBOL(netdev_state_change); |
1da177e4 | 1087 | |
75c78500 | 1088 | void netdev_bonding_change(struct net_device *dev, unsigned long event) |
c1da4ac7 | 1089 | { |
75c78500 | 1090 | call_netdevice_notifiers(event, dev); |
c1da4ac7 OG |
1091 | } |
1092 | EXPORT_SYMBOL(netdev_bonding_change); | |
1093 | ||
1da177e4 LT |
1094 | /** |
1095 | * dev_load - load a network module | |
c4ea43c5 | 1096 | * @net: the applicable net namespace |
1da177e4 LT |
1097 | * @name: name of interface |
1098 | * | |
1099 | * If a network interface is not present and the process has suitable | |
1100 | * privileges this function loads the module. If module loading is not | |
1101 | * available in this kernel then it becomes a nop. | |
1102 | */ | |
1103 | ||
881d966b | 1104 | void dev_load(struct net *net, const char *name) |
1da177e4 | 1105 | { |
4ec93edb | 1106 | struct net_device *dev; |
1da177e4 | 1107 | |
72c9528b ED |
1108 | rcu_read_lock(); |
1109 | dev = dev_get_by_name_rcu(net, name); | |
1110 | rcu_read_unlock(); | |
1da177e4 | 1111 | |
a8f80e8f | 1112 | if (!dev && capable(CAP_NET_ADMIN)) |
1da177e4 LT |
1113 | request_module("%s", name); |
1114 | } | |
d1b19dff | 1115 | EXPORT_SYMBOL(dev_load); |
1da177e4 | 1116 | |
bd380811 | 1117 | static int __dev_open(struct net_device *dev) |
1da177e4 | 1118 | { |
d314774c | 1119 | const struct net_device_ops *ops = dev->netdev_ops; |
3b8bcfd5 | 1120 | int ret; |
1da177e4 | 1121 | |
e46b66bc BH |
1122 | ASSERT_RTNL(); |
1123 | ||
1da177e4 LT |
1124 | /* |
1125 | * Is it even present? | |
1126 | */ | |
1127 | if (!netif_device_present(dev)) | |
1128 | return -ENODEV; | |
1129 | ||
3b8bcfd5 JB |
1130 | ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev); |
1131 | ret = notifier_to_errno(ret); | |
1132 | if (ret) | |
1133 | return ret; | |
1134 | ||
1da177e4 LT |
1135 | /* |
1136 | * Call device private open method | |
1137 | */ | |
1138 | set_bit(__LINK_STATE_START, &dev->state); | |
bada339b | 1139 | |
d314774c SH |
1140 | if (ops->ndo_validate_addr) |
1141 | ret = ops->ndo_validate_addr(dev); | |
bada339b | 1142 | |
d314774c SH |
1143 | if (!ret && ops->ndo_open) |
1144 | ret = ops->ndo_open(dev); | |
1da177e4 | 1145 | |
4ec93edb | 1146 | /* |
1da177e4 LT |
1147 | * If it went open OK then: |
1148 | */ | |
1149 | ||
bada339b JG |
1150 | if (ret) |
1151 | clear_bit(__LINK_STATE_START, &dev->state); | |
1152 | else { | |
1da177e4 LT |
1153 | /* |
1154 | * Set the flags. | |
1155 | */ | |
1156 | dev->flags |= IFF_UP; | |
1157 | ||
649274d9 DW |
1158 | /* |
1159 | * Enable NET_DMA | |
1160 | */ | |
b4bd07c2 | 1161 | net_dmaengine_get(); |
649274d9 | 1162 | |
1da177e4 LT |
1163 | /* |
1164 | * Initialize multicasting status | |
1165 | */ | |
4417da66 | 1166 | dev_set_rx_mode(dev); |
1da177e4 LT |
1167 | |
1168 | /* | |
1169 | * Wakeup transmit queue engine | |
1170 | */ | |
1171 | dev_activate(dev); | |
1da177e4 | 1172 | } |
bada339b | 1173 | |
1da177e4 LT |
1174 | return ret; |
1175 | } | |
1176 | ||
1177 | /** | |
bd380811 PM |
1178 | * dev_open - prepare an interface for use. |
1179 | * @dev: device to open | |
1da177e4 | 1180 | * |
bd380811 PM |
1181 | * Takes a device from down to up state. The device's private open |
1182 | * function is invoked and then the multicast lists are loaded. Finally | |
1183 | * the device is moved into the up state and a %NETDEV_UP message is | |
1184 | * sent to the netdev notifier chain. | |
1185 | * | |
1186 | * Calling this function on an active interface is a nop. On a failure | |
1187 | * a negative errno code is returned. | |
1da177e4 | 1188 | */ |
bd380811 PM |
1189 | int dev_open(struct net_device *dev) |
1190 | { | |
1191 | int ret; | |
1192 | ||
1193 | /* | |
1194 | * Is it already up? | |
1195 | */ | |
1196 | if (dev->flags & IFF_UP) | |
1197 | return 0; | |
1198 | ||
1199 | /* | |
1200 | * Open device | |
1201 | */ | |
1202 | ret = __dev_open(dev); | |
1203 | if (ret < 0) | |
1204 | return ret; | |
1205 | ||
1206 | /* | |
1207 | * ... and announce new interface. | |
1208 | */ | |
1209 | rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING); | |
1210 | call_netdevice_notifiers(NETDEV_UP, dev); | |
1211 | ||
1212 | return ret; | |
1213 | } | |
1214 | EXPORT_SYMBOL(dev_open); | |
1215 | ||
1216 | static int __dev_close(struct net_device *dev) | |
1da177e4 | 1217 | { |
d314774c | 1218 | const struct net_device_ops *ops = dev->netdev_ops; |
e46b66bc | 1219 | |
bd380811 | 1220 | ASSERT_RTNL(); |
9d5010db DM |
1221 | might_sleep(); |
1222 | ||
1da177e4 LT |
1223 | /* |
1224 | * Tell people we are going down, so that they can | |
1225 | * prepare to death, when device is still operating. | |
1226 | */ | |
056925ab | 1227 | call_netdevice_notifiers(NETDEV_GOING_DOWN, dev); |
1da177e4 | 1228 | |
1da177e4 LT |
1229 | clear_bit(__LINK_STATE_START, &dev->state); |
1230 | ||
1231 | /* Synchronize to scheduled poll. We cannot touch poll list, | |
bea3348e SH |
1232 | * it can be even on different cpu. So just clear netif_running(). |
1233 | * | |
1234 | * dev->stop() will invoke napi_disable() on all of it's | |
1235 | * napi_struct instances on this device. | |
1236 | */ | |
1da177e4 | 1237 | smp_mb__after_clear_bit(); /* Commit netif_running(). */ |
1da177e4 | 1238 | |
d8b2a4d2 ML |
1239 | dev_deactivate(dev); |
1240 | ||
1da177e4 LT |
1241 | /* |
1242 | * Call the device specific close. This cannot fail. | |
1243 | * Only if device is UP | |
1244 | * | |
1245 | * We allow it to be called even after a DETACH hot-plug | |
1246 | * event. | |
1247 | */ | |
d314774c SH |
1248 | if (ops->ndo_stop) |
1249 | ops->ndo_stop(dev); | |
1da177e4 LT |
1250 | |
1251 | /* | |
1252 | * Device is now down. | |
1253 | */ | |
1254 | ||
1255 | dev->flags &= ~IFF_UP; | |
1256 | ||
1257 | /* | |
bd380811 | 1258 | * Shutdown NET_DMA |
1da177e4 | 1259 | */ |
bd380811 PM |
1260 | net_dmaengine_put(); |
1261 | ||
1262 | return 0; | |
1263 | } | |
1264 | ||
1265 | /** | |
1266 | * dev_close - shutdown an interface. | |
1267 | * @dev: device to shutdown | |
1268 | * | |
1269 | * This function moves an active device into down state. A | |
1270 | * %NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device | |
1271 | * is then deactivated and finally a %NETDEV_DOWN is sent to the notifier | |
1272 | * chain. | |
1273 | */ | |
1274 | int dev_close(struct net_device *dev) | |
1275 | { | |
1276 | if (!(dev->flags & IFF_UP)) | |
1277 | return 0; | |
1278 | ||
1279 | __dev_close(dev); | |
1da177e4 | 1280 | |
649274d9 | 1281 | /* |
bd380811 | 1282 | * Tell people we are down |
649274d9 | 1283 | */ |
bd380811 PM |
1284 | rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING); |
1285 | call_netdevice_notifiers(NETDEV_DOWN, dev); | |
649274d9 | 1286 | |
1da177e4 LT |
1287 | return 0; |
1288 | } | |
d1b19dff | 1289 | EXPORT_SYMBOL(dev_close); |
1da177e4 LT |
1290 | |
1291 | ||
0187bdfb BH |
1292 | /** |
1293 | * dev_disable_lro - disable Large Receive Offload on a device | |
1294 | * @dev: device | |
1295 | * | |
1296 | * Disable Large Receive Offload (LRO) on a net device. Must be | |
1297 | * called under RTNL. This is needed if received packets may be | |
1298 | * forwarded to another interface. | |
1299 | */ | |
1300 | void dev_disable_lro(struct net_device *dev) | |
1301 | { | |
1302 | if (dev->ethtool_ops && dev->ethtool_ops->get_flags && | |
1303 | dev->ethtool_ops->set_flags) { | |
1304 | u32 flags = dev->ethtool_ops->get_flags(dev); | |
1305 | if (flags & ETH_FLAG_LRO) { | |
1306 | flags &= ~ETH_FLAG_LRO; | |
1307 | dev->ethtool_ops->set_flags(dev, flags); | |
1308 | } | |
1309 | } | |
1310 | WARN_ON(dev->features & NETIF_F_LRO); | |
1311 | } | |
1312 | EXPORT_SYMBOL(dev_disable_lro); | |
1313 | ||
1314 | ||
881d966b EB |
1315 | static int dev_boot_phase = 1; |
1316 | ||
1da177e4 LT |
1317 | /* |
1318 | * Device change register/unregister. These are not inline or static | |
1319 | * as we export them to the world. | |
1320 | */ | |
1321 | ||
1322 | /** | |
1323 | * register_netdevice_notifier - register a network notifier block | |
1324 | * @nb: notifier | |
1325 | * | |
1326 | * Register a notifier to be called when network device events occur. | |
1327 | * The notifier passed is linked into the kernel structures and must | |
1328 | * not be reused until it has been unregistered. A negative errno code | |
1329 | * is returned on a failure. | |
1330 | * | |
1331 | * When registered all registration and up events are replayed | |
4ec93edb | 1332 | * to the new notifier to allow device to have a race free |
1da177e4 LT |
1333 | * view of the network device list. |
1334 | */ | |
1335 | ||
1336 | int register_netdevice_notifier(struct notifier_block *nb) | |
1337 | { | |
1338 | struct net_device *dev; | |
fcc5a03a | 1339 | struct net_device *last; |
881d966b | 1340 | struct net *net; |
1da177e4 LT |
1341 | int err; |
1342 | ||
1343 | rtnl_lock(); | |
f07d5b94 | 1344 | err = raw_notifier_chain_register(&netdev_chain, nb); |
fcc5a03a HX |
1345 | if (err) |
1346 | goto unlock; | |
881d966b EB |
1347 | if (dev_boot_phase) |
1348 | goto unlock; | |
1349 | for_each_net(net) { | |
1350 | for_each_netdev(net, dev) { | |
1351 | err = nb->notifier_call(nb, NETDEV_REGISTER, dev); | |
1352 | err = notifier_to_errno(err); | |
1353 | if (err) | |
1354 | goto rollback; | |
1355 | ||
1356 | if (!(dev->flags & IFF_UP)) | |
1357 | continue; | |
1da177e4 | 1358 | |
881d966b EB |
1359 | nb->notifier_call(nb, NETDEV_UP, dev); |
1360 | } | |
1da177e4 | 1361 | } |
fcc5a03a HX |
1362 | |
1363 | unlock: | |
1da177e4 LT |
1364 | rtnl_unlock(); |
1365 | return err; | |
fcc5a03a HX |
1366 | |
1367 | rollback: | |
1368 | last = dev; | |
881d966b EB |
1369 | for_each_net(net) { |
1370 | for_each_netdev(net, dev) { | |
1371 | if (dev == last) | |
1372 | break; | |
fcc5a03a | 1373 | |
881d966b EB |
1374 | if (dev->flags & IFF_UP) { |
1375 | nb->notifier_call(nb, NETDEV_GOING_DOWN, dev); | |
1376 | nb->notifier_call(nb, NETDEV_DOWN, dev); | |
1377 | } | |
1378 | nb->notifier_call(nb, NETDEV_UNREGISTER, dev); | |
a5ee1551 | 1379 | nb->notifier_call(nb, NETDEV_UNREGISTER_BATCH, dev); |
fcc5a03a | 1380 | } |
fcc5a03a | 1381 | } |
c67625a1 PE |
1382 | |
1383 | raw_notifier_chain_unregister(&netdev_chain, nb); | |
fcc5a03a | 1384 | goto unlock; |
1da177e4 | 1385 | } |
d1b19dff | 1386 | EXPORT_SYMBOL(register_netdevice_notifier); |
1da177e4 LT |
1387 | |
1388 | /** | |
1389 | * unregister_netdevice_notifier - unregister a network notifier block | |
1390 | * @nb: notifier | |
1391 | * | |
1392 | * Unregister a notifier previously registered by | |
1393 | * register_netdevice_notifier(). The notifier is unlinked into the | |
1394 | * kernel structures and may then be reused. A negative errno code | |
1395 | * is returned on a failure. | |
1396 | */ | |
1397 | ||
1398 | int unregister_netdevice_notifier(struct notifier_block *nb) | |
1399 | { | |
9f514950 HX |
1400 | int err; |
1401 | ||
1402 | rtnl_lock(); | |
f07d5b94 | 1403 | err = raw_notifier_chain_unregister(&netdev_chain, nb); |
9f514950 HX |
1404 | rtnl_unlock(); |
1405 | return err; | |
1da177e4 | 1406 | } |
d1b19dff | 1407 | EXPORT_SYMBOL(unregister_netdevice_notifier); |
1da177e4 LT |
1408 | |
1409 | /** | |
1410 | * call_netdevice_notifiers - call all network notifier blocks | |
1411 | * @val: value passed unmodified to notifier function | |
c4ea43c5 | 1412 | * @dev: net_device pointer passed unmodified to notifier function |
1da177e4 LT |
1413 | * |
1414 | * Call all network notifier blocks. Parameters and return value | |
f07d5b94 | 1415 | * are as for raw_notifier_call_chain(). |
1da177e4 LT |
1416 | */ |
1417 | ||
ad7379d4 | 1418 | int call_netdevice_notifiers(unsigned long val, struct net_device *dev) |
1da177e4 | 1419 | { |
ad7379d4 | 1420 | return raw_notifier_call_chain(&netdev_chain, val, dev); |
1da177e4 LT |
1421 | } |
1422 | ||
1423 | /* When > 0 there are consumers of rx skb time stamps */ | |
1424 | static atomic_t netstamp_needed = ATOMIC_INIT(0); | |
1425 | ||
1426 | void net_enable_timestamp(void) | |
1427 | { | |
1428 | atomic_inc(&netstamp_needed); | |
1429 | } | |
d1b19dff | 1430 | EXPORT_SYMBOL(net_enable_timestamp); |
1da177e4 LT |
1431 | |
1432 | void net_disable_timestamp(void) | |
1433 | { | |
1434 | atomic_dec(&netstamp_needed); | |
1435 | } | |
d1b19dff | 1436 | EXPORT_SYMBOL(net_disable_timestamp); |
1da177e4 | 1437 | |
a61bbcf2 | 1438 | static inline void net_timestamp(struct sk_buff *skb) |
1da177e4 LT |
1439 | { |
1440 | if (atomic_read(&netstamp_needed)) | |
a61bbcf2 | 1441 | __net_timestamp(skb); |
b7aa0bf7 ED |
1442 | else |
1443 | skb->tstamp.tv64 = 0; | |
1da177e4 LT |
1444 | } |
1445 | ||
44540960 AB |
1446 | /** |
1447 | * dev_forward_skb - loopback an skb to another netif | |
1448 | * | |
1449 | * @dev: destination network device | |
1450 | * @skb: buffer to forward | |
1451 | * | |
1452 | * return values: | |
1453 | * NET_RX_SUCCESS (no congestion) | |
6ec82562 | 1454 | * NET_RX_DROP (packet was dropped, but freed) |
44540960 AB |
1455 | * |
1456 | * dev_forward_skb can be used for injecting an skb from the | |
1457 | * start_xmit function of one device into the receive queue | |
1458 | * of another device. | |
1459 | * | |
1460 | * The receiving device may be in another namespace, so | |
1461 | * we have to clear all information in the skb that could | |
1462 | * impact namespace isolation. | |
1463 | */ | |
1464 | int dev_forward_skb(struct net_device *dev, struct sk_buff *skb) | |
1465 | { | |
1466 | skb_orphan(skb); | |
1467 | ||
6ec82562 ED |
1468 | if (!(dev->flags & IFF_UP) || |
1469 | (skb->len > (dev->mtu + dev->hard_header_len))) { | |
1470 | kfree_skb(skb); | |
44540960 | 1471 | return NET_RX_DROP; |
6ec82562 | 1472 | } |
8a83a00b | 1473 | skb_set_dev(skb, dev); |
44540960 AB |
1474 | skb->tstamp.tv64 = 0; |
1475 | skb->pkt_type = PACKET_HOST; | |
1476 | skb->protocol = eth_type_trans(skb, dev); | |
44540960 AB |
1477 | return netif_rx(skb); |
1478 | } | |
1479 | EXPORT_SYMBOL_GPL(dev_forward_skb); | |
1480 | ||
1da177e4 LT |
1481 | /* |
1482 | * Support routine. Sends outgoing frames to any network | |
1483 | * taps currently in use. | |
1484 | */ | |
1485 | ||
f6a78bfc | 1486 | static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
1487 | { |
1488 | struct packet_type *ptype; | |
a61bbcf2 | 1489 | |
8caf1539 JP |
1490 | #ifdef CONFIG_NET_CLS_ACT |
1491 | if (!(skb->tstamp.tv64 && (G_TC_FROM(skb->tc_verd) & AT_INGRESS))) | |
1492 | net_timestamp(skb); | |
1493 | #else | |
a61bbcf2 | 1494 | net_timestamp(skb); |
8caf1539 | 1495 | #endif |
1da177e4 LT |
1496 | |
1497 | rcu_read_lock(); | |
1498 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
1499 | /* Never send packets back to the socket | |
1500 | * they originated from - MvS (miquels@drinkel.ow.org) | |
1501 | */ | |
1502 | if ((ptype->dev == dev || !ptype->dev) && | |
1503 | (ptype->af_packet_priv == NULL || | |
1504 | (struct sock *)ptype->af_packet_priv != skb->sk)) { | |
d1b19dff | 1505 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); |
1da177e4 LT |
1506 | if (!skb2) |
1507 | break; | |
1508 | ||
1509 | /* skb->nh should be correctly | |
1510 | set by sender, so that the second statement is | |
1511 | just protection against buggy protocols. | |
1512 | */ | |
459a98ed | 1513 | skb_reset_mac_header(skb2); |
1da177e4 | 1514 | |
d56f90a7 | 1515 | if (skb_network_header(skb2) < skb2->data || |
27a884dc | 1516 | skb2->network_header > skb2->tail) { |
1da177e4 LT |
1517 | if (net_ratelimit()) |
1518 | printk(KERN_CRIT "protocol %04x is " | |
1519 | "buggy, dev %s\n", | |
1520 | skb2->protocol, dev->name); | |
c1d2bbe1 | 1521 | skb_reset_network_header(skb2); |
1da177e4 LT |
1522 | } |
1523 | ||
b0e380b1 | 1524 | skb2->transport_header = skb2->network_header; |
1da177e4 | 1525 | skb2->pkt_type = PACKET_OUTGOING; |
f2ccd8fa | 1526 | ptype->func(skb2, skb->dev, ptype, skb->dev); |
1da177e4 LT |
1527 | } |
1528 | } | |
1529 | rcu_read_unlock(); | |
1530 | } | |
1531 | ||
56079431 | 1532 | |
def82a1d | 1533 | static inline void __netif_reschedule(struct Qdisc *q) |
56079431 | 1534 | { |
def82a1d JP |
1535 | struct softnet_data *sd; |
1536 | unsigned long flags; | |
56079431 | 1537 | |
def82a1d JP |
1538 | local_irq_save(flags); |
1539 | sd = &__get_cpu_var(softnet_data); | |
1540 | q->next_sched = sd->output_queue; | |
1541 | sd->output_queue = q; | |
1542 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
1543 | local_irq_restore(flags); | |
1544 | } | |
1545 | ||
1546 | void __netif_schedule(struct Qdisc *q) | |
1547 | { | |
1548 | if (!test_and_set_bit(__QDISC_STATE_SCHED, &q->state)) | |
1549 | __netif_reschedule(q); | |
56079431 DV |
1550 | } |
1551 | EXPORT_SYMBOL(__netif_schedule); | |
1552 | ||
bea3348e | 1553 | void dev_kfree_skb_irq(struct sk_buff *skb) |
56079431 | 1554 | { |
bea3348e SH |
1555 | if (atomic_dec_and_test(&skb->users)) { |
1556 | struct softnet_data *sd; | |
1557 | unsigned long flags; | |
56079431 | 1558 | |
bea3348e SH |
1559 | local_irq_save(flags); |
1560 | sd = &__get_cpu_var(softnet_data); | |
1561 | skb->next = sd->completion_queue; | |
1562 | sd->completion_queue = skb; | |
1563 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
1564 | local_irq_restore(flags); | |
1565 | } | |
56079431 | 1566 | } |
bea3348e | 1567 | EXPORT_SYMBOL(dev_kfree_skb_irq); |
56079431 DV |
1568 | |
1569 | void dev_kfree_skb_any(struct sk_buff *skb) | |
1570 | { | |
1571 | if (in_irq() || irqs_disabled()) | |
1572 | dev_kfree_skb_irq(skb); | |
1573 | else | |
1574 | dev_kfree_skb(skb); | |
1575 | } | |
1576 | EXPORT_SYMBOL(dev_kfree_skb_any); | |
1577 | ||
1578 | ||
bea3348e SH |
1579 | /** |
1580 | * netif_device_detach - mark device as removed | |
1581 | * @dev: network device | |
1582 | * | |
1583 | * Mark device as removed from system and therefore no longer available. | |
1584 | */ | |
56079431 DV |
1585 | void netif_device_detach(struct net_device *dev) |
1586 | { | |
1587 | if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) && | |
1588 | netif_running(dev)) { | |
d543103a | 1589 | netif_tx_stop_all_queues(dev); |
56079431 DV |
1590 | } |
1591 | } | |
1592 | EXPORT_SYMBOL(netif_device_detach); | |
1593 | ||
bea3348e SH |
1594 | /** |
1595 | * netif_device_attach - mark device as attached | |
1596 | * @dev: network device | |
1597 | * | |
1598 | * Mark device as attached from system and restart if needed. | |
1599 | */ | |
56079431 DV |
1600 | void netif_device_attach(struct net_device *dev) |
1601 | { | |
1602 | if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) && | |
1603 | netif_running(dev)) { | |
d543103a | 1604 | netif_tx_wake_all_queues(dev); |
4ec93edb | 1605 | __netdev_watchdog_up(dev); |
56079431 DV |
1606 | } |
1607 | } | |
1608 | EXPORT_SYMBOL(netif_device_attach); | |
1609 | ||
6de329e2 BH |
1610 | static bool can_checksum_protocol(unsigned long features, __be16 protocol) |
1611 | { | |
1612 | return ((features & NETIF_F_GEN_CSUM) || | |
1613 | ((features & NETIF_F_IP_CSUM) && | |
1614 | protocol == htons(ETH_P_IP)) || | |
1615 | ((features & NETIF_F_IPV6_CSUM) && | |
1c8dbcf6 YZ |
1616 | protocol == htons(ETH_P_IPV6)) || |
1617 | ((features & NETIF_F_FCOE_CRC) && | |
1618 | protocol == htons(ETH_P_FCOE))); | |
6de329e2 BH |
1619 | } |
1620 | ||
1621 | static bool dev_can_checksum(struct net_device *dev, struct sk_buff *skb) | |
1622 | { | |
1623 | if (can_checksum_protocol(dev->features, skb->protocol)) | |
1624 | return true; | |
1625 | ||
1626 | if (skb->protocol == htons(ETH_P_8021Q)) { | |
1627 | struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data; | |
1628 | if (can_checksum_protocol(dev->features & dev->vlan_features, | |
1629 | veh->h_vlan_encapsulated_proto)) | |
1630 | return true; | |
1631 | } | |
1632 | ||
1633 | return false; | |
1634 | } | |
56079431 | 1635 | |
8a83a00b AB |
1636 | /** |
1637 | * skb_dev_set -- assign a new device to a buffer | |
1638 | * @skb: buffer for the new device | |
1639 | * @dev: network device | |
1640 | * | |
1641 | * If an skb is owned by a device already, we have to reset | |
1642 | * all data private to the namespace a device belongs to | |
1643 | * before assigning it a new device. | |
1644 | */ | |
1645 | #ifdef CONFIG_NET_NS | |
1646 | void skb_set_dev(struct sk_buff *skb, struct net_device *dev) | |
1647 | { | |
1648 | skb_dst_drop(skb); | |
1649 | if (skb->dev && !net_eq(dev_net(skb->dev), dev_net(dev))) { | |
1650 | secpath_reset(skb); | |
1651 | nf_reset(skb); | |
1652 | skb_init_secmark(skb); | |
1653 | skb->mark = 0; | |
1654 | skb->priority = 0; | |
1655 | skb->nf_trace = 0; | |
1656 | skb->ipvs_property = 0; | |
1657 | #ifdef CONFIG_NET_SCHED | |
1658 | skb->tc_index = 0; | |
1659 | #endif | |
1660 | } | |
1661 | skb->dev = dev; | |
1662 | } | |
1663 | EXPORT_SYMBOL(skb_set_dev); | |
1664 | #endif /* CONFIG_NET_NS */ | |
1665 | ||
1da177e4 LT |
1666 | /* |
1667 | * Invalidate hardware checksum when packet is to be mangled, and | |
1668 | * complete checksum manually on outgoing path. | |
1669 | */ | |
84fa7933 | 1670 | int skb_checksum_help(struct sk_buff *skb) |
1da177e4 | 1671 | { |
d3bc23e7 | 1672 | __wsum csum; |
663ead3b | 1673 | int ret = 0, offset; |
1da177e4 | 1674 | |
84fa7933 | 1675 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
a430a43d HX |
1676 | goto out_set_summed; |
1677 | ||
1678 | if (unlikely(skb_shinfo(skb)->gso_size)) { | |
a430a43d HX |
1679 | /* Let GSO fix up the checksum. */ |
1680 | goto out_set_summed; | |
1da177e4 LT |
1681 | } |
1682 | ||
a030847e HX |
1683 | offset = skb->csum_start - skb_headroom(skb); |
1684 | BUG_ON(offset >= skb_headlen(skb)); | |
1685 | csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
1686 | ||
1687 | offset += skb->csum_offset; | |
1688 | BUG_ON(offset + sizeof(__sum16) > skb_headlen(skb)); | |
1689 | ||
1690 | if (skb_cloned(skb) && | |
1691 | !skb_clone_writable(skb, offset + sizeof(__sum16))) { | |
1da177e4 LT |
1692 | ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC); |
1693 | if (ret) | |
1694 | goto out; | |
1695 | } | |
1696 | ||
a030847e | 1697 | *(__sum16 *)(skb->data + offset) = csum_fold(csum); |
a430a43d | 1698 | out_set_summed: |
1da177e4 | 1699 | skb->ip_summed = CHECKSUM_NONE; |
4ec93edb | 1700 | out: |
1da177e4 LT |
1701 | return ret; |
1702 | } | |
d1b19dff | 1703 | EXPORT_SYMBOL(skb_checksum_help); |
1da177e4 | 1704 | |
f6a78bfc HX |
1705 | /** |
1706 | * skb_gso_segment - Perform segmentation on skb. | |
1707 | * @skb: buffer to segment | |
576a30eb | 1708 | * @features: features for the output path (see dev->features) |
f6a78bfc HX |
1709 | * |
1710 | * This function segments the given skb and returns a list of segments. | |
576a30eb HX |
1711 | * |
1712 | * It may return NULL if the skb requires no segmentation. This is | |
1713 | * only possible when GSO is used for verifying header integrity. | |
f6a78bfc | 1714 | */ |
576a30eb | 1715 | struct sk_buff *skb_gso_segment(struct sk_buff *skb, int features) |
f6a78bfc HX |
1716 | { |
1717 | struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT); | |
1718 | struct packet_type *ptype; | |
252e3346 | 1719 | __be16 type = skb->protocol; |
a430a43d | 1720 | int err; |
f6a78bfc | 1721 | |
459a98ed | 1722 | skb_reset_mac_header(skb); |
b0e380b1 | 1723 | skb->mac_len = skb->network_header - skb->mac_header; |
f6a78bfc HX |
1724 | __skb_pull(skb, skb->mac_len); |
1725 | ||
67fd1a73 HX |
1726 | if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { |
1727 | struct net_device *dev = skb->dev; | |
1728 | struct ethtool_drvinfo info = {}; | |
1729 | ||
1730 | if (dev && dev->ethtool_ops && dev->ethtool_ops->get_drvinfo) | |
1731 | dev->ethtool_ops->get_drvinfo(dev, &info); | |
1732 | ||
1733 | WARN(1, "%s: caps=(0x%lx, 0x%lx) len=%d data_len=%d " | |
1734 | "ip_summed=%d", | |
1735 | info.driver, dev ? dev->features : 0L, | |
1736 | skb->sk ? skb->sk->sk_route_caps : 0L, | |
1737 | skb->len, skb->data_len, skb->ip_summed); | |
1738 | ||
a430a43d HX |
1739 | if (skb_header_cloned(skb) && |
1740 | (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) | |
1741 | return ERR_PTR(err); | |
1742 | } | |
1743 | ||
f6a78bfc | 1744 | rcu_read_lock(); |
82d8a867 PE |
1745 | list_for_each_entry_rcu(ptype, |
1746 | &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { | |
f6a78bfc | 1747 | if (ptype->type == type && !ptype->dev && ptype->gso_segment) { |
84fa7933 | 1748 | if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { |
a430a43d HX |
1749 | err = ptype->gso_send_check(skb); |
1750 | segs = ERR_PTR(err); | |
1751 | if (err || skb_gso_ok(skb, features)) | |
1752 | break; | |
d56f90a7 ACM |
1753 | __skb_push(skb, (skb->data - |
1754 | skb_network_header(skb))); | |
a430a43d | 1755 | } |
576a30eb | 1756 | segs = ptype->gso_segment(skb, features); |
f6a78bfc HX |
1757 | break; |
1758 | } | |
1759 | } | |
1760 | rcu_read_unlock(); | |
1761 | ||
98e399f8 | 1762 | __skb_push(skb, skb->data - skb_mac_header(skb)); |
576a30eb | 1763 | |
f6a78bfc HX |
1764 | return segs; |
1765 | } | |
f6a78bfc HX |
1766 | EXPORT_SYMBOL(skb_gso_segment); |
1767 | ||
fb286bb2 HX |
1768 | /* Take action when hardware reception checksum errors are detected. */ |
1769 | #ifdef CONFIG_BUG | |
1770 | void netdev_rx_csum_fault(struct net_device *dev) | |
1771 | { | |
1772 | if (net_ratelimit()) { | |
4ec93edb | 1773 | printk(KERN_ERR "%s: hw csum failure.\n", |
246a4212 | 1774 | dev ? dev->name : "<unknown>"); |
fb286bb2 HX |
1775 | dump_stack(); |
1776 | } | |
1777 | } | |
1778 | EXPORT_SYMBOL(netdev_rx_csum_fault); | |
1779 | #endif | |
1780 | ||
1da177e4 LT |
1781 | /* Actually, we should eliminate this check as soon as we know, that: |
1782 | * 1. IOMMU is present and allows to map all the memory. | |
1783 | * 2. No high memory really exists on this machine. | |
1784 | */ | |
1785 | ||
1786 | static inline int illegal_highdma(struct net_device *dev, struct sk_buff *skb) | |
1787 | { | |
3d3a8533 | 1788 | #ifdef CONFIG_HIGHMEM |
1da177e4 LT |
1789 | int i; |
1790 | ||
1791 | if (dev->features & NETIF_F_HIGHDMA) | |
1792 | return 0; | |
1793 | ||
1794 | for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) | |
1795 | if (PageHighMem(skb_shinfo(skb)->frags[i].page)) | |
1796 | return 1; | |
1797 | ||
3d3a8533 | 1798 | #endif |
1da177e4 LT |
1799 | return 0; |
1800 | } | |
1da177e4 | 1801 | |
f6a78bfc HX |
1802 | struct dev_gso_cb { |
1803 | void (*destructor)(struct sk_buff *skb); | |
1804 | }; | |
1805 | ||
1806 | #define DEV_GSO_CB(skb) ((struct dev_gso_cb *)(skb)->cb) | |
1807 | ||
1808 | static void dev_gso_skb_destructor(struct sk_buff *skb) | |
1809 | { | |
1810 | struct dev_gso_cb *cb; | |
1811 | ||
1812 | do { | |
1813 | struct sk_buff *nskb = skb->next; | |
1814 | ||
1815 | skb->next = nskb->next; | |
1816 | nskb->next = NULL; | |
1817 | kfree_skb(nskb); | |
1818 | } while (skb->next); | |
1819 | ||
1820 | cb = DEV_GSO_CB(skb); | |
1821 | if (cb->destructor) | |
1822 | cb->destructor(skb); | |
1823 | } | |
1824 | ||
1825 | /** | |
1826 | * dev_gso_segment - Perform emulated hardware segmentation on skb. | |
1827 | * @skb: buffer to segment | |
1828 | * | |
1829 | * This function segments the given skb and stores the list of segments | |
1830 | * in skb->next. | |
1831 | */ | |
1832 | static int dev_gso_segment(struct sk_buff *skb) | |
1833 | { | |
1834 | struct net_device *dev = skb->dev; | |
1835 | struct sk_buff *segs; | |
576a30eb HX |
1836 | int features = dev->features & ~(illegal_highdma(dev, skb) ? |
1837 | NETIF_F_SG : 0); | |
1838 | ||
1839 | segs = skb_gso_segment(skb, features); | |
1840 | ||
1841 | /* Verifying header integrity only. */ | |
1842 | if (!segs) | |
1843 | return 0; | |
f6a78bfc | 1844 | |
801678c5 | 1845 | if (IS_ERR(segs)) |
f6a78bfc HX |
1846 | return PTR_ERR(segs); |
1847 | ||
1848 | skb->next = segs; | |
1849 | DEV_GSO_CB(skb)->destructor = skb->destructor; | |
1850 | skb->destructor = dev_gso_skb_destructor; | |
1851 | ||
1852 | return 0; | |
1853 | } | |
1854 | ||
fd2ea0a7 DM |
1855 | int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, |
1856 | struct netdev_queue *txq) | |
f6a78bfc | 1857 | { |
00829823 | 1858 | const struct net_device_ops *ops = dev->netdev_ops; |
572a9d7b | 1859 | int rc = NETDEV_TX_OK; |
00829823 | 1860 | |
f6a78bfc | 1861 | if (likely(!skb->next)) { |
9be9a6b9 | 1862 | if (!list_empty(&ptype_all)) |
f6a78bfc HX |
1863 | dev_queue_xmit_nit(skb, dev); |
1864 | ||
576a30eb HX |
1865 | if (netif_needs_gso(dev, skb)) { |
1866 | if (unlikely(dev_gso_segment(skb))) | |
1867 | goto out_kfree_skb; | |
1868 | if (skb->next) | |
1869 | goto gso; | |
1870 | } | |
f6a78bfc | 1871 | |
93f154b5 ED |
1872 | /* |
1873 | * If device doesnt need skb->dst, release it right now while | |
1874 | * its hot in this cpu cache | |
1875 | */ | |
adf30907 ED |
1876 | if (dev->priv_flags & IFF_XMIT_DST_RELEASE) |
1877 | skb_dst_drop(skb); | |
1878 | ||
ac45f602 | 1879 | rc = ops->ndo_start_xmit(skb, dev); |
ec634fe3 | 1880 | if (rc == NETDEV_TX_OK) |
08baf561 | 1881 | txq_trans_update(txq); |
ac45f602 PO |
1882 | /* |
1883 | * TODO: if skb_orphan() was called by | |
1884 | * dev->hard_start_xmit() (for example, the unmodified | |
1885 | * igb driver does that; bnx2 doesn't), then | |
1886 | * skb_tx_software_timestamp() will be unable to send | |
1887 | * back the time stamp. | |
1888 | * | |
1889 | * How can this be prevented? Always create another | |
1890 | * reference to the socket before calling | |
1891 | * dev->hard_start_xmit()? Prevent that skb_orphan() | |
1892 | * does anything in dev->hard_start_xmit() by clearing | |
1893 | * the skb destructor before the call and restoring it | |
1894 | * afterwards, then doing the skb_orphan() ourselves? | |
1895 | */ | |
ac45f602 | 1896 | return rc; |
f6a78bfc HX |
1897 | } |
1898 | ||
576a30eb | 1899 | gso: |
f6a78bfc HX |
1900 | do { |
1901 | struct sk_buff *nskb = skb->next; | |
f6a78bfc HX |
1902 | |
1903 | skb->next = nskb->next; | |
1904 | nskb->next = NULL; | |
068a2de5 KK |
1905 | |
1906 | /* | |
1907 | * If device doesnt need nskb->dst, release it right now while | |
1908 | * its hot in this cpu cache | |
1909 | */ | |
1910 | if (dev->priv_flags & IFF_XMIT_DST_RELEASE) | |
1911 | skb_dst_drop(nskb); | |
1912 | ||
00829823 | 1913 | rc = ops->ndo_start_xmit(nskb, dev); |
ec634fe3 | 1914 | if (unlikely(rc != NETDEV_TX_OK)) { |
572a9d7b PM |
1915 | if (rc & ~NETDEV_TX_MASK) |
1916 | goto out_kfree_gso_skb; | |
f54d9e8d | 1917 | nskb->next = skb->next; |
f6a78bfc HX |
1918 | skb->next = nskb; |
1919 | return rc; | |
1920 | } | |
08baf561 | 1921 | txq_trans_update(txq); |
fd2ea0a7 | 1922 | if (unlikely(netif_tx_queue_stopped(txq) && skb->next)) |
f54d9e8d | 1923 | return NETDEV_TX_BUSY; |
f6a78bfc | 1924 | } while (skb->next); |
4ec93edb | 1925 | |
572a9d7b PM |
1926 | out_kfree_gso_skb: |
1927 | if (likely(skb->next == NULL)) | |
1928 | skb->destructor = DEV_GSO_CB(skb)->destructor; | |
f6a78bfc HX |
1929 | out_kfree_skb: |
1930 | kfree_skb(skb); | |
572a9d7b | 1931 | return rc; |
f6a78bfc HX |
1932 | } |
1933 | ||
7019298a | 1934 | static u32 skb_tx_hashrnd; |
b6b2fed1 | 1935 | |
9247744e | 1936 | u16 skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb) |
8f0f2223 | 1937 | { |
7019298a | 1938 | u32 hash; |
b6b2fed1 | 1939 | |
513de11b DM |
1940 | if (skb_rx_queue_recorded(skb)) { |
1941 | hash = skb_get_rx_queue(skb); | |
d1b19dff | 1942 | while (unlikely(hash >= dev->real_num_tx_queues)) |
513de11b DM |
1943 | hash -= dev->real_num_tx_queues; |
1944 | return hash; | |
1945 | } | |
ec581f6a ED |
1946 | |
1947 | if (skb->sk && skb->sk->sk_hash) | |
7019298a | 1948 | hash = skb->sk->sk_hash; |
ec581f6a | 1949 | else |
7019298a | 1950 | hash = skb->protocol; |
d5a9e24a | 1951 | |
7019298a | 1952 | hash = jhash_1word(hash, skb_tx_hashrnd); |
b6b2fed1 DM |
1953 | |
1954 | return (u16) (((u64) hash * dev->real_num_tx_queues) >> 32); | |
8f0f2223 | 1955 | } |
9247744e | 1956 | EXPORT_SYMBOL(skb_tx_hash); |
8f0f2223 | 1957 | |
ed04642f ED |
1958 | static inline u16 dev_cap_txqueue(struct net_device *dev, u16 queue_index) |
1959 | { | |
1960 | if (unlikely(queue_index >= dev->real_num_tx_queues)) { | |
1961 | if (net_ratelimit()) { | |
1962 | WARN(1, "%s selects TX queue %d, but " | |
1963 | "real number of TX queues is %d\n", | |
1964 | dev->name, queue_index, | |
1965 | dev->real_num_tx_queues); | |
1966 | } | |
1967 | return 0; | |
1968 | } | |
1969 | return queue_index; | |
1970 | } | |
1971 | ||
e8a0464c DM |
1972 | static struct netdev_queue *dev_pick_tx(struct net_device *dev, |
1973 | struct sk_buff *skb) | |
1974 | { | |
a4ee3ce3 KK |
1975 | u16 queue_index; |
1976 | struct sock *sk = skb->sk; | |
1977 | ||
1978 | if (sk_tx_queue_recorded(sk)) { | |
1979 | queue_index = sk_tx_queue_get(sk); | |
1980 | } else { | |
1981 | const struct net_device_ops *ops = dev->netdev_ops; | |
1982 | ||
1983 | if (ops->ndo_select_queue) { | |
1984 | queue_index = ops->ndo_select_queue(dev, skb); | |
ed04642f | 1985 | queue_index = dev_cap_txqueue(dev, queue_index); |
a4ee3ce3 KK |
1986 | } else { |
1987 | queue_index = 0; | |
1988 | if (dev->real_num_tx_queues > 1) | |
1989 | queue_index = skb_tx_hash(dev, skb); | |
fd2ea0a7 | 1990 | |
8728c544 | 1991 | if (sk) { |
05d17608 | 1992 | struct dst_entry *dst = rcu_dereference_bh(sk->sk_dst_cache); |
8728c544 ED |
1993 | |
1994 | if (dst && skb_dst(skb) == dst) | |
1995 | sk_tx_queue_set(sk, queue_index); | |
1996 | } | |
a4ee3ce3 KK |
1997 | } |
1998 | } | |
eae792b7 | 1999 | |
fd2ea0a7 DM |
2000 | skb_set_queue_mapping(skb, queue_index); |
2001 | return netdev_get_tx_queue(dev, queue_index); | |
e8a0464c DM |
2002 | } |
2003 | ||
bbd8a0d3 KK |
2004 | static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q, |
2005 | struct net_device *dev, | |
2006 | struct netdev_queue *txq) | |
2007 | { | |
2008 | spinlock_t *root_lock = qdisc_lock(q); | |
2009 | int rc; | |
2010 | ||
2011 | spin_lock(root_lock); | |
2012 | if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) { | |
2013 | kfree_skb(skb); | |
2014 | rc = NET_XMIT_DROP; | |
2015 | } else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) && | |
2016 | !test_and_set_bit(__QDISC_STATE_RUNNING, &q->state)) { | |
2017 | /* | |
2018 | * This is a work-conserving queue; there are no old skbs | |
2019 | * waiting to be sent out; and the qdisc is not running - | |
2020 | * xmit the skb directly. | |
2021 | */ | |
2022 | __qdisc_update_bstats(q, skb->len); | |
2023 | if (sch_direct_xmit(skb, q, dev, txq, root_lock)) | |
2024 | __qdisc_run(q); | |
2025 | else | |
2026 | clear_bit(__QDISC_STATE_RUNNING, &q->state); | |
2027 | ||
2028 | rc = NET_XMIT_SUCCESS; | |
2029 | } else { | |
2030 | rc = qdisc_enqueue_root(skb, q); | |
2031 | qdisc_run(q); | |
2032 | } | |
2033 | spin_unlock(root_lock); | |
2034 | ||
2035 | return rc; | |
2036 | } | |
2037 | ||
4b258461 KK |
2038 | /* |
2039 | * Returns true if either: | |
2040 | * 1. skb has frag_list and the device doesn't support FRAGLIST, or | |
2041 | * 2. skb is fragmented and the device does not support SG, or if | |
2042 | * at least one of fragments is in highmem and device does not | |
2043 | * support DMA from it. | |
2044 | */ | |
2045 | static inline int skb_needs_linearize(struct sk_buff *skb, | |
2046 | struct net_device *dev) | |
2047 | { | |
2048 | return (skb_has_frags(skb) && !(dev->features & NETIF_F_FRAGLIST)) || | |
2049 | (skb_shinfo(skb)->nr_frags && (!(dev->features & NETIF_F_SG) || | |
2050 | illegal_highdma(dev, skb))); | |
2051 | } | |
2052 | ||
d29f749e DJ |
2053 | /** |
2054 | * dev_queue_xmit - transmit a buffer | |
2055 | * @skb: buffer to transmit | |
2056 | * | |
2057 | * Queue a buffer for transmission to a network device. The caller must | |
2058 | * have set the device and priority and built the buffer before calling | |
2059 | * this function. The function can be called from an interrupt. | |
2060 | * | |
2061 | * A negative errno code is returned on a failure. A success does not | |
2062 | * guarantee the frame will be transmitted as it may be dropped due | |
2063 | * to congestion or traffic shaping. | |
2064 | * | |
2065 | * ----------------------------------------------------------------------------------- | |
2066 | * I notice this method can also return errors from the queue disciplines, | |
2067 | * including NET_XMIT_DROP, which is a positive value. So, errors can also | |
2068 | * be positive. | |
2069 | * | |
2070 | * Regardless of the return value, the skb is consumed, so it is currently | |
2071 | * difficult to retry a send to this method. (You can bump the ref count | |
2072 | * before sending to hold a reference for retry if you are careful.) | |
2073 | * | |
2074 | * When calling this method, interrupts MUST be enabled. This is because | |
2075 | * the BH enable code must have IRQs enabled so that it will not deadlock. | |
2076 | * --BLG | |
2077 | */ | |
1da177e4 LT |
2078 | int dev_queue_xmit(struct sk_buff *skb) |
2079 | { | |
2080 | struct net_device *dev = skb->dev; | |
dc2b4847 | 2081 | struct netdev_queue *txq; |
1da177e4 LT |
2082 | struct Qdisc *q; |
2083 | int rc = -ENOMEM; | |
2084 | ||
f6a78bfc HX |
2085 | /* GSO will handle the following emulations directly. */ |
2086 | if (netif_needs_gso(dev, skb)) | |
2087 | goto gso; | |
2088 | ||
4b258461 KK |
2089 | /* Convert a paged skb to linear, if required */ |
2090 | if (skb_needs_linearize(skb, dev) && __skb_linearize(skb)) | |
1da177e4 LT |
2091 | goto out_kfree_skb; |
2092 | ||
2093 | /* If packet is not checksummed and device does not support | |
2094 | * checksumming for this protocol, complete checksumming here. | |
2095 | */ | |
663ead3b HX |
2096 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
2097 | skb_set_transport_header(skb, skb->csum_start - | |
2098 | skb_headroom(skb)); | |
6de329e2 BH |
2099 | if (!dev_can_checksum(dev, skb) && skb_checksum_help(skb)) |
2100 | goto out_kfree_skb; | |
663ead3b | 2101 | } |
1da177e4 | 2102 | |
f6a78bfc | 2103 | gso: |
4ec93edb YH |
2104 | /* Disable soft irqs for various locks below. Also |
2105 | * stops preemption for RCU. | |
1da177e4 | 2106 | */ |
4ec93edb | 2107 | rcu_read_lock_bh(); |
1da177e4 | 2108 | |
eae792b7 | 2109 | txq = dev_pick_tx(dev, skb); |
a898def2 | 2110 | q = rcu_dereference_bh(txq->qdisc); |
37437bb2 | 2111 | |
1da177e4 | 2112 | #ifdef CONFIG_NET_CLS_ACT |
d1b19dff | 2113 | skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS); |
1da177e4 LT |
2114 | #endif |
2115 | if (q->enqueue) { | |
bbd8a0d3 | 2116 | rc = __dev_xmit_skb(skb, q, dev, txq); |
37437bb2 | 2117 | goto out; |
1da177e4 LT |
2118 | } |
2119 | ||
2120 | /* The device has no queue. Common case for software devices: | |
2121 | loopback, all the sorts of tunnels... | |
2122 | ||
932ff279 HX |
2123 | Really, it is unlikely that netif_tx_lock protection is necessary |
2124 | here. (f.e. loopback and IP tunnels are clean ignoring statistics | |
1da177e4 LT |
2125 | counters.) |
2126 | However, it is possible, that they rely on protection | |
2127 | made by us here. | |
2128 | ||
2129 | Check this and shot the lock. It is not prone from deadlocks. | |
2130 | Either shot noqueue qdisc, it is even simpler 8) | |
2131 | */ | |
2132 | if (dev->flags & IFF_UP) { | |
2133 | int cpu = smp_processor_id(); /* ok because BHs are off */ | |
2134 | ||
c773e847 | 2135 | if (txq->xmit_lock_owner != cpu) { |
1da177e4 | 2136 | |
c773e847 | 2137 | HARD_TX_LOCK(dev, txq, cpu); |
1da177e4 | 2138 | |
fd2ea0a7 | 2139 | if (!netif_tx_queue_stopped(txq)) { |
572a9d7b PM |
2140 | rc = dev_hard_start_xmit(skb, dev, txq); |
2141 | if (dev_xmit_complete(rc)) { | |
c773e847 | 2142 | HARD_TX_UNLOCK(dev, txq); |
1da177e4 LT |
2143 | goto out; |
2144 | } | |
2145 | } | |
c773e847 | 2146 | HARD_TX_UNLOCK(dev, txq); |
1da177e4 LT |
2147 | if (net_ratelimit()) |
2148 | printk(KERN_CRIT "Virtual device %s asks to " | |
2149 | "queue packet!\n", dev->name); | |
2150 | } else { | |
2151 | /* Recursion is detected! It is possible, | |
2152 | * unfortunately */ | |
2153 | if (net_ratelimit()) | |
2154 | printk(KERN_CRIT "Dead loop on virtual device " | |
2155 | "%s, fix it urgently!\n", dev->name); | |
2156 | } | |
2157 | } | |
2158 | ||
2159 | rc = -ENETDOWN; | |
d4828d85 | 2160 | rcu_read_unlock_bh(); |
1da177e4 LT |
2161 | |
2162 | out_kfree_skb: | |
2163 | kfree_skb(skb); | |
2164 | return rc; | |
2165 | out: | |
d4828d85 | 2166 | rcu_read_unlock_bh(); |
1da177e4 LT |
2167 | return rc; |
2168 | } | |
d1b19dff | 2169 | EXPORT_SYMBOL(dev_queue_xmit); |
1da177e4 LT |
2170 | |
2171 | ||
2172 | /*======================================================================= | |
2173 | Receiver routines | |
2174 | =======================================================================*/ | |
2175 | ||
6b2bedc3 SH |
2176 | int netdev_max_backlog __read_mostly = 1000; |
2177 | int netdev_budget __read_mostly = 300; | |
2178 | int weight_p __read_mostly = 64; /* old backlog weight */ | |
1da177e4 LT |
2179 | |
2180 | DEFINE_PER_CPU(struct netif_rx_stats, netdev_rx_stat) = { 0, }; | |
2181 | ||
2182 | ||
1da177e4 LT |
2183 | /** |
2184 | * netif_rx - post buffer to the network code | |
2185 | * @skb: buffer to post | |
2186 | * | |
2187 | * This function receives a packet from a device driver and queues it for | |
2188 | * the upper (protocol) levels to process. It always succeeds. The buffer | |
2189 | * may be dropped during processing for congestion control or by the | |
2190 | * protocol layers. | |
2191 | * | |
2192 | * return values: | |
2193 | * NET_RX_SUCCESS (no congestion) | |
1da177e4 LT |
2194 | * NET_RX_DROP (packet was dropped) |
2195 | * | |
2196 | */ | |
2197 | ||
2198 | int netif_rx(struct sk_buff *skb) | |
2199 | { | |
1da177e4 LT |
2200 | struct softnet_data *queue; |
2201 | unsigned long flags; | |
2202 | ||
2203 | /* if netpoll wants it, pretend we never saw it */ | |
2204 | if (netpoll_rx(skb)) | |
2205 | return NET_RX_DROP; | |
2206 | ||
b7aa0bf7 | 2207 | if (!skb->tstamp.tv64) |
a61bbcf2 | 2208 | net_timestamp(skb); |
1da177e4 LT |
2209 | |
2210 | /* | |
2211 | * The code is rearranged so that the path is the most | |
2212 | * short when CPU is congested, but is still operating. | |
2213 | */ | |
2214 | local_irq_save(flags); | |
1da177e4 LT |
2215 | queue = &__get_cpu_var(softnet_data); |
2216 | ||
2217 | __get_cpu_var(netdev_rx_stat).total++; | |
2218 | if (queue->input_pkt_queue.qlen <= netdev_max_backlog) { | |
2219 | if (queue->input_pkt_queue.qlen) { | |
1da177e4 | 2220 | enqueue: |
1da177e4 | 2221 | __skb_queue_tail(&queue->input_pkt_queue, skb); |
1da177e4 | 2222 | local_irq_restore(flags); |
34008d8c | 2223 | return NET_RX_SUCCESS; |
1da177e4 LT |
2224 | } |
2225 | ||
bea3348e | 2226 | napi_schedule(&queue->backlog); |
1da177e4 LT |
2227 | goto enqueue; |
2228 | } | |
2229 | ||
1da177e4 LT |
2230 | __get_cpu_var(netdev_rx_stat).dropped++; |
2231 | local_irq_restore(flags); | |
2232 | ||
2233 | kfree_skb(skb); | |
2234 | return NET_RX_DROP; | |
2235 | } | |
d1b19dff | 2236 | EXPORT_SYMBOL(netif_rx); |
1da177e4 LT |
2237 | |
2238 | int netif_rx_ni(struct sk_buff *skb) | |
2239 | { | |
2240 | int err; | |
2241 | ||
2242 | preempt_disable(); | |
2243 | err = netif_rx(skb); | |
2244 | if (local_softirq_pending()) | |
2245 | do_softirq(); | |
2246 | preempt_enable(); | |
2247 | ||
2248 | return err; | |
2249 | } | |
1da177e4 LT |
2250 | EXPORT_SYMBOL(netif_rx_ni); |
2251 | ||
1da177e4 LT |
2252 | static void net_tx_action(struct softirq_action *h) |
2253 | { | |
2254 | struct softnet_data *sd = &__get_cpu_var(softnet_data); | |
2255 | ||
2256 | if (sd->completion_queue) { | |
2257 | struct sk_buff *clist; | |
2258 | ||
2259 | local_irq_disable(); | |
2260 | clist = sd->completion_queue; | |
2261 | sd->completion_queue = NULL; | |
2262 | local_irq_enable(); | |
2263 | ||
2264 | while (clist) { | |
2265 | struct sk_buff *skb = clist; | |
2266 | clist = clist->next; | |
2267 | ||
547b792c | 2268 | WARN_ON(atomic_read(&skb->users)); |
1da177e4 LT |
2269 | __kfree_skb(skb); |
2270 | } | |
2271 | } | |
2272 | ||
2273 | if (sd->output_queue) { | |
37437bb2 | 2274 | struct Qdisc *head; |
1da177e4 LT |
2275 | |
2276 | local_irq_disable(); | |
2277 | head = sd->output_queue; | |
2278 | sd->output_queue = NULL; | |
2279 | local_irq_enable(); | |
2280 | ||
2281 | while (head) { | |
37437bb2 DM |
2282 | struct Qdisc *q = head; |
2283 | spinlock_t *root_lock; | |
2284 | ||
1da177e4 LT |
2285 | head = head->next_sched; |
2286 | ||
5fb66229 | 2287 | root_lock = qdisc_lock(q); |
37437bb2 | 2288 | if (spin_trylock(root_lock)) { |
def82a1d JP |
2289 | smp_mb__before_clear_bit(); |
2290 | clear_bit(__QDISC_STATE_SCHED, | |
2291 | &q->state); | |
37437bb2 DM |
2292 | qdisc_run(q); |
2293 | spin_unlock(root_lock); | |
1da177e4 | 2294 | } else { |
195648bb | 2295 | if (!test_bit(__QDISC_STATE_DEACTIVATED, |
e8a83e10 | 2296 | &q->state)) { |
195648bb | 2297 | __netif_reschedule(q); |
e8a83e10 JP |
2298 | } else { |
2299 | smp_mb__before_clear_bit(); | |
2300 | clear_bit(__QDISC_STATE_SCHED, | |
2301 | &q->state); | |
2302 | } | |
1da177e4 LT |
2303 | } |
2304 | } | |
2305 | } | |
2306 | } | |
2307 | ||
6f05f629 SH |
2308 | static inline int deliver_skb(struct sk_buff *skb, |
2309 | struct packet_type *pt_prev, | |
2310 | struct net_device *orig_dev) | |
1da177e4 LT |
2311 | { |
2312 | atomic_inc(&skb->users); | |
f2ccd8fa | 2313 | return pt_prev->func(skb, skb->dev, pt_prev, orig_dev); |
1da177e4 LT |
2314 | } |
2315 | ||
2316 | #if defined(CONFIG_BRIDGE) || defined (CONFIG_BRIDGE_MODULE) | |
da678292 MM |
2317 | |
2318 | #if defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE) | |
2319 | /* This hook is defined here for ATM LANE */ | |
2320 | int (*br_fdb_test_addr_hook)(struct net_device *dev, | |
2321 | unsigned char *addr) __read_mostly; | |
4fb019a0 | 2322 | EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook); |
da678292 | 2323 | #endif |
1da177e4 | 2324 | |
6229e362 SH |
2325 | /* |
2326 | * If bridge module is loaded call bridging hook. | |
2327 | * returns NULL if packet was consumed. | |
2328 | */ | |
2329 | struct sk_buff *(*br_handle_frame_hook)(struct net_bridge_port *p, | |
2330 | struct sk_buff *skb) __read_mostly; | |
4fb019a0 | 2331 | EXPORT_SYMBOL_GPL(br_handle_frame_hook); |
da678292 | 2332 | |
6229e362 SH |
2333 | static inline struct sk_buff *handle_bridge(struct sk_buff *skb, |
2334 | struct packet_type **pt_prev, int *ret, | |
2335 | struct net_device *orig_dev) | |
1da177e4 LT |
2336 | { |
2337 | struct net_bridge_port *port; | |
2338 | ||
6229e362 SH |
2339 | if (skb->pkt_type == PACKET_LOOPBACK || |
2340 | (port = rcu_dereference(skb->dev->br_port)) == NULL) | |
2341 | return skb; | |
1da177e4 LT |
2342 | |
2343 | if (*pt_prev) { | |
6229e362 | 2344 | *ret = deliver_skb(skb, *pt_prev, orig_dev); |
1da177e4 | 2345 | *pt_prev = NULL; |
4ec93edb YH |
2346 | } |
2347 | ||
6229e362 | 2348 | return br_handle_frame_hook(port, skb); |
1da177e4 LT |
2349 | } |
2350 | #else | |
6229e362 | 2351 | #define handle_bridge(skb, pt_prev, ret, orig_dev) (skb) |
1da177e4 LT |
2352 | #endif |
2353 | ||
b863ceb7 PM |
2354 | #if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) |
2355 | struct sk_buff *(*macvlan_handle_frame_hook)(struct sk_buff *skb) __read_mostly; | |
2356 | EXPORT_SYMBOL_GPL(macvlan_handle_frame_hook); | |
2357 | ||
2358 | static inline struct sk_buff *handle_macvlan(struct sk_buff *skb, | |
2359 | struct packet_type **pt_prev, | |
2360 | int *ret, | |
2361 | struct net_device *orig_dev) | |
2362 | { | |
2363 | if (skb->dev->macvlan_port == NULL) | |
2364 | return skb; | |
2365 | ||
2366 | if (*pt_prev) { | |
2367 | *ret = deliver_skb(skb, *pt_prev, orig_dev); | |
2368 | *pt_prev = NULL; | |
2369 | } | |
2370 | return macvlan_handle_frame_hook(skb); | |
2371 | } | |
2372 | #else | |
2373 | #define handle_macvlan(skb, pt_prev, ret, orig_dev) (skb) | |
2374 | #endif | |
2375 | ||
1da177e4 LT |
2376 | #ifdef CONFIG_NET_CLS_ACT |
2377 | /* TODO: Maybe we should just force sch_ingress to be compiled in | |
2378 | * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions | |
2379 | * a compare and 2 stores extra right now if we dont have it on | |
2380 | * but have CONFIG_NET_CLS_ACT | |
4ec93edb | 2381 | * NOTE: This doesnt stop any functionality; if you dont have |
1da177e4 LT |
2382 | * the ingress scheduler, you just cant add policies on ingress. |
2383 | * | |
2384 | */ | |
4ec93edb | 2385 | static int ing_filter(struct sk_buff *skb) |
1da177e4 | 2386 | { |
1da177e4 | 2387 | struct net_device *dev = skb->dev; |
f697c3e8 | 2388 | u32 ttl = G_TC_RTTL(skb->tc_verd); |
555353cf DM |
2389 | struct netdev_queue *rxq; |
2390 | int result = TC_ACT_OK; | |
2391 | struct Qdisc *q; | |
4ec93edb | 2392 | |
f697c3e8 HX |
2393 | if (MAX_RED_LOOP < ttl++) { |
2394 | printk(KERN_WARNING | |
2395 | "Redir loop detected Dropping packet (%d->%d)\n", | |
8964be4a | 2396 | skb->skb_iif, dev->ifindex); |
f697c3e8 HX |
2397 | return TC_ACT_SHOT; |
2398 | } | |
1da177e4 | 2399 | |
f697c3e8 HX |
2400 | skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl); |
2401 | skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS); | |
1da177e4 | 2402 | |
555353cf DM |
2403 | rxq = &dev->rx_queue; |
2404 | ||
83874000 | 2405 | q = rxq->qdisc; |
8d50b53d | 2406 | if (q != &noop_qdisc) { |
83874000 | 2407 | spin_lock(qdisc_lock(q)); |
a9312ae8 DM |
2408 | if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) |
2409 | result = qdisc_enqueue_root(skb, q); | |
83874000 DM |
2410 | spin_unlock(qdisc_lock(q)); |
2411 | } | |
f697c3e8 HX |
2412 | |
2413 | return result; | |
2414 | } | |
86e65da9 | 2415 | |
f697c3e8 HX |
2416 | static inline struct sk_buff *handle_ing(struct sk_buff *skb, |
2417 | struct packet_type **pt_prev, | |
2418 | int *ret, struct net_device *orig_dev) | |
2419 | { | |
8d50b53d | 2420 | if (skb->dev->rx_queue.qdisc == &noop_qdisc) |
f697c3e8 | 2421 | goto out; |
1da177e4 | 2422 | |
f697c3e8 HX |
2423 | if (*pt_prev) { |
2424 | *ret = deliver_skb(skb, *pt_prev, orig_dev); | |
2425 | *pt_prev = NULL; | |
2426 | } else { | |
2427 | /* Huh? Why does turning on AF_PACKET affect this? */ | |
2428 | skb->tc_verd = SET_TC_OK2MUNGE(skb->tc_verd); | |
1da177e4 LT |
2429 | } |
2430 | ||
f697c3e8 HX |
2431 | switch (ing_filter(skb)) { |
2432 | case TC_ACT_SHOT: | |
2433 | case TC_ACT_STOLEN: | |
2434 | kfree_skb(skb); | |
2435 | return NULL; | |
2436 | } | |
2437 | ||
2438 | out: | |
2439 | skb->tc_verd = 0; | |
2440 | return skb; | |
1da177e4 LT |
2441 | } |
2442 | #endif | |
2443 | ||
bc1d0411 PM |
2444 | /* |
2445 | * netif_nit_deliver - deliver received packets to network taps | |
2446 | * @skb: buffer | |
2447 | * | |
2448 | * This function is used to deliver incoming packets to network | |
2449 | * taps. It should be used when the normal netif_receive_skb path | |
2450 | * is bypassed, for example because of VLAN acceleration. | |
2451 | */ | |
2452 | void netif_nit_deliver(struct sk_buff *skb) | |
2453 | { | |
2454 | struct packet_type *ptype; | |
2455 | ||
2456 | if (list_empty(&ptype_all)) | |
2457 | return; | |
2458 | ||
2459 | skb_reset_network_header(skb); | |
2460 | skb_reset_transport_header(skb); | |
2461 | skb->mac_len = skb->network_header - skb->mac_header; | |
2462 | ||
2463 | rcu_read_lock(); | |
2464 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
2465 | if (!ptype->dev || ptype->dev == skb->dev) | |
2466 | deliver_skb(skb, ptype, skb->dev); | |
2467 | } | |
2468 | rcu_read_unlock(); | |
2469 | } | |
2470 | ||
3b582cc1 SH |
2471 | /** |
2472 | * netif_receive_skb - process receive buffer from network | |
2473 | * @skb: buffer to process | |
2474 | * | |
2475 | * netif_receive_skb() is the main receive data processing function. | |
2476 | * It always succeeds. The buffer may be dropped during processing | |
2477 | * for congestion control or by the protocol layers. | |
2478 | * | |
2479 | * This function may only be called from softirq context and interrupts | |
2480 | * should be enabled. | |
2481 | * | |
2482 | * Return values (usually ignored): | |
2483 | * NET_RX_SUCCESS: no congestion | |
2484 | * NET_RX_DROP: packet was dropped | |
2485 | */ | |
1da177e4 LT |
2486 | int netif_receive_skb(struct sk_buff *skb) |
2487 | { | |
2488 | struct packet_type *ptype, *pt_prev; | |
f2ccd8fa | 2489 | struct net_device *orig_dev; |
0641e4fb | 2490 | struct net_device *master; |
0d7a3681 | 2491 | struct net_device *null_or_orig; |
ca8d9ea3 | 2492 | struct net_device *null_or_bond; |
1da177e4 | 2493 | int ret = NET_RX_DROP; |
252e3346 | 2494 | __be16 type; |
1da177e4 | 2495 | |
81bbb3d4 ED |
2496 | if (!skb->tstamp.tv64) |
2497 | net_timestamp(skb); | |
2498 | ||
05423b24 | 2499 | if (vlan_tx_tag_present(skb) && vlan_hwaccel_do_receive(skb)) |
9b22ea56 PM |
2500 | return NET_RX_SUCCESS; |
2501 | ||
1da177e4 | 2502 | /* if we've gotten here through NAPI, check netpoll */ |
bea3348e | 2503 | if (netpoll_receive_skb(skb)) |
1da177e4 LT |
2504 | return NET_RX_DROP; |
2505 | ||
8964be4a ED |
2506 | if (!skb->skb_iif) |
2507 | skb->skb_iif = skb->dev->ifindex; | |
86e65da9 | 2508 | |
0d7a3681 | 2509 | null_or_orig = NULL; |
cc9bd5ce | 2510 | orig_dev = skb->dev; |
0641e4fb ED |
2511 | master = ACCESS_ONCE(orig_dev->master); |
2512 | if (master) { | |
2513 | if (skb_bond_should_drop(skb, master)) | |
0d7a3681 JE |
2514 | null_or_orig = orig_dev; /* deliver only exact match */ |
2515 | else | |
0641e4fb | 2516 | skb->dev = master; |
cc9bd5ce | 2517 | } |
8f903c70 | 2518 | |
1da177e4 LT |
2519 | __get_cpu_var(netdev_rx_stat).total++; |
2520 | ||
c1d2bbe1 | 2521 | skb_reset_network_header(skb); |
badff6d0 | 2522 | skb_reset_transport_header(skb); |
b0e380b1 | 2523 | skb->mac_len = skb->network_header - skb->mac_header; |
1da177e4 LT |
2524 | |
2525 | pt_prev = NULL; | |
2526 | ||
2527 | rcu_read_lock(); | |
2528 | ||
2529 | #ifdef CONFIG_NET_CLS_ACT | |
2530 | if (skb->tc_verd & TC_NCLS) { | |
2531 | skb->tc_verd = CLR_TC_NCLS(skb->tc_verd); | |
2532 | goto ncls; | |
2533 | } | |
2534 | #endif | |
2535 | ||
2536 | list_for_each_entry_rcu(ptype, &ptype_all, list) { | |
f982307f JE |
2537 | if (ptype->dev == null_or_orig || ptype->dev == skb->dev || |
2538 | ptype->dev == orig_dev) { | |
4ec93edb | 2539 | if (pt_prev) |
f2ccd8fa | 2540 | ret = deliver_skb(skb, pt_prev, orig_dev); |
1da177e4 LT |
2541 | pt_prev = ptype; |
2542 | } | |
2543 | } | |
2544 | ||
2545 | #ifdef CONFIG_NET_CLS_ACT | |
f697c3e8 HX |
2546 | skb = handle_ing(skb, &pt_prev, &ret, orig_dev); |
2547 | if (!skb) | |
1da177e4 | 2548 | goto out; |
1da177e4 LT |
2549 | ncls: |
2550 | #endif | |
2551 | ||
6229e362 | 2552 | skb = handle_bridge(skb, &pt_prev, &ret, orig_dev); |
b863ceb7 PM |
2553 | if (!skb) |
2554 | goto out; | |
2555 | skb = handle_macvlan(skb, &pt_prev, &ret, orig_dev); | |
6229e362 | 2556 | if (!skb) |
1da177e4 LT |
2557 | goto out; |
2558 | ||
1f3c8804 AG |
2559 | /* |
2560 | * Make sure frames received on VLAN interfaces stacked on | |
2561 | * bonding interfaces still make their way to any base bonding | |
2562 | * device that may have registered for a specific ptype. The | |
2563 | * handler may have to adjust skb->dev and orig_dev. | |
1f3c8804 | 2564 | */ |
ca8d9ea3 | 2565 | null_or_bond = NULL; |
1f3c8804 AG |
2566 | if ((skb->dev->priv_flags & IFF_802_1Q_VLAN) && |
2567 | (vlan_dev_real_dev(skb->dev)->priv_flags & IFF_BONDING)) { | |
ca8d9ea3 | 2568 | null_or_bond = vlan_dev_real_dev(skb->dev); |
1f3c8804 AG |
2569 | } |
2570 | ||
1da177e4 | 2571 | type = skb->protocol; |
82d8a867 PE |
2572 | list_for_each_entry_rcu(ptype, |
2573 | &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { | |
1f3c8804 | 2574 | if (ptype->type == type && (ptype->dev == null_or_orig || |
ca8d9ea3 AG |
2575 | ptype->dev == skb->dev || ptype->dev == orig_dev || |
2576 | ptype->dev == null_or_bond)) { | |
4ec93edb | 2577 | if (pt_prev) |
f2ccd8fa | 2578 | ret = deliver_skb(skb, pt_prev, orig_dev); |
1da177e4 LT |
2579 | pt_prev = ptype; |
2580 | } | |
2581 | } | |
2582 | ||
2583 | if (pt_prev) { | |
f2ccd8fa | 2584 | ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev); |
1da177e4 LT |
2585 | } else { |
2586 | kfree_skb(skb); | |
2587 | /* Jamal, now you will not able to escape explaining | |
2588 | * me how you were going to use this. :-) | |
2589 | */ | |
2590 | ret = NET_RX_DROP; | |
2591 | } | |
2592 | ||
2593 | out: | |
2594 | rcu_read_unlock(); | |
2595 | return ret; | |
2596 | } | |
d1b19dff | 2597 | EXPORT_SYMBOL(netif_receive_skb); |
1da177e4 | 2598 | |
6e583ce5 SH |
2599 | /* Network device is going away, flush any packets still pending */ |
2600 | static void flush_backlog(void *arg) | |
2601 | { | |
2602 | struct net_device *dev = arg; | |
2603 | struct softnet_data *queue = &__get_cpu_var(softnet_data); | |
2604 | struct sk_buff *skb, *tmp; | |
2605 | ||
2606 | skb_queue_walk_safe(&queue->input_pkt_queue, skb, tmp) | |
2607 | if (skb->dev == dev) { | |
2608 | __skb_unlink(skb, &queue->input_pkt_queue); | |
2609 | kfree_skb(skb); | |
2610 | } | |
2611 | } | |
2612 | ||
d565b0a1 HX |
2613 | static int napi_gro_complete(struct sk_buff *skb) |
2614 | { | |
2615 | struct packet_type *ptype; | |
2616 | __be16 type = skb->protocol; | |
2617 | struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK]; | |
2618 | int err = -ENOENT; | |
2619 | ||
fc59f9a3 HX |
2620 | if (NAPI_GRO_CB(skb)->count == 1) { |
2621 | skb_shinfo(skb)->gso_size = 0; | |
d565b0a1 | 2622 | goto out; |
fc59f9a3 | 2623 | } |
d565b0a1 HX |
2624 | |
2625 | rcu_read_lock(); | |
2626 | list_for_each_entry_rcu(ptype, head, list) { | |
2627 | if (ptype->type != type || ptype->dev || !ptype->gro_complete) | |
2628 | continue; | |
2629 | ||
2630 | err = ptype->gro_complete(skb); | |
2631 | break; | |
2632 | } | |
2633 | rcu_read_unlock(); | |
2634 | ||
2635 | if (err) { | |
2636 | WARN_ON(&ptype->list == head); | |
2637 | kfree_skb(skb); | |
2638 | return NET_RX_SUCCESS; | |
2639 | } | |
2640 | ||
2641 | out: | |
d565b0a1 HX |
2642 | return netif_receive_skb(skb); |
2643 | } | |
2644 | ||
11380a4b | 2645 | static void napi_gro_flush(struct napi_struct *napi) |
d565b0a1 HX |
2646 | { |
2647 | struct sk_buff *skb, *next; | |
2648 | ||
2649 | for (skb = napi->gro_list; skb; skb = next) { | |
2650 | next = skb->next; | |
2651 | skb->next = NULL; | |
2652 | napi_gro_complete(skb); | |
2653 | } | |
2654 | ||
4ae5544f | 2655 | napi->gro_count = 0; |
d565b0a1 HX |
2656 | napi->gro_list = NULL; |
2657 | } | |
d565b0a1 | 2658 | |
5b252f0c | 2659 | enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb) |
d565b0a1 HX |
2660 | { |
2661 | struct sk_buff **pp = NULL; | |
2662 | struct packet_type *ptype; | |
2663 | __be16 type = skb->protocol; | |
2664 | struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK]; | |
0da2afd5 | 2665 | int same_flow; |
d565b0a1 | 2666 | int mac_len; |
5b252f0c | 2667 | enum gro_result ret; |
d565b0a1 HX |
2668 | |
2669 | if (!(skb->dev->features & NETIF_F_GRO)) | |
2670 | goto normal; | |
2671 | ||
4cf704fb | 2672 | if (skb_is_gso(skb) || skb_has_frags(skb)) |
f17f5c91 HX |
2673 | goto normal; |
2674 | ||
d565b0a1 HX |
2675 | rcu_read_lock(); |
2676 | list_for_each_entry_rcu(ptype, head, list) { | |
d565b0a1 HX |
2677 | if (ptype->type != type || ptype->dev || !ptype->gro_receive) |
2678 | continue; | |
2679 | ||
86911732 | 2680 | skb_set_network_header(skb, skb_gro_offset(skb)); |
d565b0a1 HX |
2681 | mac_len = skb->network_header - skb->mac_header; |
2682 | skb->mac_len = mac_len; | |
2683 | NAPI_GRO_CB(skb)->same_flow = 0; | |
2684 | NAPI_GRO_CB(skb)->flush = 0; | |
5d38a079 | 2685 | NAPI_GRO_CB(skb)->free = 0; |
d565b0a1 | 2686 | |
d565b0a1 HX |
2687 | pp = ptype->gro_receive(&napi->gro_list, skb); |
2688 | break; | |
2689 | } | |
2690 | rcu_read_unlock(); | |
2691 | ||
2692 | if (&ptype->list == head) | |
2693 | goto normal; | |
2694 | ||
0da2afd5 | 2695 | same_flow = NAPI_GRO_CB(skb)->same_flow; |
5d0d9be8 | 2696 | ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED; |
0da2afd5 | 2697 | |
d565b0a1 HX |
2698 | if (pp) { |
2699 | struct sk_buff *nskb = *pp; | |
2700 | ||
2701 | *pp = nskb->next; | |
2702 | nskb->next = NULL; | |
2703 | napi_gro_complete(nskb); | |
4ae5544f | 2704 | napi->gro_count--; |
d565b0a1 HX |
2705 | } |
2706 | ||
0da2afd5 | 2707 | if (same_flow) |
d565b0a1 HX |
2708 | goto ok; |
2709 | ||
4ae5544f | 2710 | if (NAPI_GRO_CB(skb)->flush || napi->gro_count >= MAX_GRO_SKBS) |
d565b0a1 | 2711 | goto normal; |
d565b0a1 | 2712 | |
4ae5544f | 2713 | napi->gro_count++; |
d565b0a1 | 2714 | NAPI_GRO_CB(skb)->count = 1; |
86911732 | 2715 | skb_shinfo(skb)->gso_size = skb_gro_len(skb); |
d565b0a1 HX |
2716 | skb->next = napi->gro_list; |
2717 | napi->gro_list = skb; | |
5d0d9be8 | 2718 | ret = GRO_HELD; |
d565b0a1 | 2719 | |
ad0f9904 | 2720 | pull: |
cb18978c HX |
2721 | if (skb_headlen(skb) < skb_gro_offset(skb)) { |
2722 | int grow = skb_gro_offset(skb) - skb_headlen(skb); | |
2723 | ||
2724 | BUG_ON(skb->end - skb->tail < grow); | |
2725 | ||
2726 | memcpy(skb_tail_pointer(skb), NAPI_GRO_CB(skb)->frag0, grow); | |
2727 | ||
2728 | skb->tail += grow; | |
2729 | skb->data_len -= grow; | |
2730 | ||
2731 | skb_shinfo(skb)->frags[0].page_offset += grow; | |
2732 | skb_shinfo(skb)->frags[0].size -= grow; | |
2733 | ||
2734 | if (unlikely(!skb_shinfo(skb)->frags[0].size)) { | |
2735 | put_page(skb_shinfo(skb)->frags[0].page); | |
2736 | memmove(skb_shinfo(skb)->frags, | |
2737 | skb_shinfo(skb)->frags + 1, | |
2738 | --skb_shinfo(skb)->nr_frags); | |
2739 | } | |
ad0f9904 HX |
2740 | } |
2741 | ||
d565b0a1 | 2742 | ok: |
5d0d9be8 | 2743 | return ret; |
d565b0a1 HX |
2744 | |
2745 | normal: | |
ad0f9904 HX |
2746 | ret = GRO_NORMAL; |
2747 | goto pull; | |
5d38a079 | 2748 | } |
96e93eab HX |
2749 | EXPORT_SYMBOL(dev_gro_receive); |
2750 | ||
5b252f0c BH |
2751 | static gro_result_t |
2752 | __napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb) | |
96e93eab HX |
2753 | { |
2754 | struct sk_buff *p; | |
2755 | ||
d1c76af9 HX |
2756 | if (netpoll_rx_on(skb)) |
2757 | return GRO_NORMAL; | |
2758 | ||
96e93eab | 2759 | for (p = napi->gro_list; p; p = p->next) { |
f64f9e71 JP |
2760 | NAPI_GRO_CB(p)->same_flow = |
2761 | (p->dev == skb->dev) && | |
2762 | !compare_ether_header(skb_mac_header(p), | |
2763 | skb_gro_mac_header(skb)); | |
96e93eab HX |
2764 | NAPI_GRO_CB(p)->flush = 0; |
2765 | } | |
2766 | ||
2767 | return dev_gro_receive(napi, skb); | |
2768 | } | |
5d38a079 | 2769 | |
c7c4b3b6 | 2770 | gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb) |
5d38a079 | 2771 | { |
5d0d9be8 HX |
2772 | switch (ret) { |
2773 | case GRO_NORMAL: | |
c7c4b3b6 BH |
2774 | if (netif_receive_skb(skb)) |
2775 | ret = GRO_DROP; | |
2776 | break; | |
5d38a079 | 2777 | |
5d0d9be8 | 2778 | case GRO_DROP: |
5d0d9be8 | 2779 | case GRO_MERGED_FREE: |
5d38a079 HX |
2780 | kfree_skb(skb); |
2781 | break; | |
5b252f0c BH |
2782 | |
2783 | case GRO_HELD: | |
2784 | case GRO_MERGED: | |
2785 | break; | |
5d38a079 HX |
2786 | } |
2787 | ||
c7c4b3b6 | 2788 | return ret; |
5d0d9be8 HX |
2789 | } |
2790 | EXPORT_SYMBOL(napi_skb_finish); | |
2791 | ||
78a478d0 HX |
2792 | void skb_gro_reset_offset(struct sk_buff *skb) |
2793 | { | |
2794 | NAPI_GRO_CB(skb)->data_offset = 0; | |
2795 | NAPI_GRO_CB(skb)->frag0 = NULL; | |
7489594c | 2796 | NAPI_GRO_CB(skb)->frag0_len = 0; |
78a478d0 | 2797 | |
78d3fd0b | 2798 | if (skb->mac_header == skb->tail && |
7489594c | 2799 | !PageHighMem(skb_shinfo(skb)->frags[0].page)) { |
78a478d0 HX |
2800 | NAPI_GRO_CB(skb)->frag0 = |
2801 | page_address(skb_shinfo(skb)->frags[0].page) + | |
2802 | skb_shinfo(skb)->frags[0].page_offset; | |
7489594c HX |
2803 | NAPI_GRO_CB(skb)->frag0_len = skb_shinfo(skb)->frags[0].size; |
2804 | } | |
78a478d0 HX |
2805 | } |
2806 | EXPORT_SYMBOL(skb_gro_reset_offset); | |
2807 | ||
c7c4b3b6 | 2808 | gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb) |
5d0d9be8 | 2809 | { |
86911732 HX |
2810 | skb_gro_reset_offset(skb); |
2811 | ||
5d0d9be8 | 2812 | return napi_skb_finish(__napi_gro_receive(napi, skb), skb); |
d565b0a1 HX |
2813 | } |
2814 | EXPORT_SYMBOL(napi_gro_receive); | |
2815 | ||
96e93eab HX |
2816 | void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb) |
2817 | { | |
96e93eab HX |
2818 | __skb_pull(skb, skb_headlen(skb)); |
2819 | skb_reserve(skb, NET_IP_ALIGN - skb_headroom(skb)); | |
2820 | ||
2821 | napi->skb = skb; | |
2822 | } | |
2823 | EXPORT_SYMBOL(napi_reuse_skb); | |
2824 | ||
76620aaf | 2825 | struct sk_buff *napi_get_frags(struct napi_struct *napi) |
5d38a079 | 2826 | { |
5d38a079 | 2827 | struct sk_buff *skb = napi->skb; |
5d38a079 HX |
2828 | |
2829 | if (!skb) { | |
89d71a66 ED |
2830 | skb = netdev_alloc_skb_ip_align(napi->dev, GRO_MAX_HEAD); |
2831 | if (skb) | |
2832 | napi->skb = skb; | |
80595d59 | 2833 | } |
96e93eab HX |
2834 | return skb; |
2835 | } | |
76620aaf | 2836 | EXPORT_SYMBOL(napi_get_frags); |
96e93eab | 2837 | |
c7c4b3b6 BH |
2838 | gro_result_t napi_frags_finish(struct napi_struct *napi, struct sk_buff *skb, |
2839 | gro_result_t ret) | |
96e93eab | 2840 | { |
5d0d9be8 HX |
2841 | switch (ret) { |
2842 | case GRO_NORMAL: | |
86911732 | 2843 | case GRO_HELD: |
e76b69cc | 2844 | skb->protocol = eth_type_trans(skb, skb->dev); |
86911732 | 2845 | |
c7c4b3b6 BH |
2846 | if (ret == GRO_HELD) |
2847 | skb_gro_pull(skb, -ETH_HLEN); | |
2848 | else if (netif_receive_skb(skb)) | |
2849 | ret = GRO_DROP; | |
86911732 | 2850 | break; |
5d38a079 | 2851 | |
5d0d9be8 | 2852 | case GRO_DROP: |
5d0d9be8 HX |
2853 | case GRO_MERGED_FREE: |
2854 | napi_reuse_skb(napi, skb); | |
2855 | break; | |
5b252f0c BH |
2856 | |
2857 | case GRO_MERGED: | |
2858 | break; | |
5d0d9be8 | 2859 | } |
5d38a079 | 2860 | |
c7c4b3b6 | 2861 | return ret; |
5d38a079 | 2862 | } |
5d0d9be8 HX |
2863 | EXPORT_SYMBOL(napi_frags_finish); |
2864 | ||
76620aaf HX |
2865 | struct sk_buff *napi_frags_skb(struct napi_struct *napi) |
2866 | { | |
2867 | struct sk_buff *skb = napi->skb; | |
2868 | struct ethhdr *eth; | |
a5b1cf28 HX |
2869 | unsigned int hlen; |
2870 | unsigned int off; | |
76620aaf HX |
2871 | |
2872 | napi->skb = NULL; | |
2873 | ||
2874 | skb_reset_mac_header(skb); | |
2875 | skb_gro_reset_offset(skb); | |
2876 | ||
a5b1cf28 HX |
2877 | off = skb_gro_offset(skb); |
2878 | hlen = off + sizeof(*eth); | |
2879 | eth = skb_gro_header_fast(skb, off); | |
2880 | if (skb_gro_header_hard(skb, hlen)) { | |
2881 | eth = skb_gro_header_slow(skb, hlen, off); | |
2882 | if (unlikely(!eth)) { | |
2883 | napi_reuse_skb(napi, skb); | |
2884 | skb = NULL; | |
2885 | goto out; | |
2886 | } | |
76620aaf HX |
2887 | } |
2888 | ||
2889 | skb_gro_pull(skb, sizeof(*eth)); | |
2890 | ||
2891 | /* | |
2892 | * This works because the only protocols we care about don't require | |
2893 | * special handling. We'll fix it up properly at the end. | |
2894 | */ | |
2895 | skb->protocol = eth->h_proto; | |
2896 | ||
2897 | out: | |
2898 | return skb; | |
2899 | } | |
2900 | EXPORT_SYMBOL(napi_frags_skb); | |
2901 | ||
c7c4b3b6 | 2902 | gro_result_t napi_gro_frags(struct napi_struct *napi) |
5d0d9be8 | 2903 | { |
76620aaf | 2904 | struct sk_buff *skb = napi_frags_skb(napi); |
5d0d9be8 HX |
2905 | |
2906 | if (!skb) | |
c7c4b3b6 | 2907 | return GRO_DROP; |
5d0d9be8 HX |
2908 | |
2909 | return napi_frags_finish(napi, skb, __napi_gro_receive(napi, skb)); | |
2910 | } | |
5d38a079 HX |
2911 | EXPORT_SYMBOL(napi_gro_frags); |
2912 | ||
bea3348e | 2913 | static int process_backlog(struct napi_struct *napi, int quota) |
1da177e4 LT |
2914 | { |
2915 | int work = 0; | |
1da177e4 LT |
2916 | struct softnet_data *queue = &__get_cpu_var(softnet_data); |
2917 | unsigned long start_time = jiffies; | |
2918 | ||
bea3348e SH |
2919 | napi->weight = weight_p; |
2920 | do { | |
1da177e4 | 2921 | struct sk_buff *skb; |
1da177e4 LT |
2922 | |
2923 | local_irq_disable(); | |
2924 | skb = __skb_dequeue(&queue->input_pkt_queue); | |
bea3348e | 2925 | if (!skb) { |
8f1ead2d | 2926 | __napi_complete(napi); |
bea3348e | 2927 | local_irq_enable(); |
8f1ead2d | 2928 | break; |
bea3348e | 2929 | } |
1da177e4 LT |
2930 | local_irq_enable(); |
2931 | ||
8f1ead2d | 2932 | netif_receive_skb(skb); |
bea3348e | 2933 | } while (++work < quota && jiffies == start_time); |
1da177e4 | 2934 | |
bea3348e SH |
2935 | return work; |
2936 | } | |
1da177e4 | 2937 | |
bea3348e SH |
2938 | /** |
2939 | * __napi_schedule - schedule for receive | |
c4ea43c5 | 2940 | * @n: entry to schedule |
bea3348e SH |
2941 | * |
2942 | * The entry's receive function will be scheduled to run | |
2943 | */ | |
b5606c2d | 2944 | void __napi_schedule(struct napi_struct *n) |
bea3348e SH |
2945 | { |
2946 | unsigned long flags; | |
1da177e4 | 2947 | |
bea3348e SH |
2948 | local_irq_save(flags); |
2949 | list_add_tail(&n->poll_list, &__get_cpu_var(softnet_data).poll_list); | |
2950 | __raise_softirq_irqoff(NET_RX_SOFTIRQ); | |
2951 | local_irq_restore(flags); | |
1da177e4 | 2952 | } |
bea3348e SH |
2953 | EXPORT_SYMBOL(__napi_schedule); |
2954 | ||
d565b0a1 HX |
2955 | void __napi_complete(struct napi_struct *n) |
2956 | { | |
2957 | BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); | |
2958 | BUG_ON(n->gro_list); | |
2959 | ||
2960 | list_del(&n->poll_list); | |
2961 | smp_mb__before_clear_bit(); | |
2962 | clear_bit(NAPI_STATE_SCHED, &n->state); | |
2963 | } | |
2964 | EXPORT_SYMBOL(__napi_complete); | |
2965 | ||
2966 | void napi_complete(struct napi_struct *n) | |
2967 | { | |
2968 | unsigned long flags; | |
2969 | ||
2970 | /* | |
2971 | * don't let napi dequeue from the cpu poll list | |
2972 | * just in case its running on a different cpu | |
2973 | */ | |
2974 | if (unlikely(test_bit(NAPI_STATE_NPSVC, &n->state))) | |
2975 | return; | |
2976 | ||
2977 | napi_gro_flush(n); | |
2978 | local_irq_save(flags); | |
2979 | __napi_complete(n); | |
2980 | local_irq_restore(flags); | |
2981 | } | |
2982 | EXPORT_SYMBOL(napi_complete); | |
2983 | ||
2984 | void netif_napi_add(struct net_device *dev, struct napi_struct *napi, | |
2985 | int (*poll)(struct napi_struct *, int), int weight) | |
2986 | { | |
2987 | INIT_LIST_HEAD(&napi->poll_list); | |
4ae5544f | 2988 | napi->gro_count = 0; |
d565b0a1 | 2989 | napi->gro_list = NULL; |
5d38a079 | 2990 | napi->skb = NULL; |
d565b0a1 HX |
2991 | napi->poll = poll; |
2992 | napi->weight = weight; | |
2993 | list_add(&napi->dev_list, &dev->napi_list); | |
d565b0a1 | 2994 | napi->dev = dev; |
5d38a079 | 2995 | #ifdef CONFIG_NETPOLL |
d565b0a1 HX |
2996 | spin_lock_init(&napi->poll_lock); |
2997 | napi->poll_owner = -1; | |
2998 | #endif | |
2999 | set_bit(NAPI_STATE_SCHED, &napi->state); | |
3000 | } | |
3001 | EXPORT_SYMBOL(netif_napi_add); | |
3002 | ||
3003 | void netif_napi_del(struct napi_struct *napi) | |
3004 | { | |
3005 | struct sk_buff *skb, *next; | |
3006 | ||
d7b06636 | 3007 | list_del_init(&napi->dev_list); |
76620aaf | 3008 | napi_free_frags(napi); |
d565b0a1 HX |
3009 | |
3010 | for (skb = napi->gro_list; skb; skb = next) { | |
3011 | next = skb->next; | |
3012 | skb->next = NULL; | |
3013 | kfree_skb(skb); | |
3014 | } | |
3015 | ||
3016 | napi->gro_list = NULL; | |
4ae5544f | 3017 | napi->gro_count = 0; |
d565b0a1 HX |
3018 | } |
3019 | EXPORT_SYMBOL(netif_napi_del); | |
3020 | ||
1da177e4 LT |
3021 | |
3022 | static void net_rx_action(struct softirq_action *h) | |
3023 | { | |
bea3348e | 3024 | struct list_head *list = &__get_cpu_var(softnet_data).poll_list; |
24f8b238 | 3025 | unsigned long time_limit = jiffies + 2; |
51b0bded | 3026 | int budget = netdev_budget; |
53fb95d3 MM |
3027 | void *have; |
3028 | ||
1da177e4 LT |
3029 | local_irq_disable(); |
3030 | ||
bea3348e SH |
3031 | while (!list_empty(list)) { |
3032 | struct napi_struct *n; | |
3033 | int work, weight; | |
1da177e4 | 3034 | |
bea3348e | 3035 | /* If softirq window is exhuasted then punt. |
24f8b238 SH |
3036 | * Allow this to run for 2 jiffies since which will allow |
3037 | * an average latency of 1.5/HZ. | |
bea3348e | 3038 | */ |
24f8b238 | 3039 | if (unlikely(budget <= 0 || time_after(jiffies, time_limit))) |
1da177e4 LT |
3040 | goto softnet_break; |
3041 | ||
3042 | local_irq_enable(); | |
3043 | ||
bea3348e SH |
3044 | /* Even though interrupts have been re-enabled, this |
3045 | * access is safe because interrupts can only add new | |
3046 | * entries to the tail of this list, and only ->poll() | |
3047 | * calls can remove this head entry from the list. | |
3048 | */ | |
e5e26d75 | 3049 | n = list_first_entry(list, struct napi_struct, poll_list); |
1da177e4 | 3050 | |
bea3348e SH |
3051 | have = netpoll_poll_lock(n); |
3052 | ||
3053 | weight = n->weight; | |
3054 | ||
0a7606c1 DM |
3055 | /* This NAPI_STATE_SCHED test is for avoiding a race |
3056 | * with netpoll's poll_napi(). Only the entity which | |
3057 | * obtains the lock and sees NAPI_STATE_SCHED set will | |
3058 | * actually make the ->poll() call. Therefore we avoid | |
3059 | * accidently calling ->poll() when NAPI is not scheduled. | |
3060 | */ | |
3061 | work = 0; | |
4ea7e386 | 3062 | if (test_bit(NAPI_STATE_SCHED, &n->state)) { |
0a7606c1 | 3063 | work = n->poll(n, weight); |
4ea7e386 NH |
3064 | trace_napi_poll(n); |
3065 | } | |
bea3348e SH |
3066 | |
3067 | WARN_ON_ONCE(work > weight); | |
3068 | ||
3069 | budget -= work; | |
3070 | ||
3071 | local_irq_disable(); | |
3072 | ||
3073 | /* Drivers must not modify the NAPI state if they | |
3074 | * consume the entire weight. In such cases this code | |
3075 | * still "owns" the NAPI instance and therefore can | |
3076 | * move the instance around on the list at-will. | |
3077 | */ | |
fed17f30 | 3078 | if (unlikely(work == weight)) { |
ff780cd8 HX |
3079 | if (unlikely(napi_disable_pending(n))) { |
3080 | local_irq_enable(); | |
3081 | napi_complete(n); | |
3082 | local_irq_disable(); | |
3083 | } else | |
fed17f30 DM |
3084 | list_move_tail(&n->poll_list, list); |
3085 | } | |
bea3348e SH |
3086 | |
3087 | netpoll_poll_unlock(have); | |
1da177e4 LT |
3088 | } |
3089 | out: | |
515e06c4 | 3090 | local_irq_enable(); |
bea3348e | 3091 | |
db217334 CL |
3092 | #ifdef CONFIG_NET_DMA |
3093 | /* | |
3094 | * There may not be any more sk_buffs coming right now, so push | |
3095 | * any pending DMA copies to hardware | |
3096 | */ | |
2ba05622 | 3097 | dma_issue_pending_all(); |
db217334 | 3098 | #endif |
bea3348e | 3099 | |
1da177e4 LT |
3100 | return; |
3101 | ||
3102 | softnet_break: | |
3103 | __get_cpu_var(netdev_rx_stat).time_squeeze++; | |
3104 | __raise_softirq_irqoff(NET_RX_SOFTIRQ); | |
3105 | goto out; | |
3106 | } | |
3107 | ||
d1b19dff | 3108 | static gifconf_func_t *gifconf_list[NPROTO]; |
1da177e4 LT |
3109 | |
3110 | /** | |
3111 | * register_gifconf - register a SIOCGIF handler | |
3112 | * @family: Address family | |
3113 | * @gifconf: Function handler | |
3114 | * | |
3115 | * Register protocol dependent address dumping routines. The handler | |
3116 | * that is passed must not be freed or reused until it has been replaced | |
3117 | * by another handler. | |
3118 | */ | |
d1b19dff | 3119 | int register_gifconf(unsigned int family, gifconf_func_t *gifconf) |
1da177e4 LT |
3120 | { |
3121 | if (family >= NPROTO) | |
3122 | return -EINVAL; | |
3123 | gifconf_list[family] = gifconf; | |
3124 | return 0; | |
3125 | } | |
d1b19dff | 3126 | EXPORT_SYMBOL(register_gifconf); |
1da177e4 LT |
3127 | |
3128 | ||
3129 | /* | |
3130 | * Map an interface index to its name (SIOCGIFNAME) | |
3131 | */ | |
3132 | ||
3133 | /* | |
3134 | * We need this ioctl for efficient implementation of the | |
3135 | * if_indextoname() function required by the IPv6 API. Without | |
3136 | * it, we would have to search all the interfaces to find a | |
3137 | * match. --pb | |
3138 | */ | |
3139 | ||
881d966b | 3140 | static int dev_ifname(struct net *net, struct ifreq __user *arg) |
1da177e4 LT |
3141 | { |
3142 | struct net_device *dev; | |
3143 | struct ifreq ifr; | |
3144 | ||
3145 | /* | |
3146 | * Fetch the caller's info block. | |
3147 | */ | |
3148 | ||
3149 | if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) | |
3150 | return -EFAULT; | |
3151 | ||
fb699dfd ED |
3152 | rcu_read_lock(); |
3153 | dev = dev_get_by_index_rcu(net, ifr.ifr_ifindex); | |
1da177e4 | 3154 | if (!dev) { |
fb699dfd | 3155 | rcu_read_unlock(); |
1da177e4 LT |
3156 | return -ENODEV; |
3157 | } | |
3158 | ||
3159 | strcpy(ifr.ifr_name, dev->name); | |
fb699dfd | 3160 | rcu_read_unlock(); |
1da177e4 LT |
3161 | |
3162 | if (copy_to_user(arg, &ifr, sizeof(struct ifreq))) | |
3163 | return -EFAULT; | |
3164 | return 0; | |
3165 | } | |
3166 | ||
3167 | /* | |
3168 | * Perform a SIOCGIFCONF call. This structure will change | |
3169 | * size eventually, and there is nothing I can do about it. | |
3170 | * Thus we will need a 'compatibility mode'. | |
3171 | */ | |
3172 | ||
881d966b | 3173 | static int dev_ifconf(struct net *net, char __user *arg) |
1da177e4 LT |
3174 | { |
3175 | struct ifconf ifc; | |
3176 | struct net_device *dev; | |
3177 | char __user *pos; | |
3178 | int len; | |
3179 | int total; | |
3180 | int i; | |
3181 | ||
3182 | /* | |
3183 | * Fetch the caller's info block. | |
3184 | */ | |
3185 | ||
3186 | if (copy_from_user(&ifc, arg, sizeof(struct ifconf))) | |
3187 | return -EFAULT; | |
3188 | ||
3189 | pos = ifc.ifc_buf; | |
3190 | len = ifc.ifc_len; | |
3191 | ||
3192 | /* | |
3193 | * Loop over the interfaces, and write an info block for each. | |
3194 | */ | |
3195 | ||
3196 | total = 0; | |
881d966b | 3197 | for_each_netdev(net, dev) { |
1da177e4 LT |
3198 | for (i = 0; i < NPROTO; i++) { |
3199 | if (gifconf_list[i]) { | |
3200 | int done; | |
3201 | if (!pos) | |
3202 | done = gifconf_list[i](dev, NULL, 0); | |
3203 | else | |
3204 | done = gifconf_list[i](dev, pos + total, | |
3205 | len - total); | |
3206 | if (done < 0) | |
3207 | return -EFAULT; | |
3208 | total += done; | |
3209 | } | |
3210 | } | |
4ec93edb | 3211 | } |
1da177e4 LT |
3212 | |
3213 | /* | |
3214 | * All done. Write the updated control block back to the caller. | |
3215 | */ | |
3216 | ifc.ifc_len = total; | |
3217 | ||
3218 | /* | |
3219 | * Both BSD and Solaris return 0 here, so we do too. | |
3220 | */ | |
3221 | return copy_to_user(arg, &ifc, sizeof(struct ifconf)) ? -EFAULT : 0; | |
3222 | } | |
3223 | ||
3224 | #ifdef CONFIG_PROC_FS | |
3225 | /* | |
3226 | * This is invoked by the /proc filesystem handler to display a device | |
3227 | * in detail. | |
3228 | */ | |
7562f876 | 3229 | void *dev_seq_start(struct seq_file *seq, loff_t *pos) |
c6d14c84 | 3230 | __acquires(RCU) |
1da177e4 | 3231 | { |
e372c414 | 3232 | struct net *net = seq_file_net(seq); |
7562f876 | 3233 | loff_t off; |
1da177e4 | 3234 | struct net_device *dev; |
1da177e4 | 3235 | |
c6d14c84 | 3236 | rcu_read_lock(); |
7562f876 PE |
3237 | if (!*pos) |
3238 | return SEQ_START_TOKEN; | |
1da177e4 | 3239 | |
7562f876 | 3240 | off = 1; |
c6d14c84 | 3241 | for_each_netdev_rcu(net, dev) |
7562f876 PE |
3242 | if (off++ == *pos) |
3243 | return dev; | |
1da177e4 | 3244 | |
7562f876 | 3245 | return NULL; |
1da177e4 LT |
3246 | } |
3247 | ||
3248 | void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3249 | { | |
c6d14c84 ED |
3250 | struct net_device *dev = (v == SEQ_START_TOKEN) ? |
3251 | first_net_device(seq_file_net(seq)) : | |
3252 | next_net_device((struct net_device *)v); | |
3253 | ||
1da177e4 | 3254 | ++*pos; |
c6d14c84 | 3255 | return rcu_dereference(dev); |
1da177e4 LT |
3256 | } |
3257 | ||
3258 | void dev_seq_stop(struct seq_file *seq, void *v) | |
c6d14c84 | 3259 | __releases(RCU) |
1da177e4 | 3260 | { |
c6d14c84 | 3261 | rcu_read_unlock(); |
1da177e4 LT |
3262 | } |
3263 | ||
3264 | static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev) | |
3265 | { | |
eeda3fd6 | 3266 | const struct net_device_stats *stats = dev_get_stats(dev); |
1da177e4 | 3267 | |
2d13bafe | 3268 | seq_printf(seq, "%6s: %7lu %7lu %4lu %4lu %4lu %5lu %10lu %9lu " |
5a1b5898 RR |
3269 | "%8lu %7lu %4lu %4lu %4lu %5lu %7lu %10lu\n", |
3270 | dev->name, stats->rx_bytes, stats->rx_packets, | |
3271 | stats->rx_errors, | |
3272 | stats->rx_dropped + stats->rx_missed_errors, | |
3273 | stats->rx_fifo_errors, | |
3274 | stats->rx_length_errors + stats->rx_over_errors + | |
3275 | stats->rx_crc_errors + stats->rx_frame_errors, | |
3276 | stats->rx_compressed, stats->multicast, | |
3277 | stats->tx_bytes, stats->tx_packets, | |
3278 | stats->tx_errors, stats->tx_dropped, | |
3279 | stats->tx_fifo_errors, stats->collisions, | |
3280 | stats->tx_carrier_errors + | |
3281 | stats->tx_aborted_errors + | |
3282 | stats->tx_window_errors + | |
3283 | stats->tx_heartbeat_errors, | |
3284 | stats->tx_compressed); | |
1da177e4 LT |
3285 | } |
3286 | ||
3287 | /* | |
3288 | * Called from the PROCfs module. This now uses the new arbitrary sized | |
3289 | * /proc/net interface to create /proc/net/dev | |
3290 | */ | |
3291 | static int dev_seq_show(struct seq_file *seq, void *v) | |
3292 | { | |
3293 | if (v == SEQ_START_TOKEN) | |
3294 | seq_puts(seq, "Inter-| Receive " | |
3295 | " | Transmit\n" | |
3296 | " face |bytes packets errs drop fifo frame " | |
3297 | "compressed multicast|bytes packets errs " | |
3298 | "drop fifo colls carrier compressed\n"); | |
3299 | else | |
3300 | dev_seq_printf_stats(seq, v); | |
3301 | return 0; | |
3302 | } | |
3303 | ||
3304 | static struct netif_rx_stats *softnet_get_online(loff_t *pos) | |
3305 | { | |
3306 | struct netif_rx_stats *rc = NULL; | |
3307 | ||
0c0b0aca | 3308 | while (*pos < nr_cpu_ids) |
4ec93edb | 3309 | if (cpu_online(*pos)) { |
1da177e4 LT |
3310 | rc = &per_cpu(netdev_rx_stat, *pos); |
3311 | break; | |
3312 | } else | |
3313 | ++*pos; | |
3314 | return rc; | |
3315 | } | |
3316 | ||
3317 | static void *softnet_seq_start(struct seq_file *seq, loff_t *pos) | |
3318 | { | |
3319 | return softnet_get_online(pos); | |
3320 | } | |
3321 | ||
3322 | static void *softnet_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3323 | { | |
3324 | ++*pos; | |
3325 | return softnet_get_online(pos); | |
3326 | } | |
3327 | ||
3328 | static void softnet_seq_stop(struct seq_file *seq, void *v) | |
3329 | { | |
3330 | } | |
3331 | ||
3332 | static int softnet_seq_show(struct seq_file *seq, void *v) | |
3333 | { | |
3334 | struct netif_rx_stats *s = v; | |
3335 | ||
3336 | seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n", | |
31aa02c5 | 3337 | s->total, s->dropped, s->time_squeeze, 0, |
c1ebcdb8 | 3338 | 0, 0, 0, 0, /* was fastroute */ |
d1b19dff | 3339 | s->cpu_collision); |
1da177e4 LT |
3340 | return 0; |
3341 | } | |
3342 | ||
f690808e | 3343 | static const struct seq_operations dev_seq_ops = { |
1da177e4 LT |
3344 | .start = dev_seq_start, |
3345 | .next = dev_seq_next, | |
3346 | .stop = dev_seq_stop, | |
3347 | .show = dev_seq_show, | |
3348 | }; | |
3349 | ||
3350 | static int dev_seq_open(struct inode *inode, struct file *file) | |
3351 | { | |
e372c414 DL |
3352 | return seq_open_net(inode, file, &dev_seq_ops, |
3353 | sizeof(struct seq_net_private)); | |
1da177e4 LT |
3354 | } |
3355 | ||
9a32144e | 3356 | static const struct file_operations dev_seq_fops = { |
1da177e4 LT |
3357 | .owner = THIS_MODULE, |
3358 | .open = dev_seq_open, | |
3359 | .read = seq_read, | |
3360 | .llseek = seq_lseek, | |
e372c414 | 3361 | .release = seq_release_net, |
1da177e4 LT |
3362 | }; |
3363 | ||
f690808e | 3364 | static const struct seq_operations softnet_seq_ops = { |
1da177e4 LT |
3365 | .start = softnet_seq_start, |
3366 | .next = softnet_seq_next, | |
3367 | .stop = softnet_seq_stop, | |
3368 | .show = softnet_seq_show, | |
3369 | }; | |
3370 | ||
3371 | static int softnet_seq_open(struct inode *inode, struct file *file) | |
3372 | { | |
3373 | return seq_open(file, &softnet_seq_ops); | |
3374 | } | |
3375 | ||
9a32144e | 3376 | static const struct file_operations softnet_seq_fops = { |
1da177e4 LT |
3377 | .owner = THIS_MODULE, |
3378 | .open = softnet_seq_open, | |
3379 | .read = seq_read, | |
3380 | .llseek = seq_lseek, | |
3381 | .release = seq_release, | |
3382 | }; | |
3383 | ||
0e1256ff SH |
3384 | static void *ptype_get_idx(loff_t pos) |
3385 | { | |
3386 | struct packet_type *pt = NULL; | |
3387 | loff_t i = 0; | |
3388 | int t; | |
3389 | ||
3390 | list_for_each_entry_rcu(pt, &ptype_all, list) { | |
3391 | if (i == pos) | |
3392 | return pt; | |
3393 | ++i; | |
3394 | } | |
3395 | ||
82d8a867 | 3396 | for (t = 0; t < PTYPE_HASH_SIZE; t++) { |
0e1256ff SH |
3397 | list_for_each_entry_rcu(pt, &ptype_base[t], list) { |
3398 | if (i == pos) | |
3399 | return pt; | |
3400 | ++i; | |
3401 | } | |
3402 | } | |
3403 | return NULL; | |
3404 | } | |
3405 | ||
3406 | static void *ptype_seq_start(struct seq_file *seq, loff_t *pos) | |
72348a42 | 3407 | __acquires(RCU) |
0e1256ff SH |
3408 | { |
3409 | rcu_read_lock(); | |
3410 | return *pos ? ptype_get_idx(*pos - 1) : SEQ_START_TOKEN; | |
3411 | } | |
3412 | ||
3413 | static void *ptype_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3414 | { | |
3415 | struct packet_type *pt; | |
3416 | struct list_head *nxt; | |
3417 | int hash; | |
3418 | ||
3419 | ++*pos; | |
3420 | if (v == SEQ_START_TOKEN) | |
3421 | return ptype_get_idx(0); | |
3422 | ||
3423 | pt = v; | |
3424 | nxt = pt->list.next; | |
3425 | if (pt->type == htons(ETH_P_ALL)) { | |
3426 | if (nxt != &ptype_all) | |
3427 | goto found; | |
3428 | hash = 0; | |
3429 | nxt = ptype_base[0].next; | |
3430 | } else | |
82d8a867 | 3431 | hash = ntohs(pt->type) & PTYPE_HASH_MASK; |
0e1256ff SH |
3432 | |
3433 | while (nxt == &ptype_base[hash]) { | |
82d8a867 | 3434 | if (++hash >= PTYPE_HASH_SIZE) |
0e1256ff SH |
3435 | return NULL; |
3436 | nxt = ptype_base[hash].next; | |
3437 | } | |
3438 | found: | |
3439 | return list_entry(nxt, struct packet_type, list); | |
3440 | } | |
3441 | ||
3442 | static void ptype_seq_stop(struct seq_file *seq, void *v) | |
72348a42 | 3443 | __releases(RCU) |
0e1256ff SH |
3444 | { |
3445 | rcu_read_unlock(); | |
3446 | } | |
3447 | ||
0e1256ff SH |
3448 | static int ptype_seq_show(struct seq_file *seq, void *v) |
3449 | { | |
3450 | struct packet_type *pt = v; | |
3451 | ||
3452 | if (v == SEQ_START_TOKEN) | |
3453 | seq_puts(seq, "Type Device Function\n"); | |
c346dca1 | 3454 | else if (pt->dev == NULL || dev_net(pt->dev) == seq_file_net(seq)) { |
0e1256ff SH |
3455 | if (pt->type == htons(ETH_P_ALL)) |
3456 | seq_puts(seq, "ALL "); | |
3457 | else | |
3458 | seq_printf(seq, "%04x", ntohs(pt->type)); | |
3459 | ||
908cd2da AD |
3460 | seq_printf(seq, " %-8s %pF\n", |
3461 | pt->dev ? pt->dev->name : "", pt->func); | |
0e1256ff SH |
3462 | } |
3463 | ||
3464 | return 0; | |
3465 | } | |
3466 | ||
3467 | static const struct seq_operations ptype_seq_ops = { | |
3468 | .start = ptype_seq_start, | |
3469 | .next = ptype_seq_next, | |
3470 | .stop = ptype_seq_stop, | |
3471 | .show = ptype_seq_show, | |
3472 | }; | |
3473 | ||
3474 | static int ptype_seq_open(struct inode *inode, struct file *file) | |
3475 | { | |
2feb27db PE |
3476 | return seq_open_net(inode, file, &ptype_seq_ops, |
3477 | sizeof(struct seq_net_private)); | |
0e1256ff SH |
3478 | } |
3479 | ||
3480 | static const struct file_operations ptype_seq_fops = { | |
3481 | .owner = THIS_MODULE, | |
3482 | .open = ptype_seq_open, | |
3483 | .read = seq_read, | |
3484 | .llseek = seq_lseek, | |
2feb27db | 3485 | .release = seq_release_net, |
0e1256ff SH |
3486 | }; |
3487 | ||
3488 | ||
4665079c | 3489 | static int __net_init dev_proc_net_init(struct net *net) |
1da177e4 LT |
3490 | { |
3491 | int rc = -ENOMEM; | |
3492 | ||
881d966b | 3493 | if (!proc_net_fops_create(net, "dev", S_IRUGO, &dev_seq_fops)) |
1da177e4 | 3494 | goto out; |
881d966b | 3495 | if (!proc_net_fops_create(net, "softnet_stat", S_IRUGO, &softnet_seq_fops)) |
1da177e4 | 3496 | goto out_dev; |
881d966b | 3497 | if (!proc_net_fops_create(net, "ptype", S_IRUGO, &ptype_seq_fops)) |
457c4cbc | 3498 | goto out_softnet; |
0e1256ff | 3499 | |
881d966b | 3500 | if (wext_proc_init(net)) |
457c4cbc | 3501 | goto out_ptype; |
1da177e4 LT |
3502 | rc = 0; |
3503 | out: | |
3504 | return rc; | |
457c4cbc | 3505 | out_ptype: |
881d966b | 3506 | proc_net_remove(net, "ptype"); |
1da177e4 | 3507 | out_softnet: |
881d966b | 3508 | proc_net_remove(net, "softnet_stat"); |
1da177e4 | 3509 | out_dev: |
881d966b | 3510 | proc_net_remove(net, "dev"); |
1da177e4 LT |
3511 | goto out; |
3512 | } | |
881d966b | 3513 | |
4665079c | 3514 | static void __net_exit dev_proc_net_exit(struct net *net) |
881d966b EB |
3515 | { |
3516 | wext_proc_exit(net); | |
3517 | ||
3518 | proc_net_remove(net, "ptype"); | |
3519 | proc_net_remove(net, "softnet_stat"); | |
3520 | proc_net_remove(net, "dev"); | |
3521 | } | |
3522 | ||
022cbae6 | 3523 | static struct pernet_operations __net_initdata dev_proc_ops = { |
881d966b EB |
3524 | .init = dev_proc_net_init, |
3525 | .exit = dev_proc_net_exit, | |
3526 | }; | |
3527 | ||
3528 | static int __init dev_proc_init(void) | |
3529 | { | |
3530 | return register_pernet_subsys(&dev_proc_ops); | |
3531 | } | |
1da177e4 LT |
3532 | #else |
3533 | #define dev_proc_init() 0 | |
3534 | #endif /* CONFIG_PROC_FS */ | |
3535 | ||
3536 | ||
3537 | /** | |
3538 | * netdev_set_master - set up master/slave pair | |
3539 | * @slave: slave device | |
3540 | * @master: new master device | |
3541 | * | |
3542 | * Changes the master device of the slave. Pass %NULL to break the | |
3543 | * bonding. The caller must hold the RTNL semaphore. On a failure | |
3544 | * a negative errno code is returned. On success the reference counts | |
3545 | * are adjusted, %RTM_NEWLINK is sent to the routing socket and the | |
3546 | * function returns zero. | |
3547 | */ | |
3548 | int netdev_set_master(struct net_device *slave, struct net_device *master) | |
3549 | { | |
3550 | struct net_device *old = slave->master; | |
3551 | ||
3552 | ASSERT_RTNL(); | |
3553 | ||
3554 | if (master) { | |
3555 | if (old) | |
3556 | return -EBUSY; | |
3557 | dev_hold(master); | |
3558 | } | |
3559 | ||
3560 | slave->master = master; | |
4ec93edb | 3561 | |
1da177e4 LT |
3562 | synchronize_net(); |
3563 | ||
3564 | if (old) | |
3565 | dev_put(old); | |
3566 | ||
3567 | if (master) | |
3568 | slave->flags |= IFF_SLAVE; | |
3569 | else | |
3570 | slave->flags &= ~IFF_SLAVE; | |
3571 | ||
3572 | rtmsg_ifinfo(RTM_NEWLINK, slave, IFF_SLAVE); | |
3573 | return 0; | |
3574 | } | |
d1b19dff | 3575 | EXPORT_SYMBOL(netdev_set_master); |
1da177e4 | 3576 | |
b6c40d68 PM |
3577 | static void dev_change_rx_flags(struct net_device *dev, int flags) |
3578 | { | |
d314774c SH |
3579 | const struct net_device_ops *ops = dev->netdev_ops; |
3580 | ||
3581 | if ((dev->flags & IFF_UP) && ops->ndo_change_rx_flags) | |
3582 | ops->ndo_change_rx_flags(dev, flags); | |
b6c40d68 PM |
3583 | } |
3584 | ||
dad9b335 | 3585 | static int __dev_set_promiscuity(struct net_device *dev, int inc) |
1da177e4 LT |
3586 | { |
3587 | unsigned short old_flags = dev->flags; | |
8192b0c4 DH |
3588 | uid_t uid; |
3589 | gid_t gid; | |
1da177e4 | 3590 | |
24023451 PM |
3591 | ASSERT_RTNL(); |
3592 | ||
dad9b335 WC |
3593 | dev->flags |= IFF_PROMISC; |
3594 | dev->promiscuity += inc; | |
3595 | if (dev->promiscuity == 0) { | |
3596 | /* | |
3597 | * Avoid overflow. | |
3598 | * If inc causes overflow, untouch promisc and return error. | |
3599 | */ | |
3600 | if (inc < 0) | |
3601 | dev->flags &= ~IFF_PROMISC; | |
3602 | else { | |
3603 | dev->promiscuity -= inc; | |
3604 | printk(KERN_WARNING "%s: promiscuity touches roof, " | |
3605 | "set promiscuity failed, promiscuity feature " | |
3606 | "of device might be broken.\n", dev->name); | |
3607 | return -EOVERFLOW; | |
3608 | } | |
3609 | } | |
52609c0b | 3610 | if (dev->flags != old_flags) { |
1da177e4 LT |
3611 | printk(KERN_INFO "device %s %s promiscuous mode\n", |
3612 | dev->name, (dev->flags & IFF_PROMISC) ? "entered" : | |
4ec93edb | 3613 | "left"); |
8192b0c4 DH |
3614 | if (audit_enabled) { |
3615 | current_uid_gid(&uid, &gid); | |
7759db82 KHK |
3616 | audit_log(current->audit_context, GFP_ATOMIC, |
3617 | AUDIT_ANOM_PROMISCUOUS, | |
3618 | "dev=%s prom=%d old_prom=%d auid=%u uid=%u gid=%u ses=%u", | |
3619 | dev->name, (dev->flags & IFF_PROMISC), | |
3620 | (old_flags & IFF_PROMISC), | |
3621 | audit_get_loginuid(current), | |
8192b0c4 | 3622 | uid, gid, |
7759db82 | 3623 | audit_get_sessionid(current)); |
8192b0c4 | 3624 | } |
24023451 | 3625 | |
b6c40d68 | 3626 | dev_change_rx_flags(dev, IFF_PROMISC); |
1da177e4 | 3627 | } |
dad9b335 | 3628 | return 0; |
1da177e4 LT |
3629 | } |
3630 | ||
4417da66 PM |
3631 | /** |
3632 | * dev_set_promiscuity - update promiscuity count on a device | |
3633 | * @dev: device | |
3634 | * @inc: modifier | |
3635 | * | |
3636 | * Add or remove promiscuity from a device. While the count in the device | |
3637 | * remains above zero the interface remains promiscuous. Once it hits zero | |
3638 | * the device reverts back to normal filtering operation. A negative inc | |
3639 | * value is used to drop promiscuity on the device. | |
dad9b335 | 3640 | * Return 0 if successful or a negative errno code on error. |
4417da66 | 3641 | */ |
dad9b335 | 3642 | int dev_set_promiscuity(struct net_device *dev, int inc) |
4417da66 PM |
3643 | { |
3644 | unsigned short old_flags = dev->flags; | |
dad9b335 | 3645 | int err; |
4417da66 | 3646 | |
dad9b335 | 3647 | err = __dev_set_promiscuity(dev, inc); |
4b5a698e | 3648 | if (err < 0) |
dad9b335 | 3649 | return err; |
4417da66 PM |
3650 | if (dev->flags != old_flags) |
3651 | dev_set_rx_mode(dev); | |
dad9b335 | 3652 | return err; |
4417da66 | 3653 | } |
d1b19dff | 3654 | EXPORT_SYMBOL(dev_set_promiscuity); |
4417da66 | 3655 | |
1da177e4 LT |
3656 | /** |
3657 | * dev_set_allmulti - update allmulti count on a device | |
3658 | * @dev: device | |
3659 | * @inc: modifier | |
3660 | * | |
3661 | * Add or remove reception of all multicast frames to a device. While the | |
3662 | * count in the device remains above zero the interface remains listening | |
3663 | * to all interfaces. Once it hits zero the device reverts back to normal | |
3664 | * filtering operation. A negative @inc value is used to drop the counter | |
3665 | * when releasing a resource needing all multicasts. | |
dad9b335 | 3666 | * Return 0 if successful or a negative errno code on error. |
1da177e4 LT |
3667 | */ |
3668 | ||
dad9b335 | 3669 | int dev_set_allmulti(struct net_device *dev, int inc) |
1da177e4 LT |
3670 | { |
3671 | unsigned short old_flags = dev->flags; | |
3672 | ||
24023451 PM |
3673 | ASSERT_RTNL(); |
3674 | ||
1da177e4 | 3675 | dev->flags |= IFF_ALLMULTI; |
dad9b335 WC |
3676 | dev->allmulti += inc; |
3677 | if (dev->allmulti == 0) { | |
3678 | /* | |
3679 | * Avoid overflow. | |
3680 | * If inc causes overflow, untouch allmulti and return error. | |
3681 | */ | |
3682 | if (inc < 0) | |
3683 | dev->flags &= ~IFF_ALLMULTI; | |
3684 | else { | |
3685 | dev->allmulti -= inc; | |
3686 | printk(KERN_WARNING "%s: allmulti touches roof, " | |
3687 | "set allmulti failed, allmulti feature of " | |
3688 | "device might be broken.\n", dev->name); | |
3689 | return -EOVERFLOW; | |
3690 | } | |
3691 | } | |
24023451 | 3692 | if (dev->flags ^ old_flags) { |
b6c40d68 | 3693 | dev_change_rx_flags(dev, IFF_ALLMULTI); |
4417da66 | 3694 | dev_set_rx_mode(dev); |
24023451 | 3695 | } |
dad9b335 | 3696 | return 0; |
4417da66 | 3697 | } |
d1b19dff | 3698 | EXPORT_SYMBOL(dev_set_allmulti); |
4417da66 PM |
3699 | |
3700 | /* | |
3701 | * Upload unicast and multicast address lists to device and | |
3702 | * configure RX filtering. When the device doesn't support unicast | |
53ccaae1 | 3703 | * filtering it is put in promiscuous mode while unicast addresses |
4417da66 PM |
3704 | * are present. |
3705 | */ | |
3706 | void __dev_set_rx_mode(struct net_device *dev) | |
3707 | { | |
d314774c SH |
3708 | const struct net_device_ops *ops = dev->netdev_ops; |
3709 | ||
4417da66 PM |
3710 | /* dev_open will call this function so the list will stay sane. */ |
3711 | if (!(dev->flags&IFF_UP)) | |
3712 | return; | |
3713 | ||
3714 | if (!netif_device_present(dev)) | |
40b77c94 | 3715 | return; |
4417da66 | 3716 | |
d314774c SH |
3717 | if (ops->ndo_set_rx_mode) |
3718 | ops->ndo_set_rx_mode(dev); | |
4417da66 PM |
3719 | else { |
3720 | /* Unicast addresses changes may only happen under the rtnl, | |
3721 | * therefore calling __dev_set_promiscuity here is safe. | |
3722 | */ | |
32e7bfc4 | 3723 | if (!netdev_uc_empty(dev) && !dev->uc_promisc) { |
4417da66 PM |
3724 | __dev_set_promiscuity(dev, 1); |
3725 | dev->uc_promisc = 1; | |
32e7bfc4 | 3726 | } else if (netdev_uc_empty(dev) && dev->uc_promisc) { |
4417da66 PM |
3727 | __dev_set_promiscuity(dev, -1); |
3728 | dev->uc_promisc = 0; | |
3729 | } | |
3730 | ||
d314774c SH |
3731 | if (ops->ndo_set_multicast_list) |
3732 | ops->ndo_set_multicast_list(dev); | |
4417da66 PM |
3733 | } |
3734 | } | |
3735 | ||
3736 | void dev_set_rx_mode(struct net_device *dev) | |
3737 | { | |
b9e40857 | 3738 | netif_addr_lock_bh(dev); |
4417da66 | 3739 | __dev_set_rx_mode(dev); |
b9e40857 | 3740 | netif_addr_unlock_bh(dev); |
1da177e4 LT |
3741 | } |
3742 | ||
f001fde5 JP |
3743 | /* hw addresses list handling functions */ |
3744 | ||
31278e71 JP |
3745 | static int __hw_addr_add(struct netdev_hw_addr_list *list, unsigned char *addr, |
3746 | int addr_len, unsigned char addr_type) | |
f001fde5 JP |
3747 | { |
3748 | struct netdev_hw_addr *ha; | |
3749 | int alloc_size; | |
3750 | ||
3751 | if (addr_len > MAX_ADDR_LEN) | |
3752 | return -EINVAL; | |
3753 | ||
31278e71 | 3754 | list_for_each_entry(ha, &list->list, list) { |
ccffad25 JP |
3755 | if (!memcmp(ha->addr, addr, addr_len) && |
3756 | ha->type == addr_type) { | |
3757 | ha->refcount++; | |
3758 | return 0; | |
3759 | } | |
3760 | } | |
3761 | ||
3762 | ||
f001fde5 JP |
3763 | alloc_size = sizeof(*ha); |
3764 | if (alloc_size < L1_CACHE_BYTES) | |
3765 | alloc_size = L1_CACHE_BYTES; | |
3766 | ha = kmalloc(alloc_size, GFP_ATOMIC); | |
3767 | if (!ha) | |
3768 | return -ENOMEM; | |
3769 | memcpy(ha->addr, addr, addr_len); | |
3770 | ha->type = addr_type; | |
ccffad25 JP |
3771 | ha->refcount = 1; |
3772 | ha->synced = false; | |
31278e71 JP |
3773 | list_add_tail_rcu(&ha->list, &list->list); |
3774 | list->count++; | |
f001fde5 JP |
3775 | return 0; |
3776 | } | |
3777 | ||
3778 | static void ha_rcu_free(struct rcu_head *head) | |
3779 | { | |
3780 | struct netdev_hw_addr *ha; | |
3781 | ||
3782 | ha = container_of(head, struct netdev_hw_addr, rcu_head); | |
3783 | kfree(ha); | |
3784 | } | |
3785 | ||
31278e71 JP |
3786 | static int __hw_addr_del(struct netdev_hw_addr_list *list, unsigned char *addr, |
3787 | int addr_len, unsigned char addr_type) | |
f001fde5 JP |
3788 | { |
3789 | struct netdev_hw_addr *ha; | |
f001fde5 | 3790 | |
31278e71 | 3791 | list_for_each_entry(ha, &list->list, list) { |
ccffad25 | 3792 | if (!memcmp(ha->addr, addr, addr_len) && |
f001fde5 | 3793 | (ha->type == addr_type || !addr_type)) { |
ccffad25 JP |
3794 | if (--ha->refcount) |
3795 | return 0; | |
f001fde5 JP |
3796 | list_del_rcu(&ha->list); |
3797 | call_rcu(&ha->rcu_head, ha_rcu_free); | |
31278e71 | 3798 | list->count--; |
f001fde5 JP |
3799 | return 0; |
3800 | } | |
3801 | } | |
3802 | return -ENOENT; | |
3803 | } | |
3804 | ||
31278e71 JP |
3805 | static int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list, |
3806 | struct netdev_hw_addr_list *from_list, | |
3807 | int addr_len, | |
ccffad25 | 3808 | unsigned char addr_type) |
f001fde5 JP |
3809 | { |
3810 | int err; | |
3811 | struct netdev_hw_addr *ha, *ha2; | |
3812 | unsigned char type; | |
3813 | ||
31278e71 | 3814 | list_for_each_entry(ha, &from_list->list, list) { |
f001fde5 | 3815 | type = addr_type ? addr_type : ha->type; |
31278e71 | 3816 | err = __hw_addr_add(to_list, ha->addr, addr_len, type); |
f001fde5 JP |
3817 | if (err) |
3818 | goto unroll; | |
3819 | } | |
3820 | return 0; | |
3821 | ||
3822 | unroll: | |
31278e71 | 3823 | list_for_each_entry(ha2, &from_list->list, list) { |
f001fde5 JP |
3824 | if (ha2 == ha) |
3825 | break; | |
3826 | type = addr_type ? addr_type : ha2->type; | |
31278e71 | 3827 | __hw_addr_del(to_list, ha2->addr, addr_len, type); |
f001fde5 JP |
3828 | } |
3829 | return err; | |
3830 | } | |
3831 | ||
31278e71 JP |
3832 | static void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list, |
3833 | struct netdev_hw_addr_list *from_list, | |
3834 | int addr_len, | |
ccffad25 | 3835 | unsigned char addr_type) |
f001fde5 JP |
3836 | { |
3837 | struct netdev_hw_addr *ha; | |
3838 | unsigned char type; | |
3839 | ||
31278e71 | 3840 | list_for_each_entry(ha, &from_list->list, list) { |
f001fde5 | 3841 | type = addr_type ? addr_type : ha->type; |
31278e71 | 3842 | __hw_addr_del(to_list, ha->addr, addr_len, addr_type); |
ccffad25 JP |
3843 | } |
3844 | } | |
3845 | ||
31278e71 JP |
3846 | static int __hw_addr_sync(struct netdev_hw_addr_list *to_list, |
3847 | struct netdev_hw_addr_list *from_list, | |
ccffad25 JP |
3848 | int addr_len) |
3849 | { | |
3850 | int err = 0; | |
3851 | struct netdev_hw_addr *ha, *tmp; | |
3852 | ||
31278e71 | 3853 | list_for_each_entry_safe(ha, tmp, &from_list->list, list) { |
ccffad25 | 3854 | if (!ha->synced) { |
31278e71 | 3855 | err = __hw_addr_add(to_list, ha->addr, |
ccffad25 JP |
3856 | addr_len, ha->type); |
3857 | if (err) | |
3858 | break; | |
3859 | ha->synced = true; | |
3860 | ha->refcount++; | |
3861 | } else if (ha->refcount == 1) { | |
31278e71 JP |
3862 | __hw_addr_del(to_list, ha->addr, addr_len, ha->type); |
3863 | __hw_addr_del(from_list, ha->addr, addr_len, ha->type); | |
ccffad25 | 3864 | } |
f001fde5 | 3865 | } |
ccffad25 | 3866 | return err; |
f001fde5 JP |
3867 | } |
3868 | ||
31278e71 JP |
3869 | static void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, |
3870 | struct netdev_hw_addr_list *from_list, | |
ccffad25 JP |
3871 | int addr_len) |
3872 | { | |
3873 | struct netdev_hw_addr *ha, *tmp; | |
3874 | ||
31278e71 | 3875 | list_for_each_entry_safe(ha, tmp, &from_list->list, list) { |
ccffad25 | 3876 | if (ha->synced) { |
31278e71 | 3877 | __hw_addr_del(to_list, ha->addr, |
ccffad25 JP |
3878 | addr_len, ha->type); |
3879 | ha->synced = false; | |
31278e71 | 3880 | __hw_addr_del(from_list, ha->addr, |
ccffad25 JP |
3881 | addr_len, ha->type); |
3882 | } | |
3883 | } | |
3884 | } | |
3885 | ||
31278e71 | 3886 | static void __hw_addr_flush(struct netdev_hw_addr_list *list) |
f001fde5 JP |
3887 | { |
3888 | struct netdev_hw_addr *ha, *tmp; | |
3889 | ||
31278e71 | 3890 | list_for_each_entry_safe(ha, tmp, &list->list, list) { |
f001fde5 JP |
3891 | list_del_rcu(&ha->list); |
3892 | call_rcu(&ha->rcu_head, ha_rcu_free); | |
3893 | } | |
31278e71 JP |
3894 | list->count = 0; |
3895 | } | |
3896 | ||
3897 | static void __hw_addr_init(struct netdev_hw_addr_list *list) | |
3898 | { | |
3899 | INIT_LIST_HEAD(&list->list); | |
3900 | list->count = 0; | |
f001fde5 JP |
3901 | } |
3902 | ||
3903 | /* Device addresses handling functions */ | |
3904 | ||
3905 | static void dev_addr_flush(struct net_device *dev) | |
3906 | { | |
3907 | /* rtnl_mutex must be held here */ | |
3908 | ||
31278e71 | 3909 | __hw_addr_flush(&dev->dev_addrs); |
f001fde5 JP |
3910 | dev->dev_addr = NULL; |
3911 | } | |
3912 | ||
3913 | static int dev_addr_init(struct net_device *dev) | |
3914 | { | |
3915 | unsigned char addr[MAX_ADDR_LEN]; | |
3916 | struct netdev_hw_addr *ha; | |
3917 | int err; | |
3918 | ||
3919 | /* rtnl_mutex must be held here */ | |
3920 | ||
31278e71 | 3921 | __hw_addr_init(&dev->dev_addrs); |
0c27922e | 3922 | memset(addr, 0, sizeof(addr)); |
31278e71 | 3923 | err = __hw_addr_add(&dev->dev_addrs, addr, sizeof(addr), |
f001fde5 JP |
3924 | NETDEV_HW_ADDR_T_LAN); |
3925 | if (!err) { | |
3926 | /* | |
3927 | * Get the first (previously created) address from the list | |
3928 | * and set dev_addr pointer to this location. | |
3929 | */ | |
31278e71 | 3930 | ha = list_first_entry(&dev->dev_addrs.list, |
f001fde5 JP |
3931 | struct netdev_hw_addr, list); |
3932 | dev->dev_addr = ha->addr; | |
3933 | } | |
3934 | return err; | |
3935 | } | |
3936 | ||
3937 | /** | |
3938 | * dev_addr_add - Add a device address | |
3939 | * @dev: device | |
3940 | * @addr: address to add | |
3941 | * @addr_type: address type | |
3942 | * | |
3943 | * Add a device address to the device or increase the reference count if | |
3944 | * it already exists. | |
3945 | * | |
3946 | * The caller must hold the rtnl_mutex. | |
3947 | */ | |
3948 | int dev_addr_add(struct net_device *dev, unsigned char *addr, | |
3949 | unsigned char addr_type) | |
3950 | { | |
3951 | int err; | |
3952 | ||
3953 | ASSERT_RTNL(); | |
3954 | ||
31278e71 | 3955 | err = __hw_addr_add(&dev->dev_addrs, addr, dev->addr_len, addr_type); |
f001fde5 JP |
3956 | if (!err) |
3957 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); | |
3958 | return err; | |
3959 | } | |
3960 | EXPORT_SYMBOL(dev_addr_add); | |
3961 | ||
3962 | /** | |
3963 | * dev_addr_del - Release a device address. | |
3964 | * @dev: device | |
3965 | * @addr: address to delete | |
3966 | * @addr_type: address type | |
3967 | * | |
3968 | * Release reference to a device address and remove it from the device | |
3969 | * if the reference count drops to zero. | |
3970 | * | |
3971 | * The caller must hold the rtnl_mutex. | |
3972 | */ | |
3973 | int dev_addr_del(struct net_device *dev, unsigned char *addr, | |
3974 | unsigned char addr_type) | |
3975 | { | |
3976 | int err; | |
ccffad25 | 3977 | struct netdev_hw_addr *ha; |
f001fde5 JP |
3978 | |
3979 | ASSERT_RTNL(); | |
3980 | ||
ccffad25 JP |
3981 | /* |
3982 | * We can not remove the first address from the list because | |
3983 | * dev->dev_addr points to that. | |
3984 | */ | |
31278e71 JP |
3985 | ha = list_first_entry(&dev->dev_addrs.list, |
3986 | struct netdev_hw_addr, list); | |
ccffad25 JP |
3987 | if (ha->addr == dev->dev_addr && ha->refcount == 1) |
3988 | return -ENOENT; | |
3989 | ||
31278e71 | 3990 | err = __hw_addr_del(&dev->dev_addrs, addr, dev->addr_len, |
ccffad25 | 3991 | addr_type); |
f001fde5 JP |
3992 | if (!err) |
3993 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); | |
3994 | return err; | |
3995 | } | |
3996 | EXPORT_SYMBOL(dev_addr_del); | |
3997 | ||
3998 | /** | |
3999 | * dev_addr_add_multiple - Add device addresses from another device | |
4000 | * @to_dev: device to which addresses will be added | |
4001 | * @from_dev: device from which addresses will be added | |
4002 | * @addr_type: address type - 0 means type will be used from from_dev | |
4003 | * | |
4004 | * Add device addresses of the one device to another. | |
4005 | ** | |
4006 | * The caller must hold the rtnl_mutex. | |
4007 | */ | |
4008 | int dev_addr_add_multiple(struct net_device *to_dev, | |
4009 | struct net_device *from_dev, | |
4010 | unsigned char addr_type) | |
4011 | { | |
4012 | int err; | |
4013 | ||
4014 | ASSERT_RTNL(); | |
4015 | ||
4016 | if (from_dev->addr_len != to_dev->addr_len) | |
4017 | return -EINVAL; | |
31278e71 | 4018 | err = __hw_addr_add_multiple(&to_dev->dev_addrs, &from_dev->dev_addrs, |
ccffad25 | 4019 | to_dev->addr_len, addr_type); |
f001fde5 JP |
4020 | if (!err) |
4021 | call_netdevice_notifiers(NETDEV_CHANGEADDR, to_dev); | |
4022 | return err; | |
4023 | } | |
4024 | EXPORT_SYMBOL(dev_addr_add_multiple); | |
4025 | ||
4026 | /** | |
4027 | * dev_addr_del_multiple - Delete device addresses by another device | |
4028 | * @to_dev: device where the addresses will be deleted | |
4029 | * @from_dev: device by which addresses the addresses will be deleted | |
4030 | * @addr_type: address type - 0 means type will used from from_dev | |
4031 | * | |
4032 | * Deletes addresses in to device by the list of addresses in from device. | |
4033 | * | |
4034 | * The caller must hold the rtnl_mutex. | |
4035 | */ | |
4036 | int dev_addr_del_multiple(struct net_device *to_dev, | |
4037 | struct net_device *from_dev, | |
4038 | unsigned char addr_type) | |
4039 | { | |
4040 | ASSERT_RTNL(); | |
4041 | ||
4042 | if (from_dev->addr_len != to_dev->addr_len) | |
4043 | return -EINVAL; | |
31278e71 | 4044 | __hw_addr_del_multiple(&to_dev->dev_addrs, &from_dev->dev_addrs, |
ccffad25 | 4045 | to_dev->addr_len, addr_type); |
f001fde5 JP |
4046 | call_netdevice_notifiers(NETDEV_CHANGEADDR, to_dev); |
4047 | return 0; | |
4048 | } | |
4049 | EXPORT_SYMBOL(dev_addr_del_multiple); | |
4050 | ||
31278e71 | 4051 | /* multicast addresses handling functions */ |
f001fde5 | 4052 | |
61cbc2fc PM |
4053 | int __dev_addr_delete(struct dev_addr_list **list, int *count, |
4054 | void *addr, int alen, int glbl) | |
bf742482 PM |
4055 | { |
4056 | struct dev_addr_list *da; | |
4057 | ||
4058 | for (; (da = *list) != NULL; list = &da->next) { | |
4059 | if (memcmp(da->da_addr, addr, da->da_addrlen) == 0 && | |
4060 | alen == da->da_addrlen) { | |
4061 | if (glbl) { | |
4062 | int old_glbl = da->da_gusers; | |
4063 | da->da_gusers = 0; | |
4064 | if (old_glbl == 0) | |
4065 | break; | |
4066 | } | |
4067 | if (--da->da_users) | |
4068 | return 0; | |
4069 | ||
4070 | *list = da->next; | |
4071 | kfree(da); | |
61cbc2fc | 4072 | (*count)--; |
bf742482 PM |
4073 | return 0; |
4074 | } | |
4075 | } | |
4076 | return -ENOENT; | |
4077 | } | |
4078 | ||
61cbc2fc PM |
4079 | int __dev_addr_add(struct dev_addr_list **list, int *count, |
4080 | void *addr, int alen, int glbl) | |
bf742482 PM |
4081 | { |
4082 | struct dev_addr_list *da; | |
4083 | ||
4084 | for (da = *list; da != NULL; da = da->next) { | |
4085 | if (memcmp(da->da_addr, addr, da->da_addrlen) == 0 && | |
4086 | da->da_addrlen == alen) { | |
4087 | if (glbl) { | |
4088 | int old_glbl = da->da_gusers; | |
4089 | da->da_gusers = 1; | |
4090 | if (old_glbl) | |
4091 | return 0; | |
4092 | } | |
4093 | da->da_users++; | |
4094 | return 0; | |
4095 | } | |
4096 | } | |
4097 | ||
12aa343a | 4098 | da = kzalloc(sizeof(*da), GFP_ATOMIC); |
bf742482 PM |
4099 | if (da == NULL) |
4100 | return -ENOMEM; | |
4101 | memcpy(da->da_addr, addr, alen); | |
4102 | da->da_addrlen = alen; | |
4103 | da->da_users = 1; | |
4104 | da->da_gusers = glbl ? 1 : 0; | |
4105 | da->next = *list; | |
4106 | *list = da; | |
61cbc2fc | 4107 | (*count)++; |
bf742482 PM |
4108 | return 0; |
4109 | } | |
4110 | ||
4417da66 PM |
4111 | /** |
4112 | * dev_unicast_delete - Release secondary unicast address. | |
4113 | * @dev: device | |
0ed72ec4 | 4114 | * @addr: address to delete |
4417da66 PM |
4115 | * |
4116 | * Release reference to a secondary unicast address and remove it | |
0ed72ec4 | 4117 | * from the device if the reference count drops to zero. |
4417da66 PM |
4118 | * |
4119 | * The caller must hold the rtnl_mutex. | |
4120 | */ | |
ccffad25 | 4121 | int dev_unicast_delete(struct net_device *dev, void *addr) |
4417da66 PM |
4122 | { |
4123 | int err; | |
4124 | ||
4125 | ASSERT_RTNL(); | |
4126 | ||
a6ac65db | 4127 | netif_addr_lock_bh(dev); |
31278e71 JP |
4128 | err = __hw_addr_del(&dev->uc, addr, dev->addr_len, |
4129 | NETDEV_HW_ADDR_T_UNICAST); | |
61cbc2fc | 4130 | if (!err) |
4417da66 | 4131 | __dev_set_rx_mode(dev); |
a6ac65db | 4132 | netif_addr_unlock_bh(dev); |
4417da66 PM |
4133 | return err; |
4134 | } | |
4135 | EXPORT_SYMBOL(dev_unicast_delete); | |
4136 | ||
4137 | /** | |
4138 | * dev_unicast_add - add a secondary unicast address | |
4139 | * @dev: device | |
5dbaec5d | 4140 | * @addr: address to add |
4417da66 PM |
4141 | * |
4142 | * Add a secondary unicast address to the device or increase | |
4143 | * the reference count if it already exists. | |
4144 | * | |
4145 | * The caller must hold the rtnl_mutex. | |
4146 | */ | |
ccffad25 | 4147 | int dev_unicast_add(struct net_device *dev, void *addr) |
4417da66 PM |
4148 | { |
4149 | int err; | |
4150 | ||
4151 | ASSERT_RTNL(); | |
4152 | ||
a6ac65db | 4153 | netif_addr_lock_bh(dev); |
31278e71 JP |
4154 | err = __hw_addr_add(&dev->uc, addr, dev->addr_len, |
4155 | NETDEV_HW_ADDR_T_UNICAST); | |
61cbc2fc | 4156 | if (!err) |
4417da66 | 4157 | __dev_set_rx_mode(dev); |
a6ac65db | 4158 | netif_addr_unlock_bh(dev); |
4417da66 PM |
4159 | return err; |
4160 | } | |
4161 | EXPORT_SYMBOL(dev_unicast_add); | |
4162 | ||
e83a2ea8 CL |
4163 | int __dev_addr_sync(struct dev_addr_list **to, int *to_count, |
4164 | struct dev_addr_list **from, int *from_count) | |
4165 | { | |
4166 | struct dev_addr_list *da, *next; | |
4167 | int err = 0; | |
4168 | ||
4169 | da = *from; | |
4170 | while (da != NULL) { | |
4171 | next = da->next; | |
4172 | if (!da->da_synced) { | |
4173 | err = __dev_addr_add(to, to_count, | |
4174 | da->da_addr, da->da_addrlen, 0); | |
4175 | if (err < 0) | |
4176 | break; | |
4177 | da->da_synced = 1; | |
4178 | da->da_users++; | |
4179 | } else if (da->da_users == 1) { | |
4180 | __dev_addr_delete(to, to_count, | |
4181 | da->da_addr, da->da_addrlen, 0); | |
4182 | __dev_addr_delete(from, from_count, | |
4183 | da->da_addr, da->da_addrlen, 0); | |
4184 | } | |
4185 | da = next; | |
4186 | } | |
4187 | return err; | |
4188 | } | |
c4029083 | 4189 | EXPORT_SYMBOL_GPL(__dev_addr_sync); |
e83a2ea8 CL |
4190 | |
4191 | void __dev_addr_unsync(struct dev_addr_list **to, int *to_count, | |
4192 | struct dev_addr_list **from, int *from_count) | |
4193 | { | |
4194 | struct dev_addr_list *da, *next; | |
4195 | ||
4196 | da = *from; | |
4197 | while (da != NULL) { | |
4198 | next = da->next; | |
4199 | if (da->da_synced) { | |
4200 | __dev_addr_delete(to, to_count, | |
4201 | da->da_addr, da->da_addrlen, 0); | |
4202 | da->da_synced = 0; | |
4203 | __dev_addr_delete(from, from_count, | |
4204 | da->da_addr, da->da_addrlen, 0); | |
4205 | } | |
4206 | da = next; | |
4207 | } | |
4208 | } | |
c4029083 | 4209 | EXPORT_SYMBOL_GPL(__dev_addr_unsync); |
e83a2ea8 CL |
4210 | |
4211 | /** | |
4212 | * dev_unicast_sync - Synchronize device's unicast list to another device | |
4213 | * @to: destination device | |
4214 | * @from: source device | |
4215 | * | |
4216 | * Add newly added addresses to the destination device and release | |
a6ac65db JP |
4217 | * addresses that have no users left. The source device must be |
4218 | * locked by netif_tx_lock_bh. | |
e83a2ea8 CL |
4219 | * |
4220 | * This function is intended to be called from the dev->set_rx_mode | |
4221 | * function of layered software devices. | |
4222 | */ | |
4223 | int dev_unicast_sync(struct net_device *to, struct net_device *from) | |
4224 | { | |
4225 | int err = 0; | |
4226 | ||
ccffad25 JP |
4227 | if (to->addr_len != from->addr_len) |
4228 | return -EINVAL; | |
4229 | ||
a6ac65db | 4230 | netif_addr_lock_bh(to); |
31278e71 | 4231 | err = __hw_addr_sync(&to->uc, &from->uc, to->addr_len); |
e83a2ea8 CL |
4232 | if (!err) |
4233 | __dev_set_rx_mode(to); | |
a6ac65db | 4234 | netif_addr_unlock_bh(to); |
e83a2ea8 CL |
4235 | return err; |
4236 | } | |
4237 | EXPORT_SYMBOL(dev_unicast_sync); | |
4238 | ||
4239 | /** | |
bc2cda1e | 4240 | * dev_unicast_unsync - Remove synchronized addresses from the destination device |
e83a2ea8 CL |
4241 | * @to: destination device |
4242 | * @from: source device | |
4243 | * | |
4244 | * Remove all addresses that were added to the destination device by | |
4245 | * dev_unicast_sync(). This function is intended to be called from the | |
4246 | * dev->stop function of layered software devices. | |
4247 | */ | |
4248 | void dev_unicast_unsync(struct net_device *to, struct net_device *from) | |
4249 | { | |
ccffad25 JP |
4250 | if (to->addr_len != from->addr_len) |
4251 | return; | |
e83a2ea8 | 4252 | |
a6ac65db JP |
4253 | netif_addr_lock_bh(from); |
4254 | netif_addr_lock(to); | |
31278e71 | 4255 | __hw_addr_unsync(&to->uc, &from->uc, to->addr_len); |
ccffad25 | 4256 | __dev_set_rx_mode(to); |
a6ac65db JP |
4257 | netif_addr_unlock(to); |
4258 | netif_addr_unlock_bh(from); | |
e83a2ea8 CL |
4259 | } |
4260 | EXPORT_SYMBOL(dev_unicast_unsync); | |
4261 | ||
ccffad25 JP |
4262 | static void dev_unicast_flush(struct net_device *dev) |
4263 | { | |
a6ac65db | 4264 | netif_addr_lock_bh(dev); |
31278e71 | 4265 | __hw_addr_flush(&dev->uc); |
a6ac65db | 4266 | netif_addr_unlock_bh(dev); |
ccffad25 JP |
4267 | } |
4268 | ||
4269 | static void dev_unicast_init(struct net_device *dev) | |
4270 | { | |
31278e71 | 4271 | __hw_addr_init(&dev->uc); |
ccffad25 JP |
4272 | } |
4273 | ||
4274 | ||
12972621 DC |
4275 | static void __dev_addr_discard(struct dev_addr_list **list) |
4276 | { | |
4277 | struct dev_addr_list *tmp; | |
4278 | ||
4279 | while (*list != NULL) { | |
4280 | tmp = *list; | |
4281 | *list = tmp->next; | |
4282 | if (tmp->da_users > tmp->da_gusers) | |
4283 | printk("__dev_addr_discard: address leakage! " | |
4284 | "da_users=%d\n", tmp->da_users); | |
4285 | kfree(tmp); | |
4286 | } | |
4287 | } | |
4288 | ||
26cc2522 | 4289 | static void dev_addr_discard(struct net_device *dev) |
4417da66 | 4290 | { |
b9e40857 | 4291 | netif_addr_lock_bh(dev); |
26cc2522 | 4292 | |
456ad75c | 4293 | __dev_addr_discard(&dev->mc_list); |
4cd24eaf | 4294 | netdev_mc_count(dev) = 0; |
26cc2522 | 4295 | |
b9e40857 | 4296 | netif_addr_unlock_bh(dev); |
456ad75c DC |
4297 | } |
4298 | ||
f0db275a SH |
4299 | /** |
4300 | * dev_get_flags - get flags reported to userspace | |
4301 | * @dev: device | |
4302 | * | |
4303 | * Get the combination of flag bits exported through APIs to userspace. | |
4304 | */ | |
1da177e4 LT |
4305 | unsigned dev_get_flags(const struct net_device *dev) |
4306 | { | |
4307 | unsigned flags; | |
4308 | ||
4309 | flags = (dev->flags & ~(IFF_PROMISC | | |
4310 | IFF_ALLMULTI | | |
b00055aa SR |
4311 | IFF_RUNNING | |
4312 | IFF_LOWER_UP | | |
4313 | IFF_DORMANT)) | | |
1da177e4 LT |
4314 | (dev->gflags & (IFF_PROMISC | |
4315 | IFF_ALLMULTI)); | |
4316 | ||
b00055aa SR |
4317 | if (netif_running(dev)) { |
4318 | if (netif_oper_up(dev)) | |
4319 | flags |= IFF_RUNNING; | |
4320 | if (netif_carrier_ok(dev)) | |
4321 | flags |= IFF_LOWER_UP; | |
4322 | if (netif_dormant(dev)) | |
4323 | flags |= IFF_DORMANT; | |
4324 | } | |
1da177e4 LT |
4325 | |
4326 | return flags; | |
4327 | } | |
d1b19dff | 4328 | EXPORT_SYMBOL(dev_get_flags); |
1da177e4 | 4329 | |
bd380811 | 4330 | int __dev_change_flags(struct net_device *dev, unsigned int flags) |
1da177e4 | 4331 | { |
1da177e4 | 4332 | int old_flags = dev->flags; |
bd380811 | 4333 | int ret; |
1da177e4 | 4334 | |
24023451 PM |
4335 | ASSERT_RTNL(); |
4336 | ||
1da177e4 LT |
4337 | /* |
4338 | * Set the flags on our device. | |
4339 | */ | |
4340 | ||
4341 | dev->flags = (flags & (IFF_DEBUG | IFF_NOTRAILERS | IFF_NOARP | | |
4342 | IFF_DYNAMIC | IFF_MULTICAST | IFF_PORTSEL | | |
4343 | IFF_AUTOMEDIA)) | | |
4344 | (dev->flags & (IFF_UP | IFF_VOLATILE | IFF_PROMISC | | |
4345 | IFF_ALLMULTI)); | |
4346 | ||
4347 | /* | |
4348 | * Load in the correct multicast list now the flags have changed. | |
4349 | */ | |
4350 | ||
b6c40d68 PM |
4351 | if ((old_flags ^ flags) & IFF_MULTICAST) |
4352 | dev_change_rx_flags(dev, IFF_MULTICAST); | |
24023451 | 4353 | |
4417da66 | 4354 | dev_set_rx_mode(dev); |
1da177e4 LT |
4355 | |
4356 | /* | |
4357 | * Have we downed the interface. We handle IFF_UP ourselves | |
4358 | * according to user attempts to set it, rather than blindly | |
4359 | * setting it. | |
4360 | */ | |
4361 | ||
4362 | ret = 0; | |
4363 | if ((old_flags ^ flags) & IFF_UP) { /* Bit is different ? */ | |
bd380811 | 4364 | ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev); |
1da177e4 LT |
4365 | |
4366 | if (!ret) | |
4417da66 | 4367 | dev_set_rx_mode(dev); |
1da177e4 LT |
4368 | } |
4369 | ||
1da177e4 | 4370 | if ((flags ^ dev->gflags) & IFF_PROMISC) { |
d1b19dff ED |
4371 | int inc = (flags & IFF_PROMISC) ? 1 : -1; |
4372 | ||
1da177e4 LT |
4373 | dev->gflags ^= IFF_PROMISC; |
4374 | dev_set_promiscuity(dev, inc); | |
4375 | } | |
4376 | ||
4377 | /* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI | |
4378 | is important. Some (broken) drivers set IFF_PROMISC, when | |
4379 | IFF_ALLMULTI is requested not asking us and not reporting. | |
4380 | */ | |
4381 | if ((flags ^ dev->gflags) & IFF_ALLMULTI) { | |
d1b19dff ED |
4382 | int inc = (flags & IFF_ALLMULTI) ? 1 : -1; |
4383 | ||
1da177e4 LT |
4384 | dev->gflags ^= IFF_ALLMULTI; |
4385 | dev_set_allmulti(dev, inc); | |
4386 | } | |
4387 | ||
bd380811 PM |
4388 | return ret; |
4389 | } | |
4390 | ||
4391 | void __dev_notify_flags(struct net_device *dev, unsigned int old_flags) | |
4392 | { | |
4393 | unsigned int changes = dev->flags ^ old_flags; | |
4394 | ||
4395 | if (changes & IFF_UP) { | |
4396 | if (dev->flags & IFF_UP) | |
4397 | call_netdevice_notifiers(NETDEV_UP, dev); | |
4398 | else | |
4399 | call_netdevice_notifiers(NETDEV_DOWN, dev); | |
4400 | } | |
4401 | ||
4402 | if (dev->flags & IFF_UP && | |
4403 | (changes & ~(IFF_UP | IFF_PROMISC | IFF_ALLMULTI | IFF_VOLATILE))) | |
4404 | call_netdevice_notifiers(NETDEV_CHANGE, dev); | |
4405 | } | |
4406 | ||
4407 | /** | |
4408 | * dev_change_flags - change device settings | |
4409 | * @dev: device | |
4410 | * @flags: device state flags | |
4411 | * | |
4412 | * Change settings on device based state flags. The flags are | |
4413 | * in the userspace exported format. | |
4414 | */ | |
4415 | int dev_change_flags(struct net_device *dev, unsigned flags) | |
4416 | { | |
4417 | int ret, changes; | |
4418 | int old_flags = dev->flags; | |
4419 | ||
4420 | ret = __dev_change_flags(dev, flags); | |
4421 | if (ret < 0) | |
4422 | return ret; | |
4423 | ||
4424 | changes = old_flags ^ dev->flags; | |
7c355f53 TG |
4425 | if (changes) |
4426 | rtmsg_ifinfo(RTM_NEWLINK, dev, changes); | |
1da177e4 | 4427 | |
bd380811 | 4428 | __dev_notify_flags(dev, old_flags); |
1da177e4 LT |
4429 | return ret; |
4430 | } | |
d1b19dff | 4431 | EXPORT_SYMBOL(dev_change_flags); |
1da177e4 | 4432 | |
f0db275a SH |
4433 | /** |
4434 | * dev_set_mtu - Change maximum transfer unit | |
4435 | * @dev: device | |
4436 | * @new_mtu: new transfer unit | |
4437 | * | |
4438 | * Change the maximum transfer size of the network device. | |
4439 | */ | |
1da177e4 LT |
4440 | int dev_set_mtu(struct net_device *dev, int new_mtu) |
4441 | { | |
d314774c | 4442 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
4443 | int err; |
4444 | ||
4445 | if (new_mtu == dev->mtu) | |
4446 | return 0; | |
4447 | ||
4448 | /* MTU must be positive. */ | |
4449 | if (new_mtu < 0) | |
4450 | return -EINVAL; | |
4451 | ||
4452 | if (!netif_device_present(dev)) | |
4453 | return -ENODEV; | |
4454 | ||
4455 | err = 0; | |
d314774c SH |
4456 | if (ops->ndo_change_mtu) |
4457 | err = ops->ndo_change_mtu(dev, new_mtu); | |
1da177e4 LT |
4458 | else |
4459 | dev->mtu = new_mtu; | |
d314774c | 4460 | |
1da177e4 | 4461 | if (!err && dev->flags & IFF_UP) |
056925ab | 4462 | call_netdevice_notifiers(NETDEV_CHANGEMTU, dev); |
1da177e4 LT |
4463 | return err; |
4464 | } | |
d1b19dff | 4465 | EXPORT_SYMBOL(dev_set_mtu); |
1da177e4 | 4466 | |
f0db275a SH |
4467 | /** |
4468 | * dev_set_mac_address - Change Media Access Control Address | |
4469 | * @dev: device | |
4470 | * @sa: new address | |
4471 | * | |
4472 | * Change the hardware (MAC) address of the device | |
4473 | */ | |
1da177e4 LT |
4474 | int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa) |
4475 | { | |
d314774c | 4476 | const struct net_device_ops *ops = dev->netdev_ops; |
1da177e4 LT |
4477 | int err; |
4478 | ||
d314774c | 4479 | if (!ops->ndo_set_mac_address) |
1da177e4 LT |
4480 | return -EOPNOTSUPP; |
4481 | if (sa->sa_family != dev->type) | |
4482 | return -EINVAL; | |
4483 | if (!netif_device_present(dev)) | |
4484 | return -ENODEV; | |
d314774c | 4485 | err = ops->ndo_set_mac_address(dev, sa); |
1da177e4 | 4486 | if (!err) |
056925ab | 4487 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); |
1da177e4 LT |
4488 | return err; |
4489 | } | |
d1b19dff | 4490 | EXPORT_SYMBOL(dev_set_mac_address); |
1da177e4 LT |
4491 | |
4492 | /* | |
3710becf | 4493 | * Perform the SIOCxIFxxx calls, inside rcu_read_lock() |
1da177e4 | 4494 | */ |
14e3e079 | 4495 | static int dev_ifsioc_locked(struct net *net, struct ifreq *ifr, unsigned int cmd) |
1da177e4 LT |
4496 | { |
4497 | int err; | |
3710becf | 4498 | struct net_device *dev = dev_get_by_name_rcu(net, ifr->ifr_name); |
1da177e4 LT |
4499 | |
4500 | if (!dev) | |
4501 | return -ENODEV; | |
4502 | ||
4503 | switch (cmd) { | |
d1b19dff ED |
4504 | case SIOCGIFFLAGS: /* Get interface flags */ |
4505 | ifr->ifr_flags = (short) dev_get_flags(dev); | |
4506 | return 0; | |
1da177e4 | 4507 | |
d1b19dff ED |
4508 | case SIOCGIFMETRIC: /* Get the metric on the interface |
4509 | (currently unused) */ | |
4510 | ifr->ifr_metric = 0; | |
4511 | return 0; | |
1da177e4 | 4512 | |
d1b19dff ED |
4513 | case SIOCGIFMTU: /* Get the MTU of a device */ |
4514 | ifr->ifr_mtu = dev->mtu; | |
4515 | return 0; | |
1da177e4 | 4516 | |
d1b19dff ED |
4517 | case SIOCGIFHWADDR: |
4518 | if (!dev->addr_len) | |
4519 | memset(ifr->ifr_hwaddr.sa_data, 0, sizeof ifr->ifr_hwaddr.sa_data); | |
4520 | else | |
4521 | memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr, | |
4522 | min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); | |
4523 | ifr->ifr_hwaddr.sa_family = dev->type; | |
4524 | return 0; | |
1da177e4 | 4525 | |
d1b19dff ED |
4526 | case SIOCGIFSLAVE: |
4527 | err = -EINVAL; | |
4528 | break; | |
14e3e079 | 4529 | |
d1b19dff ED |
4530 | case SIOCGIFMAP: |
4531 | ifr->ifr_map.mem_start = dev->mem_start; | |
4532 | ifr->ifr_map.mem_end = dev->mem_end; | |
4533 | ifr->ifr_map.base_addr = dev->base_addr; | |
4534 | ifr->ifr_map.irq = dev->irq; | |
4535 | ifr->ifr_map.dma = dev->dma; | |
4536 | ifr->ifr_map.port = dev->if_port; | |
4537 | return 0; | |
14e3e079 | 4538 | |
d1b19dff ED |
4539 | case SIOCGIFINDEX: |
4540 | ifr->ifr_ifindex = dev->ifindex; | |
4541 | return 0; | |
14e3e079 | 4542 | |
d1b19dff ED |
4543 | case SIOCGIFTXQLEN: |
4544 | ifr->ifr_qlen = dev->tx_queue_len; | |
4545 | return 0; | |
14e3e079 | 4546 | |
d1b19dff ED |
4547 | default: |
4548 | /* dev_ioctl() should ensure this case | |
4549 | * is never reached | |
4550 | */ | |
4551 | WARN_ON(1); | |
4552 | err = -EINVAL; | |
4553 | break; | |
14e3e079 JG |
4554 | |
4555 | } | |
4556 | return err; | |
4557 | } | |
4558 | ||
4559 | /* | |
4560 | * Perform the SIOCxIFxxx calls, inside rtnl_lock() | |
4561 | */ | |
4562 | static int dev_ifsioc(struct net *net, struct ifreq *ifr, unsigned int cmd) | |
4563 | { | |
4564 | int err; | |
4565 | struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name); | |
5f2f6da7 | 4566 | const struct net_device_ops *ops; |
14e3e079 JG |
4567 | |
4568 | if (!dev) | |
4569 | return -ENODEV; | |
4570 | ||
5f2f6da7 JP |
4571 | ops = dev->netdev_ops; |
4572 | ||
14e3e079 | 4573 | switch (cmd) { |
d1b19dff ED |
4574 | case SIOCSIFFLAGS: /* Set interface flags */ |
4575 | return dev_change_flags(dev, ifr->ifr_flags); | |
14e3e079 | 4576 | |
d1b19dff ED |
4577 | case SIOCSIFMETRIC: /* Set the metric on the interface |
4578 | (currently unused) */ | |
4579 | return -EOPNOTSUPP; | |
14e3e079 | 4580 | |
d1b19dff ED |
4581 | case SIOCSIFMTU: /* Set the MTU of a device */ |
4582 | return dev_set_mtu(dev, ifr->ifr_mtu); | |
1da177e4 | 4583 | |
d1b19dff ED |
4584 | case SIOCSIFHWADDR: |
4585 | return dev_set_mac_address(dev, &ifr->ifr_hwaddr); | |
1da177e4 | 4586 | |
d1b19dff ED |
4587 | case SIOCSIFHWBROADCAST: |
4588 | if (ifr->ifr_hwaddr.sa_family != dev->type) | |
4589 | return -EINVAL; | |
4590 | memcpy(dev->broadcast, ifr->ifr_hwaddr.sa_data, | |
4591 | min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); | |
4592 | call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); | |
4593 | return 0; | |
1da177e4 | 4594 | |
d1b19dff ED |
4595 | case SIOCSIFMAP: |
4596 | if (ops->ndo_set_config) { | |
1da177e4 LT |
4597 | if (!netif_device_present(dev)) |
4598 | return -ENODEV; | |
d1b19dff ED |
4599 | return ops->ndo_set_config(dev, &ifr->ifr_map); |
4600 | } | |
4601 | return -EOPNOTSUPP; | |
1da177e4 | 4602 | |
d1b19dff ED |
4603 | case SIOCADDMULTI: |
4604 | if ((!ops->ndo_set_multicast_list && !ops->ndo_set_rx_mode) || | |
4605 | ifr->ifr_hwaddr.sa_family != AF_UNSPEC) | |
4606 | return -EINVAL; | |
4607 | if (!netif_device_present(dev)) | |
4608 | return -ENODEV; | |
4609 | return dev_mc_add(dev, ifr->ifr_hwaddr.sa_data, | |
4610 | dev->addr_len, 1); | |
4611 | ||
4612 | case SIOCDELMULTI: | |
4613 | if ((!ops->ndo_set_multicast_list && !ops->ndo_set_rx_mode) || | |
4614 | ifr->ifr_hwaddr.sa_family != AF_UNSPEC) | |
4615 | return -EINVAL; | |
4616 | if (!netif_device_present(dev)) | |
4617 | return -ENODEV; | |
4618 | return dev_mc_delete(dev, ifr->ifr_hwaddr.sa_data, | |
4619 | dev->addr_len, 1); | |
1da177e4 | 4620 | |
d1b19dff ED |
4621 | case SIOCSIFTXQLEN: |
4622 | if (ifr->ifr_qlen < 0) | |
4623 | return -EINVAL; | |
4624 | dev->tx_queue_len = ifr->ifr_qlen; | |
4625 | return 0; | |
1da177e4 | 4626 | |
d1b19dff ED |
4627 | case SIOCSIFNAME: |
4628 | ifr->ifr_newname[IFNAMSIZ-1] = '\0'; | |
4629 | return dev_change_name(dev, ifr->ifr_newname); | |
1da177e4 | 4630 | |
d1b19dff ED |
4631 | /* |
4632 | * Unknown or private ioctl | |
4633 | */ | |
4634 | default: | |
4635 | if ((cmd >= SIOCDEVPRIVATE && | |
4636 | cmd <= SIOCDEVPRIVATE + 15) || | |
4637 | cmd == SIOCBONDENSLAVE || | |
4638 | cmd == SIOCBONDRELEASE || | |
4639 | cmd == SIOCBONDSETHWADDR || | |
4640 | cmd == SIOCBONDSLAVEINFOQUERY || | |
4641 | cmd == SIOCBONDINFOQUERY || | |
4642 | cmd == SIOCBONDCHANGEACTIVE || | |
4643 | cmd == SIOCGMIIPHY || | |
4644 | cmd == SIOCGMIIREG || | |
4645 | cmd == SIOCSMIIREG || | |
4646 | cmd == SIOCBRADDIF || | |
4647 | cmd == SIOCBRDELIF || | |
4648 | cmd == SIOCSHWTSTAMP || | |
4649 | cmd == SIOCWANDEV) { | |
4650 | err = -EOPNOTSUPP; | |
4651 | if (ops->ndo_do_ioctl) { | |
4652 | if (netif_device_present(dev)) | |
4653 | err = ops->ndo_do_ioctl(dev, ifr, cmd); | |
4654 | else | |
4655 | err = -ENODEV; | |
4656 | } | |
4657 | } else | |
4658 | err = -EINVAL; | |
1da177e4 LT |
4659 | |
4660 | } | |
4661 | return err; | |
4662 | } | |
4663 | ||
4664 | /* | |
4665 | * This function handles all "interface"-type I/O control requests. The actual | |
4666 | * 'doing' part of this is dev_ifsioc above. | |
4667 | */ | |
4668 | ||
4669 | /** | |
4670 | * dev_ioctl - network device ioctl | |
c4ea43c5 | 4671 | * @net: the applicable net namespace |
1da177e4 LT |
4672 | * @cmd: command to issue |
4673 | * @arg: pointer to a struct ifreq in user space | |
4674 | * | |
4675 | * Issue ioctl functions to devices. This is normally called by the | |
4676 | * user space syscall interfaces but can sometimes be useful for | |
4677 | * other purposes. The return value is the return from the syscall if | |
4678 | * positive or a negative errno code on error. | |
4679 | */ | |
4680 | ||
881d966b | 4681 | int dev_ioctl(struct net *net, unsigned int cmd, void __user *arg) |
1da177e4 LT |
4682 | { |
4683 | struct ifreq ifr; | |
4684 | int ret; | |
4685 | char *colon; | |
4686 | ||
4687 | /* One special case: SIOCGIFCONF takes ifconf argument | |
4688 | and requires shared lock, because it sleeps writing | |
4689 | to user space. | |
4690 | */ | |
4691 | ||
4692 | if (cmd == SIOCGIFCONF) { | |
6756ae4b | 4693 | rtnl_lock(); |
881d966b | 4694 | ret = dev_ifconf(net, (char __user *) arg); |
6756ae4b | 4695 | rtnl_unlock(); |
1da177e4 LT |
4696 | return ret; |
4697 | } | |
4698 | if (cmd == SIOCGIFNAME) | |
881d966b | 4699 | return dev_ifname(net, (struct ifreq __user *)arg); |
1da177e4 LT |
4700 | |
4701 | if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) | |
4702 | return -EFAULT; | |
4703 | ||
4704 | ifr.ifr_name[IFNAMSIZ-1] = 0; | |
4705 | ||
4706 | colon = strchr(ifr.ifr_name, ':'); | |
4707 | if (colon) | |
4708 | *colon = 0; | |
4709 | ||
4710 | /* | |
4711 | * See which interface the caller is talking about. | |
4712 | */ | |
4713 | ||
4714 | switch (cmd) { | |
d1b19dff ED |
4715 | /* |
4716 | * These ioctl calls: | |
4717 | * - can be done by all. | |
4718 | * - atomic and do not require locking. | |
4719 | * - return a value | |
4720 | */ | |
4721 | case SIOCGIFFLAGS: | |
4722 | case SIOCGIFMETRIC: | |
4723 | case SIOCGIFMTU: | |
4724 | case SIOCGIFHWADDR: | |
4725 | case SIOCGIFSLAVE: | |
4726 | case SIOCGIFMAP: | |
4727 | case SIOCGIFINDEX: | |
4728 | case SIOCGIFTXQLEN: | |
4729 | dev_load(net, ifr.ifr_name); | |
3710becf | 4730 | rcu_read_lock(); |
d1b19dff | 4731 | ret = dev_ifsioc_locked(net, &ifr, cmd); |
3710becf | 4732 | rcu_read_unlock(); |
d1b19dff ED |
4733 | if (!ret) { |
4734 | if (colon) | |
4735 | *colon = ':'; | |
4736 | if (copy_to_user(arg, &ifr, | |
4737 | sizeof(struct ifreq))) | |
4738 | ret = -EFAULT; | |
4739 | } | |
4740 | return ret; | |
1da177e4 | 4741 | |
d1b19dff ED |
4742 | case SIOCETHTOOL: |
4743 | dev_load(net, ifr.ifr_name); | |
4744 | rtnl_lock(); | |
4745 | ret = dev_ethtool(net, &ifr); | |
4746 | rtnl_unlock(); | |
4747 | if (!ret) { | |
4748 | if (colon) | |
4749 | *colon = ':'; | |
4750 | if (copy_to_user(arg, &ifr, | |
4751 | sizeof(struct ifreq))) | |
4752 | ret = -EFAULT; | |
4753 | } | |
4754 | return ret; | |
1da177e4 | 4755 | |
d1b19dff ED |
4756 | /* |
4757 | * These ioctl calls: | |
4758 | * - require superuser power. | |
4759 | * - require strict serialization. | |
4760 | * - return a value | |
4761 | */ | |
4762 | case SIOCGMIIPHY: | |
4763 | case SIOCGMIIREG: | |
4764 | case SIOCSIFNAME: | |
4765 | if (!capable(CAP_NET_ADMIN)) | |
4766 | return -EPERM; | |
4767 | dev_load(net, ifr.ifr_name); | |
4768 | rtnl_lock(); | |
4769 | ret = dev_ifsioc(net, &ifr, cmd); | |
4770 | rtnl_unlock(); | |
4771 | if (!ret) { | |
4772 | if (colon) | |
4773 | *colon = ':'; | |
4774 | if (copy_to_user(arg, &ifr, | |
4775 | sizeof(struct ifreq))) | |
4776 | ret = -EFAULT; | |
4777 | } | |
4778 | return ret; | |
1da177e4 | 4779 | |
d1b19dff ED |
4780 | /* |
4781 | * These ioctl calls: | |
4782 | * - require superuser power. | |
4783 | * - require strict serialization. | |
4784 | * - do not return a value | |
4785 | */ | |
4786 | case SIOCSIFFLAGS: | |
4787 | case SIOCSIFMETRIC: | |
4788 | case SIOCSIFMTU: | |
4789 | case SIOCSIFMAP: | |
4790 | case SIOCSIFHWADDR: | |
4791 | case SIOCSIFSLAVE: | |
4792 | case SIOCADDMULTI: | |
4793 | case SIOCDELMULTI: | |
4794 | case SIOCSIFHWBROADCAST: | |
4795 | case SIOCSIFTXQLEN: | |
4796 | case SIOCSMIIREG: | |
4797 | case SIOCBONDENSLAVE: | |
4798 | case SIOCBONDRELEASE: | |
4799 | case SIOCBONDSETHWADDR: | |
4800 | case SIOCBONDCHANGEACTIVE: | |
4801 | case SIOCBRADDIF: | |
4802 | case SIOCBRDELIF: | |
4803 | case SIOCSHWTSTAMP: | |
4804 | if (!capable(CAP_NET_ADMIN)) | |
4805 | return -EPERM; | |
4806 | /* fall through */ | |
4807 | case SIOCBONDSLAVEINFOQUERY: | |
4808 | case SIOCBONDINFOQUERY: | |
4809 | dev_load(net, ifr.ifr_name); | |
4810 | rtnl_lock(); | |
4811 | ret = dev_ifsioc(net, &ifr, cmd); | |
4812 | rtnl_unlock(); | |
4813 | return ret; | |
4814 | ||
4815 | case SIOCGIFMEM: | |
4816 | /* Get the per device memory space. We can add this but | |
4817 | * currently do not support it */ | |
4818 | case SIOCSIFMEM: | |
4819 | /* Set the per device memory buffer space. | |
4820 | * Not applicable in our case */ | |
4821 | case SIOCSIFLINK: | |
4822 | return -EINVAL; | |
4823 | ||
4824 | /* | |
4825 | * Unknown or private ioctl. | |
4826 | */ | |
4827 | default: | |
4828 | if (cmd == SIOCWANDEV || | |
4829 | (cmd >= SIOCDEVPRIVATE && | |
4830 | cmd <= SIOCDEVPRIVATE + 15)) { | |
881d966b | 4831 | dev_load(net, ifr.ifr_name); |
1da177e4 | 4832 | rtnl_lock(); |
881d966b | 4833 | ret = dev_ifsioc(net, &ifr, cmd); |
1da177e4 | 4834 | rtnl_unlock(); |
d1b19dff ED |
4835 | if (!ret && copy_to_user(arg, &ifr, |
4836 | sizeof(struct ifreq))) | |
4837 | ret = -EFAULT; | |
1da177e4 | 4838 | return ret; |
d1b19dff ED |
4839 | } |
4840 | /* Take care of Wireless Extensions */ | |
4841 | if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) | |
4842 | return wext_handle_ioctl(net, &ifr, cmd, arg); | |
4843 | return -EINVAL; | |
1da177e4 LT |
4844 | } |
4845 | } | |
4846 | ||
4847 | ||
4848 | /** | |
4849 | * dev_new_index - allocate an ifindex | |
c4ea43c5 | 4850 | * @net: the applicable net namespace |
1da177e4 LT |
4851 | * |
4852 | * Returns a suitable unique value for a new device interface | |
4853 | * number. The caller must hold the rtnl semaphore or the | |
4854 | * dev_base_lock to be sure it remains unique. | |
4855 | */ | |
881d966b | 4856 | static int dev_new_index(struct net *net) |
1da177e4 LT |
4857 | { |
4858 | static int ifindex; | |
4859 | for (;;) { | |
4860 | if (++ifindex <= 0) | |
4861 | ifindex = 1; | |
881d966b | 4862 | if (!__dev_get_by_index(net, ifindex)) |
1da177e4 LT |
4863 | return ifindex; |
4864 | } | |
4865 | } | |
4866 | ||
1da177e4 | 4867 | /* Delayed registration/unregisteration */ |
3b5b34fd | 4868 | static LIST_HEAD(net_todo_list); |
1da177e4 | 4869 | |
6f05f629 | 4870 | static void net_set_todo(struct net_device *dev) |
1da177e4 | 4871 | { |
1da177e4 | 4872 | list_add_tail(&dev->todo_list, &net_todo_list); |
1da177e4 LT |
4873 | } |
4874 | ||
9b5e383c | 4875 | static void rollback_registered_many(struct list_head *head) |
93ee31f1 | 4876 | { |
e93737b0 | 4877 | struct net_device *dev, *tmp; |
9b5e383c | 4878 | |
93ee31f1 DL |
4879 | BUG_ON(dev_boot_phase); |
4880 | ASSERT_RTNL(); | |
4881 | ||
e93737b0 | 4882 | list_for_each_entry_safe(dev, tmp, head, unreg_list) { |
9b5e383c | 4883 | /* Some devices call without registering |
e93737b0 KK |
4884 | * for initialization unwind. Remove those |
4885 | * devices and proceed with the remaining. | |
9b5e383c ED |
4886 | */ |
4887 | if (dev->reg_state == NETREG_UNINITIALIZED) { | |
4888 | pr_debug("unregister_netdevice: device %s/%p never " | |
4889 | "was registered\n", dev->name, dev); | |
93ee31f1 | 4890 | |
9b5e383c | 4891 | WARN_ON(1); |
e93737b0 KK |
4892 | list_del(&dev->unreg_list); |
4893 | continue; | |
9b5e383c | 4894 | } |
93ee31f1 | 4895 | |
9b5e383c | 4896 | BUG_ON(dev->reg_state != NETREG_REGISTERED); |
93ee31f1 | 4897 | |
9b5e383c ED |
4898 | /* If device is running, close it first. */ |
4899 | dev_close(dev); | |
93ee31f1 | 4900 | |
9b5e383c ED |
4901 | /* And unlink it from device chain. */ |
4902 | unlist_netdevice(dev); | |
93ee31f1 | 4903 | |
9b5e383c ED |
4904 | dev->reg_state = NETREG_UNREGISTERING; |
4905 | } | |
93ee31f1 DL |
4906 | |
4907 | synchronize_net(); | |
4908 | ||
9b5e383c ED |
4909 | list_for_each_entry(dev, head, unreg_list) { |
4910 | /* Shutdown queueing discipline. */ | |
4911 | dev_shutdown(dev); | |
93ee31f1 DL |
4912 | |
4913 | ||
9b5e383c ED |
4914 | /* Notify protocols, that we are about to destroy |
4915 | this device. They should clean all the things. | |
4916 | */ | |
4917 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); | |
93ee31f1 | 4918 | |
a2835763 PM |
4919 | if (!dev->rtnl_link_ops || |
4920 | dev->rtnl_link_state == RTNL_LINK_INITIALIZED) | |
4921 | rtmsg_ifinfo(RTM_DELLINK, dev, ~0U); | |
4922 | ||
9b5e383c ED |
4923 | /* |
4924 | * Flush the unicast and multicast chains | |
4925 | */ | |
4926 | dev_unicast_flush(dev); | |
4927 | dev_addr_discard(dev); | |
93ee31f1 | 4928 | |
9b5e383c ED |
4929 | if (dev->netdev_ops->ndo_uninit) |
4930 | dev->netdev_ops->ndo_uninit(dev); | |
93ee31f1 | 4931 | |
9b5e383c ED |
4932 | /* Notifier chain MUST detach us from master device. */ |
4933 | WARN_ON(dev->master); | |
93ee31f1 | 4934 | |
9b5e383c ED |
4935 | /* Remove entries from kobject tree */ |
4936 | netdev_unregister_kobject(dev); | |
4937 | } | |
93ee31f1 | 4938 | |
a5ee1551 | 4939 | /* Process any work delayed until the end of the batch */ |
e5e26d75 | 4940 | dev = list_first_entry(head, struct net_device, unreg_list); |
a5ee1551 | 4941 | call_netdevice_notifiers(NETDEV_UNREGISTER_BATCH, dev); |
93ee31f1 | 4942 | |
a5ee1551 | 4943 | synchronize_net(); |
395264d5 | 4944 | |
a5ee1551 | 4945 | list_for_each_entry(dev, head, unreg_list) |
9b5e383c ED |
4946 | dev_put(dev); |
4947 | } | |
4948 | ||
4949 | static void rollback_registered(struct net_device *dev) | |
4950 | { | |
4951 | LIST_HEAD(single); | |
4952 | ||
4953 | list_add(&dev->unreg_list, &single); | |
4954 | rollback_registered_many(&single); | |
93ee31f1 DL |
4955 | } |
4956 | ||
e8a0464c DM |
4957 | static void __netdev_init_queue_locks_one(struct net_device *dev, |
4958 | struct netdev_queue *dev_queue, | |
4959 | void *_unused) | |
c773e847 DM |
4960 | { |
4961 | spin_lock_init(&dev_queue->_xmit_lock); | |
cf508b12 | 4962 | netdev_set_xmit_lockdep_class(&dev_queue->_xmit_lock, dev->type); |
c773e847 DM |
4963 | dev_queue->xmit_lock_owner = -1; |
4964 | } | |
4965 | ||
4966 | static void netdev_init_queue_locks(struct net_device *dev) | |
4967 | { | |
e8a0464c DM |
4968 | netdev_for_each_tx_queue(dev, __netdev_init_queue_locks_one, NULL); |
4969 | __netdev_init_queue_locks_one(dev, &dev->rx_queue, NULL); | |
c773e847 DM |
4970 | } |
4971 | ||
b63365a2 HX |
4972 | unsigned long netdev_fix_features(unsigned long features, const char *name) |
4973 | { | |
4974 | /* Fix illegal SG+CSUM combinations. */ | |
4975 | if ((features & NETIF_F_SG) && | |
4976 | !(features & NETIF_F_ALL_CSUM)) { | |
4977 | if (name) | |
4978 | printk(KERN_NOTICE "%s: Dropping NETIF_F_SG since no " | |
4979 | "checksum feature.\n", name); | |
4980 | features &= ~NETIF_F_SG; | |
4981 | } | |
4982 | ||
4983 | /* TSO requires that SG is present as well. */ | |
4984 | if ((features & NETIF_F_TSO) && !(features & NETIF_F_SG)) { | |
4985 | if (name) | |
4986 | printk(KERN_NOTICE "%s: Dropping NETIF_F_TSO since no " | |
4987 | "SG feature.\n", name); | |
4988 | features &= ~NETIF_F_TSO; | |
4989 | } | |
4990 | ||
4991 | if (features & NETIF_F_UFO) { | |
4992 | if (!(features & NETIF_F_GEN_CSUM)) { | |
4993 | if (name) | |
4994 | printk(KERN_ERR "%s: Dropping NETIF_F_UFO " | |
4995 | "since no NETIF_F_HW_CSUM feature.\n", | |
4996 | name); | |
4997 | features &= ~NETIF_F_UFO; | |
4998 | } | |
4999 | ||
5000 | if (!(features & NETIF_F_SG)) { | |
5001 | if (name) | |
5002 | printk(KERN_ERR "%s: Dropping NETIF_F_UFO " | |
5003 | "since no NETIF_F_SG feature.\n", name); | |
5004 | features &= ~NETIF_F_UFO; | |
5005 | } | |
5006 | } | |
5007 | ||
5008 | return features; | |
5009 | } | |
5010 | EXPORT_SYMBOL(netdev_fix_features); | |
5011 | ||
fc4a7489 PM |
5012 | /** |
5013 | * netif_stacked_transfer_operstate - transfer operstate | |
5014 | * @rootdev: the root or lower level device to transfer state from | |
5015 | * @dev: the device to transfer operstate to | |
5016 | * | |
5017 | * Transfer operational state from root to device. This is normally | |
5018 | * called when a stacking relationship exists between the root | |
5019 | * device and the device(a leaf device). | |
5020 | */ | |
5021 | void netif_stacked_transfer_operstate(const struct net_device *rootdev, | |
5022 | struct net_device *dev) | |
5023 | { | |
5024 | if (rootdev->operstate == IF_OPER_DORMANT) | |
5025 | netif_dormant_on(dev); | |
5026 | else | |
5027 | netif_dormant_off(dev); | |
5028 | ||
5029 | if (netif_carrier_ok(rootdev)) { | |
5030 | if (!netif_carrier_ok(dev)) | |
5031 | netif_carrier_on(dev); | |
5032 | } else { | |
5033 | if (netif_carrier_ok(dev)) | |
5034 | netif_carrier_off(dev); | |
5035 | } | |
5036 | } | |
5037 | EXPORT_SYMBOL(netif_stacked_transfer_operstate); | |
5038 | ||
1da177e4 LT |
5039 | /** |
5040 | * register_netdevice - register a network device | |
5041 | * @dev: device to register | |
5042 | * | |
5043 | * Take a completed network device structure and add it to the kernel | |
5044 | * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier | |
5045 | * chain. 0 is returned on success. A negative errno code is returned | |
5046 | * on a failure to set up the device, or if the name is a duplicate. | |
5047 | * | |
5048 | * Callers must hold the rtnl semaphore. You may want | |
5049 | * register_netdev() instead of this. | |
5050 | * | |
5051 | * BUGS: | |
5052 | * The locking appears insufficient to guarantee two parallel registers | |
5053 | * will not get the same name. | |
5054 | */ | |
5055 | ||
5056 | int register_netdevice(struct net_device *dev) | |
5057 | { | |
1da177e4 | 5058 | int ret; |
d314774c | 5059 | struct net *net = dev_net(dev); |
1da177e4 LT |
5060 | |
5061 | BUG_ON(dev_boot_phase); | |
5062 | ASSERT_RTNL(); | |
5063 | ||
b17a7c17 SH |
5064 | might_sleep(); |
5065 | ||
1da177e4 LT |
5066 | /* When net_device's are persistent, this will be fatal. */ |
5067 | BUG_ON(dev->reg_state != NETREG_UNINITIALIZED); | |
d314774c | 5068 | BUG_ON(!net); |
1da177e4 | 5069 | |
f1f28aa3 | 5070 | spin_lock_init(&dev->addr_list_lock); |
cf508b12 | 5071 | netdev_set_addr_lockdep_class(dev); |
c773e847 | 5072 | netdev_init_queue_locks(dev); |
1da177e4 | 5073 | |
1da177e4 LT |
5074 | dev->iflink = -1; |
5075 | ||
5076 | /* Init, if this function is available */ | |
d314774c SH |
5077 | if (dev->netdev_ops->ndo_init) { |
5078 | ret = dev->netdev_ops->ndo_init(dev); | |
1da177e4 LT |
5079 | if (ret) { |
5080 | if (ret > 0) | |
5081 | ret = -EIO; | |
90833aa4 | 5082 | goto out; |
1da177e4 LT |
5083 | } |
5084 | } | |
4ec93edb | 5085 | |
d9031024 OP |
5086 | ret = dev_get_valid_name(net, dev->name, dev->name, 0); |
5087 | if (ret) | |
7ce1b0ed | 5088 | goto err_uninit; |
1da177e4 | 5089 | |
881d966b | 5090 | dev->ifindex = dev_new_index(net); |
1da177e4 LT |
5091 | if (dev->iflink == -1) |
5092 | dev->iflink = dev->ifindex; | |
5093 | ||
d212f87b SH |
5094 | /* Fix illegal checksum combinations */ |
5095 | if ((dev->features & NETIF_F_HW_CSUM) && | |
5096 | (dev->features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { | |
5097 | printk(KERN_NOTICE "%s: mixed HW and IP checksum settings.\n", | |
5098 | dev->name); | |
5099 | dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); | |
5100 | } | |
5101 | ||
5102 | if ((dev->features & NETIF_F_NO_CSUM) && | |
5103 | (dev->features & (NETIF_F_HW_CSUM|NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { | |
5104 | printk(KERN_NOTICE "%s: mixed no checksumming and other settings.\n", | |
5105 | dev->name); | |
5106 | dev->features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM|NETIF_F_HW_CSUM); | |
5107 | } | |
5108 | ||
b63365a2 | 5109 | dev->features = netdev_fix_features(dev->features, dev->name); |
1da177e4 | 5110 | |
e5a4a72d LB |
5111 | /* Enable software GSO if SG is supported. */ |
5112 | if (dev->features & NETIF_F_SG) | |
5113 | dev->features |= NETIF_F_GSO; | |
5114 | ||
aaf8cdc3 | 5115 | netdev_initialize_kobject(dev); |
7ffbe3fd JB |
5116 | |
5117 | ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev); | |
5118 | ret = notifier_to_errno(ret); | |
5119 | if (ret) | |
5120 | goto err_uninit; | |
5121 | ||
8b41d188 | 5122 | ret = netdev_register_kobject(dev); |
b17a7c17 | 5123 | if (ret) |
7ce1b0ed | 5124 | goto err_uninit; |
b17a7c17 SH |
5125 | dev->reg_state = NETREG_REGISTERED; |
5126 | ||
1da177e4 LT |
5127 | /* |
5128 | * Default initial state at registry is that the | |
5129 | * device is present. | |
5130 | */ | |
5131 | ||
5132 | set_bit(__LINK_STATE_PRESENT, &dev->state); | |
5133 | ||
1da177e4 | 5134 | dev_init_scheduler(dev); |
1da177e4 | 5135 | dev_hold(dev); |
ce286d32 | 5136 | list_netdevice(dev); |
1da177e4 LT |
5137 | |
5138 | /* Notify protocols, that a new device appeared. */ | |
056925ab | 5139 | ret = call_netdevice_notifiers(NETDEV_REGISTER, dev); |
fcc5a03a | 5140 | ret = notifier_to_errno(ret); |
93ee31f1 DL |
5141 | if (ret) { |
5142 | rollback_registered(dev); | |
5143 | dev->reg_state = NETREG_UNREGISTERED; | |
5144 | } | |
d90a909e EB |
5145 | /* |
5146 | * Prevent userspace races by waiting until the network | |
5147 | * device is fully setup before sending notifications. | |
5148 | */ | |
a2835763 PM |
5149 | if (!dev->rtnl_link_ops || |
5150 | dev->rtnl_link_state == RTNL_LINK_INITIALIZED) | |
5151 | rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U); | |
1da177e4 LT |
5152 | |
5153 | out: | |
5154 | return ret; | |
7ce1b0ed HX |
5155 | |
5156 | err_uninit: | |
d314774c SH |
5157 | if (dev->netdev_ops->ndo_uninit) |
5158 | dev->netdev_ops->ndo_uninit(dev); | |
7ce1b0ed | 5159 | goto out; |
1da177e4 | 5160 | } |
d1b19dff | 5161 | EXPORT_SYMBOL(register_netdevice); |
1da177e4 | 5162 | |
937f1ba5 BH |
5163 | /** |
5164 | * init_dummy_netdev - init a dummy network device for NAPI | |
5165 | * @dev: device to init | |
5166 | * | |
5167 | * This takes a network device structure and initialize the minimum | |
5168 | * amount of fields so it can be used to schedule NAPI polls without | |
5169 | * registering a full blown interface. This is to be used by drivers | |
5170 | * that need to tie several hardware interfaces to a single NAPI | |
5171 | * poll scheduler due to HW limitations. | |
5172 | */ | |
5173 | int init_dummy_netdev(struct net_device *dev) | |
5174 | { | |
5175 | /* Clear everything. Note we don't initialize spinlocks | |
5176 | * are they aren't supposed to be taken by any of the | |
5177 | * NAPI code and this dummy netdev is supposed to be | |
5178 | * only ever used for NAPI polls | |
5179 | */ | |
5180 | memset(dev, 0, sizeof(struct net_device)); | |
5181 | ||
5182 | /* make sure we BUG if trying to hit standard | |
5183 | * register/unregister code path | |
5184 | */ | |
5185 | dev->reg_state = NETREG_DUMMY; | |
5186 | ||
5187 | /* initialize the ref count */ | |
5188 | atomic_set(&dev->refcnt, 1); | |
5189 | ||
5190 | /* NAPI wants this */ | |
5191 | INIT_LIST_HEAD(&dev->napi_list); | |
5192 | ||
5193 | /* a dummy interface is started by default */ | |
5194 | set_bit(__LINK_STATE_PRESENT, &dev->state); | |
5195 | set_bit(__LINK_STATE_START, &dev->state); | |
5196 | ||
5197 | return 0; | |
5198 | } | |
5199 | EXPORT_SYMBOL_GPL(init_dummy_netdev); | |
5200 | ||
5201 | ||
1da177e4 LT |
5202 | /** |
5203 | * register_netdev - register a network device | |
5204 | * @dev: device to register | |
5205 | * | |
5206 | * Take a completed network device structure and add it to the kernel | |
5207 | * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier | |
5208 | * chain. 0 is returned on success. A negative errno code is returned | |
5209 | * on a failure to set up the device, or if the name is a duplicate. | |
5210 | * | |
38b4da38 | 5211 | * This is a wrapper around register_netdevice that takes the rtnl semaphore |
1da177e4 LT |
5212 | * and expands the device name if you passed a format string to |
5213 | * alloc_netdev. | |
5214 | */ | |
5215 | int register_netdev(struct net_device *dev) | |
5216 | { | |
5217 | int err; | |
5218 | ||
5219 | rtnl_lock(); | |
5220 | ||
5221 | /* | |
5222 | * If the name is a format string the caller wants us to do a | |
5223 | * name allocation. | |
5224 | */ | |
5225 | if (strchr(dev->name, '%')) { | |
5226 | err = dev_alloc_name(dev, dev->name); | |
5227 | if (err < 0) | |
5228 | goto out; | |
5229 | } | |
4ec93edb | 5230 | |
1da177e4 LT |
5231 | err = register_netdevice(dev); |
5232 | out: | |
5233 | rtnl_unlock(); | |
5234 | return err; | |
5235 | } | |
5236 | EXPORT_SYMBOL(register_netdev); | |
5237 | ||
5238 | /* | |
5239 | * netdev_wait_allrefs - wait until all references are gone. | |
5240 | * | |
5241 | * This is called when unregistering network devices. | |
5242 | * | |
5243 | * Any protocol or device that holds a reference should register | |
5244 | * for netdevice notification, and cleanup and put back the | |
5245 | * reference if they receive an UNREGISTER event. | |
5246 | * We can get stuck here if buggy protocols don't correctly | |
4ec93edb | 5247 | * call dev_put. |
1da177e4 LT |
5248 | */ |
5249 | static void netdev_wait_allrefs(struct net_device *dev) | |
5250 | { | |
5251 | unsigned long rebroadcast_time, warning_time; | |
5252 | ||
e014debe ED |
5253 | linkwatch_forget_dev(dev); |
5254 | ||
1da177e4 LT |
5255 | rebroadcast_time = warning_time = jiffies; |
5256 | while (atomic_read(&dev->refcnt) != 0) { | |
5257 | if (time_after(jiffies, rebroadcast_time + 1 * HZ)) { | |
6756ae4b | 5258 | rtnl_lock(); |
1da177e4 LT |
5259 | |
5260 | /* Rebroadcast unregister notification */ | |
056925ab | 5261 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); |
a5ee1551 | 5262 | /* don't resend NETDEV_UNREGISTER_BATCH, _BATCH users |
395264d5 | 5263 | * should have already handle it the first time */ |
1da177e4 LT |
5264 | |
5265 | if (test_bit(__LINK_STATE_LINKWATCH_PENDING, | |
5266 | &dev->state)) { | |
5267 | /* We must not have linkwatch events | |
5268 | * pending on unregister. If this | |
5269 | * happens, we simply run the queue | |
5270 | * unscheduled, resulting in a noop | |
5271 | * for this device. | |
5272 | */ | |
5273 | linkwatch_run_queue(); | |
5274 | } | |
5275 | ||
6756ae4b | 5276 | __rtnl_unlock(); |
1da177e4 LT |
5277 | |
5278 | rebroadcast_time = jiffies; | |
5279 | } | |
5280 | ||
5281 | msleep(250); | |
5282 | ||
5283 | if (time_after(jiffies, warning_time + 10 * HZ)) { | |
5284 | printk(KERN_EMERG "unregister_netdevice: " | |
5285 | "waiting for %s to become free. Usage " | |
5286 | "count = %d\n", | |
5287 | dev->name, atomic_read(&dev->refcnt)); | |
5288 | warning_time = jiffies; | |
5289 | } | |
5290 | } | |
5291 | } | |
5292 | ||
5293 | /* The sequence is: | |
5294 | * | |
5295 | * rtnl_lock(); | |
5296 | * ... | |
5297 | * register_netdevice(x1); | |
5298 | * register_netdevice(x2); | |
5299 | * ... | |
5300 | * unregister_netdevice(y1); | |
5301 | * unregister_netdevice(y2); | |
5302 | * ... | |
5303 | * rtnl_unlock(); | |
5304 | * free_netdev(y1); | |
5305 | * free_netdev(y2); | |
5306 | * | |
58ec3b4d | 5307 | * We are invoked by rtnl_unlock(). |
1da177e4 | 5308 | * This allows us to deal with problems: |
b17a7c17 | 5309 | * 1) We can delete sysfs objects which invoke hotplug |
1da177e4 LT |
5310 | * without deadlocking with linkwatch via keventd. |
5311 | * 2) Since we run with the RTNL semaphore not held, we can sleep | |
5312 | * safely in order to wait for the netdev refcnt to drop to zero. | |
58ec3b4d HX |
5313 | * |
5314 | * We must not return until all unregister events added during | |
5315 | * the interval the lock was held have been completed. | |
1da177e4 | 5316 | */ |
1da177e4 LT |
5317 | void netdev_run_todo(void) |
5318 | { | |
626ab0e6 | 5319 | struct list_head list; |
1da177e4 | 5320 | |
1da177e4 | 5321 | /* Snapshot list, allow later requests */ |
626ab0e6 | 5322 | list_replace_init(&net_todo_list, &list); |
58ec3b4d HX |
5323 | |
5324 | __rtnl_unlock(); | |
626ab0e6 | 5325 | |
1da177e4 LT |
5326 | while (!list_empty(&list)) { |
5327 | struct net_device *dev | |
e5e26d75 | 5328 | = list_first_entry(&list, struct net_device, todo_list); |
1da177e4 LT |
5329 | list_del(&dev->todo_list); |
5330 | ||
b17a7c17 SH |
5331 | if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) { |
5332 | printk(KERN_ERR "network todo '%s' but state %d\n", | |
5333 | dev->name, dev->reg_state); | |
5334 | dump_stack(); | |
5335 | continue; | |
5336 | } | |
1da177e4 | 5337 | |
b17a7c17 | 5338 | dev->reg_state = NETREG_UNREGISTERED; |
1da177e4 | 5339 | |
6e583ce5 SH |
5340 | on_each_cpu(flush_backlog, dev, 1); |
5341 | ||
b17a7c17 | 5342 | netdev_wait_allrefs(dev); |
1da177e4 | 5343 | |
b17a7c17 SH |
5344 | /* paranoia */ |
5345 | BUG_ON(atomic_read(&dev->refcnt)); | |
547b792c IJ |
5346 | WARN_ON(dev->ip_ptr); |
5347 | WARN_ON(dev->ip6_ptr); | |
5348 | WARN_ON(dev->dn_ptr); | |
1da177e4 | 5349 | |
b17a7c17 SH |
5350 | if (dev->destructor) |
5351 | dev->destructor(dev); | |
9093bbb2 SH |
5352 | |
5353 | /* Free network device */ | |
5354 | kobject_put(&dev->dev.kobj); | |
1da177e4 | 5355 | } |
1da177e4 LT |
5356 | } |
5357 | ||
d83345ad ED |
5358 | /** |
5359 | * dev_txq_stats_fold - fold tx_queues stats | |
5360 | * @dev: device to get statistics from | |
5361 | * @stats: struct net_device_stats to hold results | |
5362 | */ | |
5363 | void dev_txq_stats_fold(const struct net_device *dev, | |
5364 | struct net_device_stats *stats) | |
5365 | { | |
5366 | unsigned long tx_bytes = 0, tx_packets = 0, tx_dropped = 0; | |
5367 | unsigned int i; | |
5368 | struct netdev_queue *txq; | |
5369 | ||
5370 | for (i = 0; i < dev->num_tx_queues; i++) { | |
5371 | txq = netdev_get_tx_queue(dev, i); | |
5372 | tx_bytes += txq->tx_bytes; | |
5373 | tx_packets += txq->tx_packets; | |
5374 | tx_dropped += txq->tx_dropped; | |
5375 | } | |
5376 | if (tx_bytes || tx_packets || tx_dropped) { | |
5377 | stats->tx_bytes = tx_bytes; | |
5378 | stats->tx_packets = tx_packets; | |
5379 | stats->tx_dropped = tx_dropped; | |
5380 | } | |
5381 | } | |
5382 | EXPORT_SYMBOL(dev_txq_stats_fold); | |
5383 | ||
eeda3fd6 SH |
5384 | /** |
5385 | * dev_get_stats - get network device statistics | |
5386 | * @dev: device to get statistics from | |
5387 | * | |
5388 | * Get network statistics from device. The device driver may provide | |
5389 | * its own method by setting dev->netdev_ops->get_stats; otherwise | |
5390 | * the internal statistics structure is used. | |
5391 | */ | |
5392 | const struct net_device_stats *dev_get_stats(struct net_device *dev) | |
7004bf25 | 5393 | { |
eeda3fd6 SH |
5394 | const struct net_device_ops *ops = dev->netdev_ops; |
5395 | ||
5396 | if (ops->ndo_get_stats) | |
5397 | return ops->ndo_get_stats(dev); | |
d83345ad ED |
5398 | |
5399 | dev_txq_stats_fold(dev, &dev->stats); | |
5400 | return &dev->stats; | |
c45d286e | 5401 | } |
eeda3fd6 | 5402 | EXPORT_SYMBOL(dev_get_stats); |
c45d286e | 5403 | |
dc2b4847 | 5404 | static void netdev_init_one_queue(struct net_device *dev, |
e8a0464c DM |
5405 | struct netdev_queue *queue, |
5406 | void *_unused) | |
dc2b4847 | 5407 | { |
dc2b4847 DM |
5408 | queue->dev = dev; |
5409 | } | |
5410 | ||
bb949fbd DM |
5411 | static void netdev_init_queues(struct net_device *dev) |
5412 | { | |
e8a0464c DM |
5413 | netdev_init_one_queue(dev, &dev->rx_queue, NULL); |
5414 | netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL); | |
c3f26a26 | 5415 | spin_lock_init(&dev->tx_global_lock); |
bb949fbd DM |
5416 | } |
5417 | ||
1da177e4 | 5418 | /** |
f25f4e44 | 5419 | * alloc_netdev_mq - allocate network device |
1da177e4 LT |
5420 | * @sizeof_priv: size of private data to allocate space for |
5421 | * @name: device name format string | |
5422 | * @setup: callback to initialize device | |
f25f4e44 | 5423 | * @queue_count: the number of subqueues to allocate |
1da177e4 LT |
5424 | * |
5425 | * Allocates a struct net_device with private data area for driver use | |
f25f4e44 PWJ |
5426 | * and performs basic initialization. Also allocates subquue structs |
5427 | * for each queue on the device at the end of the netdevice. | |
1da177e4 | 5428 | */ |
f25f4e44 PWJ |
5429 | struct net_device *alloc_netdev_mq(int sizeof_priv, const char *name, |
5430 | void (*setup)(struct net_device *), unsigned int queue_count) | |
1da177e4 | 5431 | { |
e8a0464c | 5432 | struct netdev_queue *tx; |
1da177e4 | 5433 | struct net_device *dev; |
7943986c | 5434 | size_t alloc_size; |
1ce8e7b5 | 5435 | struct net_device *p; |
1da177e4 | 5436 | |
b6fe17d6 SH |
5437 | BUG_ON(strlen(name) >= sizeof(dev->name)); |
5438 | ||
fd2ea0a7 | 5439 | alloc_size = sizeof(struct net_device); |
d1643d24 AD |
5440 | if (sizeof_priv) { |
5441 | /* ensure 32-byte alignment of private area */ | |
1ce8e7b5 | 5442 | alloc_size = ALIGN(alloc_size, NETDEV_ALIGN); |
d1643d24 AD |
5443 | alloc_size += sizeof_priv; |
5444 | } | |
5445 | /* ensure 32-byte alignment of whole construct */ | |
1ce8e7b5 | 5446 | alloc_size += NETDEV_ALIGN - 1; |
1da177e4 | 5447 | |
31380de9 | 5448 | p = kzalloc(alloc_size, GFP_KERNEL); |
1da177e4 | 5449 | if (!p) { |
b6fe17d6 | 5450 | printk(KERN_ERR "alloc_netdev: Unable to allocate device.\n"); |
1da177e4 LT |
5451 | return NULL; |
5452 | } | |
1da177e4 | 5453 | |
7943986c | 5454 | tx = kcalloc(queue_count, sizeof(struct netdev_queue), GFP_KERNEL); |
e8a0464c DM |
5455 | if (!tx) { |
5456 | printk(KERN_ERR "alloc_netdev: Unable to allocate " | |
5457 | "tx qdiscs.\n"); | |
ab9c73cc | 5458 | goto free_p; |
e8a0464c DM |
5459 | } |
5460 | ||
1ce8e7b5 | 5461 | dev = PTR_ALIGN(p, NETDEV_ALIGN); |
1da177e4 | 5462 | dev->padded = (char *)dev - (char *)p; |
ab9c73cc JP |
5463 | |
5464 | if (dev_addr_init(dev)) | |
5465 | goto free_tx; | |
5466 | ||
ccffad25 JP |
5467 | dev_unicast_init(dev); |
5468 | ||
c346dca1 | 5469 | dev_net_set(dev, &init_net); |
1da177e4 | 5470 | |
e8a0464c DM |
5471 | dev->_tx = tx; |
5472 | dev->num_tx_queues = queue_count; | |
fd2ea0a7 | 5473 | dev->real_num_tx_queues = queue_count; |
e8a0464c | 5474 | |
82cc1a7a | 5475 | dev->gso_max_size = GSO_MAX_SIZE; |
1da177e4 | 5476 | |
bb949fbd DM |
5477 | netdev_init_queues(dev); |
5478 | ||
15682bc4 PWJ |
5479 | INIT_LIST_HEAD(&dev->ethtool_ntuple_list.list); |
5480 | dev->ethtool_ntuple_list.count = 0; | |
d565b0a1 | 5481 | INIT_LIST_HEAD(&dev->napi_list); |
9fdce099 | 5482 | INIT_LIST_HEAD(&dev->unreg_list); |
e014debe | 5483 | INIT_LIST_HEAD(&dev->link_watch_list); |
93f154b5 | 5484 | dev->priv_flags = IFF_XMIT_DST_RELEASE; |
1da177e4 LT |
5485 | setup(dev); |
5486 | strcpy(dev->name, name); | |
5487 | return dev; | |
ab9c73cc JP |
5488 | |
5489 | free_tx: | |
5490 | kfree(tx); | |
5491 | ||
5492 | free_p: | |
5493 | kfree(p); | |
5494 | return NULL; | |
1da177e4 | 5495 | } |
f25f4e44 | 5496 | EXPORT_SYMBOL(alloc_netdev_mq); |
1da177e4 LT |
5497 | |
5498 | /** | |
5499 | * free_netdev - free network device | |
5500 | * @dev: device | |
5501 | * | |
4ec93edb YH |
5502 | * This function does the last stage of destroying an allocated device |
5503 | * interface. The reference to the device object is released. | |
1da177e4 LT |
5504 | * If this is the last reference then it will be freed. |
5505 | */ | |
5506 | void free_netdev(struct net_device *dev) | |
5507 | { | |
d565b0a1 HX |
5508 | struct napi_struct *p, *n; |
5509 | ||
f3005d7f DL |
5510 | release_net(dev_net(dev)); |
5511 | ||
e8a0464c DM |
5512 | kfree(dev->_tx); |
5513 | ||
f001fde5 JP |
5514 | /* Flush device addresses */ |
5515 | dev_addr_flush(dev); | |
5516 | ||
15682bc4 PWJ |
5517 | /* Clear ethtool n-tuple list */ |
5518 | ethtool_ntuple_flush(dev); | |
5519 | ||
d565b0a1 HX |
5520 | list_for_each_entry_safe(p, n, &dev->napi_list, dev_list) |
5521 | netif_napi_del(p); | |
5522 | ||
3041a069 | 5523 | /* Compatibility with error handling in drivers */ |
1da177e4 LT |
5524 | if (dev->reg_state == NETREG_UNINITIALIZED) { |
5525 | kfree((char *)dev - dev->padded); | |
5526 | return; | |
5527 | } | |
5528 | ||
5529 | BUG_ON(dev->reg_state != NETREG_UNREGISTERED); | |
5530 | dev->reg_state = NETREG_RELEASED; | |
5531 | ||
43cb76d9 GKH |
5532 | /* will free via device release */ |
5533 | put_device(&dev->dev); | |
1da177e4 | 5534 | } |
d1b19dff | 5535 | EXPORT_SYMBOL(free_netdev); |
4ec93edb | 5536 | |
f0db275a SH |
5537 | /** |
5538 | * synchronize_net - Synchronize with packet receive processing | |
5539 | * | |
5540 | * Wait for packets currently being received to be done. | |
5541 | * Does not block later packets from starting. | |
5542 | */ | |
4ec93edb | 5543 | void synchronize_net(void) |
1da177e4 LT |
5544 | { |
5545 | might_sleep(); | |
fbd568a3 | 5546 | synchronize_rcu(); |
1da177e4 | 5547 | } |
d1b19dff | 5548 | EXPORT_SYMBOL(synchronize_net); |
1da177e4 LT |
5549 | |
5550 | /** | |
44a0873d | 5551 | * unregister_netdevice_queue - remove device from the kernel |
1da177e4 | 5552 | * @dev: device |
44a0873d | 5553 | * @head: list |
6ebfbc06 | 5554 | * |
1da177e4 | 5555 | * This function shuts down a device interface and removes it |
d59b54b1 | 5556 | * from the kernel tables. |
44a0873d | 5557 | * If head not NULL, device is queued to be unregistered later. |
1da177e4 LT |
5558 | * |
5559 | * Callers must hold the rtnl semaphore. You may want | |
5560 | * unregister_netdev() instead of this. | |
5561 | */ | |
5562 | ||
44a0873d | 5563 | void unregister_netdevice_queue(struct net_device *dev, struct list_head *head) |
1da177e4 | 5564 | { |
a6620712 HX |
5565 | ASSERT_RTNL(); |
5566 | ||
44a0873d | 5567 | if (head) { |
9fdce099 | 5568 | list_move_tail(&dev->unreg_list, head); |
44a0873d ED |
5569 | } else { |
5570 | rollback_registered(dev); | |
5571 | /* Finish processing unregister after unlock */ | |
5572 | net_set_todo(dev); | |
5573 | } | |
1da177e4 | 5574 | } |
44a0873d | 5575 | EXPORT_SYMBOL(unregister_netdevice_queue); |
1da177e4 | 5576 | |
9b5e383c ED |
5577 | /** |
5578 | * unregister_netdevice_many - unregister many devices | |
5579 | * @head: list of devices | |
9b5e383c ED |
5580 | */ |
5581 | void unregister_netdevice_many(struct list_head *head) | |
5582 | { | |
5583 | struct net_device *dev; | |
5584 | ||
5585 | if (!list_empty(head)) { | |
5586 | rollback_registered_many(head); | |
5587 | list_for_each_entry(dev, head, unreg_list) | |
5588 | net_set_todo(dev); | |
5589 | } | |
5590 | } | |
63c8099d | 5591 | EXPORT_SYMBOL(unregister_netdevice_many); |
9b5e383c | 5592 | |
1da177e4 LT |
5593 | /** |
5594 | * unregister_netdev - remove device from the kernel | |
5595 | * @dev: device | |
5596 | * | |
5597 | * This function shuts down a device interface and removes it | |
d59b54b1 | 5598 | * from the kernel tables. |
1da177e4 LT |
5599 | * |
5600 | * This is just a wrapper for unregister_netdevice that takes | |
5601 | * the rtnl semaphore. In general you want to use this and not | |
5602 | * unregister_netdevice. | |
5603 | */ | |
5604 | void unregister_netdev(struct net_device *dev) | |
5605 | { | |
5606 | rtnl_lock(); | |
5607 | unregister_netdevice(dev); | |
5608 | rtnl_unlock(); | |
5609 | } | |
1da177e4 LT |
5610 | EXPORT_SYMBOL(unregister_netdev); |
5611 | ||
ce286d32 EB |
5612 | /** |
5613 | * dev_change_net_namespace - move device to different nethost namespace | |
5614 | * @dev: device | |
5615 | * @net: network namespace | |
5616 | * @pat: If not NULL name pattern to try if the current device name | |
5617 | * is already taken in the destination network namespace. | |
5618 | * | |
5619 | * This function shuts down a device interface and moves it | |
5620 | * to a new network namespace. On success 0 is returned, on | |
5621 | * a failure a netagive errno code is returned. | |
5622 | * | |
5623 | * Callers must hold the rtnl semaphore. | |
5624 | */ | |
5625 | ||
5626 | int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat) | |
5627 | { | |
ce286d32 EB |
5628 | int err; |
5629 | ||
5630 | ASSERT_RTNL(); | |
5631 | ||
5632 | /* Don't allow namespace local devices to be moved. */ | |
5633 | err = -EINVAL; | |
5634 | if (dev->features & NETIF_F_NETNS_LOCAL) | |
5635 | goto out; | |
5636 | ||
3891845e EB |
5637 | #ifdef CONFIG_SYSFS |
5638 | /* Don't allow real devices to be moved when sysfs | |
5639 | * is enabled. | |
5640 | */ | |
5641 | err = -EINVAL; | |
5642 | if (dev->dev.parent) | |
5643 | goto out; | |
5644 | #endif | |
5645 | ||
ce286d32 EB |
5646 | /* Ensure the device has been registrered */ |
5647 | err = -EINVAL; | |
5648 | if (dev->reg_state != NETREG_REGISTERED) | |
5649 | goto out; | |
5650 | ||
5651 | /* Get out if there is nothing todo */ | |
5652 | err = 0; | |
878628fb | 5653 | if (net_eq(dev_net(dev), net)) |
ce286d32 EB |
5654 | goto out; |
5655 | ||
5656 | /* Pick the destination device name, and ensure | |
5657 | * we can use it in the destination network namespace. | |
5658 | */ | |
5659 | err = -EEXIST; | |
d9031024 | 5660 | if (__dev_get_by_name(net, dev->name)) { |
ce286d32 EB |
5661 | /* We get here if we can't use the current device name */ |
5662 | if (!pat) | |
5663 | goto out; | |
d9031024 | 5664 | if (dev_get_valid_name(net, pat, dev->name, 1)) |
ce286d32 EB |
5665 | goto out; |
5666 | } | |
5667 | ||
5668 | /* | |
5669 | * And now a mini version of register_netdevice unregister_netdevice. | |
5670 | */ | |
5671 | ||
5672 | /* If device is running close it first. */ | |
9b772652 | 5673 | dev_close(dev); |
ce286d32 EB |
5674 | |
5675 | /* And unlink it from device chain */ | |
5676 | err = -ENODEV; | |
5677 | unlist_netdevice(dev); | |
5678 | ||
5679 | synchronize_net(); | |
5680 | ||
5681 | /* Shutdown queueing discipline. */ | |
5682 | dev_shutdown(dev); | |
5683 | ||
5684 | /* Notify protocols, that we are about to destroy | |
5685 | this device. They should clean all the things. | |
5686 | */ | |
5687 | call_netdevice_notifiers(NETDEV_UNREGISTER, dev); | |
a5ee1551 | 5688 | call_netdevice_notifiers(NETDEV_UNREGISTER_BATCH, dev); |
ce286d32 EB |
5689 | |
5690 | /* | |
5691 | * Flush the unicast and multicast chains | |
5692 | */ | |
ccffad25 | 5693 | dev_unicast_flush(dev); |
ce286d32 EB |
5694 | dev_addr_discard(dev); |
5695 | ||
3891845e EB |
5696 | netdev_unregister_kobject(dev); |
5697 | ||
ce286d32 | 5698 | /* Actually switch the network namespace */ |
c346dca1 | 5699 | dev_net_set(dev, net); |
ce286d32 | 5700 | |
ce286d32 EB |
5701 | /* If there is an ifindex conflict assign a new one */ |
5702 | if (__dev_get_by_index(net, dev->ifindex)) { | |
5703 | int iflink = (dev->iflink == dev->ifindex); | |
5704 | dev->ifindex = dev_new_index(net); | |
5705 | if (iflink) | |
5706 | dev->iflink = dev->ifindex; | |
5707 | } | |
5708 | ||
8b41d188 | 5709 | /* Fixup kobjects */ |
aaf8cdc3 | 5710 | err = netdev_register_kobject(dev); |
8b41d188 | 5711 | WARN_ON(err); |
ce286d32 EB |
5712 | |
5713 | /* Add the device back in the hashes */ | |
5714 | list_netdevice(dev); | |
5715 | ||
5716 | /* Notify protocols, that a new device appeared. */ | |
5717 | call_netdevice_notifiers(NETDEV_REGISTER, dev); | |
5718 | ||
d90a909e EB |
5719 | /* |
5720 | * Prevent userspace races by waiting until the network | |
5721 | * device is fully setup before sending notifications. | |
5722 | */ | |
5723 | rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U); | |
5724 | ||
ce286d32 EB |
5725 | synchronize_net(); |
5726 | err = 0; | |
5727 | out: | |
5728 | return err; | |
5729 | } | |
463d0183 | 5730 | EXPORT_SYMBOL_GPL(dev_change_net_namespace); |
ce286d32 | 5731 | |
1da177e4 LT |
5732 | static int dev_cpu_callback(struct notifier_block *nfb, |
5733 | unsigned long action, | |
5734 | void *ocpu) | |
5735 | { | |
5736 | struct sk_buff **list_skb; | |
37437bb2 | 5737 | struct Qdisc **list_net; |
1da177e4 LT |
5738 | struct sk_buff *skb; |
5739 | unsigned int cpu, oldcpu = (unsigned long)ocpu; | |
5740 | struct softnet_data *sd, *oldsd; | |
5741 | ||
8bb78442 | 5742 | if (action != CPU_DEAD && action != CPU_DEAD_FROZEN) |
1da177e4 LT |
5743 | return NOTIFY_OK; |
5744 | ||
5745 | local_irq_disable(); | |
5746 | cpu = smp_processor_id(); | |
5747 | sd = &per_cpu(softnet_data, cpu); | |
5748 | oldsd = &per_cpu(softnet_data, oldcpu); | |
5749 | ||
5750 | /* Find end of our completion_queue. */ | |
5751 | list_skb = &sd->completion_queue; | |
5752 | while (*list_skb) | |
5753 | list_skb = &(*list_skb)->next; | |
5754 | /* Append completion queue from offline CPU. */ | |
5755 | *list_skb = oldsd->completion_queue; | |
5756 | oldsd->completion_queue = NULL; | |
5757 | ||
5758 | /* Find end of our output_queue. */ | |
5759 | list_net = &sd->output_queue; | |
5760 | while (*list_net) | |
5761 | list_net = &(*list_net)->next_sched; | |
5762 | /* Append output queue from offline CPU. */ | |
5763 | *list_net = oldsd->output_queue; | |
5764 | oldsd->output_queue = NULL; | |
5765 | ||
5766 | raise_softirq_irqoff(NET_TX_SOFTIRQ); | |
5767 | local_irq_enable(); | |
5768 | ||
5769 | /* Process offline CPU's input_pkt_queue */ | |
5770 | while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) | |
5771 | netif_rx(skb); | |
5772 | ||
5773 | return NOTIFY_OK; | |
5774 | } | |
1da177e4 LT |
5775 | |
5776 | ||
7f353bf2 | 5777 | /** |
b63365a2 HX |
5778 | * netdev_increment_features - increment feature set by one |
5779 | * @all: current feature set | |
5780 | * @one: new feature set | |
5781 | * @mask: mask feature set | |
7f353bf2 HX |
5782 | * |
5783 | * Computes a new feature set after adding a device with feature set | |
b63365a2 HX |
5784 | * @one to the master device with current feature set @all. Will not |
5785 | * enable anything that is off in @mask. Returns the new feature set. | |
7f353bf2 | 5786 | */ |
b63365a2 HX |
5787 | unsigned long netdev_increment_features(unsigned long all, unsigned long one, |
5788 | unsigned long mask) | |
5789 | { | |
5790 | /* If device needs checksumming, downgrade to it. */ | |
d1b19dff | 5791 | if (all & NETIF_F_NO_CSUM && !(one & NETIF_F_NO_CSUM)) |
b63365a2 HX |
5792 | all ^= NETIF_F_NO_CSUM | (one & NETIF_F_ALL_CSUM); |
5793 | else if (mask & NETIF_F_ALL_CSUM) { | |
5794 | /* If one device supports v4/v6 checksumming, set for all. */ | |
5795 | if (one & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM) && | |
5796 | !(all & NETIF_F_GEN_CSUM)) { | |
5797 | all &= ~NETIF_F_ALL_CSUM; | |
5798 | all |= one & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM); | |
5799 | } | |
e2a6b852 | 5800 | |
b63365a2 HX |
5801 | /* If one device supports hw checksumming, set for all. */ |
5802 | if (one & NETIF_F_GEN_CSUM && !(all & NETIF_F_GEN_CSUM)) { | |
5803 | all &= ~NETIF_F_ALL_CSUM; | |
5804 | all |= NETIF_F_HW_CSUM; | |
5805 | } | |
5806 | } | |
7f353bf2 | 5807 | |
b63365a2 | 5808 | one |= NETIF_F_ALL_CSUM; |
7f353bf2 | 5809 | |
b63365a2 | 5810 | one |= all & NETIF_F_ONE_FOR_ALL; |
d9f5950f | 5811 | all &= one | NETIF_F_LLTX | NETIF_F_GSO | NETIF_F_UFO; |
b63365a2 | 5812 | all |= one & mask & NETIF_F_ONE_FOR_ALL; |
7f353bf2 HX |
5813 | |
5814 | return all; | |
5815 | } | |
b63365a2 | 5816 | EXPORT_SYMBOL(netdev_increment_features); |
7f353bf2 | 5817 | |
30d97d35 PE |
5818 | static struct hlist_head *netdev_create_hash(void) |
5819 | { | |
5820 | int i; | |
5821 | struct hlist_head *hash; | |
5822 | ||
5823 | hash = kmalloc(sizeof(*hash) * NETDEV_HASHENTRIES, GFP_KERNEL); | |
5824 | if (hash != NULL) | |
5825 | for (i = 0; i < NETDEV_HASHENTRIES; i++) | |
5826 | INIT_HLIST_HEAD(&hash[i]); | |
5827 | ||
5828 | return hash; | |
5829 | } | |
5830 | ||
881d966b | 5831 | /* Initialize per network namespace state */ |
4665079c | 5832 | static int __net_init netdev_init(struct net *net) |
881d966b | 5833 | { |
881d966b | 5834 | INIT_LIST_HEAD(&net->dev_base_head); |
881d966b | 5835 | |
30d97d35 PE |
5836 | net->dev_name_head = netdev_create_hash(); |
5837 | if (net->dev_name_head == NULL) | |
5838 | goto err_name; | |
881d966b | 5839 | |
30d97d35 PE |
5840 | net->dev_index_head = netdev_create_hash(); |
5841 | if (net->dev_index_head == NULL) | |
5842 | goto err_idx; | |
881d966b EB |
5843 | |
5844 | return 0; | |
30d97d35 PE |
5845 | |
5846 | err_idx: | |
5847 | kfree(net->dev_name_head); | |
5848 | err_name: | |
5849 | return -ENOMEM; | |
881d966b EB |
5850 | } |
5851 | ||
f0db275a SH |
5852 | /** |
5853 | * netdev_drivername - network driver for the device | |
5854 | * @dev: network device | |
5855 | * @buffer: buffer for resulting name | |
5856 | * @len: size of buffer | |
5857 | * | |
5858 | * Determine network driver for device. | |
5859 | */ | |
cf04a4c7 | 5860 | char *netdev_drivername(const struct net_device *dev, char *buffer, int len) |
6579e57b | 5861 | { |
cf04a4c7 SH |
5862 | const struct device_driver *driver; |
5863 | const struct device *parent; | |
6579e57b AV |
5864 | |
5865 | if (len <= 0 || !buffer) | |
5866 | return buffer; | |
5867 | buffer[0] = 0; | |
5868 | ||
5869 | parent = dev->dev.parent; | |
5870 | ||
5871 | if (!parent) | |
5872 | return buffer; | |
5873 | ||
5874 | driver = parent->driver; | |
5875 | if (driver && driver->name) | |
5876 | strlcpy(buffer, driver->name, len); | |
5877 | return buffer; | |
5878 | } | |
5879 | ||
4665079c | 5880 | static void __net_exit netdev_exit(struct net *net) |
881d966b EB |
5881 | { |
5882 | kfree(net->dev_name_head); | |
5883 | kfree(net->dev_index_head); | |
5884 | } | |
5885 | ||
022cbae6 | 5886 | static struct pernet_operations __net_initdata netdev_net_ops = { |
881d966b EB |
5887 | .init = netdev_init, |
5888 | .exit = netdev_exit, | |
5889 | }; | |
5890 | ||
4665079c | 5891 | static void __net_exit default_device_exit(struct net *net) |
ce286d32 | 5892 | { |
e008b5fc | 5893 | struct net_device *dev, *aux; |
ce286d32 | 5894 | /* |
e008b5fc | 5895 | * Push all migratable network devices back to the |
ce286d32 EB |
5896 | * initial network namespace |
5897 | */ | |
5898 | rtnl_lock(); | |
e008b5fc | 5899 | for_each_netdev_safe(net, dev, aux) { |
ce286d32 | 5900 | int err; |
aca51397 | 5901 | char fb_name[IFNAMSIZ]; |
ce286d32 EB |
5902 | |
5903 | /* Ignore unmoveable devices (i.e. loopback) */ | |
5904 | if (dev->features & NETIF_F_NETNS_LOCAL) | |
5905 | continue; | |
5906 | ||
e008b5fc EB |
5907 | /* Leave virtual devices for the generic cleanup */ |
5908 | if (dev->rtnl_link_ops) | |
5909 | continue; | |
d0c082ce | 5910 | |
ce286d32 | 5911 | /* Push remaing network devices to init_net */ |
aca51397 PE |
5912 | snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex); |
5913 | err = dev_change_net_namespace(dev, &init_net, fb_name); | |
ce286d32 | 5914 | if (err) { |
aca51397 | 5915 | printk(KERN_EMERG "%s: failed to move %s to init_net: %d\n", |
ce286d32 | 5916 | __func__, dev->name, err); |
aca51397 | 5917 | BUG(); |
ce286d32 EB |
5918 | } |
5919 | } | |
5920 | rtnl_unlock(); | |
5921 | } | |
5922 | ||
04dc7f6b EB |
5923 | static void __net_exit default_device_exit_batch(struct list_head *net_list) |
5924 | { | |
5925 | /* At exit all network devices most be removed from a network | |
5926 | * namespace. Do this in the reverse order of registeration. | |
5927 | * Do this across as many network namespaces as possible to | |
5928 | * improve batching efficiency. | |
5929 | */ | |
5930 | struct net_device *dev; | |
5931 | struct net *net; | |
5932 | LIST_HEAD(dev_kill_list); | |
5933 | ||
5934 | rtnl_lock(); | |
5935 | list_for_each_entry(net, net_list, exit_list) { | |
5936 | for_each_netdev_reverse(net, dev) { | |
5937 | if (dev->rtnl_link_ops) | |
5938 | dev->rtnl_link_ops->dellink(dev, &dev_kill_list); | |
5939 | else | |
5940 | unregister_netdevice_queue(dev, &dev_kill_list); | |
5941 | } | |
5942 | } | |
5943 | unregister_netdevice_many(&dev_kill_list); | |
5944 | rtnl_unlock(); | |
5945 | } | |
5946 | ||
022cbae6 | 5947 | static struct pernet_operations __net_initdata default_device_ops = { |
ce286d32 | 5948 | .exit = default_device_exit, |
04dc7f6b | 5949 | .exit_batch = default_device_exit_batch, |
ce286d32 EB |
5950 | }; |
5951 | ||
1da177e4 LT |
5952 | /* |
5953 | * Initialize the DEV module. At boot time this walks the device list and | |
5954 | * unhooks any devices that fail to initialise (normally hardware not | |
5955 | * present) and leaves us with a valid list of present and active devices. | |
5956 | * | |
5957 | */ | |
5958 | ||
5959 | /* | |
5960 | * This is called single threaded during boot, so no need | |
5961 | * to take the rtnl semaphore. | |
5962 | */ | |
5963 | static int __init net_dev_init(void) | |
5964 | { | |
5965 | int i, rc = -ENOMEM; | |
5966 | ||
5967 | BUG_ON(!dev_boot_phase); | |
5968 | ||
1da177e4 LT |
5969 | if (dev_proc_init()) |
5970 | goto out; | |
5971 | ||
8b41d188 | 5972 | if (netdev_kobject_init()) |
1da177e4 LT |
5973 | goto out; |
5974 | ||
5975 | INIT_LIST_HEAD(&ptype_all); | |
82d8a867 | 5976 | for (i = 0; i < PTYPE_HASH_SIZE; i++) |
1da177e4 LT |
5977 | INIT_LIST_HEAD(&ptype_base[i]); |
5978 | ||
881d966b EB |
5979 | if (register_pernet_subsys(&netdev_net_ops)) |
5980 | goto out; | |
1da177e4 LT |
5981 | |
5982 | /* | |
5983 | * Initialise the packet receive queues. | |
5984 | */ | |
5985 | ||
6f912042 | 5986 | for_each_possible_cpu(i) { |
1da177e4 LT |
5987 | struct softnet_data *queue; |
5988 | ||
5989 | queue = &per_cpu(softnet_data, i); | |
5990 | skb_queue_head_init(&queue->input_pkt_queue); | |
1da177e4 LT |
5991 | queue->completion_queue = NULL; |
5992 | INIT_LIST_HEAD(&queue->poll_list); | |
bea3348e SH |
5993 | |
5994 | queue->backlog.poll = process_backlog; | |
5995 | queue->backlog.weight = weight_p; | |
d565b0a1 | 5996 | queue->backlog.gro_list = NULL; |
4ae5544f | 5997 | queue->backlog.gro_count = 0; |
1da177e4 LT |
5998 | } |
5999 | ||
1da177e4 LT |
6000 | dev_boot_phase = 0; |
6001 | ||
505d4f73 EB |
6002 | /* The loopback device is special if any other network devices |
6003 | * is present in a network namespace the loopback device must | |
6004 | * be present. Since we now dynamically allocate and free the | |
6005 | * loopback device ensure this invariant is maintained by | |
6006 | * keeping the loopback device as the first device on the | |
6007 | * list of network devices. Ensuring the loopback devices | |
6008 | * is the first device that appears and the last network device | |
6009 | * that disappears. | |
6010 | */ | |
6011 | if (register_pernet_device(&loopback_net_ops)) | |
6012 | goto out; | |
6013 | ||
6014 | if (register_pernet_device(&default_device_ops)) | |
6015 | goto out; | |
6016 | ||
962cf36c CM |
6017 | open_softirq(NET_TX_SOFTIRQ, net_tx_action); |
6018 | open_softirq(NET_RX_SOFTIRQ, net_rx_action); | |
1da177e4 LT |
6019 | |
6020 | hotcpu_notifier(dev_cpu_callback, 0); | |
6021 | dst_init(); | |
6022 | dev_mcast_init(); | |
6023 | rc = 0; | |
6024 | out: | |
6025 | return rc; | |
6026 | } | |
6027 | ||
6028 | subsys_initcall(net_dev_init); | |
6029 | ||
e88721f8 KK |
6030 | static int __init initialize_hashrnd(void) |
6031 | { | |
6032 | get_random_bytes(&skb_tx_hashrnd, sizeof(skb_tx_hashrnd)); | |
6033 | return 0; | |
6034 | } | |
6035 | ||
6036 | late_initcall_sync(initialize_hashrnd); | |
6037 |