Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * ip6_flowlabel.c IPv6 flowlabel manager. | |
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 | * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> | |
10 | */ | |
11 | ||
4fc268d2 | 12 | #include <linux/capability.h> |
1da177e4 LT |
13 | #include <linux/errno.h> |
14 | #include <linux/types.h> | |
15 | #include <linux/socket.h> | |
16 | #include <linux/net.h> | |
17 | #include <linux/netdevice.h> | |
18 | #include <linux/if_arp.h> | |
19 | #include <linux/in6.h> | |
20 | #include <linux/route.h> | |
21 | #include <linux/proc_fs.h> | |
22 | #include <linux/seq_file.h> | |
5a0e3ad6 | 23 | #include <linux/slab.h> |
1da177e4 | 24 | |
457c4cbc | 25 | #include <net/net_namespace.h> |
1da177e4 LT |
26 | #include <net/sock.h> |
27 | ||
28 | #include <net/ipv6.h> | |
29 | #include <net/ndisc.h> | |
30 | #include <net/protocol.h> | |
31 | #include <net/ip6_route.h> | |
32 | #include <net/addrconf.h> | |
33 | #include <net/rawv6.h> | |
34 | #include <net/icmp.h> | |
35 | #include <net/transp_v6.h> | |
36 | ||
37 | #include <asm/uaccess.h> | |
38 | ||
39 | #define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified | |
40 | in old IPv6 RFC. Well, it was reasonable value. | |
41 | */ | |
42 | #define FL_MAX_LINGER 60 /* Maximal linger timeout */ | |
43 | ||
44 | /* FL hash table */ | |
45 | ||
46 | #define FL_MAX_PER_SOCK 32 | |
47 | #define FL_MAX_SIZE 4096 | |
48 | #define FL_HASH_MASK 255 | |
49 | #define FL_HASH(l) (ntohl(l)&FL_HASH_MASK) | |
50 | ||
51 | static atomic_t fl_size = ATOMIC_INIT(0); | |
52 | static struct ip6_flowlabel *fl_ht[FL_HASH_MASK+1]; | |
53 | ||
54 | static void ip6_fl_gc(unsigned long dummy); | |
8d06afab | 55 | static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0); |
1da177e4 LT |
56 | |
57 | /* FL hash table lock: it protects only of GC */ | |
58 | ||
59 | static DEFINE_RWLOCK(ip6_fl_lock); | |
60 | ||
61 | /* Big socket sock */ | |
62 | ||
63 | static DEFINE_RWLOCK(ip6_sk_fl_lock); | |
64 | ||
65 | ||
60e8fbc4 | 66 | static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label) |
1da177e4 LT |
67 | { |
68 | struct ip6_flowlabel *fl; | |
69 | ||
70 | for (fl=fl_ht[FL_HASH(label)]; fl; fl = fl->next) { | |
09ad9bc7 | 71 | if (fl->label == label && net_eq(fl->fl_net, net)) |
1da177e4 LT |
72 | return fl; |
73 | } | |
74 | return NULL; | |
75 | } | |
76 | ||
60e8fbc4 | 77 | static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label) |
1da177e4 LT |
78 | { |
79 | struct ip6_flowlabel *fl; | |
80 | ||
81 | read_lock_bh(&ip6_fl_lock); | |
60e8fbc4 | 82 | fl = __fl_lookup(net, label); |
1da177e4 LT |
83 | if (fl) |
84 | atomic_inc(&fl->users); | |
85 | read_unlock_bh(&ip6_fl_lock); | |
86 | return fl; | |
87 | } | |
88 | ||
89 | ||
90 | static void fl_free(struct ip6_flowlabel *fl) | |
91 | { | |
60e8fbc4 BT |
92 | if (fl) { |
93 | release_net(fl->fl_net); | |
1da177e4 | 94 | kfree(fl->opt); |
60e8fbc4 | 95 | } |
1da177e4 LT |
96 | kfree(fl); |
97 | } | |
98 | ||
99 | static void fl_release(struct ip6_flowlabel *fl) | |
100 | { | |
101 | write_lock_bh(&ip6_fl_lock); | |
102 | ||
103 | fl->lastuse = jiffies; | |
104 | if (atomic_dec_and_test(&fl->users)) { | |
105 | unsigned long ttd = fl->lastuse + fl->linger; | |
106 | if (time_after(ttd, fl->expires)) | |
107 | fl->expires = ttd; | |
108 | ttd = fl->expires; | |
109 | if (fl->opt && fl->share == IPV6_FL_S_EXCL) { | |
110 | struct ipv6_txoptions *opt = fl->opt; | |
111 | fl->opt = NULL; | |
112 | kfree(opt); | |
113 | } | |
114 | if (!timer_pending(&ip6_fl_gc_timer) || | |
115 | time_after(ip6_fl_gc_timer.expires, ttd)) | |
116 | mod_timer(&ip6_fl_gc_timer, ttd); | |
117 | } | |
1da177e4 LT |
118 | write_unlock_bh(&ip6_fl_lock); |
119 | } | |
120 | ||
121 | static void ip6_fl_gc(unsigned long dummy) | |
122 | { | |
123 | int i; | |
124 | unsigned long now = jiffies; | |
125 | unsigned long sched = 0; | |
126 | ||
127 | write_lock(&ip6_fl_lock); | |
128 | ||
129 | for (i=0; i<=FL_HASH_MASK; i++) { | |
130 | struct ip6_flowlabel *fl, **flp; | |
131 | flp = &fl_ht[i]; | |
132 | while ((fl=*flp) != NULL) { | |
133 | if (atomic_read(&fl->users) == 0) { | |
134 | unsigned long ttd = fl->lastuse + fl->linger; | |
135 | if (time_after(ttd, fl->expires)) | |
136 | fl->expires = ttd; | |
137 | ttd = fl->expires; | |
138 | if (time_after_eq(now, ttd)) { | |
139 | *flp = fl->next; | |
140 | fl_free(fl); | |
141 | atomic_dec(&fl_size); | |
142 | continue; | |
143 | } | |
144 | if (!sched || time_before(ttd, sched)) | |
145 | sched = ttd; | |
146 | } | |
147 | flp = &fl->next; | |
148 | } | |
149 | } | |
150 | if (!sched && atomic_read(&fl_size)) | |
151 | sched = now + FL_MAX_LINGER; | |
152 | if (sched) { | |
60e8fbc4 | 153 | mod_timer(&ip6_fl_gc_timer, sched); |
1da177e4 LT |
154 | } |
155 | write_unlock(&ip6_fl_lock); | |
156 | } | |
157 | ||
2c8c1e72 | 158 | static void __net_exit ip6_fl_purge(struct net *net) |
60e8fbc4 BT |
159 | { |
160 | int i; | |
161 | ||
162 | write_lock(&ip6_fl_lock); | |
163 | for (i = 0; i <= FL_HASH_MASK; i++) { | |
164 | struct ip6_flowlabel *fl, **flp; | |
165 | flp = &fl_ht[i]; | |
166 | while ((fl = *flp) != NULL) { | |
09ad9bc7 OP |
167 | if (net_eq(fl->fl_net, net) && |
168 | atomic_read(&fl->users) == 0) { | |
60e8fbc4 BT |
169 | *flp = fl->next; |
170 | fl_free(fl); | |
171 | atomic_dec(&fl_size); | |
172 | continue; | |
173 | } | |
174 | flp = &fl->next; | |
175 | } | |
176 | } | |
177 | write_unlock(&ip6_fl_lock); | |
178 | } | |
179 | ||
180 | static struct ip6_flowlabel *fl_intern(struct net *net, | |
181 | struct ip6_flowlabel *fl, __be32 label) | |
1da177e4 | 182 | { |
78c2e502 PE |
183 | struct ip6_flowlabel *lfl; |
184 | ||
1da177e4 LT |
185 | fl->label = label & IPV6_FLOWLABEL_MASK; |
186 | ||
187 | write_lock_bh(&ip6_fl_lock); | |
188 | if (label == 0) { | |
189 | for (;;) { | |
190 | fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK; | |
191 | if (fl->label) { | |
60e8fbc4 | 192 | lfl = __fl_lookup(net, fl->label); |
1da177e4 LT |
193 | if (lfl == NULL) |
194 | break; | |
195 | } | |
196 | } | |
78c2e502 PE |
197 | } else { |
198 | /* | |
199 | * we dropper the ip6_fl_lock, so this entry could reappear | |
200 | * and we need to recheck with it. | |
201 | * | |
202 | * OTOH no need to search the active socket first, like it is | |
203 | * done in ipv6_flowlabel_opt - sock is locked, so new entry | |
204 | * with the same label can only appear on another sock | |
205 | */ | |
60e8fbc4 | 206 | lfl = __fl_lookup(net, fl->label); |
78c2e502 PE |
207 | if (lfl != NULL) { |
208 | atomic_inc(&lfl->users); | |
209 | write_unlock_bh(&ip6_fl_lock); | |
210 | return lfl; | |
211 | } | |
1da177e4 LT |
212 | } |
213 | ||
214 | fl->lastuse = jiffies; | |
215 | fl->next = fl_ht[FL_HASH(fl->label)]; | |
216 | fl_ht[FL_HASH(fl->label)] = fl; | |
217 | atomic_inc(&fl_size); | |
218 | write_unlock_bh(&ip6_fl_lock); | |
78c2e502 | 219 | return NULL; |
1da177e4 LT |
220 | } |
221 | ||
222 | ||
223 | ||
224 | /* Socket flowlabel lists */ | |
225 | ||
90bcaf7b | 226 | struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label) |
1da177e4 LT |
227 | { |
228 | struct ipv6_fl_socklist *sfl; | |
229 | struct ipv6_pinfo *np = inet6_sk(sk); | |
230 | ||
231 | label &= IPV6_FLOWLABEL_MASK; | |
232 | ||
bd0bf577 | 233 | read_lock_bh(&ip6_sk_fl_lock); |
1da177e4 LT |
234 | for (sfl=np->ipv6_fl_list; sfl; sfl = sfl->next) { |
235 | struct ip6_flowlabel *fl = sfl->fl; | |
236 | if (fl->label == label) { | |
237 | fl->lastuse = jiffies; | |
238 | atomic_inc(&fl->users); | |
52f095ee | 239 | read_unlock_bh(&ip6_sk_fl_lock); |
1da177e4 LT |
240 | return fl; |
241 | } | |
242 | } | |
bd0bf577 | 243 | read_unlock_bh(&ip6_sk_fl_lock); |
1da177e4 LT |
244 | return NULL; |
245 | } | |
246 | ||
3cf3dc6c ACM |
247 | EXPORT_SYMBOL_GPL(fl6_sock_lookup); |
248 | ||
1da177e4 LT |
249 | void fl6_free_socklist(struct sock *sk) |
250 | { | |
251 | struct ipv6_pinfo *np = inet6_sk(sk); | |
252 | struct ipv6_fl_socklist *sfl; | |
253 | ||
254 | while ((sfl = np->ipv6_fl_list) != NULL) { | |
255 | np->ipv6_fl_list = sfl->next; | |
256 | fl_release(sfl->fl); | |
257 | kfree(sfl); | |
258 | } | |
259 | } | |
260 | ||
261 | /* Service routines */ | |
262 | ||
263 | ||
264 | /* | |
265 | It is the only difficult place. flowlabel enforces equal headers | |
266 | before and including routing header, however user may supply options | |
267 | following rthdr. | |
268 | */ | |
269 | ||
270 | struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space, | |
271 | struct ip6_flowlabel * fl, | |
272 | struct ipv6_txoptions * fopt) | |
273 | { | |
df9890c3 | 274 | struct ipv6_txoptions * fl_opt = fl->opt; |
1ab1457c | 275 | |
df9890c3 YH |
276 | if (fopt == NULL || fopt->opt_flen == 0) |
277 | return fl_opt; | |
1ab1457c | 278 | |
1da177e4 LT |
279 | if (fl_opt != NULL) { |
280 | opt_space->hopopt = fl_opt->hopopt; | |
df9890c3 | 281 | opt_space->dst0opt = fl_opt->dst0opt; |
1da177e4 LT |
282 | opt_space->srcrt = fl_opt->srcrt; |
283 | opt_space->opt_nflen = fl_opt->opt_nflen; | |
284 | } else { | |
285 | if (fopt->opt_nflen == 0) | |
286 | return fopt; | |
287 | opt_space->hopopt = NULL; | |
288 | opt_space->dst0opt = NULL; | |
289 | opt_space->srcrt = NULL; | |
290 | opt_space->opt_nflen = 0; | |
291 | } | |
292 | opt_space->dst1opt = fopt->dst1opt; | |
1da177e4 LT |
293 | opt_space->opt_flen = fopt->opt_flen; |
294 | return opt_space; | |
295 | } | |
296 | ||
297 | static unsigned long check_linger(unsigned long ttl) | |
298 | { | |
299 | if (ttl < FL_MIN_LINGER) | |
300 | return FL_MIN_LINGER*HZ; | |
301 | if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN)) | |
302 | return 0; | |
303 | return ttl*HZ; | |
304 | } | |
305 | ||
306 | static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires) | |
307 | { | |
308 | linger = check_linger(linger); | |
309 | if (!linger) | |
310 | return -EPERM; | |
311 | expires = check_linger(expires); | |
312 | if (!expires) | |
313 | return -EPERM; | |
314 | fl->lastuse = jiffies; | |
315 | if (time_before(fl->linger, linger)) | |
316 | fl->linger = linger; | |
317 | if (time_before(expires, fl->linger)) | |
318 | expires = fl->linger; | |
319 | if (time_before(fl->expires, fl->lastuse + expires)) | |
320 | fl->expires = fl->lastuse + expires; | |
321 | return 0; | |
322 | } | |
323 | ||
324 | static struct ip6_flowlabel * | |
60e8fbc4 BT |
325 | fl_create(struct net *net, struct in6_flowlabel_req *freq, char __user *optval, |
326 | int optlen, int *err_p) | |
1da177e4 | 327 | { |
684de409 | 328 | struct ip6_flowlabel *fl = NULL; |
1da177e4 LT |
329 | int olen; |
330 | int addr_type; | |
331 | int err; | |
332 | ||
684de409 DM |
333 | olen = optlen - CMSG_ALIGN(sizeof(*freq)); |
334 | err = -EINVAL; | |
335 | if (olen > 64 * 1024) | |
336 | goto done; | |
337 | ||
1da177e4 | 338 | err = -ENOMEM; |
0c600eda | 339 | fl = kzalloc(sizeof(*fl), GFP_KERNEL); |
1da177e4 LT |
340 | if (fl == NULL) |
341 | goto done; | |
1da177e4 | 342 | |
1da177e4 LT |
343 | if (olen > 0) { |
344 | struct msghdr msg; | |
345 | struct flowi flowi; | |
346 | int junk; | |
347 | ||
348 | err = -ENOMEM; | |
349 | fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL); | |
350 | if (fl->opt == NULL) | |
351 | goto done; | |
352 | ||
353 | memset(fl->opt, 0, sizeof(*fl->opt)); | |
354 | fl->opt->tot_len = sizeof(*fl->opt) + olen; | |
355 | err = -EFAULT; | |
356 | if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen)) | |
357 | goto done; | |
358 | ||
359 | msg.msg_controllen = olen; | |
360 | msg.msg_control = (void*)(fl->opt+1); | |
361 | flowi.oif = 0; | |
362 | ||
91e1908f | 363 | err = datagram_send_ctl(net, &msg, &flowi, fl->opt, &junk, &junk); |
1da177e4 LT |
364 | if (err) |
365 | goto done; | |
366 | err = -EINVAL; | |
367 | if (fl->opt->opt_flen) | |
368 | goto done; | |
369 | if (fl->opt->opt_nflen == 0) { | |
370 | kfree(fl->opt); | |
371 | fl->opt = NULL; | |
372 | } | |
373 | } | |
374 | ||
60e8fbc4 | 375 | fl->fl_net = hold_net(net); |
1da177e4 LT |
376 | fl->expires = jiffies; |
377 | err = fl6_renew(fl, freq->flr_linger, freq->flr_expires); | |
378 | if (err) | |
379 | goto done; | |
380 | fl->share = freq->flr_share; | |
381 | addr_type = ipv6_addr_type(&freq->flr_dst); | |
35700212 JP |
382 | if ((addr_type & IPV6_ADDR_MAPPED) || |
383 | addr_type == IPV6_ADDR_ANY) { | |
c6817e4c | 384 | err = -EINVAL; |
1da177e4 | 385 | goto done; |
c6817e4c | 386 | } |
1da177e4 LT |
387 | ipv6_addr_copy(&fl->dst, &freq->flr_dst); |
388 | atomic_set(&fl->users, 1); | |
389 | switch (fl->share) { | |
390 | case IPV6_FL_S_EXCL: | |
391 | case IPV6_FL_S_ANY: | |
392 | break; | |
393 | case IPV6_FL_S_PROCESS: | |
394 | fl->owner = current->pid; | |
395 | break; | |
396 | case IPV6_FL_S_USER: | |
f82b3590 | 397 | fl->owner = current_euid(); |
1da177e4 LT |
398 | break; |
399 | default: | |
400 | err = -EINVAL; | |
401 | goto done; | |
402 | } | |
403 | return fl; | |
404 | ||
405 | done: | |
406 | fl_free(fl); | |
407 | *err_p = err; | |
408 | return NULL; | |
409 | } | |
410 | ||
411 | static int mem_check(struct sock *sk) | |
412 | { | |
413 | struct ipv6_pinfo *np = inet6_sk(sk); | |
414 | struct ipv6_fl_socklist *sfl; | |
415 | int room = FL_MAX_SIZE - atomic_read(&fl_size); | |
416 | int count = 0; | |
417 | ||
418 | if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK) | |
419 | return 0; | |
420 | ||
421 | for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) | |
422 | count++; | |
423 | ||
424 | if (room <= 0 || | |
425 | ((count >= FL_MAX_PER_SOCK || | |
35700212 JP |
426 | (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) && |
427 | !capable(CAP_NET_ADMIN))) | |
1da177e4 LT |
428 | return -ENOBUFS; |
429 | ||
430 | return 0; | |
431 | } | |
432 | ||
433 | static int ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2) | |
434 | { | |
435 | if (h1 == h2) | |
436 | return 0; | |
437 | if (h1 == NULL || h2 == NULL) | |
438 | return 1; | |
439 | if (h1->hdrlen != h2->hdrlen) | |
440 | return 1; | |
441 | return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1)); | |
442 | } | |
443 | ||
444 | static int ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2) | |
445 | { | |
446 | if (o1 == o2) | |
447 | return 0; | |
448 | if (o1 == NULL || o2 == NULL) | |
449 | return 1; | |
450 | if (o1->opt_nflen != o2->opt_nflen) | |
451 | return 1; | |
452 | if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt)) | |
453 | return 1; | |
454 | if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt)) | |
455 | return 1; | |
456 | if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt)) | |
457 | return 1; | |
458 | return 0; | |
459 | } | |
460 | ||
04028045 PE |
461 | static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl, |
462 | struct ip6_flowlabel *fl) | |
463 | { | |
464 | write_lock_bh(&ip6_sk_fl_lock); | |
465 | sfl->fl = fl; | |
466 | sfl->next = np->ipv6_fl_list; | |
467 | np->ipv6_fl_list = sfl; | |
468 | write_unlock_bh(&ip6_sk_fl_lock); | |
469 | } | |
470 | ||
1da177e4 LT |
471 | int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen) |
472 | { | |
55205d40 | 473 | int uninitialized_var(err); |
60e8fbc4 | 474 | struct net *net = sock_net(sk); |
1da177e4 LT |
475 | struct ipv6_pinfo *np = inet6_sk(sk); |
476 | struct in6_flowlabel_req freq; | |
477 | struct ipv6_fl_socklist *sfl1=NULL; | |
478 | struct ipv6_fl_socklist *sfl, **sflp; | |
78c2e502 PE |
479 | struct ip6_flowlabel *fl, *fl1 = NULL; |
480 | ||
1da177e4 LT |
481 | |
482 | if (optlen < sizeof(freq)) | |
483 | return -EINVAL; | |
484 | ||
485 | if (copy_from_user(&freq, optval, sizeof(freq))) | |
486 | return -EFAULT; | |
487 | ||
488 | switch (freq.flr_action) { | |
489 | case IPV6_FL_A_PUT: | |
490 | write_lock_bh(&ip6_sk_fl_lock); | |
491 | for (sflp = &np->ipv6_fl_list; (sfl=*sflp)!=NULL; sflp = &sfl->next) { | |
492 | if (sfl->fl->label == freq.flr_label) { | |
493 | if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK)) | |
494 | np->flow_label &= ~IPV6_FLOWLABEL_MASK; | |
495 | *sflp = sfl->next; | |
496 | write_unlock_bh(&ip6_sk_fl_lock); | |
497 | fl_release(sfl->fl); | |
498 | kfree(sfl); | |
499 | return 0; | |
500 | } | |
501 | } | |
502 | write_unlock_bh(&ip6_sk_fl_lock); | |
503 | return -ESRCH; | |
504 | ||
505 | case IPV6_FL_A_RENEW: | |
506 | read_lock_bh(&ip6_sk_fl_lock); | |
507 | for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) { | |
508 | if (sfl->fl->label == freq.flr_label) { | |
509 | err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires); | |
510 | read_unlock_bh(&ip6_sk_fl_lock); | |
511 | return err; | |
512 | } | |
513 | } | |
514 | read_unlock_bh(&ip6_sk_fl_lock); | |
515 | ||
516 | if (freq.flr_share == IPV6_FL_S_NONE && capable(CAP_NET_ADMIN)) { | |
60e8fbc4 | 517 | fl = fl_lookup(net, freq.flr_label); |
1da177e4 LT |
518 | if (fl) { |
519 | err = fl6_renew(fl, freq.flr_linger, freq.flr_expires); | |
520 | fl_release(fl); | |
521 | return err; | |
522 | } | |
523 | } | |
524 | return -ESRCH; | |
525 | ||
526 | case IPV6_FL_A_GET: | |
527 | if (freq.flr_label & ~IPV6_FLOWLABEL_MASK) | |
528 | return -EINVAL; | |
529 | ||
60e8fbc4 | 530 | fl = fl_create(net, &freq, optval, optlen, &err); |
1da177e4 LT |
531 | if (fl == NULL) |
532 | return err; | |
533 | sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL); | |
534 | ||
535 | if (freq.flr_label) { | |
1da177e4 LT |
536 | err = -EEXIST; |
537 | read_lock_bh(&ip6_sk_fl_lock); | |
538 | for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) { | |
539 | if (sfl->fl->label == freq.flr_label) { | |
540 | if (freq.flr_flags&IPV6_FL_F_EXCL) { | |
541 | read_unlock_bh(&ip6_sk_fl_lock); | |
542 | goto done; | |
543 | } | |
544 | fl1 = sfl->fl; | |
4ea6a804 | 545 | atomic_inc(&fl1->users); |
1da177e4 LT |
546 | break; |
547 | } | |
548 | } | |
549 | read_unlock_bh(&ip6_sk_fl_lock); | |
550 | ||
551 | if (fl1 == NULL) | |
60e8fbc4 | 552 | fl1 = fl_lookup(net, freq.flr_label); |
1da177e4 | 553 | if (fl1) { |
78c2e502 | 554 | recheck: |
1da177e4 LT |
555 | err = -EEXIST; |
556 | if (freq.flr_flags&IPV6_FL_F_EXCL) | |
557 | goto release; | |
558 | err = -EPERM; | |
559 | if (fl1->share == IPV6_FL_S_EXCL || | |
560 | fl1->share != fl->share || | |
561 | fl1->owner != fl->owner) | |
562 | goto release; | |
563 | ||
564 | err = -EINVAL; | |
565 | if (!ipv6_addr_equal(&fl1->dst, &fl->dst) || | |
566 | ipv6_opt_cmp(fl1->opt, fl->opt)) | |
567 | goto release; | |
568 | ||
569 | err = -ENOMEM; | |
570 | if (sfl1 == NULL) | |
571 | goto release; | |
572 | if (fl->linger > fl1->linger) | |
573 | fl1->linger = fl->linger; | |
574 | if ((long)(fl->expires - fl1->expires) > 0) | |
575 | fl1->expires = fl->expires; | |
04028045 | 576 | fl_link(np, sfl1, fl1); |
1da177e4 LT |
577 | fl_free(fl); |
578 | return 0; | |
579 | ||
580 | release: | |
581 | fl_release(fl1); | |
582 | goto done; | |
583 | } | |
584 | } | |
585 | err = -ENOENT; | |
586 | if (!(freq.flr_flags&IPV6_FL_F_CREATE)) | |
587 | goto done; | |
588 | ||
589 | err = -ENOMEM; | |
590 | if (sfl1 == NULL || (err = mem_check(sk)) != 0) | |
591 | goto done; | |
592 | ||
60e8fbc4 | 593 | fl1 = fl_intern(net, fl, freq.flr_label); |
78c2e502 PE |
594 | if (fl1 != NULL) |
595 | goto recheck; | |
1da177e4 | 596 | |
6c94d361 DM |
597 | if (!freq.flr_label) { |
598 | if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label, | |
599 | &fl->label, sizeof(fl->label))) { | |
600 | /* Intentionally ignore fault. */ | |
601 | } | |
602 | } | |
1da177e4 | 603 | |
04028045 | 604 | fl_link(np, sfl1, fl); |
1da177e4 LT |
605 | return 0; |
606 | ||
607 | default: | |
608 | return -EINVAL; | |
609 | } | |
610 | ||
611 | done: | |
612 | fl_free(fl); | |
613 | kfree(sfl1); | |
614 | return err; | |
615 | } | |
616 | ||
617 | #ifdef CONFIG_PROC_FS | |
618 | ||
619 | struct ip6fl_iter_state { | |
5983a3df | 620 | struct seq_net_private p; |
1da177e4 LT |
621 | int bucket; |
622 | }; | |
623 | ||
624 | #define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private) | |
625 | ||
626 | static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq) | |
627 | { | |
628 | struct ip6_flowlabel *fl = NULL; | |
629 | struct ip6fl_iter_state *state = ip6fl_seq_private(seq); | |
5983a3df | 630 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
631 | |
632 | for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) { | |
5983a3df BT |
633 | fl = fl_ht[state->bucket]; |
634 | ||
09ad9bc7 | 635 | while (fl && !net_eq(fl->fl_net, net)) |
5983a3df BT |
636 | fl = fl->next; |
637 | if (fl) | |
1da177e4 | 638 | break; |
1da177e4 LT |
639 | } |
640 | return fl; | |
641 | } | |
642 | ||
643 | static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl) | |
644 | { | |
645 | struct ip6fl_iter_state *state = ip6fl_seq_private(seq); | |
5983a3df | 646 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
647 | |
648 | fl = fl->next; | |
5983a3df | 649 | try_again: |
09ad9bc7 | 650 | while (fl && !net_eq(fl->fl_net, net)) |
5983a3df BT |
651 | fl = fl->next; |
652 | ||
1da177e4 | 653 | while (!fl) { |
5983a3df | 654 | if (++state->bucket <= FL_HASH_MASK) { |
1da177e4 | 655 | fl = fl_ht[state->bucket]; |
5983a3df BT |
656 | goto try_again; |
657 | } else | |
bcd62075 | 658 | break; |
1da177e4 LT |
659 | } |
660 | return fl; | |
661 | } | |
662 | ||
663 | static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos) | |
664 | { | |
665 | struct ip6_flowlabel *fl = ip6fl_get_first(seq); | |
666 | if (fl) | |
667 | while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL) | |
668 | --pos; | |
669 | return pos ? NULL : fl; | |
670 | } | |
671 | ||
672 | static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos) | |
9a429c49 | 673 | __acquires(ip6_fl_lock) |
1da177e4 LT |
674 | { |
675 | read_lock_bh(&ip6_fl_lock); | |
676 | return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
677 | } | |
678 | ||
679 | static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
680 | { | |
681 | struct ip6_flowlabel *fl; | |
682 | ||
683 | if (v == SEQ_START_TOKEN) | |
684 | fl = ip6fl_get_first(seq); | |
685 | else | |
686 | fl = ip6fl_get_next(seq, v); | |
687 | ++*pos; | |
688 | return fl; | |
689 | } | |
690 | ||
691 | static void ip6fl_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 692 | __releases(ip6_fl_lock) |
1da177e4 LT |
693 | { |
694 | read_unlock_bh(&ip6_fl_lock); | |
695 | } | |
696 | ||
1b7c2dbc | 697 | static int ip6fl_seq_show(struct seq_file *seq, void *v) |
1da177e4 | 698 | { |
1b7c2dbc JM |
699 | if (v == SEQ_START_TOKEN) |
700 | seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n", | |
701 | "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt"); | |
702 | else { | |
703 | struct ip6_flowlabel *fl = v; | |
1da177e4 | 704 | seq_printf(seq, |
4b7a4274 | 705 | "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n", |
1da177e4 LT |
706 | (unsigned)ntohl(fl->label), |
707 | fl->share, | |
708 | (unsigned)fl->owner, | |
709 | atomic_read(&fl->users), | |
710 | fl->linger/HZ, | |
711 | (long)(fl->expires - jiffies)/HZ, | |
b071195d | 712 | &fl->dst, |
1da177e4 | 713 | fl->opt ? fl->opt->opt_nflen : 0); |
1da177e4 | 714 | } |
1da177e4 LT |
715 | return 0; |
716 | } | |
717 | ||
56b3d975 | 718 | static const struct seq_operations ip6fl_seq_ops = { |
1da177e4 LT |
719 | .start = ip6fl_seq_start, |
720 | .next = ip6fl_seq_next, | |
721 | .stop = ip6fl_seq_stop, | |
722 | .show = ip6fl_seq_show, | |
723 | }; | |
724 | ||
725 | static int ip6fl_seq_open(struct inode *inode, struct file *file) | |
726 | { | |
5983a3df BT |
727 | return seq_open_net(inode, file, &ip6fl_seq_ops, |
728 | sizeof(struct ip6fl_iter_state)); | |
1da177e4 LT |
729 | } |
730 | ||
9a32144e | 731 | static const struct file_operations ip6fl_seq_fops = { |
1da177e4 LT |
732 | .owner = THIS_MODULE, |
733 | .open = ip6fl_seq_open, | |
734 | .read = seq_read, | |
735 | .llseek = seq_lseek, | |
5983a3df | 736 | .release = seq_release_net, |
1da177e4 | 737 | }; |
1da177e4 | 738 | |
2c8c1e72 | 739 | static int __net_init ip6_flowlabel_proc_init(struct net *net) |
0a3e78ac | 740 | { |
5983a3df BT |
741 | if (!proc_net_fops_create(net, "ip6_flowlabel", |
742 | S_IRUGO, &ip6fl_seq_fops)) | |
0a3e78ac DL |
743 | return -ENOMEM; |
744 | return 0; | |
745 | } | |
1da177e4 | 746 | |
2c8c1e72 | 747 | static void __net_exit ip6_flowlabel_proc_fini(struct net *net) |
1da177e4 | 748 | { |
0a3e78ac DL |
749 | proc_net_remove(net, "ip6_flowlabel"); |
750 | } | |
751 | #else | |
752 | static inline int ip6_flowlabel_proc_init(struct net *net) | |
753 | { | |
754 | return 0; | |
755 | } | |
756 | static inline void ip6_flowlabel_proc_fini(struct net *net) | |
757 | { | |
0a3e78ac | 758 | } |
1da177e4 | 759 | #endif |
0a3e78ac | 760 | |
2c8c1e72 | 761 | static void __net_exit ip6_flowlabel_net_exit(struct net *net) |
60e8fbc4 BT |
762 | { |
763 | ip6_fl_purge(net); | |
5983a3df | 764 | ip6_flowlabel_proc_fini(net); |
60e8fbc4 BT |
765 | } |
766 | ||
767 | static struct pernet_operations ip6_flowlabel_net_ops = { | |
5983a3df | 768 | .init = ip6_flowlabel_proc_init, |
60e8fbc4 BT |
769 | .exit = ip6_flowlabel_net_exit, |
770 | }; | |
771 | ||
0a3e78ac DL |
772 | int ip6_flowlabel_init(void) |
773 | { | |
5983a3df | 774 | return register_pernet_subsys(&ip6_flowlabel_net_ops); |
1da177e4 LT |
775 | } |
776 | ||
777 | void ip6_flowlabel_cleanup(void) | |
778 | { | |
779 | del_timer(&ip6_fl_gc_timer); | |
60e8fbc4 | 780 | unregister_pernet_subsys(&ip6_flowlabel_net_ops); |
1da177e4 | 781 | } |