blob: c836a6a20a34fb27d8df8766a2a199b12a02fc99 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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
Randy Dunlap4fc268d2006-01-11 12:17:47 -080012#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090023#include <linux/slab.h>
Paul Gortmakerbc3b2d72011-07-15 11:47:34 -040024#include <linux/export.h>
Eric W. Biederman4f82f452012-05-24 10:37:59 -060025#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020027#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <net/sock.h>
29
30#include <net/ipv6.h>
31#include <net/ndisc.h>
32#include <net/protocol.h>
33#include <net/ip6_route.h>
34#include <net/addrconf.h>
35#include <net/rawv6.h>
36#include <net/icmp.h>
37#include <net/transp_v6.h>
38
39#include <asm/uaccess.h>
40
41#define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified
42 in old IPv6 RFC. Well, it was reasonable value.
43 */
44#define FL_MAX_LINGER 60 /* Maximal linger timeout */
45
46/* FL hash table */
47
48#define FL_MAX_PER_SOCK 32
49#define FL_MAX_SIZE 4096
50#define FL_HASH_MASK 255
51#define FL_HASH(l) (ntohl(l)&FL_HASH_MASK)
52
53static atomic_t fl_size = ATOMIC_INIT(0);
54static struct ip6_flowlabel *fl_ht[FL_HASH_MASK+1];
55
56static void ip6_fl_gc(unsigned long dummy);
Ingo Molnar8d06afa2005-09-09 13:10:40 -070057static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
59/* FL hash table lock: it protects only of GC */
60
61static DEFINE_RWLOCK(ip6_fl_lock);
62
63/* Big socket sock */
64
65static DEFINE_RWLOCK(ip6_sk_fl_lock);
66
67
Benjamin Thery60e8fbc2008-03-26 16:53:08 -070068static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label)
Linus Torvalds1da177e2005-04-16 15:20:36 -070069{
70 struct ip6_flowlabel *fl;
71
72 for (fl=fl_ht[FL_HASH(label)]; fl; fl = fl->next) {
Octavian Purdila09ad9bc2009-11-25 15:14:13 -080073 if (fl->label == label && net_eq(fl->fl_net, net))
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 return fl;
75 }
76 return NULL;
77}
78
Benjamin Thery60e8fbc2008-03-26 16:53:08 -070079static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label)
Linus Torvalds1da177e2005-04-16 15:20:36 -070080{
81 struct ip6_flowlabel *fl;
82
83 read_lock_bh(&ip6_fl_lock);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -070084 fl = __fl_lookup(net, label);
Linus Torvalds1da177e2005-04-16 15:20:36 -070085 if (fl)
86 atomic_inc(&fl->users);
87 read_unlock_bh(&ip6_fl_lock);
88 return fl;
89}
90
91
92static void fl_free(struct ip6_flowlabel *fl)
93{
Eric W. Biederman4f82f452012-05-24 10:37:59 -060094 switch (fl->share) {
95 case IPV6_FL_S_PROCESS:
96 put_pid(fl->owner.pid);
97 break;
98 }
Benjamin Thery60e8fbc2008-03-26 16:53:08 -070099 if (fl) {
100 release_net(fl->fl_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 kfree(fl->opt);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700102 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 kfree(fl);
104}
105
106static void fl_release(struct ip6_flowlabel *fl)
107{
108 write_lock_bh(&ip6_fl_lock);
109
110 fl->lastuse = jiffies;
111 if (atomic_dec_and_test(&fl->users)) {
112 unsigned long ttd = fl->lastuse + fl->linger;
113 if (time_after(ttd, fl->expires))
114 fl->expires = ttd;
115 ttd = fl->expires;
116 if (fl->opt && fl->share == IPV6_FL_S_EXCL) {
117 struct ipv6_txoptions *opt = fl->opt;
118 fl->opt = NULL;
119 kfree(opt);
120 }
121 if (!timer_pending(&ip6_fl_gc_timer) ||
122 time_after(ip6_fl_gc_timer.expires, ttd))
123 mod_timer(&ip6_fl_gc_timer, ttd);
124 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 write_unlock_bh(&ip6_fl_lock);
126}
127
128static void ip6_fl_gc(unsigned long dummy)
129{
130 int i;
131 unsigned long now = jiffies;
132 unsigned long sched = 0;
133
134 write_lock(&ip6_fl_lock);
135
136 for (i=0; i<=FL_HASH_MASK; i++) {
137 struct ip6_flowlabel *fl, **flp;
138 flp = &fl_ht[i];
139 while ((fl=*flp) != NULL) {
140 if (atomic_read(&fl->users) == 0) {
141 unsigned long ttd = fl->lastuse + fl->linger;
142 if (time_after(ttd, fl->expires))
143 fl->expires = ttd;
144 ttd = fl->expires;
145 if (time_after_eq(now, ttd)) {
146 *flp = fl->next;
147 fl_free(fl);
148 atomic_dec(&fl_size);
149 continue;
150 }
151 if (!sched || time_before(ttd, sched))
152 sched = ttd;
153 }
154 flp = &fl->next;
155 }
156 }
157 if (!sched && atomic_read(&fl_size))
158 sched = now + FL_MAX_LINGER;
159 if (sched) {
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700160 mod_timer(&ip6_fl_gc_timer, sched);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 }
162 write_unlock(&ip6_fl_lock);
163}
164
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000165static void __net_exit ip6_fl_purge(struct net *net)
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700166{
167 int i;
168
169 write_lock(&ip6_fl_lock);
170 for (i = 0; i <= FL_HASH_MASK; i++) {
171 struct ip6_flowlabel *fl, **flp;
172 flp = &fl_ht[i];
173 while ((fl = *flp) != NULL) {
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800174 if (net_eq(fl->fl_net, net) &&
175 atomic_read(&fl->users) == 0) {
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700176 *flp = fl->next;
177 fl_free(fl);
178 atomic_dec(&fl_size);
179 continue;
180 }
181 flp = &fl->next;
182 }
183 }
184 write_unlock(&ip6_fl_lock);
185}
186
187static struct ip6_flowlabel *fl_intern(struct net *net,
188 struct ip6_flowlabel *fl, __be32 label)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700190 struct ip6_flowlabel *lfl;
191
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 fl->label = label & IPV6_FLOWLABEL_MASK;
193
194 write_lock_bh(&ip6_fl_lock);
195 if (label == 0) {
196 for (;;) {
197 fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK;
198 if (fl->label) {
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700199 lfl = __fl_lookup(net, fl->label);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 if (lfl == NULL)
201 break;
202 }
203 }
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700204 } else {
205 /*
206 * we dropper the ip6_fl_lock, so this entry could reappear
207 * and we need to recheck with it.
208 *
209 * OTOH no need to search the active socket first, like it is
210 * done in ipv6_flowlabel_opt - sock is locked, so new entry
211 * with the same label can only appear on another sock
212 */
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700213 lfl = __fl_lookup(net, fl->label);
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700214 if (lfl != NULL) {
215 atomic_inc(&lfl->users);
216 write_unlock_bh(&ip6_fl_lock);
217 return lfl;
218 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 }
220
221 fl->lastuse = jiffies;
222 fl->next = fl_ht[FL_HASH(fl->label)];
223 fl_ht[FL_HASH(fl->label)] = fl;
224 atomic_inc(&fl_size);
225 write_unlock_bh(&ip6_fl_lock);
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700226 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227}
228
229
230
231/* Socket flowlabel lists */
232
Al Viro90bcaf72006-11-08 00:25:17 -0800233struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234{
235 struct ipv6_fl_socklist *sfl;
236 struct ipv6_pinfo *np = inet6_sk(sk);
237
238 label &= IPV6_FLOWLABEL_MASK;
239
Pavel Emelyanovbd0bf572007-10-18 05:15:57 -0700240 read_lock_bh(&ip6_sk_fl_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 for (sfl=np->ipv6_fl_list; sfl; sfl = sfl->next) {
242 struct ip6_flowlabel *fl = sfl->fl;
243 if (fl->label == label) {
244 fl->lastuse = jiffies;
245 atomic_inc(&fl->users);
Pavel Emelyanov52f095e2007-10-18 05:38:48 -0700246 read_unlock_bh(&ip6_sk_fl_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 return fl;
248 }
249 }
Pavel Emelyanovbd0bf572007-10-18 05:15:57 -0700250 read_unlock_bh(&ip6_sk_fl_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 return NULL;
252}
253
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -0800254EXPORT_SYMBOL_GPL(fl6_sock_lookup);
255
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256void fl6_free_socklist(struct sock *sk)
257{
258 struct ipv6_pinfo *np = inet6_sk(sk);
259 struct ipv6_fl_socklist *sfl;
260
261 while ((sfl = np->ipv6_fl_list) != NULL) {
262 np->ipv6_fl_list = sfl->next;
263 fl_release(sfl->fl);
264 kfree(sfl);
265 }
266}
267
268/* Service routines */
269
270
271/*
272 It is the only difficult place. flowlabel enforces equal headers
273 before and including routing header, however user may supply options
274 following rthdr.
275 */
276
277struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space,
278 struct ip6_flowlabel * fl,
279 struct ipv6_txoptions * fopt)
280{
YOSHIFUJI Hideakidf9890c2005-11-20 12:23:18 +0900281 struct ipv6_txoptions * fl_opt = fl->opt;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900282
YOSHIFUJI Hideakidf9890c2005-11-20 12:23:18 +0900283 if (fopt == NULL || fopt->opt_flen == 0)
284 return fl_opt;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900285
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 if (fl_opt != NULL) {
287 opt_space->hopopt = fl_opt->hopopt;
YOSHIFUJI Hideakidf9890c2005-11-20 12:23:18 +0900288 opt_space->dst0opt = fl_opt->dst0opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 opt_space->srcrt = fl_opt->srcrt;
290 opt_space->opt_nflen = fl_opt->opt_nflen;
291 } else {
292 if (fopt->opt_nflen == 0)
293 return fopt;
294 opt_space->hopopt = NULL;
295 opt_space->dst0opt = NULL;
296 opt_space->srcrt = NULL;
297 opt_space->opt_nflen = 0;
298 }
299 opt_space->dst1opt = fopt->dst1opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 opt_space->opt_flen = fopt->opt_flen;
301 return opt_space;
302}
Chris Elstona495f832012-04-29 21:48:53 +0000303EXPORT_SYMBOL_GPL(fl6_merge_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304
305static unsigned long check_linger(unsigned long ttl)
306{
307 if (ttl < FL_MIN_LINGER)
308 return FL_MIN_LINGER*HZ;
309 if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
310 return 0;
311 return ttl*HZ;
312}
313
314static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
315{
316 linger = check_linger(linger);
317 if (!linger)
318 return -EPERM;
319 expires = check_linger(expires);
320 if (!expires)
321 return -EPERM;
322 fl->lastuse = jiffies;
323 if (time_before(fl->linger, linger))
324 fl->linger = linger;
325 if (time_before(expires, fl->linger))
326 expires = fl->linger;
327 if (time_before(fl->expires, fl->lastuse + expires))
328 fl->expires = fl->lastuse + expires;
329 return 0;
330}
331
332static struct ip6_flowlabel *
Maciej Żenczykowskiec0506d2011-08-28 12:35:31 +0000333fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq,
334 char __user *optval, int optlen, int *err_p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335{
David S. Miller684de402009-02-06 00:49:55 -0800336 struct ip6_flowlabel *fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 int olen;
338 int addr_type;
339 int err;
340
David S. Miller684de402009-02-06 00:49:55 -0800341 olen = optlen - CMSG_ALIGN(sizeof(*freq));
342 err = -EINVAL;
343 if (olen > 64 * 1024)
344 goto done;
345
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 err = -ENOMEM;
Ingo Oeser0c600ed2006-03-20 23:01:32 -0800347 fl = kzalloc(sizeof(*fl), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 if (fl == NULL)
349 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 if (olen > 0) {
352 struct msghdr msg;
David S. Miller4c9483b2011-03-12 16:22:43 -0500353 struct flowi6 flowi6;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 int junk;
355
356 err = -ENOMEM;
357 fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
358 if (fl->opt == NULL)
359 goto done;
360
361 memset(fl->opt, 0, sizeof(*fl->opt));
362 fl->opt->tot_len = sizeof(*fl->opt) + olen;
363 err = -EFAULT;
364 if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
365 goto done;
366
367 msg.msg_controllen = olen;
368 msg.msg_control = (void*)(fl->opt+1);
David S. Miller4c9483b2011-03-12 16:22:43 -0500369 memset(&flowi6, 0, sizeof(flowi6));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
Maciej Żenczykowskiec0506d2011-08-28 12:35:31 +0000371 err = datagram_send_ctl(net, sk, &msg, &flowi6, fl->opt, &junk,
Brian Haley13b52cd2010-04-23 11:26:08 +0000372 &junk, &junk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 if (err)
374 goto done;
375 err = -EINVAL;
376 if (fl->opt->opt_flen)
377 goto done;
378 if (fl->opt->opt_nflen == 0) {
379 kfree(fl->opt);
380 fl->opt = NULL;
381 }
382 }
383
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700384 fl->fl_net = hold_net(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 fl->expires = jiffies;
386 err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
387 if (err)
388 goto done;
389 fl->share = freq->flr_share;
390 addr_type = ipv6_addr_type(&freq->flr_dst);
Joe Perches35700212009-11-24 14:52:52 -0800391 if ((addr_type & IPV6_ADDR_MAPPED) ||
392 addr_type == IPV6_ADDR_ANY) {
James Morrisc6817e42006-10-30 18:56:06 -0800393 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 goto done;
James Morrisc6817e42006-10-30 18:56:06 -0800395 }
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000396 fl->dst = freq->flr_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 atomic_set(&fl->users, 1);
398 switch (fl->share) {
399 case IPV6_FL_S_EXCL:
400 case IPV6_FL_S_ANY:
401 break;
402 case IPV6_FL_S_PROCESS:
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600403 fl->owner.pid = get_task_pid(current, PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 break;
405 case IPV6_FL_S_USER:
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600406 fl->owner.uid = current_euid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 break;
408 default:
409 err = -EINVAL;
410 goto done;
411 }
412 return fl;
413
414done:
415 fl_free(fl);
416 *err_p = err;
417 return NULL;
418}
419
420static int mem_check(struct sock *sk)
421{
422 struct ipv6_pinfo *np = inet6_sk(sk);
423 struct ipv6_fl_socklist *sfl;
424 int room = FL_MAX_SIZE - atomic_read(&fl_size);
425 int count = 0;
426
427 if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
428 return 0;
429
430 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next)
431 count++;
432
433 if (room <= 0 ||
434 ((count >= FL_MAX_PER_SOCK ||
Joe Perches35700212009-11-24 14:52:52 -0800435 (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) &&
436 !capable(CAP_NET_ADMIN)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 return -ENOBUFS;
438
439 return 0;
440}
441
Eric Dumazeta50feda2012-05-18 18:57:34 +0000442static bool ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443{
444 if (h1 == h2)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000445 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 if (h1 == NULL || h2 == NULL)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000447 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 if (h1->hdrlen != h2->hdrlen)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000449 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1));
451}
452
Eric Dumazeta50feda2012-05-18 18:57:34 +0000453static bool ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454{
455 if (o1 == o2)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000456 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 if (o1 == NULL || o2 == NULL)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000458 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 if (o1->opt_nflen != o2->opt_nflen)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000460 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt))
Eric Dumazeta50feda2012-05-18 18:57:34 +0000462 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt))
Eric Dumazeta50feda2012-05-18 18:57:34 +0000464 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt))
Eric Dumazeta50feda2012-05-18 18:57:34 +0000466 return true;
467 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468}
469
Pavel Emelyanov04028042007-10-18 05:14:58 -0700470static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
471 struct ip6_flowlabel *fl)
472{
473 write_lock_bh(&ip6_sk_fl_lock);
474 sfl->fl = fl;
475 sfl->next = np->ipv6_fl_list;
476 np->ipv6_fl_list = sfl;
477 write_unlock_bh(&ip6_sk_fl_lock);
478}
479
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
481{
Ingo Molnar55205d42008-11-25 16:50:30 -0800482 int uninitialized_var(err);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700483 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 struct ipv6_pinfo *np = inet6_sk(sk);
485 struct in6_flowlabel_req freq;
486 struct ipv6_fl_socklist *sfl1=NULL;
487 struct ipv6_fl_socklist *sfl, **sflp;
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700488 struct ip6_flowlabel *fl, *fl1 = NULL;
489
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
491 if (optlen < sizeof(freq))
492 return -EINVAL;
493
494 if (copy_from_user(&freq, optval, sizeof(freq)))
495 return -EFAULT;
496
497 switch (freq.flr_action) {
498 case IPV6_FL_A_PUT:
499 write_lock_bh(&ip6_sk_fl_lock);
500 for (sflp = &np->ipv6_fl_list; (sfl=*sflp)!=NULL; sflp = &sfl->next) {
501 if (sfl->fl->label == freq.flr_label) {
502 if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
503 np->flow_label &= ~IPV6_FLOWLABEL_MASK;
504 *sflp = sfl->next;
505 write_unlock_bh(&ip6_sk_fl_lock);
506 fl_release(sfl->fl);
507 kfree(sfl);
508 return 0;
509 }
510 }
511 write_unlock_bh(&ip6_sk_fl_lock);
512 return -ESRCH;
513
514 case IPV6_FL_A_RENEW:
515 read_lock_bh(&ip6_sk_fl_lock);
516 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
517 if (sfl->fl->label == freq.flr_label) {
518 err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
519 read_unlock_bh(&ip6_sk_fl_lock);
520 return err;
521 }
522 }
523 read_unlock_bh(&ip6_sk_fl_lock);
524
525 if (freq.flr_share == IPV6_FL_S_NONE && capable(CAP_NET_ADMIN)) {
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700526 fl = fl_lookup(net, freq.flr_label);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 if (fl) {
528 err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
529 fl_release(fl);
530 return err;
531 }
532 }
533 return -ESRCH;
534
535 case IPV6_FL_A_GET:
536 if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
537 return -EINVAL;
538
Maciej Żenczykowskiec0506d2011-08-28 12:35:31 +0000539 fl = fl_create(net, sk, &freq, optval, optlen, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 if (fl == NULL)
541 return err;
542 sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
543
544 if (freq.flr_label) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 err = -EEXIST;
546 read_lock_bh(&ip6_sk_fl_lock);
547 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
548 if (sfl->fl->label == freq.flr_label) {
549 if (freq.flr_flags&IPV6_FL_F_EXCL) {
550 read_unlock_bh(&ip6_sk_fl_lock);
551 goto done;
552 }
553 fl1 = sfl->fl;
Yan Zheng4ea6a802005-10-24 19:55:23 +0800554 atomic_inc(&fl1->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 break;
556 }
557 }
558 read_unlock_bh(&ip6_sk_fl_lock);
559
560 if (fl1 == NULL)
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700561 fl1 = fl_lookup(net, freq.flr_label);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 if (fl1) {
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700563recheck:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 err = -EEXIST;
565 if (freq.flr_flags&IPV6_FL_F_EXCL)
566 goto release;
567 err = -EPERM;
568 if (fl1->share == IPV6_FL_S_EXCL ||
569 fl1->share != fl->share ||
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600570 ((fl1->share == IPV6_FL_S_PROCESS) &&
571 (fl1->owner.pid == fl->owner.pid)) ||
572 ((fl1->share == IPV6_FL_S_USER) &&
573 uid_eq(fl1->owner.uid, fl->owner.uid)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 goto release;
575
576 err = -EINVAL;
577 if (!ipv6_addr_equal(&fl1->dst, &fl->dst) ||
578 ipv6_opt_cmp(fl1->opt, fl->opt))
579 goto release;
580
581 err = -ENOMEM;
582 if (sfl1 == NULL)
583 goto release;
584 if (fl->linger > fl1->linger)
585 fl1->linger = fl->linger;
586 if ((long)(fl->expires - fl1->expires) > 0)
587 fl1->expires = fl->expires;
Pavel Emelyanov04028042007-10-18 05:14:58 -0700588 fl_link(np, sfl1, fl1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 fl_free(fl);
590 return 0;
591
592release:
593 fl_release(fl1);
594 goto done;
595 }
596 }
597 err = -ENOENT;
598 if (!(freq.flr_flags&IPV6_FL_F_CREATE))
599 goto done;
600
601 err = -ENOMEM;
602 if (sfl1 == NULL || (err = mem_check(sk)) != 0)
603 goto done;
604
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700605 fl1 = fl_intern(net, fl, freq.flr_label);
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700606 if (fl1 != NULL)
607 goto recheck;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608
David S. Miller6c94d362005-05-29 20:28:01 -0700609 if (!freq.flr_label) {
610 if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
611 &fl->label, sizeof(fl->label))) {
612 /* Intentionally ignore fault. */
613 }
614 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615
Pavel Emelyanov04028042007-10-18 05:14:58 -0700616 fl_link(np, sfl1, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 return 0;
618
619 default:
620 return -EINVAL;
621 }
622
623done:
624 fl_free(fl);
625 kfree(sfl1);
626 return err;
627}
628
629#ifdef CONFIG_PROC_FS
630
631struct ip6fl_iter_state {
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700632 struct seq_net_private p;
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600633 struct pid_namespace *pid_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 int bucket;
635};
636
637#define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private)
638
639static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
640{
641 struct ip6_flowlabel *fl = NULL;
642 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700643 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644
645 for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700646 fl = fl_ht[state->bucket];
647
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800648 while (fl && !net_eq(fl->fl_net, net))
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700649 fl = fl->next;
650 if (fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 }
653 return fl;
654}
655
656static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
657{
658 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700659 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
661 fl = fl->next;
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700662try_again:
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800663 while (fl && !net_eq(fl->fl_net, net))
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700664 fl = fl->next;
665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 while (!fl) {
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700667 if (++state->bucket <= FL_HASH_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 fl = fl_ht[state->bucket];
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700669 goto try_again;
670 } else
James Morrisbcd62072006-10-30 15:08:42 -0800671 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 }
673 return fl;
674}
675
676static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
677{
678 struct ip6_flowlabel *fl = ip6fl_get_first(seq);
679 if (fl)
680 while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
681 --pos;
682 return pos ? NULL : fl;
683}
684
685static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800686 __acquires(ip6_fl_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687{
688 read_lock_bh(&ip6_fl_lock);
689 return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
690}
691
692static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
693{
694 struct ip6_flowlabel *fl;
695
696 if (v == SEQ_START_TOKEN)
697 fl = ip6fl_get_first(seq);
698 else
699 fl = ip6fl_get_next(seq, v);
700 ++*pos;
701 return fl;
702}
703
704static void ip6fl_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800705 __releases(ip6_fl_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706{
707 read_unlock_bh(&ip6_fl_lock);
708}
709
James Morris1b7c2db2006-10-31 00:43:44 -0800710static int ip6fl_seq_show(struct seq_file *seq, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711{
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600712 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
James Morris1b7c2db2006-10-31 00:43:44 -0800713 if (v == SEQ_START_TOKEN)
714 seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n",
715 "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt");
716 else {
717 struct ip6_flowlabel *fl = v;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 seq_printf(seq,
Harvey Harrison4b7a4272008-10-29 12:50:24 -0700719 "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
Eric Dumazet95c96172012-04-15 05:58:06 +0000720 (unsigned int)ntohl(fl->label),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 fl->share,
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600722 ((fl->share == IPV6_FL_S_PROCESS) ?
723 pid_nr_ns(fl->owner.pid, state->pid_ns) :
724 ((fl->share == IPV6_FL_S_USER) ?
725 from_kuid_munged(seq_user_ns(seq), fl->owner.uid) :
726 0)),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 atomic_read(&fl->users),
728 fl->linger/HZ,
729 (long)(fl->expires - jiffies)/HZ,
Harvey Harrisonb0711952008-10-28 16:05:40 -0700730 &fl->dst,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 fl->opt ? fl->opt->opt_nflen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 return 0;
734}
735
Philippe De Muyter56b3d972007-07-10 23:07:31 -0700736static const struct seq_operations ip6fl_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 .start = ip6fl_seq_start,
738 .next = ip6fl_seq_next,
739 .stop = ip6fl_seq_stop,
740 .show = ip6fl_seq_show,
741};
742
743static int ip6fl_seq_open(struct inode *inode, struct file *file)
744{
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600745 struct seq_file *seq;
746 struct ip6fl_iter_state *state;
747 int err;
748
749 err = seq_open_net(inode, file, &ip6fl_seq_ops,
750 sizeof(struct ip6fl_iter_state));
751
752 if (!err) {
753 seq = file->private_data;
754 state = ip6fl_seq_private(seq);
755 rcu_read_lock();
756 state->pid_ns = get_pid_ns(task_active_pid_ns(current));
757 rcu_read_unlock();
758 }
759 return err;
760}
761
762static int ip6fl_seq_release(struct inode *inode, struct file *file)
763{
764 struct seq_file *seq = file->private_data;
765 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
766 put_pid_ns(state->pid_ns);
767 return seq_release_net(inode, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768}
769
Arjan van de Ven9a321442007-02-12 00:55:35 -0800770static const struct file_operations ip6fl_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 .owner = THIS_MODULE,
772 .open = ip6fl_seq_open,
773 .read = seq_read,
774 .llseek = seq_lseek,
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600775 .release = ip6fl_seq_release,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000778static int __net_init ip6_flowlabel_proc_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779{
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700780 if (!proc_net_fops_create(net, "ip6_flowlabel",
781 S_IRUGO, &ip6fl_seq_fops))
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800782 return -ENOMEM;
783 return 0;
784}
785
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000786static void __net_exit ip6_flowlabel_proc_fini(struct net *net)
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800787{
788 proc_net_remove(net, "ip6_flowlabel");
789}
790#else
791static inline int ip6_flowlabel_proc_init(struct net *net)
792{
793 return 0;
794}
795static inline void ip6_flowlabel_proc_fini(struct net *net)
796{
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800797}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798#endif
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800799
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000800static void __net_exit ip6_flowlabel_net_exit(struct net *net)
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700801{
802 ip6_fl_purge(net);
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700803 ip6_flowlabel_proc_fini(net);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700804}
805
806static struct pernet_operations ip6_flowlabel_net_ops = {
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700807 .init = ip6_flowlabel_proc_init,
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700808 .exit = ip6_flowlabel_net_exit,
809};
810
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800811int ip6_flowlabel_init(void)
812{
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700813 return register_pernet_subsys(&ip6_flowlabel_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814}
815
816void ip6_flowlabel_cleanup(void)
817{
818 del_timer(&ip6_fl_gc_timer);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700819 unregister_pernet_subsys(&ip6_flowlabel_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820}