blob: 29124b7a04c8dfb335df69014b5348e22b3fb6ab [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{
Benjamin Thery60e8fbc2008-03-26 16:53:08 -070094 if (fl) {
Dan Carpenter898132a2012-08-16 16:15:02 +030095 if (fl->share == IPV6_FL_S_PROCESS)
96 put_pid(fl->owner.pid);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -070097 release_net(fl->fl_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -070098 kfree(fl->opt);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -070099 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 kfree(fl);
101}
102
103static void fl_release(struct ip6_flowlabel *fl)
104{
105 write_lock_bh(&ip6_fl_lock);
106
107 fl->lastuse = jiffies;
108 if (atomic_dec_and_test(&fl->users)) {
109 unsigned long ttd = fl->lastuse + fl->linger;
110 if (time_after(ttd, fl->expires))
111 fl->expires = ttd;
112 ttd = fl->expires;
113 if (fl->opt && fl->share == IPV6_FL_S_EXCL) {
114 struct ipv6_txoptions *opt = fl->opt;
115 fl->opt = NULL;
116 kfree(opt);
117 }
118 if (!timer_pending(&ip6_fl_gc_timer) ||
119 time_after(ip6_fl_gc_timer.expires, ttd))
120 mod_timer(&ip6_fl_gc_timer, ttd);
121 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 write_unlock_bh(&ip6_fl_lock);
123}
124
125static void ip6_fl_gc(unsigned long dummy)
126{
127 int i;
128 unsigned long now = jiffies;
129 unsigned long sched = 0;
130
131 write_lock(&ip6_fl_lock);
132
133 for (i=0; i<=FL_HASH_MASK; i++) {
134 struct ip6_flowlabel *fl, **flp;
135 flp = &fl_ht[i];
136 while ((fl=*flp) != NULL) {
137 if (atomic_read(&fl->users) == 0) {
138 unsigned long ttd = fl->lastuse + fl->linger;
139 if (time_after(ttd, fl->expires))
140 fl->expires = ttd;
141 ttd = fl->expires;
142 if (time_after_eq(now, ttd)) {
143 *flp = fl->next;
144 fl_free(fl);
145 atomic_dec(&fl_size);
146 continue;
147 }
148 if (!sched || time_before(ttd, sched))
149 sched = ttd;
150 }
151 flp = &fl->next;
152 }
153 }
154 if (!sched && atomic_read(&fl_size))
155 sched = now + FL_MAX_LINGER;
156 if (sched) {
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700157 mod_timer(&ip6_fl_gc_timer, sched);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 }
159 write_unlock(&ip6_fl_lock);
160}
161
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000162static void __net_exit ip6_fl_purge(struct net *net)
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700163{
164 int i;
165
166 write_lock(&ip6_fl_lock);
167 for (i = 0; i <= FL_HASH_MASK; i++) {
168 struct ip6_flowlabel *fl, **flp;
169 flp = &fl_ht[i];
170 while ((fl = *flp) != NULL) {
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800171 if (net_eq(fl->fl_net, net) &&
172 atomic_read(&fl->users) == 0) {
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700173 *flp = fl->next;
174 fl_free(fl);
175 atomic_dec(&fl_size);
176 continue;
177 }
178 flp = &fl->next;
179 }
180 }
181 write_unlock(&ip6_fl_lock);
182}
183
184static struct ip6_flowlabel *fl_intern(struct net *net,
185 struct ip6_flowlabel *fl, __be32 label)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186{
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700187 struct ip6_flowlabel *lfl;
188
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 fl->label = label & IPV6_FLOWLABEL_MASK;
190
191 write_lock_bh(&ip6_fl_lock);
192 if (label == 0) {
193 for (;;) {
194 fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK;
195 if (fl->label) {
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700196 lfl = __fl_lookup(net, fl->label);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 if (lfl == NULL)
198 break;
199 }
200 }
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700201 } else {
202 /*
203 * we dropper the ip6_fl_lock, so this entry could reappear
204 * and we need to recheck with it.
205 *
206 * OTOH no need to search the active socket first, like it is
207 * done in ipv6_flowlabel_opt - sock is locked, so new entry
208 * with the same label can only appear on another sock
209 */
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700210 lfl = __fl_lookup(net, fl->label);
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700211 if (lfl != NULL) {
212 atomic_inc(&lfl->users);
213 write_unlock_bh(&ip6_fl_lock);
214 return lfl;
215 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 }
217
218 fl->lastuse = jiffies;
219 fl->next = fl_ht[FL_HASH(fl->label)];
220 fl_ht[FL_HASH(fl->label)] = fl;
221 atomic_inc(&fl_size);
222 write_unlock_bh(&ip6_fl_lock);
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700223 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224}
225
226
227
228/* Socket flowlabel lists */
229
Al Viro90bcaf72006-11-08 00:25:17 -0800230struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231{
232 struct ipv6_fl_socklist *sfl;
233 struct ipv6_pinfo *np = inet6_sk(sk);
234
235 label &= IPV6_FLOWLABEL_MASK;
236
Pavel Emelyanovbd0bf572007-10-18 05:15:57 -0700237 read_lock_bh(&ip6_sk_fl_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 for (sfl=np->ipv6_fl_list; sfl; sfl = sfl->next) {
239 struct ip6_flowlabel *fl = sfl->fl;
240 if (fl->label == label) {
241 fl->lastuse = jiffies;
242 atomic_inc(&fl->users);
Pavel Emelyanov52f095e2007-10-18 05:38:48 -0700243 read_unlock_bh(&ip6_sk_fl_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 return fl;
245 }
246 }
Pavel Emelyanovbd0bf572007-10-18 05:15:57 -0700247 read_unlock_bh(&ip6_sk_fl_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 return NULL;
249}
250
Arnaldo Carvalho de Melo3cf3dc62005-12-13 23:23:20 -0800251EXPORT_SYMBOL_GPL(fl6_sock_lookup);
252
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253void fl6_free_socklist(struct sock *sk)
254{
255 struct ipv6_pinfo *np = inet6_sk(sk);
256 struct ipv6_fl_socklist *sfl;
257
258 while ((sfl = np->ipv6_fl_list) != NULL) {
259 np->ipv6_fl_list = sfl->next;
260 fl_release(sfl->fl);
261 kfree(sfl);
262 }
263}
264
265/* Service routines */
266
267
268/*
269 It is the only difficult place. flowlabel enforces equal headers
270 before and including routing header, however user may supply options
271 following rthdr.
272 */
273
274struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space,
275 struct ip6_flowlabel * fl,
276 struct ipv6_txoptions * fopt)
277{
YOSHIFUJI Hideakidf9890c2005-11-20 12:23:18 +0900278 struct ipv6_txoptions * fl_opt = fl->opt;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900279
YOSHIFUJI Hideakidf9890c2005-11-20 12:23:18 +0900280 if (fopt == NULL || fopt->opt_flen == 0)
281 return fl_opt;
YOSHIFUJI Hideaki1ab14572007-02-09 23:24:49 +0900282
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 if (fl_opt != NULL) {
284 opt_space->hopopt = fl_opt->hopopt;
YOSHIFUJI Hideakidf9890c2005-11-20 12:23:18 +0900285 opt_space->dst0opt = fl_opt->dst0opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 opt_space->srcrt = fl_opt->srcrt;
287 opt_space->opt_nflen = fl_opt->opt_nflen;
288 } else {
289 if (fopt->opt_nflen == 0)
290 return fopt;
291 opt_space->hopopt = NULL;
292 opt_space->dst0opt = NULL;
293 opt_space->srcrt = NULL;
294 opt_space->opt_nflen = 0;
295 }
296 opt_space->dst1opt = fopt->dst1opt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 opt_space->opt_flen = fopt->opt_flen;
298 return opt_space;
299}
Chris Elstona495f832012-04-29 21:48:53 +0000300EXPORT_SYMBOL_GPL(fl6_merge_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
302static unsigned long check_linger(unsigned long ttl)
303{
304 if (ttl < FL_MIN_LINGER)
305 return FL_MIN_LINGER*HZ;
306 if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
307 return 0;
308 return ttl*HZ;
309}
310
311static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
312{
313 linger = check_linger(linger);
314 if (!linger)
315 return -EPERM;
316 expires = check_linger(expires);
317 if (!expires)
318 return -EPERM;
319 fl->lastuse = jiffies;
320 if (time_before(fl->linger, linger))
321 fl->linger = linger;
322 if (time_before(expires, fl->linger))
323 expires = fl->linger;
324 if (time_before(fl->expires, fl->lastuse + expires))
325 fl->expires = fl->lastuse + expires;
326 return 0;
327}
328
329static struct ip6_flowlabel *
Maciej Żenczykowskiec0506d2011-08-28 12:35:31 +0000330fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq,
331 char __user *optval, int optlen, int *err_p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332{
David S. Miller684de402009-02-06 00:49:55 -0800333 struct ip6_flowlabel *fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 int olen;
335 int addr_type;
336 int err;
337
David S. Miller684de402009-02-06 00:49:55 -0800338 olen = optlen - CMSG_ALIGN(sizeof(*freq));
339 err = -EINVAL;
340 if (olen > 64 * 1024)
341 goto done;
342
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 err = -ENOMEM;
Ingo Oeser0c600ed2006-03-20 23:01:32 -0800344 fl = kzalloc(sizeof(*fl), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 if (fl == NULL)
346 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 if (olen > 0) {
349 struct msghdr msg;
David S. Miller4c9483b2011-03-12 16:22:43 -0500350 struct flowi6 flowi6;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 int junk;
352
353 err = -ENOMEM;
354 fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
355 if (fl->opt == NULL)
356 goto done;
357
358 memset(fl->opt, 0, sizeof(*fl->opt));
359 fl->opt->tot_len = sizeof(*fl->opt) + olen;
360 err = -EFAULT;
361 if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
362 goto done;
363
364 msg.msg_controllen = olen;
365 msg.msg_control = (void*)(fl->opt+1);
David S. Miller4c9483b2011-03-12 16:22:43 -0500366 memset(&flowi6, 0, sizeof(flowi6));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367
Maciej Żenczykowskiec0506d2011-08-28 12:35:31 +0000368 err = datagram_send_ctl(net, sk, &msg, &flowi6, fl->opt, &junk,
Brian Haley13b52cd2010-04-23 11:26:08 +0000369 &junk, &junk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 if (err)
371 goto done;
372 err = -EINVAL;
373 if (fl->opt->opt_flen)
374 goto done;
375 if (fl->opt->opt_nflen == 0) {
376 kfree(fl->opt);
377 fl->opt = NULL;
378 }
379 }
380
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700381 fl->fl_net = hold_net(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 fl->expires = jiffies;
383 err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
384 if (err)
385 goto done;
386 fl->share = freq->flr_share;
387 addr_type = ipv6_addr_type(&freq->flr_dst);
Joe Perches35700212009-11-24 14:52:52 -0800388 if ((addr_type & IPV6_ADDR_MAPPED) ||
389 addr_type == IPV6_ADDR_ANY) {
James Morrisc6817e42006-10-30 18:56:06 -0800390 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 goto done;
James Morrisc6817e42006-10-30 18:56:06 -0800392 }
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000393 fl->dst = freq->flr_dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 atomic_set(&fl->users, 1);
395 switch (fl->share) {
396 case IPV6_FL_S_EXCL:
397 case IPV6_FL_S_ANY:
398 break;
399 case IPV6_FL_S_PROCESS:
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600400 fl->owner.pid = get_task_pid(current, PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 break;
402 case IPV6_FL_S_USER:
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600403 fl->owner.uid = current_euid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 break;
405 default:
406 err = -EINVAL;
407 goto done;
408 }
409 return fl;
410
411done:
412 fl_free(fl);
413 *err_p = err;
414 return NULL;
415}
416
417static int mem_check(struct sock *sk)
418{
419 struct ipv6_pinfo *np = inet6_sk(sk);
420 struct ipv6_fl_socklist *sfl;
421 int room = FL_MAX_SIZE - atomic_read(&fl_size);
422 int count = 0;
423
424 if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
425 return 0;
426
427 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next)
428 count++;
429
430 if (room <= 0 ||
431 ((count >= FL_MAX_PER_SOCK ||
Joe Perches35700212009-11-24 14:52:52 -0800432 (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) &&
433 !capable(CAP_NET_ADMIN)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 return -ENOBUFS;
435
436 return 0;
437}
438
Eric Dumazeta50feda2012-05-18 18:57:34 +0000439static bool ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440{
441 if (h1 == h2)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000442 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 if (h1 == NULL || h2 == NULL)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000444 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 if (h1->hdrlen != h2->hdrlen)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000446 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1));
448}
449
Eric Dumazeta50feda2012-05-18 18:57:34 +0000450static bool ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451{
452 if (o1 == o2)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000453 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 if (o1 == NULL || o2 == NULL)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000455 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 if (o1->opt_nflen != o2->opt_nflen)
Eric Dumazeta50feda2012-05-18 18:57:34 +0000457 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt))
Eric Dumazeta50feda2012-05-18 18:57:34 +0000459 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt))
Eric Dumazeta50feda2012-05-18 18:57:34 +0000461 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt))
Eric Dumazeta50feda2012-05-18 18:57:34 +0000463 return true;
464 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465}
466
Pavel Emelyanov04028042007-10-18 05:14:58 -0700467static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
468 struct ip6_flowlabel *fl)
469{
470 write_lock_bh(&ip6_sk_fl_lock);
471 sfl->fl = fl;
472 sfl->next = np->ipv6_fl_list;
473 np->ipv6_fl_list = sfl;
474 write_unlock_bh(&ip6_sk_fl_lock);
475}
476
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
478{
Ingo Molnar55205d42008-11-25 16:50:30 -0800479 int uninitialized_var(err);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700480 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 struct ipv6_pinfo *np = inet6_sk(sk);
482 struct in6_flowlabel_req freq;
483 struct ipv6_fl_socklist *sfl1=NULL;
484 struct ipv6_fl_socklist *sfl, **sflp;
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700485 struct ip6_flowlabel *fl, *fl1 = NULL;
486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
488 if (optlen < sizeof(freq))
489 return -EINVAL;
490
491 if (copy_from_user(&freq, optval, sizeof(freq)))
492 return -EFAULT;
493
494 switch (freq.flr_action) {
495 case IPV6_FL_A_PUT:
496 write_lock_bh(&ip6_sk_fl_lock);
497 for (sflp = &np->ipv6_fl_list; (sfl=*sflp)!=NULL; sflp = &sfl->next) {
498 if (sfl->fl->label == freq.flr_label) {
499 if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
500 np->flow_label &= ~IPV6_FLOWLABEL_MASK;
501 *sflp = sfl->next;
502 write_unlock_bh(&ip6_sk_fl_lock);
503 fl_release(sfl->fl);
504 kfree(sfl);
505 return 0;
506 }
507 }
508 write_unlock_bh(&ip6_sk_fl_lock);
509 return -ESRCH;
510
511 case IPV6_FL_A_RENEW:
512 read_lock_bh(&ip6_sk_fl_lock);
513 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
514 if (sfl->fl->label == freq.flr_label) {
515 err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
516 read_unlock_bh(&ip6_sk_fl_lock);
517 return err;
518 }
519 }
520 read_unlock_bh(&ip6_sk_fl_lock);
521
Eric W. Biedermanaf31f412012-11-16 03:03:06 +0000522 if (freq.flr_share == IPV6_FL_S_NONE &&
523 ns_capable(net->user_ns, CAP_NET_ADMIN)) {
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700524 fl = fl_lookup(net, freq.flr_label);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 if (fl) {
526 err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
527 fl_release(fl);
528 return err;
529 }
530 }
531 return -ESRCH;
532
533 case IPV6_FL_A_GET:
534 if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
535 return -EINVAL;
536
Maciej Żenczykowskiec0506d2011-08-28 12:35:31 +0000537 fl = fl_create(net, sk, &freq, optval, optlen, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 if (fl == NULL)
539 return err;
540 sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
541
542 if (freq.flr_label) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 err = -EEXIST;
544 read_lock_bh(&ip6_sk_fl_lock);
545 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
546 if (sfl->fl->label == freq.flr_label) {
547 if (freq.flr_flags&IPV6_FL_F_EXCL) {
548 read_unlock_bh(&ip6_sk_fl_lock);
549 goto done;
550 }
551 fl1 = sfl->fl;
Yan Zheng4ea6a802005-10-24 19:55:23 +0800552 atomic_inc(&fl1->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 break;
554 }
555 }
556 read_unlock_bh(&ip6_sk_fl_lock);
557
558 if (fl1 == NULL)
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700559 fl1 = fl_lookup(net, freq.flr_label);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 if (fl1) {
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700561recheck:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 err = -EEXIST;
563 if (freq.flr_flags&IPV6_FL_F_EXCL)
564 goto release;
565 err = -EPERM;
566 if (fl1->share == IPV6_FL_S_EXCL ||
567 fl1->share != fl->share ||
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600568 ((fl1->share == IPV6_FL_S_PROCESS) &&
569 (fl1->owner.pid == fl->owner.pid)) ||
570 ((fl1->share == IPV6_FL_S_USER) &&
571 uid_eq(fl1->owner.uid, fl->owner.uid)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 goto release;
573
574 err = -EINVAL;
575 if (!ipv6_addr_equal(&fl1->dst, &fl->dst) ||
576 ipv6_opt_cmp(fl1->opt, fl->opt))
577 goto release;
578
579 err = -ENOMEM;
580 if (sfl1 == NULL)
581 goto release;
582 if (fl->linger > fl1->linger)
583 fl1->linger = fl->linger;
584 if ((long)(fl->expires - fl1->expires) > 0)
585 fl1->expires = fl->expires;
Pavel Emelyanov04028042007-10-18 05:14:58 -0700586 fl_link(np, sfl1, fl1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 fl_free(fl);
588 return 0;
589
590release:
591 fl_release(fl1);
592 goto done;
593 }
594 }
595 err = -ENOENT;
596 if (!(freq.flr_flags&IPV6_FL_F_CREATE))
597 goto done;
598
599 err = -ENOMEM;
600 if (sfl1 == NULL || (err = mem_check(sk)) != 0)
601 goto done;
602
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700603 fl1 = fl_intern(net, fl, freq.flr_label);
Pavel Emelyanov78c2e502007-10-18 05:18:56 -0700604 if (fl1 != NULL)
605 goto recheck;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
David S. Miller6c94d362005-05-29 20:28:01 -0700607 if (!freq.flr_label) {
608 if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
609 &fl->label, sizeof(fl->label))) {
610 /* Intentionally ignore fault. */
611 }
612 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
Pavel Emelyanov04028042007-10-18 05:14:58 -0700614 fl_link(np, sfl1, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 return 0;
616
617 default:
618 return -EINVAL;
619 }
620
621done:
622 fl_free(fl);
623 kfree(sfl1);
624 return err;
625}
626
627#ifdef CONFIG_PROC_FS
628
629struct ip6fl_iter_state {
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700630 struct seq_net_private p;
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600631 struct pid_namespace *pid_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 int bucket;
633};
634
635#define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private)
636
637static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
638{
639 struct ip6_flowlabel *fl = NULL;
640 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700641 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642
643 for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700644 fl = fl_ht[state->bucket];
645
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800646 while (fl && !net_eq(fl->fl_net, net))
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700647 fl = fl->next;
648 if (fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 }
651 return fl;
652}
653
654static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
655{
656 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700657 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
659 fl = fl->next;
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700660try_again:
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800661 while (fl && !net_eq(fl->fl_net, net))
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700662 fl = fl->next;
663
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 while (!fl) {
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700665 if (++state->bucket <= FL_HASH_MASK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 fl = fl_ht[state->bucket];
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700667 goto try_again;
668 } else
James Morrisbcd62072006-10-30 15:08:42 -0800669 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 }
671 return fl;
672}
673
674static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
675{
676 struct ip6_flowlabel *fl = ip6fl_get_first(seq);
677 if (fl)
678 while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
679 --pos;
680 return pos ? NULL : fl;
681}
682
683static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800684 __acquires(ip6_fl_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685{
686 read_lock_bh(&ip6_fl_lock);
687 return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
688}
689
690static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
691{
692 struct ip6_flowlabel *fl;
693
694 if (v == SEQ_START_TOKEN)
695 fl = ip6fl_get_first(seq);
696 else
697 fl = ip6fl_get_next(seq, v);
698 ++*pos;
699 return fl;
700}
701
702static void ip6fl_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800703 __releases(ip6_fl_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704{
705 read_unlock_bh(&ip6_fl_lock);
706}
707
James Morris1b7c2db2006-10-31 00:43:44 -0800708static int ip6fl_seq_show(struct seq_file *seq, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709{
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600710 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
James Morris1b7c2db2006-10-31 00:43:44 -0800711 if (v == SEQ_START_TOKEN)
712 seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n",
713 "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt");
714 else {
715 struct ip6_flowlabel *fl = v;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 seq_printf(seq,
Harvey Harrison4b7a4272008-10-29 12:50:24 -0700717 "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
Eric Dumazet95c96172012-04-15 05:58:06 +0000718 (unsigned int)ntohl(fl->label),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 fl->share,
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600720 ((fl->share == IPV6_FL_S_PROCESS) ?
721 pid_nr_ns(fl->owner.pid, state->pid_ns) :
722 ((fl->share == IPV6_FL_S_USER) ?
723 from_kuid_munged(seq_user_ns(seq), fl->owner.uid) :
724 0)),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 atomic_read(&fl->users),
726 fl->linger/HZ,
727 (long)(fl->expires - jiffies)/HZ,
Harvey Harrisonb0711952008-10-28 16:05:40 -0700728 &fl->dst,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 fl->opt ? fl->opt->opt_nflen : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 return 0;
732}
733
Philippe De Muyter56b3d972007-07-10 23:07:31 -0700734static const struct seq_operations ip6fl_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 .start = ip6fl_seq_start,
736 .next = ip6fl_seq_next,
737 .stop = ip6fl_seq_stop,
738 .show = ip6fl_seq_show,
739};
740
741static int ip6fl_seq_open(struct inode *inode, struct file *file)
742{
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600743 struct seq_file *seq;
744 struct ip6fl_iter_state *state;
745 int err;
746
747 err = seq_open_net(inode, file, &ip6fl_seq_ops,
748 sizeof(struct ip6fl_iter_state));
749
750 if (!err) {
751 seq = file->private_data;
752 state = ip6fl_seq_private(seq);
753 rcu_read_lock();
754 state->pid_ns = get_pid_ns(task_active_pid_ns(current));
755 rcu_read_unlock();
756 }
757 return err;
758}
759
760static int ip6fl_seq_release(struct inode *inode, struct file *file)
761{
762 struct seq_file *seq = file->private_data;
763 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
764 put_pid_ns(state->pid_ns);
765 return seq_release_net(inode, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766}
767
Arjan van de Ven9a321442007-02-12 00:55:35 -0800768static const struct file_operations ip6fl_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 .owner = THIS_MODULE,
770 .open = ip6fl_seq_open,
771 .read = seq_read,
772 .llseek = seq_lseek,
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600773 .release = ip6fl_seq_release,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000776static int __net_init ip6_flowlabel_proc_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777{
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700778 if (!proc_net_fops_create(net, "ip6_flowlabel",
779 S_IRUGO, &ip6fl_seq_fops))
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800780 return -ENOMEM;
781 return 0;
782}
783
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000784static void __net_exit ip6_flowlabel_proc_fini(struct net *net)
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800785{
786 proc_net_remove(net, "ip6_flowlabel");
787}
788#else
789static inline int ip6_flowlabel_proc_init(struct net *net)
790{
791 return 0;
792}
793static inline void ip6_flowlabel_proc_fini(struct net *net)
794{
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800795}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796#endif
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800797
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000798static void __net_exit ip6_flowlabel_net_exit(struct net *net)
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700799{
800 ip6_fl_purge(net);
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700801 ip6_flowlabel_proc_fini(net);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700802}
803
804static struct pernet_operations ip6_flowlabel_net_ops = {
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700805 .init = ip6_flowlabel_proc_init,
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700806 .exit = ip6_flowlabel_net_exit,
807};
808
Daniel Lezcano0a3e78a2007-12-11 02:23:18 -0800809int ip6_flowlabel_init(void)
810{
Benjamin Thery5983a3d2008-03-26 16:53:30 -0700811 return register_pernet_subsys(&ip6_flowlabel_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812}
813
814void ip6_flowlabel_cleanup(void)
815{
816 del_timer(&ip6_fl_gc_timer);
Benjamin Thery60e8fbc2008-03-26 16:53:08 -0700817 unregister_pernet_subsys(&ip6_flowlabel_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818}