]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/ipv6/ip6_flowlabel.c
[IPV6]: Support several new sockopt / ancillary data in Advanced API (RFC3542).
[net-next-2.6.git] / net / ipv6 / ip6_flowlabel.c
CommitLineData
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
12#include <linux/config.h>
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>
23
24#include <net/sock.h>
25
26#include <net/ipv6.h>
27#include <net/ndisc.h>
28#include <net/protocol.h>
29#include <net/ip6_route.h>
30#include <net/addrconf.h>
31#include <net/rawv6.h>
32#include <net/icmp.h>
33#include <net/transp_v6.h>
34
35#include <asm/uaccess.h>
36
37#define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified
38 in old IPv6 RFC. Well, it was reasonable value.
39 */
40#define FL_MAX_LINGER 60 /* Maximal linger timeout */
41
42/* FL hash table */
43
44#define FL_MAX_PER_SOCK 32
45#define FL_MAX_SIZE 4096
46#define FL_HASH_MASK 255
47#define FL_HASH(l) (ntohl(l)&FL_HASH_MASK)
48
49static atomic_t fl_size = ATOMIC_INIT(0);
50static struct ip6_flowlabel *fl_ht[FL_HASH_MASK+1];
51
52static void ip6_fl_gc(unsigned long dummy);
53static struct timer_list ip6_fl_gc_timer = TIMER_INITIALIZER(ip6_fl_gc, 0, 0);
54
55/* FL hash table lock: it protects only of GC */
56
57static DEFINE_RWLOCK(ip6_fl_lock);
58
59/* Big socket sock */
60
61static DEFINE_RWLOCK(ip6_sk_fl_lock);
62
63
64static __inline__ struct ip6_flowlabel * __fl_lookup(u32 label)
65{
66 struct ip6_flowlabel *fl;
67
68 for (fl=fl_ht[FL_HASH(label)]; fl; fl = fl->next) {
69 if (fl->label == label)
70 return fl;
71 }
72 return NULL;
73}
74
75static struct ip6_flowlabel * fl_lookup(u32 label)
76{
77 struct ip6_flowlabel *fl;
78
79 read_lock_bh(&ip6_fl_lock);
80 fl = __fl_lookup(label);
81 if (fl)
82 atomic_inc(&fl->users);
83 read_unlock_bh(&ip6_fl_lock);
84 return fl;
85}
86
87
88static void fl_free(struct ip6_flowlabel *fl)
89{
90 if (fl)
91 kfree(fl->opt);
92 kfree(fl);
93}
94
95static void fl_release(struct ip6_flowlabel *fl)
96{
97 write_lock_bh(&ip6_fl_lock);
98
99 fl->lastuse = jiffies;
100 if (atomic_dec_and_test(&fl->users)) {
101 unsigned long ttd = fl->lastuse + fl->linger;
102 if (time_after(ttd, fl->expires))
103 fl->expires = ttd;
104 ttd = fl->expires;
105 if (fl->opt && fl->share == IPV6_FL_S_EXCL) {
106 struct ipv6_txoptions *opt = fl->opt;
107 fl->opt = NULL;
108 kfree(opt);
109 }
110 if (!timer_pending(&ip6_fl_gc_timer) ||
111 time_after(ip6_fl_gc_timer.expires, ttd))
112 mod_timer(&ip6_fl_gc_timer, ttd);
113 }
114
115 write_unlock_bh(&ip6_fl_lock);
116}
117
118static void ip6_fl_gc(unsigned long dummy)
119{
120 int i;
121 unsigned long now = jiffies;
122 unsigned long sched = 0;
123
124 write_lock(&ip6_fl_lock);
125
126 for (i=0; i<=FL_HASH_MASK; i++) {
127 struct ip6_flowlabel *fl, **flp;
128 flp = &fl_ht[i];
129 while ((fl=*flp) != NULL) {
130 if (atomic_read(&fl->users) == 0) {
131 unsigned long ttd = fl->lastuse + fl->linger;
132 if (time_after(ttd, fl->expires))
133 fl->expires = ttd;
134 ttd = fl->expires;
135 if (time_after_eq(now, ttd)) {
136 *flp = fl->next;
137 fl_free(fl);
138 atomic_dec(&fl_size);
139 continue;
140 }
141 if (!sched || time_before(ttd, sched))
142 sched = ttd;
143 }
144 flp = &fl->next;
145 }
146 }
147 if (!sched && atomic_read(&fl_size))
148 sched = now + FL_MAX_LINGER;
149 if (sched) {
150 ip6_fl_gc_timer.expires = sched;
151 add_timer(&ip6_fl_gc_timer);
152 }
153 write_unlock(&ip6_fl_lock);
154}
155
156static int fl_intern(struct ip6_flowlabel *fl, __u32 label)
157{
158 fl->label = label & IPV6_FLOWLABEL_MASK;
159
160 write_lock_bh(&ip6_fl_lock);
161 if (label == 0) {
162 for (;;) {
163 fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK;
164 if (fl->label) {
165 struct ip6_flowlabel *lfl;
166 lfl = __fl_lookup(fl->label);
167 if (lfl == NULL)
168 break;
169 }
170 }
171 }
172
173 fl->lastuse = jiffies;
174 fl->next = fl_ht[FL_HASH(fl->label)];
175 fl_ht[FL_HASH(fl->label)] = fl;
176 atomic_inc(&fl_size);
177 write_unlock_bh(&ip6_fl_lock);
178 return 0;
179}
180
181
182
183/* Socket flowlabel lists */
184
185struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, u32 label)
186{
187 struct ipv6_fl_socklist *sfl;
188 struct ipv6_pinfo *np = inet6_sk(sk);
189
190 label &= IPV6_FLOWLABEL_MASK;
191
192 for (sfl=np->ipv6_fl_list; sfl; sfl = sfl->next) {
193 struct ip6_flowlabel *fl = sfl->fl;
194 if (fl->label == label) {
195 fl->lastuse = jiffies;
196 atomic_inc(&fl->users);
197 return fl;
198 }
199 }
200 return NULL;
201}
202
203void fl6_free_socklist(struct sock *sk)
204{
205 struct ipv6_pinfo *np = inet6_sk(sk);
206 struct ipv6_fl_socklist *sfl;
207
208 while ((sfl = np->ipv6_fl_list) != NULL) {
209 np->ipv6_fl_list = sfl->next;
210 fl_release(sfl->fl);
211 kfree(sfl);
212 }
213}
214
215/* Service routines */
216
217
218/*
219 It is the only difficult place. flowlabel enforces equal headers
220 before and including routing header, however user may supply options
221 following rthdr.
222 */
223
224struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space,
225 struct ip6_flowlabel * fl,
226 struct ipv6_txoptions * fopt)
227{
333fad53 228 struct ipv6_txoptions * fl_opt = fl ? fl->opt : NULL;
1da177e4 229
333fad53
YH
230 if (fopt == NULL || fopt->opt_flen == 0) {
231 if (!fl_opt || !fl_opt->dst0opt || fl_opt->srcrt)
232 return fl_opt;
233 }
1da177e4
LT
234
235 if (fl_opt != NULL) {
236 opt_space->hopopt = fl_opt->hopopt;
333fad53 237 opt_space->dst0opt = fl_opt->srcrt ? fl_opt->dst0opt : NULL;
1da177e4
LT
238 opt_space->srcrt = fl_opt->srcrt;
239 opt_space->opt_nflen = fl_opt->opt_nflen;
333fad53
YH
240 if (fl_opt->dst0opt && !fl_opt->srcrt)
241 opt_space->opt_nflen -= ipv6_optlen(fl_opt->dst0opt);
1da177e4
LT
242 } else {
243 if (fopt->opt_nflen == 0)
244 return fopt;
245 opt_space->hopopt = NULL;
246 opt_space->dst0opt = NULL;
247 opt_space->srcrt = NULL;
248 opt_space->opt_nflen = 0;
249 }
250 opt_space->dst1opt = fopt->dst1opt;
1da177e4
LT
251 opt_space->opt_flen = fopt->opt_flen;
252 return opt_space;
253}
254
255static unsigned long check_linger(unsigned long ttl)
256{
257 if (ttl < FL_MIN_LINGER)
258 return FL_MIN_LINGER*HZ;
259 if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
260 return 0;
261 return ttl*HZ;
262}
263
264static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
265{
266 linger = check_linger(linger);
267 if (!linger)
268 return -EPERM;
269 expires = check_linger(expires);
270 if (!expires)
271 return -EPERM;
272 fl->lastuse = jiffies;
273 if (time_before(fl->linger, linger))
274 fl->linger = linger;
275 if (time_before(expires, fl->linger))
276 expires = fl->linger;
277 if (time_before(fl->expires, fl->lastuse + expires))
278 fl->expires = fl->lastuse + expires;
279 return 0;
280}
281
282static struct ip6_flowlabel *
283fl_create(struct in6_flowlabel_req *freq, char __user *optval, int optlen, int *err_p)
284{
285 struct ip6_flowlabel *fl;
286 int olen;
287 int addr_type;
288 int err;
289
290 err = -ENOMEM;
291 fl = kmalloc(sizeof(*fl), GFP_KERNEL);
292 if (fl == NULL)
293 goto done;
294 memset(fl, 0, sizeof(*fl));
295
296 olen = optlen - CMSG_ALIGN(sizeof(*freq));
297 if (olen > 0) {
298 struct msghdr msg;
299 struct flowi flowi;
300 int junk;
301
302 err = -ENOMEM;
303 fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
304 if (fl->opt == NULL)
305 goto done;
306
307 memset(fl->opt, 0, sizeof(*fl->opt));
308 fl->opt->tot_len = sizeof(*fl->opt) + olen;
309 err = -EFAULT;
310 if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
311 goto done;
312
313 msg.msg_controllen = olen;
314 msg.msg_control = (void*)(fl->opt+1);
315 flowi.oif = 0;
316
317 err = datagram_send_ctl(&msg, &flowi, fl->opt, &junk);
318 if (err)
319 goto done;
320 err = -EINVAL;
321 if (fl->opt->opt_flen)
322 goto done;
323 if (fl->opt->opt_nflen == 0) {
324 kfree(fl->opt);
325 fl->opt = NULL;
326 }
327 }
328
329 fl->expires = jiffies;
330 err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
331 if (err)
332 goto done;
333 fl->share = freq->flr_share;
334 addr_type = ipv6_addr_type(&freq->flr_dst);
335 if ((addr_type&IPV6_ADDR_MAPPED)
336 || addr_type == IPV6_ADDR_ANY)
337 goto done;
338 ipv6_addr_copy(&fl->dst, &freq->flr_dst);
339 atomic_set(&fl->users, 1);
340 switch (fl->share) {
341 case IPV6_FL_S_EXCL:
342 case IPV6_FL_S_ANY:
343 break;
344 case IPV6_FL_S_PROCESS:
345 fl->owner = current->pid;
346 break;
347 case IPV6_FL_S_USER:
348 fl->owner = current->euid;
349 break;
350 default:
351 err = -EINVAL;
352 goto done;
353 }
354 return fl;
355
356done:
357 fl_free(fl);
358 *err_p = err;
359 return NULL;
360}
361
362static int mem_check(struct sock *sk)
363{
364 struct ipv6_pinfo *np = inet6_sk(sk);
365 struct ipv6_fl_socklist *sfl;
366 int room = FL_MAX_SIZE - atomic_read(&fl_size);
367 int count = 0;
368
369 if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
370 return 0;
371
372 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next)
373 count++;
374
375 if (room <= 0 ||
376 ((count >= FL_MAX_PER_SOCK ||
377 (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4)
378 && !capable(CAP_NET_ADMIN)))
379 return -ENOBUFS;
380
381 return 0;
382}
383
384static int ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2)
385{
386 if (h1 == h2)
387 return 0;
388 if (h1 == NULL || h2 == NULL)
389 return 1;
390 if (h1->hdrlen != h2->hdrlen)
391 return 1;
392 return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1));
393}
394
395static int ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2)
396{
397 if (o1 == o2)
398 return 0;
399 if (o1 == NULL || o2 == NULL)
400 return 1;
401 if (o1->opt_nflen != o2->opt_nflen)
402 return 1;
403 if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt))
404 return 1;
405 if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt))
406 return 1;
407 if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt))
408 return 1;
409 return 0;
410}
411
412int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
413{
414 int err;
415 struct ipv6_pinfo *np = inet6_sk(sk);
416 struct in6_flowlabel_req freq;
417 struct ipv6_fl_socklist *sfl1=NULL;
418 struct ipv6_fl_socklist *sfl, **sflp;
419 struct ip6_flowlabel *fl;
420
421 if (optlen < sizeof(freq))
422 return -EINVAL;
423
424 if (copy_from_user(&freq, optval, sizeof(freq)))
425 return -EFAULT;
426
427 switch (freq.flr_action) {
428 case IPV6_FL_A_PUT:
429 write_lock_bh(&ip6_sk_fl_lock);
430 for (sflp = &np->ipv6_fl_list; (sfl=*sflp)!=NULL; sflp = &sfl->next) {
431 if (sfl->fl->label == freq.flr_label) {
432 if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
433 np->flow_label &= ~IPV6_FLOWLABEL_MASK;
434 *sflp = sfl->next;
435 write_unlock_bh(&ip6_sk_fl_lock);
436 fl_release(sfl->fl);
437 kfree(sfl);
438 return 0;
439 }
440 }
441 write_unlock_bh(&ip6_sk_fl_lock);
442 return -ESRCH;
443
444 case IPV6_FL_A_RENEW:
445 read_lock_bh(&ip6_sk_fl_lock);
446 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
447 if (sfl->fl->label == freq.flr_label) {
448 err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
449 read_unlock_bh(&ip6_sk_fl_lock);
450 return err;
451 }
452 }
453 read_unlock_bh(&ip6_sk_fl_lock);
454
455 if (freq.flr_share == IPV6_FL_S_NONE && capable(CAP_NET_ADMIN)) {
456 fl = fl_lookup(freq.flr_label);
457 if (fl) {
458 err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
459 fl_release(fl);
460 return err;
461 }
462 }
463 return -ESRCH;
464
465 case IPV6_FL_A_GET:
466 if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
467 return -EINVAL;
468
469 fl = fl_create(&freq, optval, optlen, &err);
470 if (fl == NULL)
471 return err;
472 sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
473
474 if (freq.flr_label) {
475 struct ip6_flowlabel *fl1 = NULL;
476
477 err = -EEXIST;
478 read_lock_bh(&ip6_sk_fl_lock);
479 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
480 if (sfl->fl->label == freq.flr_label) {
481 if (freq.flr_flags&IPV6_FL_F_EXCL) {
482 read_unlock_bh(&ip6_sk_fl_lock);
483 goto done;
484 }
485 fl1 = sfl->fl;
486 atomic_inc(&fl->users);
487 break;
488 }
489 }
490 read_unlock_bh(&ip6_sk_fl_lock);
491
492 if (fl1 == NULL)
493 fl1 = fl_lookup(freq.flr_label);
494 if (fl1) {
495 err = -EEXIST;
496 if (freq.flr_flags&IPV6_FL_F_EXCL)
497 goto release;
498 err = -EPERM;
499 if (fl1->share == IPV6_FL_S_EXCL ||
500 fl1->share != fl->share ||
501 fl1->owner != fl->owner)
502 goto release;
503
504 err = -EINVAL;
505 if (!ipv6_addr_equal(&fl1->dst, &fl->dst) ||
506 ipv6_opt_cmp(fl1->opt, fl->opt))
507 goto release;
508
509 err = -ENOMEM;
510 if (sfl1 == NULL)
511 goto release;
512 if (fl->linger > fl1->linger)
513 fl1->linger = fl->linger;
514 if ((long)(fl->expires - fl1->expires) > 0)
515 fl1->expires = fl->expires;
516 write_lock_bh(&ip6_sk_fl_lock);
517 sfl1->fl = fl1;
518 sfl1->next = np->ipv6_fl_list;
519 np->ipv6_fl_list = sfl1;
520 write_unlock_bh(&ip6_sk_fl_lock);
521 fl_free(fl);
522 return 0;
523
524release:
525 fl_release(fl1);
526 goto done;
527 }
528 }
529 err = -ENOENT;
530 if (!(freq.flr_flags&IPV6_FL_F_CREATE))
531 goto done;
532
533 err = -ENOMEM;
534 if (sfl1 == NULL || (err = mem_check(sk)) != 0)
535 goto done;
536
537 err = fl_intern(fl, freq.flr_label);
538 if (err)
539 goto done;
540
6c94d361
DM
541 if (!freq.flr_label) {
542 if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
543 &fl->label, sizeof(fl->label))) {
544 /* Intentionally ignore fault. */
545 }
546 }
1da177e4
LT
547
548 sfl1->fl = fl;
549 sfl1->next = np->ipv6_fl_list;
550 np->ipv6_fl_list = sfl1;
551 return 0;
552
553 default:
554 return -EINVAL;
555 }
556
557done:
558 fl_free(fl);
559 kfree(sfl1);
560 return err;
561}
562
563#ifdef CONFIG_PROC_FS
564
565struct ip6fl_iter_state {
566 int bucket;
567};
568
569#define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private)
570
571static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
572{
573 struct ip6_flowlabel *fl = NULL;
574 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
575
576 for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
577 if (fl_ht[state->bucket]) {
578 fl = fl_ht[state->bucket];
579 break;
580 }
581 }
582 return fl;
583}
584
585static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
586{
587 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
588
589 fl = fl->next;
590 while (!fl) {
591 if (++state->bucket <= FL_HASH_MASK)
592 fl = fl_ht[state->bucket];
593 }
594 return fl;
595}
596
597static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
598{
599 struct ip6_flowlabel *fl = ip6fl_get_first(seq);
600 if (fl)
601 while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
602 --pos;
603 return pos ? NULL : fl;
604}
605
606static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
607{
608 read_lock_bh(&ip6_fl_lock);
609 return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
610}
611
612static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
613{
614 struct ip6_flowlabel *fl;
615
616 if (v == SEQ_START_TOKEN)
617 fl = ip6fl_get_first(seq);
618 else
619 fl = ip6fl_get_next(seq, v);
620 ++*pos;
621 return fl;
622}
623
624static void ip6fl_seq_stop(struct seq_file *seq, void *v)
625{
626 read_unlock_bh(&ip6_fl_lock);
627}
628
629static void ip6fl_fl_seq_show(struct seq_file *seq, struct ip6_flowlabel *fl)
630{
631 while(fl) {
632 seq_printf(seq,
633 "%05X %-1d %-6d %-6d %-6ld %-8ld "
634 "%02x%02x%02x%02x%02x%02x%02x%02x "
635 "%-4d\n",
636 (unsigned)ntohl(fl->label),
637 fl->share,
638 (unsigned)fl->owner,
639 atomic_read(&fl->users),
640 fl->linger/HZ,
641 (long)(fl->expires - jiffies)/HZ,
642 NIP6(fl->dst),
643 fl->opt ? fl->opt->opt_nflen : 0);
644 fl = fl->next;
645 }
646}
647
648static int ip6fl_seq_show(struct seq_file *seq, void *v)
649{
650 if (v == SEQ_START_TOKEN)
651 seq_puts(seq, "Label S Owner Users Linger Expires "
652 "Dst Opt\n");
653 else
654 ip6fl_fl_seq_show(seq, v);
655 return 0;
656}
657
658static struct seq_operations ip6fl_seq_ops = {
659 .start = ip6fl_seq_start,
660 .next = ip6fl_seq_next,
661 .stop = ip6fl_seq_stop,
662 .show = ip6fl_seq_show,
663};
664
665static int ip6fl_seq_open(struct inode *inode, struct file *file)
666{
667 struct seq_file *seq;
668 int rc = -ENOMEM;
669 struct ip6fl_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
670
671 if (!s)
672 goto out;
673
674 rc = seq_open(file, &ip6fl_seq_ops);
675 if (rc)
676 goto out_kfree;
677
678 seq = file->private_data;
679 seq->private = s;
680 memset(s, 0, sizeof(*s));
681out:
682 return rc;
683out_kfree:
684 kfree(s);
685 goto out;
686}
687
688static struct file_operations ip6fl_seq_fops = {
689 .owner = THIS_MODULE,
690 .open = ip6fl_seq_open,
691 .read = seq_read,
692 .llseek = seq_lseek,
693 .release = seq_release_private,
694};
695#endif
696
697
698void ip6_flowlabel_init(void)
699{
700#ifdef CONFIG_PROC_FS
701 proc_net_fops_create("ip6_flowlabel", S_IRUGO, &ip6fl_seq_fops);
702#endif
703}
704
705void ip6_flowlabel_cleanup(void)
706{
707 del_timer(&ip6_fl_gc_timer);
708#ifdef CONFIG_PROC_FS
709 proc_net_remove("ip6_flowlabel");
710#endif
711}