Fix missing numa_zonelist_order sysctl
[linux-2.6/mini2440.git] / net / ipv6 / ip6_flowlabel.c
blob413a4ebb195c728ea86990e038412941c7370ad7
1 /*
2 * ip6_flowlabel.c IPv6 flowlabel manager.
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.
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
12 #include <linux/capability.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>
24 #include <net/sock.h>
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>
35 #include <asm/uaccess.h>
37 #define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified
38 in old IPv6 RFC. Well, it was reasonable value.
40 #define FL_MAX_LINGER 60 /* Maximal linger timeout */
42 /* FL hash table */
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)
49 static atomic_t fl_size = ATOMIC_INIT(0);
50 static struct ip6_flowlabel *fl_ht[FL_HASH_MASK+1];
52 static void ip6_fl_gc(unsigned long dummy);
53 static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
55 /* FL hash table lock: it protects only of GC */
57 static DEFINE_RWLOCK(ip6_fl_lock);
59 /* Big socket sock */
61 static DEFINE_RWLOCK(ip6_sk_fl_lock);
64 static __inline__ struct ip6_flowlabel * __fl_lookup(__be32 label)
66 struct ip6_flowlabel *fl;
68 for (fl=fl_ht[FL_HASH(label)]; fl; fl = fl->next) {
69 if (fl->label == label)
70 return fl;
72 return NULL;
75 static struct ip6_flowlabel * fl_lookup(__be32 label)
77 struct ip6_flowlabel *fl;
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;
88 static void fl_free(struct ip6_flowlabel *fl)
90 if (fl)
91 kfree(fl->opt);
92 kfree(fl);
95 static void fl_release(struct ip6_flowlabel *fl)
97 write_lock_bh(&ip6_fl_lock);
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);
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);
115 write_unlock_bh(&ip6_fl_lock);
118 static void ip6_fl_gc(unsigned long dummy)
120 int i;
121 unsigned long now = jiffies;
122 unsigned long sched = 0;
124 write_lock(&ip6_fl_lock);
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;
141 if (!sched || time_before(ttd, sched))
142 sched = ttd;
144 flp = &fl->next;
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);
153 write_unlock(&ip6_fl_lock);
156 static int fl_intern(struct ip6_flowlabel *fl, __be32 label)
158 fl->label = label & IPV6_FLOWLABEL_MASK;
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;
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;
183 /* Socket flowlabel lists */
185 struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label)
187 struct ipv6_fl_socklist *sfl;
188 struct ipv6_pinfo *np = inet6_sk(sk);
190 label &= IPV6_FLOWLABEL_MASK;
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;
200 return NULL;
203 EXPORT_SYMBOL_GPL(fl6_sock_lookup);
205 void fl6_free_socklist(struct sock *sk)
207 struct ipv6_pinfo *np = inet6_sk(sk);
208 struct ipv6_fl_socklist *sfl;
210 while ((sfl = np->ipv6_fl_list) != NULL) {
211 np->ipv6_fl_list = sfl->next;
212 fl_release(sfl->fl);
213 kfree(sfl);
217 /* Service routines */
221 It is the only difficult place. flowlabel enforces equal headers
222 before and including routing header, however user may supply options
223 following rthdr.
226 struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space,
227 struct ip6_flowlabel * fl,
228 struct ipv6_txoptions * fopt)
230 struct ipv6_txoptions * fl_opt = fl->opt;
232 if (fopt == NULL || fopt->opt_flen == 0)
233 return fl_opt;
235 if (fl_opt != NULL) {
236 opt_space->hopopt = fl_opt->hopopt;
237 opt_space->dst0opt = fl_opt->dst0opt;
238 opt_space->srcrt = fl_opt->srcrt;
239 opt_space->opt_nflen = fl_opt->opt_nflen;
240 } else {
241 if (fopt->opt_nflen == 0)
242 return fopt;
243 opt_space->hopopt = NULL;
244 opt_space->dst0opt = NULL;
245 opt_space->srcrt = NULL;
246 opt_space->opt_nflen = 0;
248 opt_space->dst1opt = fopt->dst1opt;
249 opt_space->opt_flen = fopt->opt_flen;
250 return opt_space;
253 static unsigned long check_linger(unsigned long ttl)
255 if (ttl < FL_MIN_LINGER)
256 return FL_MIN_LINGER*HZ;
257 if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
258 return 0;
259 return ttl*HZ;
262 static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
264 linger = check_linger(linger);
265 if (!linger)
266 return -EPERM;
267 expires = check_linger(expires);
268 if (!expires)
269 return -EPERM;
270 fl->lastuse = jiffies;
271 if (time_before(fl->linger, linger))
272 fl->linger = linger;
273 if (time_before(expires, fl->linger))
274 expires = fl->linger;
275 if (time_before(fl->expires, fl->lastuse + expires))
276 fl->expires = fl->lastuse + expires;
277 return 0;
280 static struct ip6_flowlabel *
281 fl_create(struct in6_flowlabel_req *freq, char __user *optval, int optlen, int *err_p)
283 struct ip6_flowlabel *fl;
284 int olen;
285 int addr_type;
286 int err;
288 err = -ENOMEM;
289 fl = kzalloc(sizeof(*fl), GFP_KERNEL);
290 if (fl == NULL)
291 goto done;
293 olen = optlen - CMSG_ALIGN(sizeof(*freq));
294 if (olen > 0) {
295 struct msghdr msg;
296 struct flowi flowi;
297 int junk;
299 err = -ENOMEM;
300 fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
301 if (fl->opt == NULL)
302 goto done;
304 memset(fl->opt, 0, sizeof(*fl->opt));
305 fl->opt->tot_len = sizeof(*fl->opt) + olen;
306 err = -EFAULT;
307 if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
308 goto done;
310 msg.msg_controllen = olen;
311 msg.msg_control = (void*)(fl->opt+1);
312 flowi.oif = 0;
314 err = datagram_send_ctl(&msg, &flowi, fl->opt, &junk, &junk);
315 if (err)
316 goto done;
317 err = -EINVAL;
318 if (fl->opt->opt_flen)
319 goto done;
320 if (fl->opt->opt_nflen == 0) {
321 kfree(fl->opt);
322 fl->opt = NULL;
326 fl->expires = jiffies;
327 err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
328 if (err)
329 goto done;
330 fl->share = freq->flr_share;
331 addr_type = ipv6_addr_type(&freq->flr_dst);
332 if ((addr_type&IPV6_ADDR_MAPPED)
333 || addr_type == IPV6_ADDR_ANY) {
334 err = -EINVAL;
335 goto done;
337 ipv6_addr_copy(&fl->dst, &freq->flr_dst);
338 atomic_set(&fl->users, 1);
339 switch (fl->share) {
340 case IPV6_FL_S_EXCL:
341 case IPV6_FL_S_ANY:
342 break;
343 case IPV6_FL_S_PROCESS:
344 fl->owner = current->pid;
345 break;
346 case IPV6_FL_S_USER:
347 fl->owner = current->euid;
348 break;
349 default:
350 err = -EINVAL;
351 goto done;
353 return fl;
355 done:
356 fl_free(fl);
357 *err_p = err;
358 return NULL;
361 static int mem_check(struct sock *sk)
363 struct ipv6_pinfo *np = inet6_sk(sk);
364 struct ipv6_fl_socklist *sfl;
365 int room = FL_MAX_SIZE - atomic_read(&fl_size);
366 int count = 0;
368 if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
369 return 0;
371 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next)
372 count++;
374 if (room <= 0 ||
375 ((count >= FL_MAX_PER_SOCK ||
376 (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4)
377 && !capable(CAP_NET_ADMIN)))
378 return -ENOBUFS;
380 return 0;
383 static int ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2)
385 if (h1 == h2)
386 return 0;
387 if (h1 == NULL || h2 == NULL)
388 return 1;
389 if (h1->hdrlen != h2->hdrlen)
390 return 1;
391 return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1));
394 static int ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2)
396 if (o1 == o2)
397 return 0;
398 if (o1 == NULL || o2 == NULL)
399 return 1;
400 if (o1->opt_nflen != o2->opt_nflen)
401 return 1;
402 if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt))
403 return 1;
404 if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt))
405 return 1;
406 if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt))
407 return 1;
408 return 0;
411 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
413 int err;
414 struct ipv6_pinfo *np = inet6_sk(sk);
415 struct in6_flowlabel_req freq;
416 struct ipv6_fl_socklist *sfl1=NULL;
417 struct ipv6_fl_socklist *sfl, **sflp;
418 struct ip6_flowlabel *fl;
420 if (optlen < sizeof(freq))
421 return -EINVAL;
423 if (copy_from_user(&freq, optval, sizeof(freq)))
424 return -EFAULT;
426 switch (freq.flr_action) {
427 case IPV6_FL_A_PUT:
428 write_lock_bh(&ip6_sk_fl_lock);
429 for (sflp = &np->ipv6_fl_list; (sfl=*sflp)!=NULL; sflp = &sfl->next) {
430 if (sfl->fl->label == freq.flr_label) {
431 if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
432 np->flow_label &= ~IPV6_FLOWLABEL_MASK;
433 *sflp = sfl->next;
434 write_unlock_bh(&ip6_sk_fl_lock);
435 fl_release(sfl->fl);
436 kfree(sfl);
437 return 0;
440 write_unlock_bh(&ip6_sk_fl_lock);
441 return -ESRCH;
443 case IPV6_FL_A_RENEW:
444 read_lock_bh(&ip6_sk_fl_lock);
445 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
446 if (sfl->fl->label == freq.flr_label) {
447 err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
448 read_unlock_bh(&ip6_sk_fl_lock);
449 return err;
452 read_unlock_bh(&ip6_sk_fl_lock);
454 if (freq.flr_share == IPV6_FL_S_NONE && capable(CAP_NET_ADMIN)) {
455 fl = fl_lookup(freq.flr_label);
456 if (fl) {
457 err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
458 fl_release(fl);
459 return err;
462 return -ESRCH;
464 case IPV6_FL_A_GET:
465 if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
466 return -EINVAL;
468 fl = fl_create(&freq, optval, optlen, &err);
469 if (fl == NULL)
470 return err;
471 sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
473 if (freq.flr_label) {
474 struct ip6_flowlabel *fl1 = NULL;
476 err = -EEXIST;
477 read_lock_bh(&ip6_sk_fl_lock);
478 for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
479 if (sfl->fl->label == freq.flr_label) {
480 if (freq.flr_flags&IPV6_FL_F_EXCL) {
481 read_unlock_bh(&ip6_sk_fl_lock);
482 goto done;
484 fl1 = sfl->fl;
485 atomic_inc(&fl1->users);
486 break;
489 read_unlock_bh(&ip6_sk_fl_lock);
491 if (fl1 == NULL)
492 fl1 = fl_lookup(freq.flr_label);
493 if (fl1) {
494 err = -EEXIST;
495 if (freq.flr_flags&IPV6_FL_F_EXCL)
496 goto release;
497 err = -EPERM;
498 if (fl1->share == IPV6_FL_S_EXCL ||
499 fl1->share != fl->share ||
500 fl1->owner != fl->owner)
501 goto release;
503 err = -EINVAL;
504 if (!ipv6_addr_equal(&fl1->dst, &fl->dst) ||
505 ipv6_opt_cmp(fl1->opt, fl->opt))
506 goto release;
508 err = -ENOMEM;
509 if (sfl1 == NULL)
510 goto release;
511 if (fl->linger > fl1->linger)
512 fl1->linger = fl->linger;
513 if ((long)(fl->expires - fl1->expires) > 0)
514 fl1->expires = fl->expires;
515 write_lock_bh(&ip6_sk_fl_lock);
516 sfl1->fl = fl1;
517 sfl1->next = np->ipv6_fl_list;
518 np->ipv6_fl_list = sfl1;
519 write_unlock_bh(&ip6_sk_fl_lock);
520 fl_free(fl);
521 return 0;
523 release:
524 fl_release(fl1);
525 goto done;
528 err = -ENOENT;
529 if (!(freq.flr_flags&IPV6_FL_F_CREATE))
530 goto done;
532 err = -ENOMEM;
533 if (sfl1 == NULL || (err = mem_check(sk)) != 0)
534 goto done;
536 err = fl_intern(fl, freq.flr_label);
537 if (err)
538 goto done;
540 if (!freq.flr_label) {
541 if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
542 &fl->label, sizeof(fl->label))) {
543 /* Intentionally ignore fault. */
547 sfl1->fl = fl;
548 sfl1->next = np->ipv6_fl_list;
549 np->ipv6_fl_list = sfl1;
550 return 0;
552 default:
553 return -EINVAL;
556 done:
557 fl_free(fl);
558 kfree(sfl1);
559 return err;
562 #ifdef CONFIG_PROC_FS
564 struct ip6fl_iter_state {
565 int bucket;
568 #define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private)
570 static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
572 struct ip6_flowlabel *fl = NULL;
573 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
575 for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
576 if (fl_ht[state->bucket]) {
577 fl = fl_ht[state->bucket];
578 break;
581 return fl;
584 static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
586 struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
588 fl = fl->next;
589 while (!fl) {
590 if (++state->bucket <= FL_HASH_MASK)
591 fl = fl_ht[state->bucket];
592 else
593 break;
595 return fl;
598 static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
600 struct ip6_flowlabel *fl = ip6fl_get_first(seq);
601 if (fl)
602 while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
603 --pos;
604 return pos ? NULL : fl;
607 static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
609 read_lock_bh(&ip6_fl_lock);
610 return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
613 static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
615 struct ip6_flowlabel *fl;
617 if (v == SEQ_START_TOKEN)
618 fl = ip6fl_get_first(seq);
619 else
620 fl = ip6fl_get_next(seq, v);
621 ++*pos;
622 return fl;
625 static void ip6fl_seq_stop(struct seq_file *seq, void *v)
627 read_unlock_bh(&ip6_fl_lock);
630 static int ip6fl_seq_show(struct seq_file *seq, void *v)
632 if (v == SEQ_START_TOKEN)
633 seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n",
634 "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt");
635 else {
636 struct ip6_flowlabel *fl = v;
637 seq_printf(seq,
638 "%05X %-1d %-6d %-6d %-6ld %-8ld " NIP6_SEQFMT " %-4d\n",
639 (unsigned)ntohl(fl->label),
640 fl->share,
641 (unsigned)fl->owner,
642 atomic_read(&fl->users),
643 fl->linger/HZ,
644 (long)(fl->expires - jiffies)/HZ,
645 NIP6(fl->dst),
646 fl->opt ? fl->opt->opt_nflen : 0);
648 return 0;
651 static const struct seq_operations ip6fl_seq_ops = {
652 .start = ip6fl_seq_start,
653 .next = ip6fl_seq_next,
654 .stop = ip6fl_seq_stop,
655 .show = ip6fl_seq_show,
658 static int ip6fl_seq_open(struct inode *inode, struct file *file)
660 struct seq_file *seq;
661 int rc = -ENOMEM;
662 struct ip6fl_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
664 if (!s)
665 goto out;
667 rc = seq_open(file, &ip6fl_seq_ops);
668 if (rc)
669 goto out_kfree;
671 seq = file->private_data;
672 seq->private = s;
673 out:
674 return rc;
675 out_kfree:
676 kfree(s);
677 goto out;
680 static const struct file_operations ip6fl_seq_fops = {
681 .owner = THIS_MODULE,
682 .open = ip6fl_seq_open,
683 .read = seq_read,
684 .llseek = seq_lseek,
685 .release = seq_release_private,
687 #endif
690 void ip6_flowlabel_init(void)
692 #ifdef CONFIG_PROC_FS
693 proc_net_fops_create("ip6_flowlabel", S_IRUGO, &ip6fl_seq_fops);
694 #endif
697 void ip6_flowlabel_cleanup(void)
699 del_timer(&ip6_fl_gc_timer);
700 #ifdef CONFIG_PROC_FS
701 proc_net_remove("ip6_flowlabel");
702 #endif