xref: /openbmc/linux/net/ipv6/ip6_flowlabel.c (revision 54525552)
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/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>
23 #include <linux/slab.h>
24 
25 #include <net/net_namespace.h>
26 #include <net/sock.h>
27 
28 #include <net/ipv6.h>
29 #include <net/ndisc.h>
30 #include <net/protocol.h>
31 #include <net/ip6_route.h>
32 #include <net/addrconf.h>
33 #include <net/rawv6.h>
34 #include <net/icmp.h>
35 #include <net/transp_v6.h>
36 
37 #include <asm/uaccess.h>
38 
39 #define FL_MIN_LINGER	6	/* Minimal linger. It is set to 6sec specified
40 				   in old IPv6 RFC. Well, it was reasonable value.
41 				 */
42 #define FL_MAX_LINGER	60	/* Maximal linger timeout */
43 
44 /* FL hash table */
45 
46 #define FL_MAX_PER_SOCK	32
47 #define FL_MAX_SIZE	4096
48 #define FL_HASH_MASK	255
49 #define FL_HASH(l)	(ntohl(l)&FL_HASH_MASK)
50 
51 static atomic_t fl_size = ATOMIC_INIT(0);
52 static struct ip6_flowlabel *fl_ht[FL_HASH_MASK+1];
53 
54 static void ip6_fl_gc(unsigned long dummy);
55 static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
56 
57 /* FL hash table lock: it protects only of GC */
58 
59 static DEFINE_RWLOCK(ip6_fl_lock);
60 
61 /* Big socket sock */
62 
63 static DEFINE_RWLOCK(ip6_sk_fl_lock);
64 
65 
66 static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label)
67 {
68 	struct ip6_flowlabel *fl;
69 
70 	for (fl=fl_ht[FL_HASH(label)]; fl; fl = fl->next) {
71 		if (fl->label == label && net_eq(fl->fl_net, net))
72 			return fl;
73 	}
74 	return NULL;
75 }
76 
77 static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label)
78 {
79 	struct ip6_flowlabel *fl;
80 
81 	read_lock_bh(&ip6_fl_lock);
82 	fl = __fl_lookup(net, label);
83 	if (fl)
84 		atomic_inc(&fl->users);
85 	read_unlock_bh(&ip6_fl_lock);
86 	return fl;
87 }
88 
89 
90 static void fl_free(struct ip6_flowlabel *fl)
91 {
92 	if (fl) {
93 		release_net(fl->fl_net);
94 		kfree(fl->opt);
95 	}
96 	kfree(fl);
97 }
98 
99 static void fl_release(struct ip6_flowlabel *fl)
100 {
101 	write_lock_bh(&ip6_fl_lock);
102 
103 	fl->lastuse = jiffies;
104 	if (atomic_dec_and_test(&fl->users)) {
105 		unsigned long ttd = fl->lastuse + fl->linger;
106 		if (time_after(ttd, fl->expires))
107 			fl->expires = ttd;
108 		ttd = fl->expires;
109 		if (fl->opt && fl->share == IPV6_FL_S_EXCL) {
110 			struct ipv6_txoptions *opt = fl->opt;
111 			fl->opt = NULL;
112 			kfree(opt);
113 		}
114 		if (!timer_pending(&ip6_fl_gc_timer) ||
115 		    time_after(ip6_fl_gc_timer.expires, ttd))
116 			mod_timer(&ip6_fl_gc_timer, ttd);
117 	}
118 	write_unlock_bh(&ip6_fl_lock);
119 }
120 
121 static void ip6_fl_gc(unsigned long dummy)
122 {
123 	int i;
124 	unsigned long now = jiffies;
125 	unsigned long sched = 0;
126 
127 	write_lock(&ip6_fl_lock);
128 
129 	for (i=0; i<=FL_HASH_MASK; i++) {
130 		struct ip6_flowlabel *fl, **flp;
131 		flp = &fl_ht[i];
132 		while ((fl=*flp) != NULL) {
133 			if (atomic_read(&fl->users) == 0) {
134 				unsigned long ttd = fl->lastuse + fl->linger;
135 				if (time_after(ttd, fl->expires))
136 					fl->expires = ttd;
137 				ttd = fl->expires;
138 				if (time_after_eq(now, ttd)) {
139 					*flp = fl->next;
140 					fl_free(fl);
141 					atomic_dec(&fl_size);
142 					continue;
143 				}
144 				if (!sched || time_before(ttd, sched))
145 					sched = ttd;
146 			}
147 			flp = &fl->next;
148 		}
149 	}
150 	if (!sched && atomic_read(&fl_size))
151 		sched = now + FL_MAX_LINGER;
152 	if (sched) {
153 		mod_timer(&ip6_fl_gc_timer, sched);
154 	}
155 	write_unlock(&ip6_fl_lock);
156 }
157 
158 static void __net_exit ip6_fl_purge(struct net *net)
159 {
160 	int i;
161 
162 	write_lock(&ip6_fl_lock);
163 	for (i = 0; i <= FL_HASH_MASK; i++) {
164 		struct ip6_flowlabel *fl, **flp;
165 		flp = &fl_ht[i];
166 		while ((fl = *flp) != NULL) {
167 			if (net_eq(fl->fl_net, net) &&
168 			    atomic_read(&fl->users) == 0) {
169 				*flp = fl->next;
170 				fl_free(fl);
171 				atomic_dec(&fl_size);
172 				continue;
173 			}
174 			flp = &fl->next;
175 		}
176 	}
177 	write_unlock(&ip6_fl_lock);
178 }
179 
180 static struct ip6_flowlabel *fl_intern(struct net *net,
181 				       struct ip6_flowlabel *fl, __be32 label)
182 {
183 	struct ip6_flowlabel *lfl;
184 
185 	fl->label = label & IPV6_FLOWLABEL_MASK;
186 
187 	write_lock_bh(&ip6_fl_lock);
188 	if (label == 0) {
189 		for (;;) {
190 			fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK;
191 			if (fl->label) {
192 				lfl = __fl_lookup(net, fl->label);
193 				if (lfl == NULL)
194 					break;
195 			}
196 		}
197 	} else {
198 		/*
199 		 * we dropper the ip6_fl_lock, so this entry could reappear
200 		 * and we need to recheck with it.
201 		 *
202 		 * OTOH no need to search the active socket first, like it is
203 		 * done in ipv6_flowlabel_opt - sock is locked, so new entry
204 		 * with the same label can only appear on another sock
205 		 */
206 		lfl = __fl_lookup(net, fl->label);
207 		if (lfl != NULL) {
208 			atomic_inc(&lfl->users);
209 			write_unlock_bh(&ip6_fl_lock);
210 			return lfl;
211 		}
212 	}
213 
214 	fl->lastuse = jiffies;
215 	fl->next = fl_ht[FL_HASH(fl->label)];
216 	fl_ht[FL_HASH(fl->label)] = fl;
217 	atomic_inc(&fl_size);
218 	write_unlock_bh(&ip6_fl_lock);
219 	return NULL;
220 }
221 
222 
223 
224 /* Socket flowlabel lists */
225 
226 struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label)
227 {
228 	struct ipv6_fl_socklist *sfl;
229 	struct ipv6_pinfo *np = inet6_sk(sk);
230 
231 	label &= IPV6_FLOWLABEL_MASK;
232 
233 	read_lock_bh(&ip6_sk_fl_lock);
234 	for (sfl=np->ipv6_fl_list; sfl; sfl = sfl->next) {
235 		struct ip6_flowlabel *fl = sfl->fl;
236 		if (fl->label == label) {
237 			fl->lastuse = jiffies;
238 			atomic_inc(&fl->users);
239 			read_unlock_bh(&ip6_sk_fl_lock);
240 			return fl;
241 		}
242 	}
243 	read_unlock_bh(&ip6_sk_fl_lock);
244 	return NULL;
245 }
246 
247 EXPORT_SYMBOL_GPL(fl6_sock_lookup);
248 
249 void fl6_free_socklist(struct sock *sk)
250 {
251 	struct ipv6_pinfo *np = inet6_sk(sk);
252 	struct ipv6_fl_socklist *sfl;
253 
254 	while ((sfl = np->ipv6_fl_list) != NULL) {
255 		np->ipv6_fl_list = sfl->next;
256 		fl_release(sfl->fl);
257 		kfree(sfl);
258 	}
259 }
260 
261 /* Service routines */
262 
263 
264 /*
265    It is the only difficult place. flowlabel enforces equal headers
266    before and including routing header, however user may supply options
267    following rthdr.
268  */
269 
270 struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space,
271 					 struct ip6_flowlabel * fl,
272 					 struct ipv6_txoptions * fopt)
273 {
274 	struct ipv6_txoptions * fl_opt = fl->opt;
275 
276 	if (fopt == NULL || fopt->opt_flen == 0)
277 		return fl_opt;
278 
279 	if (fl_opt != NULL) {
280 		opt_space->hopopt = fl_opt->hopopt;
281 		opt_space->dst0opt = fl_opt->dst0opt;
282 		opt_space->srcrt = fl_opt->srcrt;
283 		opt_space->opt_nflen = fl_opt->opt_nflen;
284 	} else {
285 		if (fopt->opt_nflen == 0)
286 			return fopt;
287 		opt_space->hopopt = NULL;
288 		opt_space->dst0opt = NULL;
289 		opt_space->srcrt = NULL;
290 		opt_space->opt_nflen = 0;
291 	}
292 	opt_space->dst1opt = fopt->dst1opt;
293 	opt_space->opt_flen = fopt->opt_flen;
294 	return opt_space;
295 }
296 
297 static unsigned long check_linger(unsigned long ttl)
298 {
299 	if (ttl < FL_MIN_LINGER)
300 		return FL_MIN_LINGER*HZ;
301 	if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
302 		return 0;
303 	return ttl*HZ;
304 }
305 
306 static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
307 {
308 	linger = check_linger(linger);
309 	if (!linger)
310 		return -EPERM;
311 	expires = check_linger(expires);
312 	if (!expires)
313 		return -EPERM;
314 	fl->lastuse = jiffies;
315 	if (time_before(fl->linger, linger))
316 		fl->linger = linger;
317 	if (time_before(expires, fl->linger))
318 		expires = fl->linger;
319 	if (time_before(fl->expires, fl->lastuse + expires))
320 		fl->expires = fl->lastuse + expires;
321 	return 0;
322 }
323 
324 static struct ip6_flowlabel *
325 fl_create(struct net *net, struct in6_flowlabel_req *freq, char __user *optval,
326 	  int optlen, int *err_p)
327 {
328 	struct ip6_flowlabel *fl = NULL;
329 	int olen;
330 	int addr_type;
331 	int err;
332 
333 	olen = optlen - CMSG_ALIGN(sizeof(*freq));
334 	err = -EINVAL;
335 	if (olen > 64 * 1024)
336 		goto done;
337 
338 	err = -ENOMEM;
339 	fl = kzalloc(sizeof(*fl), GFP_KERNEL);
340 	if (fl == NULL)
341 		goto done;
342 
343 	if (olen > 0) {
344 		struct msghdr msg;
345 		struct flowi6 flowi6;
346 		int junk;
347 
348 		err = -ENOMEM;
349 		fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
350 		if (fl->opt == NULL)
351 			goto done;
352 
353 		memset(fl->opt, 0, sizeof(*fl->opt));
354 		fl->opt->tot_len = sizeof(*fl->opt) + olen;
355 		err = -EFAULT;
356 		if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
357 			goto done;
358 
359 		msg.msg_controllen = olen;
360 		msg.msg_control = (void*)(fl->opt+1);
361 		memset(&flowi6, 0, sizeof(flowi6));
362 
363 		err = datagram_send_ctl(net, &msg, &flowi6, fl->opt, &junk,
364 					&junk, &junk);
365 		if (err)
366 			goto done;
367 		err = -EINVAL;
368 		if (fl->opt->opt_flen)
369 			goto done;
370 		if (fl->opt->opt_nflen == 0) {
371 			kfree(fl->opt);
372 			fl->opt = NULL;
373 		}
374 	}
375 
376 	fl->fl_net = hold_net(net);
377 	fl->expires = jiffies;
378 	err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
379 	if (err)
380 		goto done;
381 	fl->share = freq->flr_share;
382 	addr_type = ipv6_addr_type(&freq->flr_dst);
383 	if ((addr_type & IPV6_ADDR_MAPPED) ||
384 	    addr_type == IPV6_ADDR_ANY) {
385 		err = -EINVAL;
386 		goto done;
387 	}
388 	ipv6_addr_copy(&fl->dst, &freq->flr_dst);
389 	atomic_set(&fl->users, 1);
390 	switch (fl->share) {
391 	case IPV6_FL_S_EXCL:
392 	case IPV6_FL_S_ANY:
393 		break;
394 	case IPV6_FL_S_PROCESS:
395 		fl->owner = current->pid;
396 		break;
397 	case IPV6_FL_S_USER:
398 		fl->owner = current_euid();
399 		break;
400 	default:
401 		err = -EINVAL;
402 		goto done;
403 	}
404 	return fl;
405 
406 done:
407 	fl_free(fl);
408 	*err_p = err;
409 	return NULL;
410 }
411 
412 static int mem_check(struct sock *sk)
413 {
414 	struct ipv6_pinfo *np = inet6_sk(sk);
415 	struct ipv6_fl_socklist *sfl;
416 	int room = FL_MAX_SIZE - atomic_read(&fl_size);
417 	int count = 0;
418 
419 	if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
420 		return 0;
421 
422 	for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next)
423 		count++;
424 
425 	if (room <= 0 ||
426 	    ((count >= FL_MAX_PER_SOCK ||
427 	      (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) &&
428 	     !capable(CAP_NET_ADMIN)))
429 		return -ENOBUFS;
430 
431 	return 0;
432 }
433 
434 static int ipv6_hdr_cmp(struct ipv6_opt_hdr *h1, struct ipv6_opt_hdr *h2)
435 {
436 	if (h1 == h2)
437 		return 0;
438 	if (h1 == NULL || h2 == NULL)
439 		return 1;
440 	if (h1->hdrlen != h2->hdrlen)
441 		return 1;
442 	return memcmp(h1+1, h2+1, ((h1->hdrlen+1)<<3) - sizeof(*h1));
443 }
444 
445 static int ipv6_opt_cmp(struct ipv6_txoptions *o1, struct ipv6_txoptions *o2)
446 {
447 	if (o1 == o2)
448 		return 0;
449 	if (o1 == NULL || o2 == NULL)
450 		return 1;
451 	if (o1->opt_nflen != o2->opt_nflen)
452 		return 1;
453 	if (ipv6_hdr_cmp(o1->hopopt, o2->hopopt))
454 		return 1;
455 	if (ipv6_hdr_cmp(o1->dst0opt, o2->dst0opt))
456 		return 1;
457 	if (ipv6_hdr_cmp((struct ipv6_opt_hdr *)o1->srcrt, (struct ipv6_opt_hdr *)o2->srcrt))
458 		return 1;
459 	return 0;
460 }
461 
462 static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
463 		struct ip6_flowlabel *fl)
464 {
465 	write_lock_bh(&ip6_sk_fl_lock);
466 	sfl->fl = fl;
467 	sfl->next = np->ipv6_fl_list;
468 	np->ipv6_fl_list = sfl;
469 	write_unlock_bh(&ip6_sk_fl_lock);
470 }
471 
472 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
473 {
474 	int uninitialized_var(err);
475 	struct net *net = sock_net(sk);
476 	struct ipv6_pinfo *np = inet6_sk(sk);
477 	struct in6_flowlabel_req freq;
478 	struct ipv6_fl_socklist *sfl1=NULL;
479 	struct ipv6_fl_socklist *sfl, **sflp;
480 	struct ip6_flowlabel *fl, *fl1 = NULL;
481 
482 
483 	if (optlen < sizeof(freq))
484 		return -EINVAL;
485 
486 	if (copy_from_user(&freq, optval, sizeof(freq)))
487 		return -EFAULT;
488 
489 	switch (freq.flr_action) {
490 	case IPV6_FL_A_PUT:
491 		write_lock_bh(&ip6_sk_fl_lock);
492 		for (sflp = &np->ipv6_fl_list; (sfl=*sflp)!=NULL; sflp = &sfl->next) {
493 			if (sfl->fl->label == freq.flr_label) {
494 				if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
495 					np->flow_label &= ~IPV6_FLOWLABEL_MASK;
496 				*sflp = sfl->next;
497 				write_unlock_bh(&ip6_sk_fl_lock);
498 				fl_release(sfl->fl);
499 				kfree(sfl);
500 				return 0;
501 			}
502 		}
503 		write_unlock_bh(&ip6_sk_fl_lock);
504 		return -ESRCH;
505 
506 	case IPV6_FL_A_RENEW:
507 		read_lock_bh(&ip6_sk_fl_lock);
508 		for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
509 			if (sfl->fl->label == freq.flr_label) {
510 				err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
511 				read_unlock_bh(&ip6_sk_fl_lock);
512 				return err;
513 			}
514 		}
515 		read_unlock_bh(&ip6_sk_fl_lock);
516 
517 		if (freq.flr_share == IPV6_FL_S_NONE && capable(CAP_NET_ADMIN)) {
518 			fl = fl_lookup(net, freq.flr_label);
519 			if (fl) {
520 				err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
521 				fl_release(fl);
522 				return err;
523 			}
524 		}
525 		return -ESRCH;
526 
527 	case IPV6_FL_A_GET:
528 		if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
529 			return -EINVAL;
530 
531 		fl = fl_create(net, &freq, optval, optlen, &err);
532 		if (fl == NULL)
533 			return err;
534 		sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
535 
536 		if (freq.flr_label) {
537 			err = -EEXIST;
538 			read_lock_bh(&ip6_sk_fl_lock);
539 			for (sfl = np->ipv6_fl_list; sfl; sfl = sfl->next) {
540 				if (sfl->fl->label == freq.flr_label) {
541 					if (freq.flr_flags&IPV6_FL_F_EXCL) {
542 						read_unlock_bh(&ip6_sk_fl_lock);
543 						goto done;
544 					}
545 					fl1 = sfl->fl;
546 					atomic_inc(&fl1->users);
547 					break;
548 				}
549 			}
550 			read_unlock_bh(&ip6_sk_fl_lock);
551 
552 			if (fl1 == NULL)
553 				fl1 = fl_lookup(net, freq.flr_label);
554 			if (fl1) {
555 recheck:
556 				err = -EEXIST;
557 				if (freq.flr_flags&IPV6_FL_F_EXCL)
558 					goto release;
559 				err = -EPERM;
560 				if (fl1->share == IPV6_FL_S_EXCL ||
561 				    fl1->share != fl->share ||
562 				    fl1->owner != fl->owner)
563 					goto release;
564 
565 				err = -EINVAL;
566 				if (!ipv6_addr_equal(&fl1->dst, &fl->dst) ||
567 				    ipv6_opt_cmp(fl1->opt, fl->opt))
568 					goto release;
569 
570 				err = -ENOMEM;
571 				if (sfl1 == NULL)
572 					goto release;
573 				if (fl->linger > fl1->linger)
574 					fl1->linger = fl->linger;
575 				if ((long)(fl->expires - fl1->expires) > 0)
576 					fl1->expires = fl->expires;
577 				fl_link(np, sfl1, fl1);
578 				fl_free(fl);
579 				return 0;
580 
581 release:
582 				fl_release(fl1);
583 				goto done;
584 			}
585 		}
586 		err = -ENOENT;
587 		if (!(freq.flr_flags&IPV6_FL_F_CREATE))
588 			goto done;
589 
590 		err = -ENOMEM;
591 		if (sfl1 == NULL || (err = mem_check(sk)) != 0)
592 			goto done;
593 
594 		fl1 = fl_intern(net, fl, freq.flr_label);
595 		if (fl1 != NULL)
596 			goto recheck;
597 
598 		if (!freq.flr_label) {
599 			if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
600 					 &fl->label, sizeof(fl->label))) {
601 				/* Intentionally ignore fault. */
602 			}
603 		}
604 
605 		fl_link(np, sfl1, fl);
606 		return 0;
607 
608 	default:
609 		return -EINVAL;
610 	}
611 
612 done:
613 	fl_free(fl);
614 	kfree(sfl1);
615 	return err;
616 }
617 
618 #ifdef CONFIG_PROC_FS
619 
620 struct ip6fl_iter_state {
621 	struct seq_net_private p;
622 	int bucket;
623 };
624 
625 #define ip6fl_seq_private(seq)	((struct ip6fl_iter_state *)(seq)->private)
626 
627 static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
628 {
629 	struct ip6_flowlabel *fl = NULL;
630 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
631 	struct net *net = seq_file_net(seq);
632 
633 	for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
634 		fl = fl_ht[state->bucket];
635 
636 		while (fl && !net_eq(fl->fl_net, net))
637 			fl = fl->next;
638 		if (fl)
639 			break;
640 	}
641 	return fl;
642 }
643 
644 static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
645 {
646 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
647 	struct net *net = seq_file_net(seq);
648 
649 	fl = fl->next;
650 try_again:
651 	while (fl && !net_eq(fl->fl_net, net))
652 		fl = fl->next;
653 
654 	while (!fl) {
655 		if (++state->bucket <= FL_HASH_MASK) {
656 			fl = fl_ht[state->bucket];
657 			goto try_again;
658 		} else
659 			break;
660 	}
661 	return fl;
662 }
663 
664 static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
665 {
666 	struct ip6_flowlabel *fl = ip6fl_get_first(seq);
667 	if (fl)
668 		while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
669 			--pos;
670 	return pos ? NULL : fl;
671 }
672 
673 static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
674 	__acquires(ip6_fl_lock)
675 {
676 	read_lock_bh(&ip6_fl_lock);
677 	return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
678 }
679 
680 static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
681 {
682 	struct ip6_flowlabel *fl;
683 
684 	if (v == SEQ_START_TOKEN)
685 		fl = ip6fl_get_first(seq);
686 	else
687 		fl = ip6fl_get_next(seq, v);
688 	++*pos;
689 	return fl;
690 }
691 
692 static void ip6fl_seq_stop(struct seq_file *seq, void *v)
693 	__releases(ip6_fl_lock)
694 {
695 	read_unlock_bh(&ip6_fl_lock);
696 }
697 
698 static int ip6fl_seq_show(struct seq_file *seq, void *v)
699 {
700 	if (v == SEQ_START_TOKEN)
701 		seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n",
702 			   "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt");
703 	else {
704 		struct ip6_flowlabel *fl = v;
705 		seq_printf(seq,
706 			   "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
707 			   (unsigned)ntohl(fl->label),
708 			   fl->share,
709 			   (unsigned)fl->owner,
710 			   atomic_read(&fl->users),
711 			   fl->linger/HZ,
712 			   (long)(fl->expires - jiffies)/HZ,
713 			   &fl->dst,
714 			   fl->opt ? fl->opt->opt_nflen : 0);
715 	}
716 	return 0;
717 }
718 
719 static const struct seq_operations ip6fl_seq_ops = {
720 	.start	=	ip6fl_seq_start,
721 	.next	=	ip6fl_seq_next,
722 	.stop	=	ip6fl_seq_stop,
723 	.show	=	ip6fl_seq_show,
724 };
725 
726 static int ip6fl_seq_open(struct inode *inode, struct file *file)
727 {
728 	return seq_open_net(inode, file, &ip6fl_seq_ops,
729 			    sizeof(struct ip6fl_iter_state));
730 }
731 
732 static const struct file_operations ip6fl_seq_fops = {
733 	.owner		=	THIS_MODULE,
734 	.open		=	ip6fl_seq_open,
735 	.read		=	seq_read,
736 	.llseek		=	seq_lseek,
737 	.release	=	seq_release_net,
738 };
739 
740 static int __net_init ip6_flowlabel_proc_init(struct net *net)
741 {
742 	if (!proc_net_fops_create(net, "ip6_flowlabel",
743 				  S_IRUGO, &ip6fl_seq_fops))
744 		return -ENOMEM;
745 	return 0;
746 }
747 
748 static void __net_exit ip6_flowlabel_proc_fini(struct net *net)
749 {
750 	proc_net_remove(net, "ip6_flowlabel");
751 }
752 #else
753 static inline int ip6_flowlabel_proc_init(struct net *net)
754 {
755 	return 0;
756 }
757 static inline void ip6_flowlabel_proc_fini(struct net *net)
758 {
759 }
760 #endif
761 
762 static void __net_exit ip6_flowlabel_net_exit(struct net *net)
763 {
764 	ip6_fl_purge(net);
765 	ip6_flowlabel_proc_fini(net);
766 }
767 
768 static struct pernet_operations ip6_flowlabel_net_ops = {
769 	.init = ip6_flowlabel_proc_init,
770 	.exit = ip6_flowlabel_net_exit,
771 };
772 
773 int ip6_flowlabel_init(void)
774 {
775 	return register_pernet_subsys(&ip6_flowlabel_net_ops);
776 }
777 
778 void ip6_flowlabel_cleanup(void)
779 {
780 	del_timer(&ip6_fl_gc_timer);
781 	unregister_pernet_subsys(&ip6_flowlabel_net_ops);
782 }
783