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/in6.h>
19#include <linux/proc_fs.h>
20#include <linux/seq_file.h>
21#include <linux/slab.h>
22#include <linux/export.h>
23#include <linux/pid_namespace.h>
24
25#include <net/net_namespace.h>
26#include <net/sock.h>
27
28#include <net/ipv6.h>
29#include <net/rawv6.h>
30#include <net/transp_v6.h>
31
32#include <asm/uaccess.h>
33
34#define FL_MIN_LINGER	6	/* Minimal linger. It is set to 6sec specified
35				   in old IPv6 RFC. Well, it was reasonable value.
36				 */
37#define FL_MAX_LINGER	150	/* Maximal linger timeout */
38
39/* FL hash table */
40
41#define FL_MAX_PER_SOCK	32
42#define FL_MAX_SIZE	4096
43#define FL_HASH_MASK	255
44#define FL_HASH(l)	(ntohl(l)&FL_HASH_MASK)
45
46static atomic_t fl_size = ATOMIC_INIT(0);
47static struct ip6_flowlabel __rcu *fl_ht[FL_HASH_MASK+1];
48
49static void ip6_fl_gc(unsigned long dummy);
50static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
51
52/* FL hash table lock: it protects only of GC */
53
54static DEFINE_SPINLOCK(ip6_fl_lock);
55
56/* Big socket sock */
57
58static DEFINE_SPINLOCK(ip6_sk_fl_lock);
59
60#define for_each_fl_rcu(hash, fl)				\
61	for (fl = rcu_dereference_bh(fl_ht[(hash)]);		\
62	     fl != NULL;					\
63	     fl = rcu_dereference_bh(fl->next))
64#define for_each_fl_continue_rcu(fl)				\
65	for (fl = rcu_dereference_bh(fl->next);			\
66	     fl != NULL;					\
67	     fl = rcu_dereference_bh(fl->next))
68
69#define for_each_sk_fl_rcu(np, sfl)				\
70	for (sfl = rcu_dereference_bh(np->ipv6_fl_list);	\
71	     sfl != NULL;					\
72	     sfl = rcu_dereference_bh(sfl->next))
73
74static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label)
75{
76	struct ip6_flowlabel *fl;
77
78	for_each_fl_rcu(FL_HASH(label), fl) {
79		if (fl->label == label && net_eq(fl->fl_net, net))
80			return fl;
81	}
82	return NULL;
83}
84
85static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label)
86{
87	struct ip6_flowlabel *fl;
88
89	rcu_read_lock_bh();
90	fl = __fl_lookup(net, label);
91	if (fl && !atomic_inc_not_zero(&fl->users))
92		fl = NULL;
93	rcu_read_unlock_bh();
94	return fl;
95}
96
97
98static void fl_free(struct ip6_flowlabel *fl)
99{
100	if (fl) {
101		if (fl->share == IPV6_FL_S_PROCESS)
102			put_pid(fl->owner.pid);
103		kfree(fl->opt);
104		kfree_rcu(fl, rcu);
105	}
106}
107
108static void fl_release(struct ip6_flowlabel *fl)
109{
110	spin_lock_bh(&ip6_fl_lock);
111
112	fl->lastuse = jiffies;
113	if (atomic_dec_and_test(&fl->users)) {
114		unsigned long ttd = fl->lastuse + fl->linger;
115		if (time_after(ttd, fl->expires))
116			fl->expires = ttd;
117		ttd = fl->expires;
118		if (fl->opt && fl->share == IPV6_FL_S_EXCL) {
119			struct ipv6_txoptions *opt = fl->opt;
120			fl->opt = NULL;
121			kfree(opt);
122		}
123		if (!timer_pending(&ip6_fl_gc_timer) ||
124		    time_after(ip6_fl_gc_timer.expires, ttd))
125			mod_timer(&ip6_fl_gc_timer, ttd);
126	}
127	spin_unlock_bh(&ip6_fl_lock);
128}
129
130static void ip6_fl_gc(unsigned long dummy)
131{
132	int i;
133	unsigned long now = jiffies;
134	unsigned long sched = 0;
135
136	spin_lock(&ip6_fl_lock);
137
138	for (i = 0; i <= FL_HASH_MASK; i++) {
139		struct ip6_flowlabel *fl;
140		struct ip6_flowlabel __rcu **flp;
141
142		flp = &fl_ht[i];
143		while ((fl = rcu_dereference_protected(*flp,
144						       lockdep_is_held(&ip6_fl_lock))) != NULL) {
145			if (atomic_read(&fl->users) == 0) {
146				unsigned long ttd = fl->lastuse + fl->linger;
147				if (time_after(ttd, fl->expires))
148					fl->expires = ttd;
149				ttd = fl->expires;
150				if (time_after_eq(now, ttd)) {
151					*flp = fl->next;
152					fl_free(fl);
153					atomic_dec(&fl_size);
154					continue;
155				}
156				if (!sched || time_before(ttd, sched))
157					sched = ttd;
158			}
159			flp = &fl->next;
160		}
161	}
162	if (!sched && atomic_read(&fl_size))
163		sched = now + FL_MAX_LINGER;
164	if (sched) {
165		mod_timer(&ip6_fl_gc_timer, sched);
166	}
167	spin_unlock(&ip6_fl_lock);
168}
169
170static void __net_exit ip6_fl_purge(struct net *net)
171{
172	int i;
173
174	spin_lock_bh(&ip6_fl_lock);
175	for (i = 0; i <= FL_HASH_MASK; i++) {
176		struct ip6_flowlabel *fl;
177		struct ip6_flowlabel __rcu **flp;
178
179		flp = &fl_ht[i];
180		while ((fl = rcu_dereference_protected(*flp,
181						       lockdep_is_held(&ip6_fl_lock))) != NULL) {
182			if (net_eq(fl->fl_net, net) &&
183			    atomic_read(&fl->users) == 0) {
184				*flp = fl->next;
185				fl_free(fl);
186				atomic_dec(&fl_size);
187				continue;
188			}
189			flp = &fl->next;
190		}
191	}
192	spin_unlock_bh(&ip6_fl_lock);
193}
194
195static struct ip6_flowlabel *fl_intern(struct net *net,
196				       struct ip6_flowlabel *fl, __be32 label)
197{
198	struct ip6_flowlabel *lfl;
199
200	fl->label = label & IPV6_FLOWLABEL_MASK;
201
202	spin_lock_bh(&ip6_fl_lock);
203	if (label == 0) {
204		for (;;) {
205			fl->label = htonl(prandom_u32())&IPV6_FLOWLABEL_MASK;
206			if (fl->label) {
207				lfl = __fl_lookup(net, fl->label);
208				if (!lfl)
209					break;
210			}
211		}
212	} else {
213		/*
214		 * we dropper the ip6_fl_lock, so this entry could reappear
215		 * and we need to recheck with it.
216		 *
217		 * OTOH no need to search the active socket first, like it is
218		 * done in ipv6_flowlabel_opt - sock is locked, so new entry
219		 * with the same label can only appear on another sock
220		 */
221		lfl = __fl_lookup(net, fl->label);
222		if (lfl) {
223			atomic_inc(&lfl->users);
224			spin_unlock_bh(&ip6_fl_lock);
225			return lfl;
226		}
227	}
228
229	fl->lastuse = jiffies;
230	fl->next = fl_ht[FL_HASH(fl->label)];
231	rcu_assign_pointer(fl_ht[FL_HASH(fl->label)], fl);
232	atomic_inc(&fl_size);
233	spin_unlock_bh(&ip6_fl_lock);
234	return NULL;
235}
236
237
238
239/* Socket flowlabel lists */
240
241struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk, __be32 label)
242{
243	struct ipv6_fl_socklist *sfl;
244	struct ipv6_pinfo *np = inet6_sk(sk);
245
246	label &= IPV6_FLOWLABEL_MASK;
247
248	rcu_read_lock_bh();
249	for_each_sk_fl_rcu(np, sfl) {
250		struct ip6_flowlabel *fl = sfl->fl;
251		if (fl->label == label) {
252			fl->lastuse = jiffies;
253			atomic_inc(&fl->users);
254			rcu_read_unlock_bh();
255			return fl;
256		}
257	}
258	rcu_read_unlock_bh();
259	return NULL;
260}
261EXPORT_SYMBOL_GPL(fl6_sock_lookup);
262
263void fl6_free_socklist(struct sock *sk)
264{
265	struct ipv6_pinfo *np = inet6_sk(sk);
266	struct ipv6_fl_socklist *sfl;
267
268	if (!rcu_access_pointer(np->ipv6_fl_list))
269		return;
270
271	spin_lock_bh(&ip6_sk_fl_lock);
272	while ((sfl = rcu_dereference_protected(np->ipv6_fl_list,
273						lockdep_is_held(&ip6_sk_fl_lock))) != NULL) {
274		np->ipv6_fl_list = sfl->next;
275		spin_unlock_bh(&ip6_sk_fl_lock);
276
277		fl_release(sfl->fl);
278		kfree_rcu(sfl, rcu);
279
280		spin_lock_bh(&ip6_sk_fl_lock);
281	}
282	spin_unlock_bh(&ip6_sk_fl_lock);
283}
284
285/* Service routines */
286
287
288/*
289   It is the only difficult place. flowlabel enforces equal headers
290   before and including routing header, however user may supply options
291   following rthdr.
292 */
293
294struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space,
295					 struct ip6_flowlabel *fl,
296					 struct ipv6_txoptions *fopt)
297{
298	struct ipv6_txoptions *fl_opt = fl->opt;
299
300	if (!fopt || fopt->opt_flen == 0)
301		return fl_opt;
302
303	if (fl_opt) {
304		opt_space->hopopt = fl_opt->hopopt;
305		opt_space->dst0opt = fl_opt->dst0opt;
306		opt_space->srcrt = fl_opt->srcrt;
307		opt_space->opt_nflen = fl_opt->opt_nflen;
308	} else {
309		if (fopt->opt_nflen == 0)
310			return fopt;
311		opt_space->hopopt = NULL;
312		opt_space->dst0opt = NULL;
313		opt_space->srcrt = NULL;
314		opt_space->opt_nflen = 0;
315	}
316	opt_space->dst1opt = fopt->dst1opt;
317	opt_space->opt_flen = fopt->opt_flen;
318	return opt_space;
319}
320EXPORT_SYMBOL_GPL(fl6_merge_options);
321
322static unsigned long check_linger(unsigned long ttl)
323{
324	if (ttl < FL_MIN_LINGER)
325		return FL_MIN_LINGER*HZ;
326	if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
327		return 0;
328	return ttl*HZ;
329}
330
331static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
332{
333	linger = check_linger(linger);
334	if (!linger)
335		return -EPERM;
336	expires = check_linger(expires);
337	if (!expires)
338		return -EPERM;
339
340	spin_lock_bh(&ip6_fl_lock);
341	fl->lastuse = jiffies;
342	if (time_before(fl->linger, linger))
343		fl->linger = linger;
344	if (time_before(expires, fl->linger))
345		expires = fl->linger;
346	if (time_before(fl->expires, fl->lastuse + expires))
347		fl->expires = fl->lastuse + expires;
348	spin_unlock_bh(&ip6_fl_lock);
349
350	return 0;
351}
352
353static struct ip6_flowlabel *
354fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq,
355	  char __user *optval, int optlen, int *err_p)
356{
357	struct ip6_flowlabel *fl = NULL;
358	int olen;
359	int addr_type;
360	int err;
361
362	olen = optlen - CMSG_ALIGN(sizeof(*freq));
363	err = -EINVAL;
364	if (olen > 64 * 1024)
365		goto done;
366
367	err = -ENOMEM;
368	fl = kzalloc(sizeof(*fl), GFP_KERNEL);
369	if (!fl)
370		goto done;
371
372	if (olen > 0) {
373		struct msghdr msg;
374		struct flowi6 flowi6;
375		int junk;
376
377		err = -ENOMEM;
378		fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
379		if (!fl->opt)
380			goto done;
381
382		memset(fl->opt, 0, sizeof(*fl->opt));
383		fl->opt->tot_len = sizeof(*fl->opt) + olen;
384		err = -EFAULT;
385		if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
386			goto done;
387
388		msg.msg_controllen = olen;
389		msg.msg_control = (void *)(fl->opt+1);
390		memset(&flowi6, 0, sizeof(flowi6));
391
392		err = ip6_datagram_send_ctl(net, sk, &msg, &flowi6, fl->opt,
393					    &junk, &junk, &junk);
394		if (err)
395			goto done;
396		err = -EINVAL;
397		if (fl->opt->opt_flen)
398			goto done;
399		if (fl->opt->opt_nflen == 0) {
400			kfree(fl->opt);
401			fl->opt = NULL;
402		}
403	}
404
405	fl->fl_net = net;
406	fl->expires = jiffies;
407	err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
408	if (err)
409		goto done;
410	fl->share = freq->flr_share;
411	addr_type = ipv6_addr_type(&freq->flr_dst);
412	if ((addr_type & IPV6_ADDR_MAPPED) ||
413	    addr_type == IPV6_ADDR_ANY) {
414		err = -EINVAL;
415		goto done;
416	}
417	fl->dst = freq->flr_dst;
418	atomic_set(&fl->users, 1);
419	switch (fl->share) {
420	case IPV6_FL_S_EXCL:
421	case IPV6_FL_S_ANY:
422		break;
423	case IPV6_FL_S_PROCESS:
424		fl->owner.pid = get_task_pid(current, PIDTYPE_PID);
425		break;
426	case IPV6_FL_S_USER:
427		fl->owner.uid = current_euid();
428		break;
429	default:
430		err = -EINVAL;
431		goto done;
432	}
433	return fl;
434
435done:
436	fl_free(fl);
437	*err_p = err;
438	return NULL;
439}
440
441static int mem_check(struct sock *sk)
442{
443	struct ipv6_pinfo *np = inet6_sk(sk);
444	struct ipv6_fl_socklist *sfl;
445	int room = FL_MAX_SIZE - atomic_read(&fl_size);
446	int count = 0;
447
448	if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
449		return 0;
450
451	rcu_read_lock_bh();
452	for_each_sk_fl_rcu(np, sfl)
453		count++;
454	rcu_read_unlock_bh();
455
456	if (room <= 0 ||
457	    ((count >= FL_MAX_PER_SOCK ||
458	      (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) &&
459	     !capable(CAP_NET_ADMIN)))
460		return -ENOBUFS;
461
462	return 0;
463}
464
465static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
466		struct ip6_flowlabel *fl)
467{
468	spin_lock_bh(&ip6_sk_fl_lock);
469	sfl->fl = fl;
470	sfl->next = np->ipv6_fl_list;
471	rcu_assign_pointer(np->ipv6_fl_list, sfl);
472	spin_unlock_bh(&ip6_sk_fl_lock);
473}
474
475int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
476			   int flags)
477{
478	struct ipv6_pinfo *np = inet6_sk(sk);
479	struct ipv6_fl_socklist *sfl;
480
481	if (flags & IPV6_FL_F_REMOTE) {
482		freq->flr_label = np->rcv_flowinfo & IPV6_FLOWLABEL_MASK;
483		return 0;
484	}
485
486	if (np->repflow) {
487		freq->flr_label = np->flow_label;
488		return 0;
489	}
490
491	rcu_read_lock_bh();
492
493	for_each_sk_fl_rcu(np, sfl) {
494		if (sfl->fl->label == (np->flow_label & IPV6_FLOWLABEL_MASK)) {
495			spin_lock_bh(&ip6_fl_lock);
496			freq->flr_label = sfl->fl->label;
497			freq->flr_dst = sfl->fl->dst;
498			freq->flr_share = sfl->fl->share;
499			freq->flr_expires = (sfl->fl->expires - jiffies) / HZ;
500			freq->flr_linger = sfl->fl->linger / HZ;
501
502			spin_unlock_bh(&ip6_fl_lock);
503			rcu_read_unlock_bh();
504			return 0;
505		}
506	}
507	rcu_read_unlock_bh();
508
509	return -ENOENT;
510}
511
512int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
513{
514	int uninitialized_var(err);
515	struct net *net = sock_net(sk);
516	struct ipv6_pinfo *np = inet6_sk(sk);
517	struct in6_flowlabel_req freq;
518	struct ipv6_fl_socklist *sfl1 = NULL;
519	struct ipv6_fl_socklist *sfl;
520	struct ipv6_fl_socklist __rcu **sflp;
521	struct ip6_flowlabel *fl, *fl1 = NULL;
522
523
524	if (optlen < sizeof(freq))
525		return -EINVAL;
526
527	if (copy_from_user(&freq, optval, sizeof(freq)))
528		return -EFAULT;
529
530	switch (freq.flr_action) {
531	case IPV6_FL_A_PUT:
532		if (freq.flr_flags & IPV6_FL_F_REFLECT) {
533			if (sk->sk_protocol != IPPROTO_TCP)
534				return -ENOPROTOOPT;
535			if (!np->repflow)
536				return -ESRCH;
537			np->flow_label = 0;
538			np->repflow = 0;
539			return 0;
540		}
541		spin_lock_bh(&ip6_sk_fl_lock);
542		for (sflp = &np->ipv6_fl_list;
543		     (sfl = rcu_dereference_protected(*sflp,
544						      lockdep_is_held(&ip6_sk_fl_lock))) != NULL;
545		     sflp = &sfl->next) {
546			if (sfl->fl->label == freq.flr_label) {
547				if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
548					np->flow_label &= ~IPV6_FLOWLABEL_MASK;
549				*sflp = sfl->next;
550				spin_unlock_bh(&ip6_sk_fl_lock);
551				fl_release(sfl->fl);
552				kfree_rcu(sfl, rcu);
553				return 0;
554			}
555		}
556		spin_unlock_bh(&ip6_sk_fl_lock);
557		return -ESRCH;
558
559	case IPV6_FL_A_RENEW:
560		rcu_read_lock_bh();
561		for_each_sk_fl_rcu(np, sfl) {
562			if (sfl->fl->label == freq.flr_label) {
563				err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
564				rcu_read_unlock_bh();
565				return err;
566			}
567		}
568		rcu_read_unlock_bh();
569
570		if (freq.flr_share == IPV6_FL_S_NONE &&
571		    ns_capable(net->user_ns, CAP_NET_ADMIN)) {
572			fl = fl_lookup(net, freq.flr_label);
573			if (fl) {
574				err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
575				fl_release(fl);
576				return err;
577			}
578		}
579		return -ESRCH;
580
581	case IPV6_FL_A_GET:
582		if (freq.flr_flags & IPV6_FL_F_REFLECT) {
583			struct net *net = sock_net(sk);
584			if (net->ipv6.sysctl.flowlabel_consistency) {
585				net_info_ratelimited("Can not set IPV6_FL_F_REFLECT if flowlabel_consistency sysctl is enable\n");
586				return -EPERM;
587			}
588
589			if (sk->sk_protocol != IPPROTO_TCP)
590				return -ENOPROTOOPT;
591
592			np->repflow = 1;
593			return 0;
594		}
595
596		if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
597			return -EINVAL;
598
599		fl = fl_create(net, sk, &freq, optval, optlen, &err);
600		if (!fl)
601			return err;
602		sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
603
604		if (freq.flr_label) {
605			err = -EEXIST;
606			rcu_read_lock_bh();
607			for_each_sk_fl_rcu(np, sfl) {
608				if (sfl->fl->label == freq.flr_label) {
609					if (freq.flr_flags&IPV6_FL_F_EXCL) {
610						rcu_read_unlock_bh();
611						goto done;
612					}
613					fl1 = sfl->fl;
614					atomic_inc(&fl1->users);
615					break;
616				}
617			}
618			rcu_read_unlock_bh();
619
620			if (!fl1)
621				fl1 = fl_lookup(net, freq.flr_label);
622			if (fl1) {
623recheck:
624				err = -EEXIST;
625				if (freq.flr_flags&IPV6_FL_F_EXCL)
626					goto release;
627				err = -EPERM;
628				if (fl1->share == IPV6_FL_S_EXCL ||
629				    fl1->share != fl->share ||
630				    ((fl1->share == IPV6_FL_S_PROCESS) &&
631				     (fl1->owner.pid == fl->owner.pid)) ||
632				    ((fl1->share == IPV6_FL_S_USER) &&
633				     uid_eq(fl1->owner.uid, fl->owner.uid)))
634					goto release;
635
636				err = -ENOMEM;
637				if (!sfl1)
638					goto release;
639				if (fl->linger > fl1->linger)
640					fl1->linger = fl->linger;
641				if ((long)(fl->expires - fl1->expires) > 0)
642					fl1->expires = fl->expires;
643				fl_link(np, sfl1, fl1);
644				fl_free(fl);
645				return 0;
646
647release:
648				fl_release(fl1);
649				goto done;
650			}
651		}
652		err = -ENOENT;
653		if (!(freq.flr_flags&IPV6_FL_F_CREATE))
654			goto done;
655
656		err = -ENOMEM;
657		if (!sfl1)
658			goto done;
659
660		err = mem_check(sk);
661		if (err != 0)
662			goto done;
663
664		fl1 = fl_intern(net, fl, freq.flr_label);
665		if (fl1)
666			goto recheck;
667
668		if (!freq.flr_label) {
669			if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
670					 &fl->label, sizeof(fl->label))) {
671				/* Intentionally ignore fault. */
672			}
673		}
674
675		fl_link(np, sfl1, fl);
676		return 0;
677
678	default:
679		return -EINVAL;
680	}
681
682done:
683	fl_free(fl);
684	kfree(sfl1);
685	return err;
686}
687
688#ifdef CONFIG_PROC_FS
689
690struct ip6fl_iter_state {
691	struct seq_net_private p;
692	struct pid_namespace *pid_ns;
693	int bucket;
694};
695
696#define ip6fl_seq_private(seq)	((struct ip6fl_iter_state *)(seq)->private)
697
698static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
699{
700	struct ip6_flowlabel *fl = NULL;
701	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
702	struct net *net = seq_file_net(seq);
703
704	for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
705		for_each_fl_rcu(state->bucket, fl) {
706			if (net_eq(fl->fl_net, net))
707				goto out;
708		}
709	}
710	fl = NULL;
711out:
712	return fl;
713}
714
715static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
716{
717	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
718	struct net *net = seq_file_net(seq);
719
720	for_each_fl_continue_rcu(fl) {
721		if (net_eq(fl->fl_net, net))
722			goto out;
723	}
724
725try_again:
726	if (++state->bucket <= FL_HASH_MASK) {
727		for_each_fl_rcu(state->bucket, fl) {
728			if (net_eq(fl->fl_net, net))
729				goto out;
730		}
731		goto try_again;
732	}
733	fl = NULL;
734
735out:
736	return fl;
737}
738
739static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
740{
741	struct ip6_flowlabel *fl = ip6fl_get_first(seq);
742	if (fl)
743		while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
744			--pos;
745	return pos ? NULL : fl;
746}
747
748static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
749	__acquires(RCU)
750{
751	rcu_read_lock_bh();
752	return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
753}
754
755static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
756{
757	struct ip6_flowlabel *fl;
758
759	if (v == SEQ_START_TOKEN)
760		fl = ip6fl_get_first(seq);
761	else
762		fl = ip6fl_get_next(seq, v);
763	++*pos;
764	return fl;
765}
766
767static void ip6fl_seq_stop(struct seq_file *seq, void *v)
768	__releases(RCU)
769{
770	rcu_read_unlock_bh();
771}
772
773static int ip6fl_seq_show(struct seq_file *seq, void *v)
774{
775	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
776	if (v == SEQ_START_TOKEN) {
777		seq_puts(seq, "Label S Owner  Users  Linger Expires  Dst                              Opt\n");
778	} else {
779		struct ip6_flowlabel *fl = v;
780		seq_printf(seq,
781			   "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
782			   (unsigned int)ntohl(fl->label),
783			   fl->share,
784			   ((fl->share == IPV6_FL_S_PROCESS) ?
785			    pid_nr_ns(fl->owner.pid, state->pid_ns) :
786			    ((fl->share == IPV6_FL_S_USER) ?
787			     from_kuid_munged(seq_user_ns(seq), fl->owner.uid) :
788			     0)),
789			   atomic_read(&fl->users),
790			   fl->linger/HZ,
791			   (long)(fl->expires - jiffies)/HZ,
792			   &fl->dst,
793			   fl->opt ? fl->opt->opt_nflen : 0);
794	}
795	return 0;
796}
797
798static const struct seq_operations ip6fl_seq_ops = {
799	.start	=	ip6fl_seq_start,
800	.next	=	ip6fl_seq_next,
801	.stop	=	ip6fl_seq_stop,
802	.show	=	ip6fl_seq_show,
803};
804
805static int ip6fl_seq_open(struct inode *inode, struct file *file)
806{
807	struct seq_file *seq;
808	struct ip6fl_iter_state *state;
809	int err;
810
811	err = seq_open_net(inode, file, &ip6fl_seq_ops,
812			   sizeof(struct ip6fl_iter_state));
813
814	if (!err) {
815		seq = file->private_data;
816		state = ip6fl_seq_private(seq);
817		rcu_read_lock();
818		state->pid_ns = get_pid_ns(task_active_pid_ns(current));
819		rcu_read_unlock();
820	}
821	return err;
822}
823
824static int ip6fl_seq_release(struct inode *inode, struct file *file)
825{
826	struct seq_file *seq = file->private_data;
827	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
828	put_pid_ns(state->pid_ns);
829	return seq_release_net(inode, file);
830}
831
832static const struct file_operations ip6fl_seq_fops = {
833	.owner		=	THIS_MODULE,
834	.open		=	ip6fl_seq_open,
835	.read		=	seq_read,
836	.llseek		=	seq_lseek,
837	.release	=	ip6fl_seq_release,
838};
839
840static int __net_init ip6_flowlabel_proc_init(struct net *net)
841{
842	if (!proc_create("ip6_flowlabel", S_IRUGO, net->proc_net,
843			 &ip6fl_seq_fops))
844		return -ENOMEM;
845	return 0;
846}
847
848static void __net_exit ip6_flowlabel_proc_fini(struct net *net)
849{
850	remove_proc_entry("ip6_flowlabel", net->proc_net);
851}
852#else
853static inline int ip6_flowlabel_proc_init(struct net *net)
854{
855	return 0;
856}
857static inline void ip6_flowlabel_proc_fini(struct net *net)
858{
859}
860#endif
861
862static void __net_exit ip6_flowlabel_net_exit(struct net *net)
863{
864	ip6_fl_purge(net);
865	ip6_flowlabel_proc_fini(net);
866}
867
868static struct pernet_operations ip6_flowlabel_net_ops = {
869	.init = ip6_flowlabel_proc_init,
870	.exit = ip6_flowlabel_net_exit,
871};
872
873int ip6_flowlabel_init(void)
874{
875	return register_pernet_subsys(&ip6_flowlabel_net_ops);
876}
877
878void ip6_flowlabel_cleanup(void)
879{
880	del_timer(&ip6_fl_gc_timer);
881	unregister_pernet_subsys(&ip6_flowlabel_net_ops);
882}
883