xref: /openbmc/linux/net/ipv6/ip6_flowlabel.c (revision afc98d90)
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 #include <linux/export.h>
25 #include <linux/pid_namespace.h>
26 
27 #include <net/net_namespace.h>
28 #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	150	/* 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 
53 static atomic_t fl_size = ATOMIC_INIT(0);
54 static struct ip6_flowlabel __rcu *fl_ht[FL_HASH_MASK+1];
55 
56 static void ip6_fl_gc(unsigned long dummy);
57 static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
58 
59 /* FL hash table lock: it protects only of GC */
60 
61 static DEFINE_SPINLOCK(ip6_fl_lock);
62 
63 /* Big socket sock */
64 
65 static DEFINE_SPINLOCK(ip6_sk_fl_lock);
66 
67 #define for_each_fl_rcu(hash, fl)				\
68 	for (fl = rcu_dereference_bh(fl_ht[(hash)]);		\
69 	     fl != NULL;					\
70 	     fl = rcu_dereference_bh(fl->next))
71 #define for_each_fl_continue_rcu(fl)				\
72 	for (fl = rcu_dereference_bh(fl->next);			\
73 	     fl != NULL;					\
74 	     fl = rcu_dereference_bh(fl->next))
75 
76 #define for_each_sk_fl_rcu(np, sfl)				\
77 	for (sfl = rcu_dereference_bh(np->ipv6_fl_list);	\
78 	     sfl != NULL;					\
79 	     sfl = rcu_dereference_bh(sfl->next))
80 
81 static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label)
82 {
83 	struct ip6_flowlabel *fl;
84 
85 	for_each_fl_rcu(FL_HASH(label), fl) {
86 		if (fl->label == label && net_eq(fl->fl_net, net))
87 			return fl;
88 	}
89 	return NULL;
90 }
91 
92 static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label)
93 {
94 	struct ip6_flowlabel *fl;
95 
96 	rcu_read_lock_bh();
97 	fl = __fl_lookup(net, label);
98 	if (fl && !atomic_inc_not_zero(&fl->users))
99 		fl = NULL;
100 	rcu_read_unlock_bh();
101 	return fl;
102 }
103 
104 
105 static void fl_free(struct ip6_flowlabel *fl)
106 {
107 	if (fl) {
108 		if (fl->share == IPV6_FL_S_PROCESS)
109 			put_pid(fl->owner.pid);
110 		release_net(fl->fl_net);
111 		kfree(fl->opt);
112 		kfree_rcu(fl, rcu);
113 	}
114 }
115 
116 static void fl_release(struct ip6_flowlabel *fl)
117 {
118 	spin_lock_bh(&ip6_fl_lock);
119 
120 	fl->lastuse = jiffies;
121 	if (atomic_dec_and_test(&fl->users)) {
122 		unsigned long ttd = fl->lastuse + fl->linger;
123 		if (time_after(ttd, fl->expires))
124 			fl->expires = ttd;
125 		ttd = fl->expires;
126 		if (fl->opt && fl->share == IPV6_FL_S_EXCL) {
127 			struct ipv6_txoptions *opt = fl->opt;
128 			fl->opt = NULL;
129 			kfree(opt);
130 		}
131 		if (!timer_pending(&ip6_fl_gc_timer) ||
132 		    time_after(ip6_fl_gc_timer.expires, ttd))
133 			mod_timer(&ip6_fl_gc_timer, ttd);
134 	}
135 	spin_unlock_bh(&ip6_fl_lock);
136 }
137 
138 static void ip6_fl_gc(unsigned long dummy)
139 {
140 	int i;
141 	unsigned long now = jiffies;
142 	unsigned long sched = 0;
143 
144 	spin_lock(&ip6_fl_lock);
145 
146 	for (i=0; i<=FL_HASH_MASK; i++) {
147 		struct ip6_flowlabel *fl;
148 		struct ip6_flowlabel __rcu **flp;
149 
150 		flp = &fl_ht[i];
151 		while ((fl = rcu_dereference_protected(*flp,
152 						       lockdep_is_held(&ip6_fl_lock))) != NULL) {
153 			if (atomic_read(&fl->users) == 0) {
154 				unsigned long ttd = fl->lastuse + fl->linger;
155 				if (time_after(ttd, fl->expires))
156 					fl->expires = ttd;
157 				ttd = fl->expires;
158 				if (time_after_eq(now, ttd)) {
159 					*flp = fl->next;
160 					fl_free(fl);
161 					atomic_dec(&fl_size);
162 					continue;
163 				}
164 				if (!sched || time_before(ttd, sched))
165 					sched = ttd;
166 			}
167 			flp = &fl->next;
168 		}
169 	}
170 	if (!sched && atomic_read(&fl_size))
171 		sched = now + FL_MAX_LINGER;
172 	if (sched) {
173 		mod_timer(&ip6_fl_gc_timer, sched);
174 	}
175 	spin_unlock(&ip6_fl_lock);
176 }
177 
178 static void __net_exit ip6_fl_purge(struct net *net)
179 {
180 	int i;
181 
182 	spin_lock(&ip6_fl_lock);
183 	for (i = 0; i <= FL_HASH_MASK; i++) {
184 		struct ip6_flowlabel *fl;
185 		struct ip6_flowlabel __rcu **flp;
186 
187 		flp = &fl_ht[i];
188 		while ((fl = rcu_dereference_protected(*flp,
189 						       lockdep_is_held(&ip6_fl_lock))) != NULL) {
190 			if (net_eq(fl->fl_net, net) &&
191 			    atomic_read(&fl->users) == 0) {
192 				*flp = fl->next;
193 				fl_free(fl);
194 				atomic_dec(&fl_size);
195 				continue;
196 			}
197 			flp = &fl->next;
198 		}
199 	}
200 	spin_unlock(&ip6_fl_lock);
201 }
202 
203 static struct ip6_flowlabel *fl_intern(struct net *net,
204 				       struct ip6_flowlabel *fl, __be32 label)
205 {
206 	struct ip6_flowlabel *lfl;
207 
208 	fl->label = label & IPV6_FLOWLABEL_MASK;
209 
210 	spin_lock_bh(&ip6_fl_lock);
211 	if (label == 0) {
212 		for (;;) {
213 			fl->label = htonl(prandom_u32())&IPV6_FLOWLABEL_MASK;
214 			if (fl->label) {
215 				lfl = __fl_lookup(net, fl->label);
216 				if (lfl == NULL)
217 					break;
218 			}
219 		}
220 	} else {
221 		/*
222 		 * we dropper the ip6_fl_lock, so this entry could reappear
223 		 * and we need to recheck with it.
224 		 *
225 		 * OTOH no need to search the active socket first, like it is
226 		 * done in ipv6_flowlabel_opt - sock is locked, so new entry
227 		 * with the same label can only appear on another sock
228 		 */
229 		lfl = __fl_lookup(net, fl->label);
230 		if (lfl != NULL) {
231 			atomic_inc(&lfl->users);
232 			spin_unlock_bh(&ip6_fl_lock);
233 			return lfl;
234 		}
235 	}
236 
237 	fl->lastuse = jiffies;
238 	fl->next = fl_ht[FL_HASH(fl->label)];
239 	rcu_assign_pointer(fl_ht[FL_HASH(fl->label)], fl);
240 	atomic_inc(&fl_size);
241 	spin_unlock_bh(&ip6_fl_lock);
242 	return NULL;
243 }
244 
245 
246 
247 /* Socket flowlabel lists */
248 
249 struct ip6_flowlabel * fl6_sock_lookup(struct sock *sk, __be32 label)
250 {
251 	struct ipv6_fl_socklist *sfl;
252 	struct ipv6_pinfo *np = inet6_sk(sk);
253 
254 	label &= IPV6_FLOWLABEL_MASK;
255 
256 	rcu_read_lock_bh();
257 	for_each_sk_fl_rcu(np, sfl) {
258 		struct ip6_flowlabel *fl = sfl->fl;
259 		if (fl->label == label) {
260 			fl->lastuse = jiffies;
261 			atomic_inc(&fl->users);
262 			rcu_read_unlock_bh();
263 			return fl;
264 		}
265 	}
266 	rcu_read_unlock_bh();
267 	return NULL;
268 }
269 
270 EXPORT_SYMBOL_GPL(fl6_sock_lookup);
271 
272 void fl6_free_socklist(struct sock *sk)
273 {
274 	struct ipv6_pinfo *np = inet6_sk(sk);
275 	struct ipv6_fl_socklist *sfl;
276 
277 	if (!rcu_access_pointer(np->ipv6_fl_list))
278 		return;
279 
280 	spin_lock_bh(&ip6_sk_fl_lock);
281 	while ((sfl = rcu_dereference_protected(np->ipv6_fl_list,
282 						lockdep_is_held(&ip6_sk_fl_lock))) != NULL) {
283 		np->ipv6_fl_list = sfl->next;
284 		spin_unlock_bh(&ip6_sk_fl_lock);
285 
286 		fl_release(sfl->fl);
287 		kfree_rcu(sfl, rcu);
288 
289 		spin_lock_bh(&ip6_sk_fl_lock);
290 	}
291 	spin_unlock_bh(&ip6_sk_fl_lock);
292 }
293 
294 /* Service routines */
295 
296 
297 /*
298    It is the only difficult place. flowlabel enforces equal headers
299    before and including routing header, however user may supply options
300    following rthdr.
301  */
302 
303 struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions * opt_space,
304 					 struct ip6_flowlabel * fl,
305 					 struct ipv6_txoptions * fopt)
306 {
307 	struct ipv6_txoptions * fl_opt = fl->opt;
308 
309 	if (fopt == NULL || fopt->opt_flen == 0)
310 		return fl_opt;
311 
312 	if (fl_opt != NULL) {
313 		opt_space->hopopt = fl_opt->hopopt;
314 		opt_space->dst0opt = fl_opt->dst0opt;
315 		opt_space->srcrt = fl_opt->srcrt;
316 		opt_space->opt_nflen = fl_opt->opt_nflen;
317 	} else {
318 		if (fopt->opt_nflen == 0)
319 			return fopt;
320 		opt_space->hopopt = NULL;
321 		opt_space->dst0opt = NULL;
322 		opt_space->srcrt = NULL;
323 		opt_space->opt_nflen = 0;
324 	}
325 	opt_space->dst1opt = fopt->dst1opt;
326 	opt_space->opt_flen = fopt->opt_flen;
327 	return opt_space;
328 }
329 EXPORT_SYMBOL_GPL(fl6_merge_options);
330 
331 static unsigned long check_linger(unsigned long ttl)
332 {
333 	if (ttl < FL_MIN_LINGER)
334 		return FL_MIN_LINGER*HZ;
335 	if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN))
336 		return 0;
337 	return ttl*HZ;
338 }
339 
340 static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires)
341 {
342 	linger = check_linger(linger);
343 	if (!linger)
344 		return -EPERM;
345 	expires = check_linger(expires);
346 	if (!expires)
347 		return -EPERM;
348 
349 	spin_lock_bh(&ip6_fl_lock);
350 	fl->lastuse = jiffies;
351 	if (time_before(fl->linger, linger))
352 		fl->linger = linger;
353 	if (time_before(expires, fl->linger))
354 		expires = fl->linger;
355 	if (time_before(fl->expires, fl->lastuse + expires))
356 		fl->expires = fl->lastuse + expires;
357 	spin_unlock_bh(&ip6_fl_lock);
358 
359 	return 0;
360 }
361 
362 static struct ip6_flowlabel *
363 fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq,
364 	  char __user *optval, int optlen, int *err_p)
365 {
366 	struct ip6_flowlabel *fl = NULL;
367 	int olen;
368 	int addr_type;
369 	int err;
370 
371 	olen = optlen - CMSG_ALIGN(sizeof(*freq));
372 	err = -EINVAL;
373 	if (olen > 64 * 1024)
374 		goto done;
375 
376 	err = -ENOMEM;
377 	fl = kzalloc(sizeof(*fl), GFP_KERNEL);
378 	if (fl == NULL)
379 		goto done;
380 
381 	if (olen > 0) {
382 		struct msghdr msg;
383 		struct flowi6 flowi6;
384 		int junk;
385 
386 		err = -ENOMEM;
387 		fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL);
388 		if (fl->opt == NULL)
389 			goto done;
390 
391 		memset(fl->opt, 0, sizeof(*fl->opt));
392 		fl->opt->tot_len = sizeof(*fl->opt) + olen;
393 		err = -EFAULT;
394 		if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
395 			goto done;
396 
397 		msg.msg_controllen = olen;
398 		msg.msg_control = (void*)(fl->opt+1);
399 		memset(&flowi6, 0, sizeof(flowi6));
400 
401 		err = ip6_datagram_send_ctl(net, sk, &msg, &flowi6, fl->opt,
402 					    &junk, &junk, &junk);
403 		if (err)
404 			goto done;
405 		err = -EINVAL;
406 		if (fl->opt->opt_flen)
407 			goto done;
408 		if (fl->opt->opt_nflen == 0) {
409 			kfree(fl->opt);
410 			fl->opt = NULL;
411 		}
412 	}
413 
414 	fl->fl_net = hold_net(net);
415 	fl->expires = jiffies;
416 	err = fl6_renew(fl, freq->flr_linger, freq->flr_expires);
417 	if (err)
418 		goto done;
419 	fl->share = freq->flr_share;
420 	addr_type = ipv6_addr_type(&freq->flr_dst);
421 	if ((addr_type & IPV6_ADDR_MAPPED) ||
422 	    addr_type == IPV6_ADDR_ANY) {
423 		err = -EINVAL;
424 		goto done;
425 	}
426 	fl->dst = freq->flr_dst;
427 	atomic_set(&fl->users, 1);
428 	switch (fl->share) {
429 	case IPV6_FL_S_EXCL:
430 	case IPV6_FL_S_ANY:
431 		break;
432 	case IPV6_FL_S_PROCESS:
433 		fl->owner.pid = get_task_pid(current, PIDTYPE_PID);
434 		break;
435 	case IPV6_FL_S_USER:
436 		fl->owner.uid = current_euid();
437 		break;
438 	default:
439 		err = -EINVAL;
440 		goto done;
441 	}
442 	return fl;
443 
444 done:
445 	fl_free(fl);
446 	*err_p = err;
447 	return NULL;
448 }
449 
450 static int mem_check(struct sock *sk)
451 {
452 	struct ipv6_pinfo *np = inet6_sk(sk);
453 	struct ipv6_fl_socklist *sfl;
454 	int room = FL_MAX_SIZE - atomic_read(&fl_size);
455 	int count = 0;
456 
457 	if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK)
458 		return 0;
459 
460 	rcu_read_lock_bh();
461 	for_each_sk_fl_rcu(np, sfl)
462 		count++;
463 	rcu_read_unlock_bh();
464 
465 	if (room <= 0 ||
466 	    ((count >= FL_MAX_PER_SOCK ||
467 	      (count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) &&
468 	     !capable(CAP_NET_ADMIN)))
469 		return -ENOBUFS;
470 
471 	return 0;
472 }
473 
474 static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
475 		struct ip6_flowlabel *fl)
476 {
477 	spin_lock_bh(&ip6_sk_fl_lock);
478 	sfl->fl = fl;
479 	sfl->next = np->ipv6_fl_list;
480 	rcu_assign_pointer(np->ipv6_fl_list, sfl);
481 	spin_unlock_bh(&ip6_sk_fl_lock);
482 }
483 
484 int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
485 			   int flags)
486 {
487 	struct ipv6_pinfo *np = inet6_sk(sk);
488 	struct ipv6_fl_socklist *sfl;
489 
490 	if (flags & IPV6_FL_F_REMOTE) {
491 		freq->flr_label = np->rcv_flowinfo & IPV6_FLOWLABEL_MASK;
492 		return 0;
493 	}
494 
495 	if (np->repflow) {
496 		freq->flr_label = np->flow_label;
497 		return 0;
498 	}
499 
500 	rcu_read_lock_bh();
501 
502 	for_each_sk_fl_rcu(np, sfl) {
503 		if (sfl->fl->label == (np->flow_label & IPV6_FLOWLABEL_MASK)) {
504 			spin_lock_bh(&ip6_fl_lock);
505 			freq->flr_label = sfl->fl->label;
506 			freq->flr_dst = sfl->fl->dst;
507 			freq->flr_share = sfl->fl->share;
508 			freq->flr_expires = (sfl->fl->expires - jiffies) / HZ;
509 			freq->flr_linger = sfl->fl->linger / HZ;
510 
511 			spin_unlock_bh(&ip6_fl_lock);
512 			rcu_read_unlock_bh();
513 			return 0;
514 		}
515 	}
516 	rcu_read_unlock_bh();
517 
518 	return -ENOENT;
519 }
520 
521 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
522 {
523 	int uninitialized_var(err);
524 	struct net *net = sock_net(sk);
525 	struct ipv6_pinfo *np = inet6_sk(sk);
526 	struct in6_flowlabel_req freq;
527 	struct ipv6_fl_socklist *sfl1=NULL;
528 	struct ipv6_fl_socklist *sfl;
529 	struct ipv6_fl_socklist __rcu **sflp;
530 	struct ip6_flowlabel *fl, *fl1 = NULL;
531 
532 
533 	if (optlen < sizeof(freq))
534 		return -EINVAL;
535 
536 	if (copy_from_user(&freq, optval, sizeof(freq)))
537 		return -EFAULT;
538 
539 	switch (freq.flr_action) {
540 	case IPV6_FL_A_PUT:
541 		if (freq.flr_flags & IPV6_FL_F_REFLECT) {
542 			if (sk->sk_protocol != IPPROTO_TCP)
543 				return -ENOPROTOOPT;
544 			if (!np->repflow)
545 				return -ESRCH;
546 			np->flow_label = 0;
547 			np->repflow = 0;
548 			return 0;
549 		}
550 		spin_lock_bh(&ip6_sk_fl_lock);
551 		for (sflp = &np->ipv6_fl_list;
552 		     (sfl = rcu_dereference(*sflp))!=NULL;
553 		     sflp = &sfl->next) {
554 			if (sfl->fl->label == freq.flr_label) {
555 				if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
556 					np->flow_label &= ~IPV6_FLOWLABEL_MASK;
557 				*sflp = rcu_dereference(sfl->next);
558 				spin_unlock_bh(&ip6_sk_fl_lock);
559 				fl_release(sfl->fl);
560 				kfree_rcu(sfl, rcu);
561 				return 0;
562 			}
563 		}
564 		spin_unlock_bh(&ip6_sk_fl_lock);
565 		return -ESRCH;
566 
567 	case IPV6_FL_A_RENEW:
568 		rcu_read_lock_bh();
569 		for_each_sk_fl_rcu(np, sfl) {
570 			if (sfl->fl->label == freq.flr_label) {
571 				err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
572 				rcu_read_unlock_bh();
573 				return err;
574 			}
575 		}
576 		rcu_read_unlock_bh();
577 
578 		if (freq.flr_share == IPV6_FL_S_NONE &&
579 		    ns_capable(net->user_ns, CAP_NET_ADMIN)) {
580 			fl = fl_lookup(net, freq.flr_label);
581 			if (fl) {
582 				err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
583 				fl_release(fl);
584 				return err;
585 			}
586 		}
587 		return -ESRCH;
588 
589 	case IPV6_FL_A_GET:
590 		if (freq.flr_flags & IPV6_FL_F_REFLECT) {
591 			struct net *net = sock_net(sk);
592 			if (net->ipv6.sysctl.flowlabel_consistency) {
593 				net_info_ratelimited("Can not set IPV6_FL_F_REFLECT if flowlabel_consistency sysctl is enable\n");
594 				return -EPERM;
595 			}
596 
597 			if (sk->sk_protocol != IPPROTO_TCP)
598 				return -ENOPROTOOPT;
599 
600 			np->repflow = 1;
601 			return 0;
602 		}
603 
604 		if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
605 			return -EINVAL;
606 
607 		fl = fl_create(net, sk, &freq, optval, optlen, &err);
608 		if (fl == NULL)
609 			return err;
610 		sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
611 
612 		if (freq.flr_label) {
613 			err = -EEXIST;
614 			rcu_read_lock_bh();
615 			for_each_sk_fl_rcu(np, sfl) {
616 				if (sfl->fl->label == freq.flr_label) {
617 					if (freq.flr_flags&IPV6_FL_F_EXCL) {
618 						rcu_read_unlock_bh();
619 						goto done;
620 					}
621 					fl1 = sfl->fl;
622 					atomic_inc(&fl1->users);
623 					break;
624 				}
625 			}
626 			rcu_read_unlock_bh();
627 
628 			if (fl1 == NULL)
629 				fl1 = fl_lookup(net, freq.flr_label);
630 			if (fl1) {
631 recheck:
632 				err = -EEXIST;
633 				if (freq.flr_flags&IPV6_FL_F_EXCL)
634 					goto release;
635 				err = -EPERM;
636 				if (fl1->share == IPV6_FL_S_EXCL ||
637 				    fl1->share != fl->share ||
638 				    ((fl1->share == IPV6_FL_S_PROCESS) &&
639 				     (fl1->owner.pid == fl->owner.pid)) ||
640 				    ((fl1->share == IPV6_FL_S_USER) &&
641 				     uid_eq(fl1->owner.uid, fl->owner.uid)))
642 					goto release;
643 
644 				err = -ENOMEM;
645 				if (sfl1 == NULL)
646 					goto release;
647 				if (fl->linger > fl1->linger)
648 					fl1->linger = fl->linger;
649 				if ((long)(fl->expires - fl1->expires) > 0)
650 					fl1->expires = fl->expires;
651 				fl_link(np, sfl1, fl1);
652 				fl_free(fl);
653 				return 0;
654 
655 release:
656 				fl_release(fl1);
657 				goto done;
658 			}
659 		}
660 		err = -ENOENT;
661 		if (!(freq.flr_flags&IPV6_FL_F_CREATE))
662 			goto done;
663 
664 		err = -ENOMEM;
665 		if (sfl1 == NULL || (err = mem_check(sk)) != 0)
666 			goto done;
667 
668 		fl1 = fl_intern(net, fl, freq.flr_label);
669 		if (fl1 != NULL)
670 			goto recheck;
671 
672 		if (!freq.flr_label) {
673 			if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
674 					 &fl->label, sizeof(fl->label))) {
675 				/* Intentionally ignore fault. */
676 			}
677 		}
678 
679 		fl_link(np, sfl1, fl);
680 		return 0;
681 
682 	default:
683 		return -EINVAL;
684 	}
685 
686 done:
687 	fl_free(fl);
688 	kfree(sfl1);
689 	return err;
690 }
691 
692 #ifdef CONFIG_PROC_FS
693 
694 struct ip6fl_iter_state {
695 	struct seq_net_private p;
696 	struct pid_namespace *pid_ns;
697 	int bucket;
698 };
699 
700 #define ip6fl_seq_private(seq)	((struct ip6fl_iter_state *)(seq)->private)
701 
702 static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
703 {
704 	struct ip6_flowlabel *fl = NULL;
705 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
706 	struct net *net = seq_file_net(seq);
707 
708 	for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
709 		for_each_fl_rcu(state->bucket, fl) {
710 			if (net_eq(fl->fl_net, net))
711 				goto out;
712 		}
713 	}
714 	fl = NULL;
715 out:
716 	return fl;
717 }
718 
719 static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
720 {
721 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
722 	struct net *net = seq_file_net(seq);
723 
724 	for_each_fl_continue_rcu(fl) {
725 		if (net_eq(fl->fl_net, net))
726 			goto out;
727 	}
728 
729 try_again:
730 	if (++state->bucket <= FL_HASH_MASK) {
731 		for_each_fl_rcu(state->bucket, fl) {
732 			if (net_eq(fl->fl_net, net))
733 				goto out;
734 		}
735 		goto try_again;
736 	}
737 	fl = NULL;
738 
739 out:
740 	return fl;
741 }
742 
743 static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
744 {
745 	struct ip6_flowlabel *fl = ip6fl_get_first(seq);
746 	if (fl)
747 		while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
748 			--pos;
749 	return pos ? NULL : fl;
750 }
751 
752 static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
753 	__acquires(RCU)
754 {
755 	rcu_read_lock_bh();
756 	return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
757 }
758 
759 static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
760 {
761 	struct ip6_flowlabel *fl;
762 
763 	if (v == SEQ_START_TOKEN)
764 		fl = ip6fl_get_first(seq);
765 	else
766 		fl = ip6fl_get_next(seq, v);
767 	++*pos;
768 	return fl;
769 }
770 
771 static void ip6fl_seq_stop(struct seq_file *seq, void *v)
772 	__releases(RCU)
773 {
774 	rcu_read_unlock_bh();
775 }
776 
777 static int ip6fl_seq_show(struct seq_file *seq, void *v)
778 {
779 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
780 	if (v == SEQ_START_TOKEN)
781 		seq_printf(seq, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n",
782 			   "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt");
783 	else {
784 		struct ip6_flowlabel *fl = v;
785 		seq_printf(seq,
786 			   "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
787 			   (unsigned int)ntohl(fl->label),
788 			   fl->share,
789 			   ((fl->share == IPV6_FL_S_PROCESS) ?
790 			    pid_nr_ns(fl->owner.pid, state->pid_ns) :
791 			    ((fl->share == IPV6_FL_S_USER) ?
792 			     from_kuid_munged(seq_user_ns(seq), fl->owner.uid) :
793 			     0)),
794 			   atomic_read(&fl->users),
795 			   fl->linger/HZ,
796 			   (long)(fl->expires - jiffies)/HZ,
797 			   &fl->dst,
798 			   fl->opt ? fl->opt->opt_nflen : 0);
799 	}
800 	return 0;
801 }
802 
803 static const struct seq_operations ip6fl_seq_ops = {
804 	.start	=	ip6fl_seq_start,
805 	.next	=	ip6fl_seq_next,
806 	.stop	=	ip6fl_seq_stop,
807 	.show	=	ip6fl_seq_show,
808 };
809 
810 static int ip6fl_seq_open(struct inode *inode, struct file *file)
811 {
812 	struct seq_file *seq;
813 	struct ip6fl_iter_state *state;
814 	int err;
815 
816 	err = seq_open_net(inode, file, &ip6fl_seq_ops,
817 			   sizeof(struct ip6fl_iter_state));
818 
819 	if (!err) {
820 		seq = file->private_data;
821 		state = ip6fl_seq_private(seq);
822 		rcu_read_lock();
823 		state->pid_ns = get_pid_ns(task_active_pid_ns(current));
824 		rcu_read_unlock();
825 	}
826 	return err;
827 }
828 
829 static int ip6fl_seq_release(struct inode *inode, struct file *file)
830 {
831 	struct seq_file *seq = file->private_data;
832 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
833 	put_pid_ns(state->pid_ns);
834 	return seq_release_net(inode, file);
835 }
836 
837 static const struct file_operations ip6fl_seq_fops = {
838 	.owner		=	THIS_MODULE,
839 	.open		=	ip6fl_seq_open,
840 	.read		=	seq_read,
841 	.llseek		=	seq_lseek,
842 	.release	=	ip6fl_seq_release,
843 };
844 
845 static int __net_init ip6_flowlabel_proc_init(struct net *net)
846 {
847 	if (!proc_create("ip6_flowlabel", S_IRUGO, net->proc_net,
848 			 &ip6fl_seq_fops))
849 		return -ENOMEM;
850 	return 0;
851 }
852 
853 static void __net_exit ip6_flowlabel_proc_fini(struct net *net)
854 {
855 	remove_proc_entry("ip6_flowlabel", net->proc_net);
856 }
857 #else
858 static inline int ip6_flowlabel_proc_init(struct net *net)
859 {
860 	return 0;
861 }
862 static inline void ip6_flowlabel_proc_fini(struct net *net)
863 {
864 }
865 #endif
866 
867 static void __net_exit ip6_flowlabel_net_exit(struct net *net)
868 {
869 	ip6_fl_purge(net);
870 	ip6_flowlabel_proc_fini(net);
871 }
872 
873 static struct pernet_operations ip6_flowlabel_net_ops = {
874 	.init = ip6_flowlabel_proc_init,
875 	.exit = ip6_flowlabel_net_exit,
876 };
877 
878 int ip6_flowlabel_init(void)
879 {
880 	return register_pernet_subsys(&ip6_flowlabel_net_ops);
881 }
882 
883 void ip6_flowlabel_cleanup(void)
884 {
885 	del_timer(&ip6_fl_gc_timer);
886 	unregister_pernet_subsys(&ip6_flowlabel_net_ops);
887 }
888