xref: /openbmc/linux/net/ipv6/ip6_flowlabel.c (revision 2c684d89)
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 
46 static atomic_t fl_size = ATOMIC_INIT(0);
47 static struct ip6_flowlabel __rcu *fl_ht[FL_HASH_MASK+1];
48 
49 static void ip6_fl_gc(unsigned long dummy);
50 static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
51 
52 /* FL hash table lock: it protects only of GC */
53 
54 static DEFINE_SPINLOCK(ip6_fl_lock);
55 
56 /* Big socket sock */
57 
58 static 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 
74 static 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 
85 static 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 
98 static 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 
108 static 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 
130 static 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 
170 static 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 
195 static 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 
241 struct 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 }
261 EXPORT_SYMBOL_GPL(fl6_sock_lookup);
262 
263 void 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 
294 struct 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 }
320 EXPORT_SYMBOL_GPL(fl6_merge_options);
321 
322 static 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 
331 static 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 
353 static struct ip6_flowlabel *
354 fl_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 
435 done:
436 	fl_free(fl);
437 	*err_p = err;
438 	return NULL;
439 }
440 
441 static 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 
465 static 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 
475 int 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 
512 int 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(*sflp)) != NULL;
544 		     sflp = &sfl->next) {
545 			if (sfl->fl->label == freq.flr_label) {
546 				if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
547 					np->flow_label &= ~IPV6_FLOWLABEL_MASK;
548 				*sflp = rcu_dereference(sfl->next);
549 				spin_unlock_bh(&ip6_sk_fl_lock);
550 				fl_release(sfl->fl);
551 				kfree_rcu(sfl, rcu);
552 				return 0;
553 			}
554 		}
555 		spin_unlock_bh(&ip6_sk_fl_lock);
556 		return -ESRCH;
557 
558 	case IPV6_FL_A_RENEW:
559 		rcu_read_lock_bh();
560 		for_each_sk_fl_rcu(np, sfl) {
561 			if (sfl->fl->label == freq.flr_label) {
562 				err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
563 				rcu_read_unlock_bh();
564 				return err;
565 			}
566 		}
567 		rcu_read_unlock_bh();
568 
569 		if (freq.flr_share == IPV6_FL_S_NONE &&
570 		    ns_capable(net->user_ns, CAP_NET_ADMIN)) {
571 			fl = fl_lookup(net, freq.flr_label);
572 			if (fl) {
573 				err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
574 				fl_release(fl);
575 				return err;
576 			}
577 		}
578 		return -ESRCH;
579 
580 	case IPV6_FL_A_GET:
581 		if (freq.flr_flags & IPV6_FL_F_REFLECT) {
582 			struct net *net = sock_net(sk);
583 			if (net->ipv6.sysctl.flowlabel_consistency) {
584 				net_info_ratelimited("Can not set IPV6_FL_F_REFLECT if flowlabel_consistency sysctl is enable\n");
585 				return -EPERM;
586 			}
587 
588 			if (sk->sk_protocol != IPPROTO_TCP)
589 				return -ENOPROTOOPT;
590 
591 			np->repflow = 1;
592 			return 0;
593 		}
594 
595 		if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
596 			return -EINVAL;
597 
598 		if (net->ipv6.sysctl.flowlabel_state_ranges &&
599 		    (freq.flr_label & IPV6_FLOWLABEL_STATELESS_FLAG))
600 			return -ERANGE;
601 
602 		fl = fl_create(net, sk, &freq, optval, optlen, &err);
603 		if (!fl)
604 			return err;
605 		sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
606 
607 		if (freq.flr_label) {
608 			err = -EEXIST;
609 			rcu_read_lock_bh();
610 			for_each_sk_fl_rcu(np, sfl) {
611 				if (sfl->fl->label == freq.flr_label) {
612 					if (freq.flr_flags&IPV6_FL_F_EXCL) {
613 						rcu_read_unlock_bh();
614 						goto done;
615 					}
616 					fl1 = sfl->fl;
617 					atomic_inc(&fl1->users);
618 					break;
619 				}
620 			}
621 			rcu_read_unlock_bh();
622 
623 			if (!fl1)
624 				fl1 = fl_lookup(net, freq.flr_label);
625 			if (fl1) {
626 recheck:
627 				err = -EEXIST;
628 				if (freq.flr_flags&IPV6_FL_F_EXCL)
629 					goto release;
630 				err = -EPERM;
631 				if (fl1->share == IPV6_FL_S_EXCL ||
632 				    fl1->share != fl->share ||
633 				    ((fl1->share == IPV6_FL_S_PROCESS) &&
634 				     (fl1->owner.pid == fl->owner.pid)) ||
635 				    ((fl1->share == IPV6_FL_S_USER) &&
636 				     uid_eq(fl1->owner.uid, fl->owner.uid)))
637 					goto release;
638 
639 				err = -ENOMEM;
640 				if (!sfl1)
641 					goto release;
642 				if (fl->linger > fl1->linger)
643 					fl1->linger = fl->linger;
644 				if ((long)(fl->expires - fl1->expires) > 0)
645 					fl1->expires = fl->expires;
646 				fl_link(np, sfl1, fl1);
647 				fl_free(fl);
648 				return 0;
649 
650 release:
651 				fl_release(fl1);
652 				goto done;
653 			}
654 		}
655 		err = -ENOENT;
656 		if (!(freq.flr_flags&IPV6_FL_F_CREATE))
657 			goto done;
658 
659 		err = -ENOMEM;
660 		if (!sfl1)
661 			goto done;
662 
663 		err = mem_check(sk);
664 		if (err != 0)
665 			goto done;
666 
667 		fl1 = fl_intern(net, fl, freq.flr_label);
668 		if (fl1)
669 			goto recheck;
670 
671 		if (!freq.flr_label) {
672 			if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
673 					 &fl->label, sizeof(fl->label))) {
674 				/* Intentionally ignore fault. */
675 			}
676 		}
677 
678 		fl_link(np, sfl1, fl);
679 		return 0;
680 
681 	default:
682 		return -EINVAL;
683 	}
684 
685 done:
686 	fl_free(fl);
687 	kfree(sfl1);
688 	return err;
689 }
690 
691 #ifdef CONFIG_PROC_FS
692 
693 struct ip6fl_iter_state {
694 	struct seq_net_private p;
695 	struct pid_namespace *pid_ns;
696 	int bucket;
697 };
698 
699 #define ip6fl_seq_private(seq)	((struct ip6fl_iter_state *)(seq)->private)
700 
701 static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
702 {
703 	struct ip6_flowlabel *fl = NULL;
704 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
705 	struct net *net = seq_file_net(seq);
706 
707 	for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
708 		for_each_fl_rcu(state->bucket, fl) {
709 			if (net_eq(fl->fl_net, net))
710 				goto out;
711 		}
712 	}
713 	fl = NULL;
714 out:
715 	return fl;
716 }
717 
718 static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl)
719 {
720 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
721 	struct net *net = seq_file_net(seq);
722 
723 	for_each_fl_continue_rcu(fl) {
724 		if (net_eq(fl->fl_net, net))
725 			goto out;
726 	}
727 
728 try_again:
729 	if (++state->bucket <= FL_HASH_MASK) {
730 		for_each_fl_rcu(state->bucket, fl) {
731 			if (net_eq(fl->fl_net, net))
732 				goto out;
733 		}
734 		goto try_again;
735 	}
736 	fl = NULL;
737 
738 out:
739 	return fl;
740 }
741 
742 static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
743 {
744 	struct ip6_flowlabel *fl = ip6fl_get_first(seq);
745 	if (fl)
746 		while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL)
747 			--pos;
748 	return pos ? NULL : fl;
749 }
750 
751 static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
752 	__acquires(RCU)
753 {
754 	rcu_read_lock_bh();
755 	return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
756 }
757 
758 static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
759 {
760 	struct ip6_flowlabel *fl;
761 
762 	if (v == SEQ_START_TOKEN)
763 		fl = ip6fl_get_first(seq);
764 	else
765 		fl = ip6fl_get_next(seq, v);
766 	++*pos;
767 	return fl;
768 }
769 
770 static void ip6fl_seq_stop(struct seq_file *seq, void *v)
771 	__releases(RCU)
772 {
773 	rcu_read_unlock_bh();
774 }
775 
776 static int ip6fl_seq_show(struct seq_file *seq, void *v)
777 {
778 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
779 	if (v == SEQ_START_TOKEN) {
780 		seq_puts(seq, "Label S Owner  Users  Linger Expires  Dst                              Opt\n");
781 	} else {
782 		struct ip6_flowlabel *fl = v;
783 		seq_printf(seq,
784 			   "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
785 			   (unsigned int)ntohl(fl->label),
786 			   fl->share,
787 			   ((fl->share == IPV6_FL_S_PROCESS) ?
788 			    pid_nr_ns(fl->owner.pid, state->pid_ns) :
789 			    ((fl->share == IPV6_FL_S_USER) ?
790 			     from_kuid_munged(seq_user_ns(seq), fl->owner.uid) :
791 			     0)),
792 			   atomic_read(&fl->users),
793 			   fl->linger/HZ,
794 			   (long)(fl->expires - jiffies)/HZ,
795 			   &fl->dst,
796 			   fl->opt ? fl->opt->opt_nflen : 0);
797 	}
798 	return 0;
799 }
800 
801 static const struct seq_operations ip6fl_seq_ops = {
802 	.start	=	ip6fl_seq_start,
803 	.next	=	ip6fl_seq_next,
804 	.stop	=	ip6fl_seq_stop,
805 	.show	=	ip6fl_seq_show,
806 };
807 
808 static int ip6fl_seq_open(struct inode *inode, struct file *file)
809 {
810 	struct seq_file *seq;
811 	struct ip6fl_iter_state *state;
812 	int err;
813 
814 	err = seq_open_net(inode, file, &ip6fl_seq_ops,
815 			   sizeof(struct ip6fl_iter_state));
816 
817 	if (!err) {
818 		seq = file->private_data;
819 		state = ip6fl_seq_private(seq);
820 		rcu_read_lock();
821 		state->pid_ns = get_pid_ns(task_active_pid_ns(current));
822 		rcu_read_unlock();
823 	}
824 	return err;
825 }
826 
827 static int ip6fl_seq_release(struct inode *inode, struct file *file)
828 {
829 	struct seq_file *seq = file->private_data;
830 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
831 	put_pid_ns(state->pid_ns);
832 	return seq_release_net(inode, file);
833 }
834 
835 static const struct file_operations ip6fl_seq_fops = {
836 	.owner		=	THIS_MODULE,
837 	.open		=	ip6fl_seq_open,
838 	.read		=	seq_read,
839 	.llseek		=	seq_lseek,
840 	.release	=	ip6fl_seq_release,
841 };
842 
843 static int __net_init ip6_flowlabel_proc_init(struct net *net)
844 {
845 	if (!proc_create("ip6_flowlabel", S_IRUGO, net->proc_net,
846 			 &ip6fl_seq_fops))
847 		return -ENOMEM;
848 	return 0;
849 }
850 
851 static void __net_exit ip6_flowlabel_proc_fini(struct net *net)
852 {
853 	remove_proc_entry("ip6_flowlabel", net->proc_net);
854 }
855 #else
856 static inline int ip6_flowlabel_proc_init(struct net *net)
857 {
858 	return 0;
859 }
860 static inline void ip6_flowlabel_proc_fini(struct net *net)
861 {
862 }
863 #endif
864 
865 static void __net_exit ip6_flowlabel_net_exit(struct net *net)
866 {
867 	ip6_fl_purge(net);
868 	ip6_flowlabel_proc_fini(net);
869 }
870 
871 static struct pernet_operations ip6_flowlabel_net_ops = {
872 	.init = ip6_flowlabel_proc_init,
873 	.exit = ip6_flowlabel_net_exit,
874 };
875 
876 int ip6_flowlabel_init(void)
877 {
878 	return register_pernet_subsys(&ip6_flowlabel_net_ops);
879 }
880 
881 void ip6_flowlabel_cleanup(void)
882 {
883 	del_timer(&ip6_fl_gc_timer);
884 	unregister_pernet_subsys(&ip6_flowlabel_net_ops);
885 }
886