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>
23 #include <linux/slab.h>
24 #include <linux/export.h>
25 #include <linux/pid_namespace.h>
27 #include <net/net_namespace.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>
37 #include <net/transp_v6.h>
39 #include <asm/uaccess.h>
41 #define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified
42 in old IPv6 RFC. Well, it was reasonable value.
44 #define FL_MAX_LINGER 60 /* Maximal linger timeout */
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)
53 static atomic_t fl_size
= ATOMIC_INIT(0);
54 static struct ip6_flowlabel
*fl_ht
[FL_HASH_MASK
+1];
56 static void ip6_fl_gc(unsigned long dummy
);
57 static DEFINE_TIMER(ip6_fl_gc_timer
, ip6_fl_gc
, 0, 0);
59 /* FL hash table lock: it protects only of GC */
61 static DEFINE_RWLOCK(ip6_fl_lock
);
65 static DEFINE_RWLOCK(ip6_sk_fl_lock
);
68 static inline struct ip6_flowlabel
*__fl_lookup(struct net
*net
, __be32 label
)
70 struct ip6_flowlabel
*fl
;
72 for (fl
=fl_ht
[FL_HASH(label
)]; fl
; fl
= fl
->next
) {
73 if (fl
->label
== label
&& net_eq(fl
->fl_net
, net
))
79 static struct ip6_flowlabel
*fl_lookup(struct net
*net
, __be32 label
)
81 struct ip6_flowlabel
*fl
;
83 read_lock_bh(&ip6_fl_lock
);
84 fl
= __fl_lookup(net
, label
);
86 atomic_inc(&fl
->users
);
87 read_unlock_bh(&ip6_fl_lock
);
92 static void fl_free(struct ip6_flowlabel
*fl
)
95 if (fl
->share
== IPV6_FL_S_PROCESS
)
96 put_pid(fl
->owner
.pid
);
97 release_net(fl
->fl_net
);
103 static void fl_release(struct ip6_flowlabel
*fl
)
105 write_lock_bh(&ip6_fl_lock
);
107 fl
->lastuse
= jiffies
;
108 if (atomic_dec_and_test(&fl
->users
)) {
109 unsigned long ttd
= fl
->lastuse
+ fl
->linger
;
110 if (time_after(ttd
, fl
->expires
))
113 if (fl
->opt
&& fl
->share
== IPV6_FL_S_EXCL
) {
114 struct ipv6_txoptions
*opt
= fl
->opt
;
118 if (!timer_pending(&ip6_fl_gc_timer
) ||
119 time_after(ip6_fl_gc_timer
.expires
, ttd
))
120 mod_timer(&ip6_fl_gc_timer
, ttd
);
122 write_unlock_bh(&ip6_fl_lock
);
125 static void ip6_fl_gc(unsigned long dummy
)
128 unsigned long now
= jiffies
;
129 unsigned long sched
= 0;
131 write_lock(&ip6_fl_lock
);
133 for (i
=0; i
<=FL_HASH_MASK
; i
++) {
134 struct ip6_flowlabel
*fl
, **flp
;
136 while ((fl
=*flp
) != NULL
) {
137 if (atomic_read(&fl
->users
) == 0) {
138 unsigned long ttd
= fl
->lastuse
+ fl
->linger
;
139 if (time_after(ttd
, fl
->expires
))
142 if (time_after_eq(now
, ttd
)) {
145 atomic_dec(&fl_size
);
148 if (!sched
|| time_before(ttd
, sched
))
154 if (!sched
&& atomic_read(&fl_size
))
155 sched
= now
+ FL_MAX_LINGER
;
157 mod_timer(&ip6_fl_gc_timer
, sched
);
159 write_unlock(&ip6_fl_lock
);
162 static void __net_exit
ip6_fl_purge(struct net
*net
)
166 write_lock(&ip6_fl_lock
);
167 for (i
= 0; i
<= FL_HASH_MASK
; i
++) {
168 struct ip6_flowlabel
*fl
, **flp
;
170 while ((fl
= *flp
) != NULL
) {
171 if (net_eq(fl
->fl_net
, net
) &&
172 atomic_read(&fl
->users
) == 0) {
175 atomic_dec(&fl_size
);
181 write_unlock(&ip6_fl_lock
);
184 static struct ip6_flowlabel
*fl_intern(struct net
*net
,
185 struct ip6_flowlabel
*fl
, __be32 label
)
187 struct ip6_flowlabel
*lfl
;
189 fl
->label
= label
& IPV6_FLOWLABEL_MASK
;
191 write_lock_bh(&ip6_fl_lock
);
194 fl
->label
= htonl(net_random())&IPV6_FLOWLABEL_MASK
;
196 lfl
= __fl_lookup(net
, fl
->label
);
203 * we dropper the ip6_fl_lock, so this entry could reappear
204 * and we need to recheck with it.
206 * OTOH no need to search the active socket first, like it is
207 * done in ipv6_flowlabel_opt - sock is locked, so new entry
208 * with the same label can only appear on another sock
210 lfl
= __fl_lookup(net
, fl
->label
);
212 atomic_inc(&lfl
->users
);
213 write_unlock_bh(&ip6_fl_lock
);
218 fl
->lastuse
= jiffies
;
219 fl
->next
= fl_ht
[FL_HASH(fl
->label
)];
220 fl_ht
[FL_HASH(fl
->label
)] = fl
;
221 atomic_inc(&fl_size
);
222 write_unlock_bh(&ip6_fl_lock
);
228 /* Socket flowlabel lists */
230 struct ip6_flowlabel
* fl6_sock_lookup(struct sock
*sk
, __be32 label
)
232 struct ipv6_fl_socklist
*sfl
;
233 struct ipv6_pinfo
*np
= inet6_sk(sk
);
235 label
&= IPV6_FLOWLABEL_MASK
;
237 read_lock_bh(&ip6_sk_fl_lock
);
238 for (sfl
=np
->ipv6_fl_list
; sfl
; sfl
= sfl
->next
) {
239 struct ip6_flowlabel
*fl
= sfl
->fl
;
240 if (fl
->label
== label
) {
241 fl
->lastuse
= jiffies
;
242 atomic_inc(&fl
->users
);
243 read_unlock_bh(&ip6_sk_fl_lock
);
247 read_unlock_bh(&ip6_sk_fl_lock
);
251 EXPORT_SYMBOL_GPL(fl6_sock_lookup
);
253 void fl6_free_socklist(struct sock
*sk
)
255 struct ipv6_pinfo
*np
= inet6_sk(sk
);
256 struct ipv6_fl_socklist
*sfl
;
258 while ((sfl
= np
->ipv6_fl_list
) != NULL
) {
259 np
->ipv6_fl_list
= sfl
->next
;
265 /* Service routines */
269 It is the only difficult place. flowlabel enforces equal headers
270 before and including routing header, however user may supply options
274 struct ipv6_txoptions
*fl6_merge_options(struct ipv6_txoptions
* opt_space
,
275 struct ip6_flowlabel
* fl
,
276 struct ipv6_txoptions
* fopt
)
278 struct ipv6_txoptions
* fl_opt
= fl
->opt
;
280 if (fopt
== NULL
|| fopt
->opt_flen
== 0)
283 if (fl_opt
!= NULL
) {
284 opt_space
->hopopt
= fl_opt
->hopopt
;
285 opt_space
->dst0opt
= fl_opt
->dst0opt
;
286 opt_space
->srcrt
= fl_opt
->srcrt
;
287 opt_space
->opt_nflen
= fl_opt
->opt_nflen
;
289 if (fopt
->opt_nflen
== 0)
291 opt_space
->hopopt
= NULL
;
292 opt_space
->dst0opt
= NULL
;
293 opt_space
->srcrt
= NULL
;
294 opt_space
->opt_nflen
= 0;
296 opt_space
->dst1opt
= fopt
->dst1opt
;
297 opt_space
->opt_flen
= fopt
->opt_flen
;
300 EXPORT_SYMBOL_GPL(fl6_merge_options
);
302 static unsigned long check_linger(unsigned long ttl
)
304 if (ttl
< FL_MIN_LINGER
)
305 return FL_MIN_LINGER
*HZ
;
306 if (ttl
> FL_MAX_LINGER
&& !capable(CAP_NET_ADMIN
))
311 static int fl6_renew(struct ip6_flowlabel
*fl
, unsigned long linger
, unsigned long expires
)
313 linger
= check_linger(linger
);
316 expires
= check_linger(expires
);
319 fl
->lastuse
= jiffies
;
320 if (time_before(fl
->linger
, linger
))
322 if (time_before(expires
, fl
->linger
))
323 expires
= fl
->linger
;
324 if (time_before(fl
->expires
, fl
->lastuse
+ expires
))
325 fl
->expires
= fl
->lastuse
+ expires
;
329 static struct ip6_flowlabel
*
330 fl_create(struct net
*net
, struct sock
*sk
, struct in6_flowlabel_req
*freq
,
331 char __user
*optval
, int optlen
, int *err_p
)
333 struct ip6_flowlabel
*fl
= NULL
;
338 olen
= optlen
- CMSG_ALIGN(sizeof(*freq
));
340 if (olen
> 64 * 1024)
344 fl
= kzalloc(sizeof(*fl
), GFP_KERNEL
);
350 struct flowi6 flowi6
;
354 fl
->opt
= kmalloc(sizeof(*fl
->opt
) + olen
, GFP_KERNEL
);
358 memset(fl
->opt
, 0, sizeof(*fl
->opt
));
359 fl
->opt
->tot_len
= sizeof(*fl
->opt
) + olen
;
361 if (copy_from_user(fl
->opt
+1, optval
+CMSG_ALIGN(sizeof(*freq
)), olen
))
364 msg
.msg_controllen
= olen
;
365 msg
.msg_control
= (void*)(fl
->opt
+1);
366 memset(&flowi6
, 0, sizeof(flowi6
));
368 err
= datagram_send_ctl(net
, sk
, &msg
, &flowi6
, fl
->opt
, &junk
,
373 if (fl
->opt
->opt_flen
)
375 if (fl
->opt
->opt_nflen
== 0) {
381 fl
->fl_net
= hold_net(net
);
382 fl
->expires
= jiffies
;
383 err
= fl6_renew(fl
, freq
->flr_linger
, freq
->flr_expires
);
386 fl
->share
= freq
->flr_share
;
387 addr_type
= ipv6_addr_type(&freq
->flr_dst
);
388 if ((addr_type
& IPV6_ADDR_MAPPED
) ||
389 addr_type
== IPV6_ADDR_ANY
) {
393 fl
->dst
= freq
->flr_dst
;
394 atomic_set(&fl
->users
, 1);
399 case IPV6_FL_S_PROCESS
:
400 fl
->owner
.pid
= get_task_pid(current
, PIDTYPE_PID
);
403 fl
->owner
.uid
= current_euid();
417 static int mem_check(struct sock
*sk
)
419 struct ipv6_pinfo
*np
= inet6_sk(sk
);
420 struct ipv6_fl_socklist
*sfl
;
421 int room
= FL_MAX_SIZE
- atomic_read(&fl_size
);
424 if (room
> FL_MAX_SIZE
- FL_MAX_PER_SOCK
)
427 for (sfl
= np
->ipv6_fl_list
; sfl
; sfl
= sfl
->next
)
431 ((count
>= FL_MAX_PER_SOCK
||
432 (count
> 0 && room
< FL_MAX_SIZE
/2) || room
< FL_MAX_SIZE
/4) &&
433 !capable(CAP_NET_ADMIN
)))
439 static bool ipv6_hdr_cmp(struct ipv6_opt_hdr
*h1
, struct ipv6_opt_hdr
*h2
)
443 if (h1
== NULL
|| h2
== NULL
)
445 if (h1
->hdrlen
!= h2
->hdrlen
)
447 return memcmp(h1
+1, h2
+1, ((h1
->hdrlen
+1)<<3) - sizeof(*h1
));
450 static bool ipv6_opt_cmp(struct ipv6_txoptions
*o1
, struct ipv6_txoptions
*o2
)
454 if (o1
== NULL
|| o2
== NULL
)
456 if (o1
->opt_nflen
!= o2
->opt_nflen
)
458 if (ipv6_hdr_cmp(o1
->hopopt
, o2
->hopopt
))
460 if (ipv6_hdr_cmp(o1
->dst0opt
, o2
->dst0opt
))
462 if (ipv6_hdr_cmp((struct ipv6_opt_hdr
*)o1
->srcrt
, (struct ipv6_opt_hdr
*)o2
->srcrt
))
467 static inline void fl_link(struct ipv6_pinfo
*np
, struct ipv6_fl_socklist
*sfl
,
468 struct ip6_flowlabel
*fl
)
470 write_lock_bh(&ip6_sk_fl_lock
);
472 sfl
->next
= np
->ipv6_fl_list
;
473 np
->ipv6_fl_list
= sfl
;
474 write_unlock_bh(&ip6_sk_fl_lock
);
477 int ipv6_flowlabel_opt(struct sock
*sk
, char __user
*optval
, int optlen
)
479 int uninitialized_var(err
);
480 struct net
*net
= sock_net(sk
);
481 struct ipv6_pinfo
*np
= inet6_sk(sk
);
482 struct in6_flowlabel_req freq
;
483 struct ipv6_fl_socklist
*sfl1
=NULL
;
484 struct ipv6_fl_socklist
*sfl
, **sflp
;
485 struct ip6_flowlabel
*fl
, *fl1
= NULL
;
488 if (optlen
< sizeof(freq
))
491 if (copy_from_user(&freq
, optval
, sizeof(freq
)))
494 switch (freq
.flr_action
) {
496 write_lock_bh(&ip6_sk_fl_lock
);
497 for (sflp
= &np
->ipv6_fl_list
; (sfl
=*sflp
)!=NULL
; sflp
= &sfl
->next
) {
498 if (sfl
->fl
->label
== freq
.flr_label
) {
499 if (freq
.flr_label
== (np
->flow_label
&IPV6_FLOWLABEL_MASK
))
500 np
->flow_label
&= ~IPV6_FLOWLABEL_MASK
;
502 write_unlock_bh(&ip6_sk_fl_lock
);
508 write_unlock_bh(&ip6_sk_fl_lock
);
511 case IPV6_FL_A_RENEW
:
512 read_lock_bh(&ip6_sk_fl_lock
);
513 for (sfl
= np
->ipv6_fl_list
; sfl
; sfl
= sfl
->next
) {
514 if (sfl
->fl
->label
== freq
.flr_label
) {
515 err
= fl6_renew(sfl
->fl
, freq
.flr_linger
, freq
.flr_expires
);
516 read_unlock_bh(&ip6_sk_fl_lock
);
520 read_unlock_bh(&ip6_sk_fl_lock
);
522 if (freq
.flr_share
== IPV6_FL_S_NONE
&&
523 ns_capable(net
->user_ns
, CAP_NET_ADMIN
)) {
524 fl
= fl_lookup(net
, freq
.flr_label
);
526 err
= fl6_renew(fl
, freq
.flr_linger
, freq
.flr_expires
);
534 if (freq
.flr_label
& ~IPV6_FLOWLABEL_MASK
)
537 fl
= fl_create(net
, sk
, &freq
, optval
, optlen
, &err
);
540 sfl1
= kmalloc(sizeof(*sfl1
), GFP_KERNEL
);
542 if (freq
.flr_label
) {
544 read_lock_bh(&ip6_sk_fl_lock
);
545 for (sfl
= np
->ipv6_fl_list
; sfl
; sfl
= sfl
->next
) {
546 if (sfl
->fl
->label
== freq
.flr_label
) {
547 if (freq
.flr_flags
&IPV6_FL_F_EXCL
) {
548 read_unlock_bh(&ip6_sk_fl_lock
);
552 atomic_inc(&fl1
->users
);
556 read_unlock_bh(&ip6_sk_fl_lock
);
559 fl1
= fl_lookup(net
, freq
.flr_label
);
563 if (freq
.flr_flags
&IPV6_FL_F_EXCL
)
566 if (fl1
->share
== IPV6_FL_S_EXCL
||
567 fl1
->share
!= fl
->share
||
568 ((fl1
->share
== IPV6_FL_S_PROCESS
) &&
569 (fl1
->owner
.pid
== fl
->owner
.pid
)) ||
570 ((fl1
->share
== IPV6_FL_S_USER
) &&
571 uid_eq(fl1
->owner
.uid
, fl
->owner
.uid
)))
575 if (!ipv6_addr_equal(&fl1
->dst
, &fl
->dst
) ||
576 ipv6_opt_cmp(fl1
->opt
, fl
->opt
))
582 if (fl
->linger
> fl1
->linger
)
583 fl1
->linger
= fl
->linger
;
584 if ((long)(fl
->expires
- fl1
->expires
) > 0)
585 fl1
->expires
= fl
->expires
;
586 fl_link(np
, sfl1
, fl1
);
596 if (!(freq
.flr_flags
&IPV6_FL_F_CREATE
))
600 if (sfl1
== NULL
|| (err
= mem_check(sk
)) != 0)
603 fl1
= fl_intern(net
, fl
, freq
.flr_label
);
607 if (!freq
.flr_label
) {
608 if (copy_to_user(&((struct in6_flowlabel_req __user
*) optval
)->flr_label
,
609 &fl
->label
, sizeof(fl
->label
))) {
610 /* Intentionally ignore fault. */
614 fl_link(np
, sfl1
, fl
);
627 #ifdef CONFIG_PROC_FS
629 struct ip6fl_iter_state
{
630 struct seq_net_private p
;
631 struct pid_namespace
*pid_ns
;
635 #define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private)
637 static struct ip6_flowlabel
*ip6fl_get_first(struct seq_file
*seq
)
639 struct ip6_flowlabel
*fl
= NULL
;
640 struct ip6fl_iter_state
*state
= ip6fl_seq_private(seq
);
641 struct net
*net
= seq_file_net(seq
);
643 for (state
->bucket
= 0; state
->bucket
<= FL_HASH_MASK
; ++state
->bucket
) {
644 fl
= fl_ht
[state
->bucket
];
646 while (fl
&& !net_eq(fl
->fl_net
, net
))
654 static struct ip6_flowlabel
*ip6fl_get_next(struct seq_file
*seq
, struct ip6_flowlabel
*fl
)
656 struct ip6fl_iter_state
*state
= ip6fl_seq_private(seq
);
657 struct net
*net
= seq_file_net(seq
);
661 while (fl
&& !net_eq(fl
->fl_net
, net
))
665 if (++state
->bucket
<= FL_HASH_MASK
) {
666 fl
= fl_ht
[state
->bucket
];
674 static struct ip6_flowlabel
*ip6fl_get_idx(struct seq_file
*seq
, loff_t pos
)
676 struct ip6_flowlabel
*fl
= ip6fl_get_first(seq
);
678 while (pos
&& (fl
= ip6fl_get_next(seq
, fl
)) != NULL
)
680 return pos
? NULL
: fl
;
683 static void *ip6fl_seq_start(struct seq_file
*seq
, loff_t
*pos
)
684 __acquires(ip6_fl_lock
)
686 read_lock_bh(&ip6_fl_lock
);
687 return *pos
? ip6fl_get_idx(seq
, *pos
- 1) : SEQ_START_TOKEN
;
690 static void *ip6fl_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
692 struct ip6_flowlabel
*fl
;
694 if (v
== SEQ_START_TOKEN
)
695 fl
= ip6fl_get_first(seq
);
697 fl
= ip6fl_get_next(seq
, v
);
702 static void ip6fl_seq_stop(struct seq_file
*seq
, void *v
)
703 __releases(ip6_fl_lock
)
705 read_unlock_bh(&ip6_fl_lock
);
708 static int ip6fl_seq_show(struct seq_file
*seq
, void *v
)
710 struct ip6fl_iter_state
*state
= ip6fl_seq_private(seq
);
711 if (v
== SEQ_START_TOKEN
)
712 seq_printf(seq
, "%-5s %-1s %-6s %-6s %-6s %-8s %-32s %s\n",
713 "Label", "S", "Owner", "Users", "Linger", "Expires", "Dst", "Opt");
715 struct ip6_flowlabel
*fl
= v
;
717 "%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n",
718 (unsigned int)ntohl(fl
->label
),
720 ((fl
->share
== IPV6_FL_S_PROCESS
) ?
721 pid_nr_ns(fl
->owner
.pid
, state
->pid_ns
) :
722 ((fl
->share
== IPV6_FL_S_USER
) ?
723 from_kuid_munged(seq_user_ns(seq
), fl
->owner
.uid
) :
725 atomic_read(&fl
->users
),
727 (long)(fl
->expires
- jiffies
)/HZ
,
729 fl
->opt
? fl
->opt
->opt_nflen
: 0);
734 static const struct seq_operations ip6fl_seq_ops
= {
735 .start
= ip6fl_seq_start
,
736 .next
= ip6fl_seq_next
,
737 .stop
= ip6fl_seq_stop
,
738 .show
= ip6fl_seq_show
,
741 static int ip6fl_seq_open(struct inode
*inode
, struct file
*file
)
743 struct seq_file
*seq
;
744 struct ip6fl_iter_state
*state
;
747 err
= seq_open_net(inode
, file
, &ip6fl_seq_ops
,
748 sizeof(struct ip6fl_iter_state
));
751 seq
= file
->private_data
;
752 state
= ip6fl_seq_private(seq
);
754 state
->pid_ns
= get_pid_ns(task_active_pid_ns(current
));
760 static int ip6fl_seq_release(struct inode
*inode
, struct file
*file
)
762 struct seq_file
*seq
= file
->private_data
;
763 struct ip6fl_iter_state
*state
= ip6fl_seq_private(seq
);
764 put_pid_ns(state
->pid_ns
);
765 return seq_release_net(inode
, file
);
768 static const struct file_operations ip6fl_seq_fops
= {
769 .owner
= THIS_MODULE
,
770 .open
= ip6fl_seq_open
,
773 .release
= ip6fl_seq_release
,
776 static int __net_init
ip6_flowlabel_proc_init(struct net
*net
)
778 if (!proc_net_fops_create(net
, "ip6_flowlabel",
779 S_IRUGO
, &ip6fl_seq_fops
))
784 static void __net_exit
ip6_flowlabel_proc_fini(struct net
*net
)
786 proc_net_remove(net
, "ip6_flowlabel");
789 static inline int ip6_flowlabel_proc_init(struct net
*net
)
793 static inline void ip6_flowlabel_proc_fini(struct net
*net
)
798 static void __net_exit
ip6_flowlabel_net_exit(struct net
*net
)
801 ip6_flowlabel_proc_fini(net
);
804 static struct pernet_operations ip6_flowlabel_net_ops
= {
805 .init
= ip6_flowlabel_proc_init
,
806 .exit
= ip6_flowlabel_net_exit
,
809 int ip6_flowlabel_init(void)
811 return register_pernet_subsys(&ip6_flowlabel_net_ops
);
814 void ip6_flowlabel_cleanup(void)
816 del_timer(&ip6_fl_gc_timer
);
817 unregister_pernet_subsys(&ip6_flowlabel_net_ops
);