blob: 1f23a57aa9e68bcafd83a00426deab83ffd49559 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * The IP fragmentation functionality.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09007 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Authors: Fred N. van Kempen <waltje@uWalt.NL.Mugnet.ORG>
Alan Cox113aa832008-10-13 19:01:08 -07009 * Alan Cox <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070010 *
11 * Fixes:
12 * Alan Cox : Split from ip.c , see ip_input.c for history.
13 * David S. Miller : Begin massive cleanup...
14 * Andi Kleen : Add sysctls.
15 * xxxx : Overlapfrag bug.
16 * Ultima : ip_expire() kernel panic.
17 * Bill Hawes : Frag accounting and evictor fixes.
18 * John McDonald : 0 length frag bug.
19 * Alexey Kuznetsov: SMP races, threading, cleanup.
20 * Patrick McHardy : LRU queue of frag heads for evictor.
21 */
22
Herbert Xu89cee8b2005-12-13 23:14:27 -080023#include <linux/compiler.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/module.h>
25#include <linux/types.h>
26#include <linux/mm.h>
27#include <linux/jiffies.h>
28#include <linux/skbuff.h>
29#include <linux/list.h>
30#include <linux/ip.h>
31#include <linux/icmp.h>
32#include <linux/netdevice.h>
33#include <linux/jhash.h>
34#include <linux/random.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090035#include <linux/slab.h>
Shan Weie9017b52010-01-23 01:57:42 -080036#include <net/route.h>
37#include <net/dst.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#include <net/sock.h>
39#include <net/ip.h>
40#include <net/icmp.h>
41#include <net/checksum.h>
Herbert Xu89cee8b2005-12-13 23:14:27 -080042#include <net/inetpeer.h>
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -070043#include <net/inet_frag.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/tcp.h>
45#include <linux/udp.h>
46#include <linux/inet.h>
47#include <linux/netfilter_ipv4.h>
Eric Dumazet6623e3b2011-01-05 07:52:55 +000048#include <net/inet_ecn.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049
50/* NOTE. Logic of IP defragmentation is parallel to corresponding IPv6
51 * code now. If you change something here, _PLEASE_ update ipv6/reassembly.c
52 * as well. Or notify me, at least. --ANK
53 */
54
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -080055static int sysctl_ipfrag_max_dist __read_mostly = 64;
Herbert Xu89cee8b2005-12-13 23:14:27 -080056
Linus Torvalds1da177e2005-04-16 15:20:36 -070057struct ipfrag_skb_cb
58{
59 struct inet_skb_parm h;
60 int offset;
61};
62
Jianjun Kongfd3f8c42008-11-03 02:47:38 -080063#define FRAG_CB(skb) ((struct ipfrag_skb_cb *)((skb)->cb))
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
65/* Describe an entry in the "incomplete datagrams" queue. */
66struct ipq {
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -070067 struct inet_frag_queue q;
68
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 u32 user;
Al Viro18277772006-09-26 22:19:02 -070070 __be32 saddr;
71 __be32 daddr;
72 __be16 id;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 u8 protocol;
Eric Dumazet6623e3b2011-01-05 07:52:55 +000074 u8 ecn; /* RFC3168 support */
Herbert Xu89cee8b2005-12-13 23:14:27 -080075 int iif;
76 unsigned int rid;
77 struct inet_peer *peer;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078};
79
Eric Dumazet5173cc02011-05-16 08:37:37 +000080/* RFC 3168 support :
81 * We want to check ECN values of all fragments, do detect invalid combinations.
82 * In ipq->ecn, we store the OR value of each ip4_frag_ecn() fragment value.
83 */
David S. Miller1d1652c2011-05-17 17:28:02 -040084#define IPFRAG_ECN_NOT_ECT 0x01 /* one frag had ECN_NOT_ECT */
85#define IPFRAG_ECN_ECT_1 0x02 /* one frag had ECN_ECT_1 */
86#define IPFRAG_ECN_ECT_0 0x04 /* one frag had ECN_ECT_0 */
87#define IPFRAG_ECN_CE 0x08 /* one frag had ECN_CE */
Eric Dumazet6623e3b2011-01-05 07:52:55 +000088
89static inline u8 ip4_frag_ecn(u8 tos)
90{
Eric Dumazet5173cc02011-05-16 08:37:37 +000091 return 1 << (tos & INET_ECN_MASK);
Eric Dumazet6623e3b2011-01-05 07:52:55 +000092}
93
Eric Dumazet5173cc02011-05-16 08:37:37 +000094/* Given the OR values of all fragments, apply RFC 3168 5.3 requirements
95 * Value : 0xff if frame should be dropped.
96 * 0 or INET_ECN_CE value, to be ORed in to final iph->tos field
97 */
98static const u8 ip4_frag_ecn_table[16] = {
99 /* at least one fragment had CE, and others ECT_0 or ECT_1 */
100 [IPFRAG_ECN_CE | IPFRAG_ECN_ECT_0] = INET_ECN_CE,
101 [IPFRAG_ECN_CE | IPFRAG_ECN_ECT_1] = INET_ECN_CE,
102 [IPFRAG_ECN_CE | IPFRAG_ECN_ECT_0 | IPFRAG_ECN_ECT_1] = INET_ECN_CE,
103
104 /* invalid combinations : drop frame */
105 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_CE] = 0xff,
106 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_ECT_0] = 0xff,
107 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_ECT_1] = 0xff,
108 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_ECT_0 | IPFRAG_ECN_ECT_1] = 0xff,
109 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_CE | IPFRAG_ECN_ECT_0] = 0xff,
110 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_CE | IPFRAG_ECN_ECT_1] = 0xff,
111 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_CE | IPFRAG_ECN_ECT_0 | IPFRAG_ECN_ECT_1] = 0xff,
112};
113
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700114static struct inet_frags ip4_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
Pavel Emelyanove5a2bb82008-01-22 06:06:23 -0800116int ip_frag_nqueues(struct net *net)
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700117{
Pavel Emelyanove5a2bb82008-01-22 06:06:23 -0800118 return net->ipv4.frags.nqueues;
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700119}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Pavel Emelyanov6ddc0822008-01-22 06:07:25 -0800121int ip_frag_mem(struct net *net)
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700122{
Pavel Emelyanov6ddc0822008-01-22 06:07:25 -0800123 return atomic_read(&net->ipv4.frags.mem);
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700124}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
Herbert Xu1706d582007-10-14 00:38:15 -0700126static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
127 struct net_device *dev);
128
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700129struct ip4_create_arg {
130 struct iphdr *iph;
131 u32 user;
132};
133
Al Viro18277772006-09-26 22:19:02 -0700134static unsigned int ipqhashfn(__be16 id, __be32 saddr, __be32 daddr, u8 prot)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135{
Al Viro18277772006-09-26 22:19:02 -0700136 return jhash_3words((__force u32)id << 16 | prot,
137 (__force u32)saddr, (__force u32)daddr,
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700138 ip4_frags.rnd) & (INETFRAGS_HASHSZ - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139}
140
Pavel Emelyanov321a3a92007-10-15 02:38:08 -0700141static unsigned int ip4_hashfn(struct inet_frag_queue *q)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142{
Pavel Emelyanov321a3a92007-10-15 02:38:08 -0700143 struct ipq *ipq;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144
Pavel Emelyanov321a3a92007-10-15 02:38:08 -0700145 ipq = container_of(q, struct ipq, q);
146 return ipqhashfn(ipq->id, ipq->saddr, ipq->daddr, ipq->protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147}
148
Pavel Emelyanovabd65232007-10-17 19:47:21 -0700149static int ip4_frag_match(struct inet_frag_queue *q, void *a)
150{
151 struct ipq *qp;
152 struct ip4_create_arg *arg = a;
153
154 qp = container_of(q, struct ipq, q);
Eric Dumazeta02cec22010-09-22 20:43:57 +0000155 return qp->id == arg->iph->id &&
Pavel Emelyanovabd65232007-10-17 19:47:21 -0700156 qp->saddr == arg->iph->saddr &&
157 qp->daddr == arg->iph->daddr &&
158 qp->protocol == arg->iph->protocol &&
Eric Dumazeta02cec22010-09-22 20:43:57 +0000159 qp->user == arg->user;
Pavel Emelyanovabd65232007-10-17 19:47:21 -0700160}
161
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162/* Memory Tracking Functions. */
Eric Dumazeta95d8c82010-06-13 23:22:43 +0000163static void frag_kfree_skb(struct netns_frags *nf, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164{
Pavel Emelyanov6ddc0822008-01-22 06:07:25 -0800165 atomic_sub(skb->truesize, &nf->mem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 kfree_skb(skb);
167}
168
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700169static void ip4_frag_init(struct inet_frag_queue *q, void *a)
170{
171 struct ipq *qp = container_of(q, struct ipq, q);
172 struct ip4_create_arg *arg = a;
173
174 qp->protocol = arg->iph->protocol;
175 qp->id = arg->iph->id;
Eric Dumazet6623e3b2011-01-05 07:52:55 +0000176 qp->ecn = ip4_frag_ecn(arg->iph->tos);
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700177 qp->saddr = arg->iph->saddr;
178 qp->daddr = arg->iph->daddr;
179 qp->user = arg->user;
180 qp->peer = sysctl_ipfrag_max_dist ?
David S. Millerb534ecf2010-11-30 11:54:19 -0800181 inet_getpeer_v4(arg->iph->saddr, 1) : NULL;
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700182}
183
Pavel Emelyanov1e4b8282007-10-15 02:39:14 -0700184static __inline__ void ip4_frag_free(struct inet_frag_queue *q)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185{
Pavel Emelyanov1e4b8282007-10-15 02:39:14 -0700186 struct ipq *qp;
187
188 qp = container_of(q, struct ipq, q);
189 if (qp->peer)
190 inet_putpeer(qp->peer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191}
192
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193
194/* Destruction primitives. */
195
Pavel Emelyanov4b6cb5d2007-10-15 02:41:09 -0700196static __inline__ void ipq_put(struct ipq *ipq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197{
Pavel Emelyanov762cc402007-10-15 02:41:56 -0700198 inet_frag_put(&ipq->q, &ip4_frags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199}
200
201/* Kill ipq entry. It is not destroyed immediately,
202 * because caller (and someone more) holds reference count.
203 */
204static void ipq_kill(struct ipq *ipq)
205{
Pavel Emelyanov277e6502007-10-15 02:37:18 -0700206 inet_frag_kill(&ipq->q, &ip4_frags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207}
208
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900209/* Memory limiting on fragments. Evictor trashes the oldest
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 * fragment queue until we are back under the threshold.
211 */
Pavel Emelyanov6ddc0822008-01-22 06:07:25 -0800212static void ip_evictor(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213{
Pavel Emelyanov8e7999c2007-10-15 02:40:06 -0700214 int evicted;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
Pavel Emelyanov6ddc0822008-01-22 06:07:25 -0800216 evicted = inet_frag_evictor(&net->ipv4.frags, &ip4_frags);
Pavel Emelyanov8e7999c2007-10-15 02:40:06 -0700217 if (evicted)
Pavel Emelyanovc5346fe2008-07-16 20:20:33 -0700218 IP_ADD_STATS_BH(net, IPSTATS_MIB_REASMFAILS, evicted);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219}
220
221/*
222 * Oops, a fragment queue timed out. Kill it and send an ICMP reply.
223 */
224static void ip_expire(unsigned long arg)
225{
Pavel Emelyanove521db92007-10-17 19:45:23 -0700226 struct ipq *qp;
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700227 struct net *net;
Pavel Emelyanove521db92007-10-17 19:45:23 -0700228
229 qp = container_of((struct inet_frag_queue *) arg, struct ipq, q);
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700230 net = container_of(qp->q.net, struct net, ipv4.frags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700232 spin_lock(&qp->q.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
Joe Perchesbc578a52008-03-28 16:35:27 -0700234 if (qp->q.last_in & INET_FRAG_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 goto out;
236
237 ipq_kill(qp);
238
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700239 IP_INC_STATS_BH(net, IPSTATS_MIB_REASMTIMEOUT);
240 IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
Joe Perchesbc578a52008-03-28 16:35:27 -0700242 if ((qp->q.last_in & INET_FRAG_FIRST_IN) && qp->q.fragments != NULL) {
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700243 struct sk_buff *head = qp->q.fragments;
Eric Dumazet64f3b9e2011-05-04 10:02:26 +0000244 const struct iphdr *iph;
245 int err;
Denis V. Lunevcb846632008-03-24 15:31:00 -0700246
Eric Dumazet69df9d52009-11-05 20:59:47 -0800247 rcu_read_lock();
248 head->dev = dev_get_by_index_rcu(net, qp->iif);
Shan Weie9017b52010-01-23 01:57:42 -0800249 if (!head->dev)
250 goto out_rcu_unlock;
251
Eric Dumazet64f3b9e2011-05-04 10:02:26 +0000252 /* skb dst is stale, drop it, and perform route lookup again */
253 skb_dst_drop(head);
254 iph = ip_hdr(head);
255 err = ip_route_input_noref(head, iph->daddr, iph->saddr,
256 iph->tos, head->dev);
257 if (err)
258 goto out_rcu_unlock;
259
Shan Weie9017b52010-01-23 01:57:42 -0800260 /*
Eric Dumazet64f3b9e2011-05-04 10:02:26 +0000261 * Only an end host needs to send an ICMP
262 * "Fragment Reassembly Timeout" message, per RFC792.
Shan Weie9017b52010-01-23 01:57:42 -0800263 */
David S. Miller595fc712011-07-05 01:05:48 -0700264 if (qp->user == IP_DEFRAG_AF_PACKET ||
265 (qp->user == IP_DEFRAG_CONNTRACK_IN &&
266 skb_rtable(head)->rt_type != RTN_LOCAL))
Eric Dumazet64f3b9e2011-05-04 10:02:26 +0000267 goto out_rcu_unlock;
Shan Weie9017b52010-01-23 01:57:42 -0800268
Shan Weie9017b52010-01-23 01:57:42 -0800269
270 /* Send an ICMP "Fragment Reassembly Timeout" message. */
271 icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0);
Shan Weie9017b52010-01-23 01:57:42 -0800272out_rcu_unlock:
Patrick McHardyd1c9ae62010-02-02 11:46:50 -0800273 rcu_read_unlock();
274 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275out:
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700276 spin_unlock(&qp->q.lock);
Pavel Emelyanov4b6cb5d2007-10-15 02:41:09 -0700277 ipq_put(qp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278}
279
Pavel Emelyanovabd65232007-10-17 19:47:21 -0700280/* Find the correct entry in the "incomplete datagrams" queue for
281 * this IP datagram, and create new one, if nothing is found.
282 */
Pavel Emelyanovac18e752008-01-22 06:02:14 -0800283static inline struct ipq *ip_find(struct net *net, struct iphdr *iph, u32 user)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284{
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700285 struct inet_frag_queue *q;
286 struct ip4_create_arg arg;
Pavel Emelyanovabd65232007-10-17 19:47:21 -0700287 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700289 arg.iph = iph;
290 arg.user = user;
Pavel Emelyanov9a375802008-06-27 20:06:08 -0700291
292 read_lock(&ip4_frags.lock);
Pavel Emelyanovabd65232007-10-17 19:47:21 -0700293 hash = ipqhashfn(iph->id, iph->saddr, iph->daddr, iph->protocol);
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700294
Pavel Emelyanovac18e752008-01-22 06:02:14 -0800295 q = inet_frag_find(&net->ipv4.frags, &ip4_frags, &arg, hash);
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700296 if (q == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 goto out_nomem;
298
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700299 return container_of(q, struct ipq, q);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300
301out_nomem:
Patrick McHardy64ce2072005-08-09 20:50:53 -0700302 LIMIT_NETDEBUG(KERN_ERR "ip_frag_create: no memory left !\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 return NULL;
304}
305
Herbert Xu89cee8b2005-12-13 23:14:27 -0800306/* Is the fragment too far ahead to be part of ipq? */
307static inline int ip_frag_too_far(struct ipq *qp)
308{
309 struct inet_peer *peer = qp->peer;
310 unsigned int max = sysctl_ipfrag_max_dist;
311 unsigned int start, end;
312
313 int rc;
314
315 if (!peer || !max)
316 return 0;
317
318 start = qp->rid;
319 end = atomic_inc_return(&peer->rid);
320 qp->rid = end;
321
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700322 rc = qp->q.fragments && (end - start) > max;
Herbert Xu89cee8b2005-12-13 23:14:27 -0800323
324 if (rc) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700325 struct net *net;
326
327 net = container_of(qp->q.net, struct net, ipv4.frags);
328 IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS);
Herbert Xu89cee8b2005-12-13 23:14:27 -0800329 }
330
331 return rc;
332}
333
334static int ip_frag_reinit(struct ipq *qp)
335{
336 struct sk_buff *fp;
337
Pavel Emelyanovb2fd5322008-01-22 06:09:37 -0800338 if (!mod_timer(&qp->q.timer, jiffies + qp->q.net->timeout)) {
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700339 atomic_inc(&qp->q.refcnt);
Herbert Xu89cee8b2005-12-13 23:14:27 -0800340 return -ETIMEDOUT;
341 }
342
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700343 fp = qp->q.fragments;
Herbert Xu89cee8b2005-12-13 23:14:27 -0800344 do {
345 struct sk_buff *xp = fp->next;
Eric Dumazeta95d8c82010-06-13 23:22:43 +0000346 frag_kfree_skb(qp->q.net, fp);
Herbert Xu89cee8b2005-12-13 23:14:27 -0800347 fp = xp;
348 } while (fp);
349
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700350 qp->q.last_in = 0;
351 qp->q.len = 0;
352 qp->q.meat = 0;
353 qp->q.fragments = NULL;
Changli Gaod6bebca2010-06-29 04:39:37 +0000354 qp->q.fragments_tail = NULL;
Herbert Xu89cee8b2005-12-13 23:14:27 -0800355 qp->iif = 0;
Eric Dumazet6623e3b2011-01-05 07:52:55 +0000356 qp->ecn = 0;
Herbert Xu89cee8b2005-12-13 23:14:27 -0800357
358 return 0;
359}
360
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361/* Add new segment to existing queue. */
Herbert Xu1706d582007-10-14 00:38:15 -0700362static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363{
364 struct sk_buff *prev, *next;
Herbert Xu1706d582007-10-14 00:38:15 -0700365 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 int flags, offset;
367 int ihl, end;
Herbert Xu1706d582007-10-14 00:38:15 -0700368 int err = -ENOENT;
Eric Dumazet6623e3b2011-01-05 07:52:55 +0000369 u8 ecn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
Joe Perchesbc578a52008-03-28 16:35:27 -0700371 if (qp->q.last_in & INET_FRAG_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 goto err;
373
Herbert Xu89cee8b2005-12-13 23:14:27 -0800374 if (!(IPCB(skb)->flags & IPSKB_FRAG_COMPLETE) &&
Herbert Xu1706d582007-10-14 00:38:15 -0700375 unlikely(ip_frag_too_far(qp)) &&
376 unlikely(err = ip_frag_reinit(qp))) {
Herbert Xu89cee8b2005-12-13 23:14:27 -0800377 ipq_kill(qp);
378 goto err;
379 }
380
Eric Dumazet6623e3b2011-01-05 07:52:55 +0000381 ecn = ip4_frag_ecn(ip_hdr(skb)->tos);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700382 offset = ntohs(ip_hdr(skb)->frag_off);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 flags = offset & ~IP_OFFSET;
384 offset &= IP_OFFSET;
385 offset <<= 3; /* offset is in 8-byte chunks */
Arnaldo Carvalho de Meloc9bdd4b2007-03-12 20:09:15 -0300386 ihl = ip_hdrlen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387
388 /* Determine the position of this fragment. */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900389 end = offset + skb->len - ihl;
Herbert Xu1706d582007-10-14 00:38:15 -0700390 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391
392 /* Is this the final fragment? */
393 if ((flags & IP_MF) == 0) {
394 /* If we already have some bits beyond end
Justin P. Mattock42b2aa82011-11-28 20:31:00 -0800395 * or have different end, the segment is corrupted.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 */
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700397 if (end < qp->q.len ||
Joe Perchesbc578a52008-03-28 16:35:27 -0700398 ((qp->q.last_in & INET_FRAG_LAST_IN) && end != qp->q.len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 goto err;
Joe Perchesbc578a52008-03-28 16:35:27 -0700400 qp->q.last_in |= INET_FRAG_LAST_IN;
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700401 qp->q.len = end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 } else {
403 if (end&7) {
404 end &= ~7;
405 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
406 skb->ip_summed = CHECKSUM_NONE;
407 }
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700408 if (end > qp->q.len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 /* Some bits beyond end -> corruption. */
Joe Perchesbc578a52008-03-28 16:35:27 -0700410 if (qp->q.last_in & INET_FRAG_LAST_IN)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 goto err;
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700412 qp->q.len = end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 }
414 }
415 if (end == offset)
416 goto err;
417
Herbert Xu1706d582007-10-14 00:38:15 -0700418 err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 if (pskb_pull(skb, ihl) == NULL)
420 goto err;
Herbert Xu1706d582007-10-14 00:38:15 -0700421
422 err = pskb_trim_rcsum(skb, end - offset);
423 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 goto err;
425
426 /* Find out which fragments are in front and at the back of us
427 * in the chain of fragments so far. We must know where to put
428 * this fragment, right?
429 */
Changli Gaod6bebca2010-06-29 04:39:37 +0000430 prev = qp->q.fragments_tail;
431 if (!prev || FRAG_CB(prev)->offset < offset) {
432 next = NULL;
433 goto found;
434 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 prev = NULL;
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700436 for (next = qp->q.fragments; next != NULL; next = next->next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 if (FRAG_CB(next)->offset >= offset)
438 break; /* bingo! */
439 prev = next;
440 }
441
Changli Gaod6bebca2010-06-29 04:39:37 +0000442found:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 /* We found where to put this one. Check for overlap with
444 * preceding fragment, and, if needed, align things so that
445 * any overlaps are eliminated.
446 */
447 if (prev) {
448 int i = (FRAG_CB(prev)->offset + prev->len) - offset;
449
450 if (i > 0) {
451 offset += i;
Herbert Xu1706d582007-10-14 00:38:15 -0700452 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 if (end <= offset)
454 goto err;
Herbert Xu1706d582007-10-14 00:38:15 -0700455 err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 if (!pskb_pull(skb, i))
457 goto err;
458 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
459 skb->ip_summed = CHECKSUM_NONE;
460 }
461 }
462
Herbert Xu1706d582007-10-14 00:38:15 -0700463 err = -ENOMEM;
464
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 while (next && FRAG_CB(next)->offset < end) {
466 int i = end - FRAG_CB(next)->offset; /* overlap is 'i' bytes */
467
468 if (i < next->len) {
469 /* Eat head of the next overlapped fragment
470 * and leave the loop. The next ones cannot overlap.
471 */
472 if (!pskb_pull(next, i))
473 goto err;
474 FRAG_CB(next)->offset += i;
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700475 qp->q.meat -= i;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 if (next->ip_summed != CHECKSUM_UNNECESSARY)
477 next->ip_summed = CHECKSUM_NONE;
478 break;
479 } else {
480 struct sk_buff *free_it = next;
481
Peter Zijlstra47c6bf772006-12-12 19:48:59 +0100482 /* Old fragment is completely overridden with
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 * new one drop it.
484 */
485 next = next->next;
486
487 if (prev)
488 prev->next = next;
489 else
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700490 qp->q.fragments = next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700492 qp->q.meat -= free_it->len;
Eric Dumazeta95d8c82010-06-13 23:22:43 +0000493 frag_kfree_skb(qp->q.net, free_it);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 }
495 }
496
497 FRAG_CB(skb)->offset = offset;
498
499 /* Insert this fragment in the chain of fragments. */
500 skb->next = next;
Changli Gaod6bebca2010-06-29 04:39:37 +0000501 if (!next)
502 qp->q.fragments_tail = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 if (prev)
504 prev->next = skb;
505 else
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700506 qp->q.fragments = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Herbert Xu1706d582007-10-14 00:38:15 -0700508 dev = skb->dev;
509 if (dev) {
510 qp->iif = dev->ifindex;
511 skb->dev = NULL;
512 }
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700513 qp->q.stamp = skb->tstamp;
514 qp->q.meat += skb->len;
Eric Dumazet6623e3b2011-01-05 07:52:55 +0000515 qp->ecn |= ecn;
Pavel Emelyanov6ddc0822008-01-22 06:07:25 -0800516 atomic_add(skb->truesize, &qp->q.net->mem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 if (offset == 0)
Joe Perchesbc578a52008-03-28 16:35:27 -0700518 qp->q.last_in |= INET_FRAG_FIRST_IN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519
Joe Perchesbc578a52008-03-28 16:35:27 -0700520 if (qp->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
521 qp->q.meat == qp->q.len)
Herbert Xu1706d582007-10-14 00:38:15 -0700522 return ip_frag_reasm(qp, prev, dev);
523
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700524 write_lock(&ip4_frags.lock);
Pavel Emelyanov3140c252008-01-22 06:11:48 -0800525 list_move_tail(&qp->q.lru_list, &qp->q.net->lru_list);
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700526 write_unlock(&ip4_frags.lock);
Herbert Xu1706d582007-10-14 00:38:15 -0700527 return -EINPROGRESS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
529err:
530 kfree_skb(skb);
Herbert Xu1706d582007-10-14 00:38:15 -0700531 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532}
533
534
535/* Build a new IP datagram from all its fragments. */
536
Herbert Xu1706d582007-10-14 00:38:15 -0700537static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
538 struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539{
Jorge Boncompte [DTI2]2bad35b2009-03-18 23:26:11 -0700540 struct net *net = container_of(qp->q.net, struct net, ipv4.frags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 struct iphdr *iph;
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700542 struct sk_buff *fp, *head = qp->q.fragments;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 int len;
544 int ihlen;
Herbert Xu1706d582007-10-14 00:38:15 -0700545 int err;
Eric Dumazet5173cc02011-05-16 08:37:37 +0000546 u8 ecn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
548 ipq_kill(qp);
549
Eric Dumazet5173cc02011-05-16 08:37:37 +0000550 ecn = ip4_frag_ecn_table[qp->ecn];
551 if (unlikely(ecn == 0xff)) {
552 err = -EINVAL;
553 goto out_fail;
554 }
Herbert Xu1706d582007-10-14 00:38:15 -0700555 /* Make the one we just received the head. */
556 if (prev) {
557 head = prev->next;
558 fp = skb_clone(head, GFP_ATOMIC);
Herbert Xu1706d582007-10-14 00:38:15 -0700559 if (!fp)
560 goto out_nomem;
561
562 fp->next = head->next;
Changli Gaod6bebca2010-06-29 04:39:37 +0000563 if (!fp->next)
564 qp->q.fragments_tail = fp;
Herbert Xu1706d582007-10-14 00:38:15 -0700565 prev->next = fp;
566
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700567 skb_morph(head, qp->q.fragments);
568 head->next = qp->q.fragments->next;
Herbert Xu1706d582007-10-14 00:38:15 -0700569
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700570 kfree_skb(qp->q.fragments);
571 qp->q.fragments = head;
Herbert Xu1706d582007-10-14 00:38:15 -0700572 }
573
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700574 WARN_ON(head == NULL);
575 WARN_ON(FRAG_CB(head)->offset != 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576
577 /* Allocate a new buffer for the datagram. */
Arnaldo Carvalho de Meloc9bdd4b2007-03-12 20:09:15 -0300578 ihlen = ip_hdrlen(head);
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700579 len = ihlen + qp->q.len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
Herbert Xu1706d582007-10-14 00:38:15 -0700581 err = -E2BIG;
Stephen Hemminger132adf52007-03-08 20:44:43 -0800582 if (len > 65535)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 goto out_oversize;
584
585 /* Head of list must not be cloned. */
586 if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
587 goto out_nomem;
588
589 /* If the first fragment is fragmented itself, we split
590 * it to two chunks: the first with data and paged part
591 * and the second, holding only fragments. */
David S. Miller21dc3302010-08-23 00:13:46 -0700592 if (skb_has_frag_list(head)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 struct sk_buff *clone;
594 int i, plen = 0;
595
596 if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
597 goto out_nomem;
598 clone->next = head->next;
599 head->next = clone;
600 skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
David S. Millerd7fcf1a2009-06-09 00:19:37 -0700601 skb_frag_list_init(head);
Eric Dumazet9e903e02011-10-18 21:00:24 +0000602 for (i = 0; i < skb_shinfo(head)->nr_frags; i++)
603 plen += skb_frag_size(&skb_shinfo(head)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 clone->len = clone->data_len = head->data_len - plen;
605 head->data_len -= clone->len;
606 head->len -= clone->len;
607 clone->csum = 0;
608 clone->ip_summed = head->ip_summed;
Pavel Emelyanov6ddc0822008-01-22 06:07:25 -0800609 atomic_add(clone->truesize, &qp->q.net->mem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 }
611
612 skb_shinfo(head)->frag_list = head->next;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -0700613 skb_push(head, head->data - skb_network_header(head));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614
615 for (fp=head->next; fp; fp = fp->next) {
616 head->data_len += fp->len;
617 head->len += fp->len;
618 if (head->ip_summed != fp->ip_summed)
619 head->ip_summed = CHECKSUM_NONE;
Patrick McHardy84fa7932006-08-29 16:44:56 -0700620 else if (head->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 head->csum = csum_add(head->csum, fp->csum);
622 head->truesize += fp->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 }
Eric Dumazetd27f9b32010-06-13 23:02:24 +0000624 atomic_sub(head->truesize, &qp->q.net->mem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
626 head->next = NULL;
627 head->dev = dev;
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700628 head->tstamp = qp->q.stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700630 iph = ip_hdr(head);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 iph->frag_off = 0;
632 iph->tot_len = htons(len);
Eric Dumazet5173cc02011-05-16 08:37:37 +0000633 iph->tos |= ecn;
Jorge Boncompte [DTI2]2bad35b2009-03-18 23:26:11 -0700634 IP_INC_STATS_BH(net, IPSTATS_MIB_REASMOKS);
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700635 qp->q.fragments = NULL;
Changli Gaod6bebca2010-06-29 04:39:37 +0000636 qp->q.fragments_tail = NULL;
Herbert Xu1706d582007-10-14 00:38:15 -0700637 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638
639out_nomem:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900640 LIMIT_NETDEBUG(KERN_ERR "IP: queue_glue: no memory for gluing "
Patrick McHardy64ce2072005-08-09 20:50:53 -0700641 "queue %p\n", qp);
David Howells45542472007-10-17 21:37:22 -0700642 err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 goto out_fail;
644out_oversize:
645 if (net_ratelimit())
Harvey Harrison673d57e2008-10-31 00:53:57 -0700646 printk(KERN_INFO "Oversized IP packet from %pI4.\n",
647 &qp->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648out_fail:
David Fordbbf31bf2009-11-29 23:02:22 -0800649 IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS);
Herbert Xu1706d582007-10-14 00:38:15 -0700650 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651}
652
653/* Process an incoming IP datagram fragment. */
Herbert Xu776c7292007-10-14 00:38:32 -0700654int ip_defrag(struct sk_buff *skb, u32 user)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 struct ipq *qp;
Pavel Emelyanovac18e752008-01-22 06:02:14 -0800657 struct net *net;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900658
Eric Dumazetadf30902009-06-02 05:19:30 +0000659 net = skb->dev ? dev_net(skb->dev) : dev_net(skb_dst(skb)->dev);
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700660 IP_INC_STATS_BH(net, IPSTATS_MIB_REASMREQDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
662 /* Start by cleaning up the memory. */
Pavel Emelyanove31e0bdc72008-01-22 06:10:13 -0800663 if (atomic_read(&net->ipv4.frags.mem) > net->ipv4.frags.high_thresh)
Pavel Emelyanov6ddc0822008-01-22 06:07:25 -0800664 ip_evictor(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 /* Lookup (or create) queue header */
Pavel Emelyanovac18e752008-01-22 06:02:14 -0800667 if ((qp = ip_find(net, ip_hdr(skb), user)) != NULL) {
Herbert Xu1706d582007-10-14 00:38:15 -0700668 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700670 spin_lock(&qp->q.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
Herbert Xu1706d582007-10-14 00:38:15 -0700672 ret = ip_frag_queue(qp, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673
Pavel Emelyanov5ab11c92007-10-15 02:24:19 -0700674 spin_unlock(&qp->q.lock);
Pavel Emelyanov4b6cb5d2007-10-15 02:41:09 -0700675 ipq_put(qp);
Herbert Xu776c7292007-10-14 00:38:32 -0700676 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 }
678
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700679 IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 kfree_skb(skb);
Herbert Xu776c7292007-10-14 00:38:32 -0700681 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000683EXPORT_SYMBOL(ip_defrag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
Eric Dumazetbc416d92011-10-06 10:28:31 +0000685struct sk_buff *ip_check_defrag(struct sk_buff *skb, u32 user)
686{
687 const struct iphdr *iph;
688 u32 len;
689
690 if (skb->protocol != htons(ETH_P_IP))
691 return skb;
692
693 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
694 return skb;
695
696 iph = ip_hdr(skb);
697 if (iph->ihl < 5 || iph->version != 4)
698 return skb;
699 if (!pskb_may_pull(skb, iph->ihl*4))
700 return skb;
701 iph = ip_hdr(skb);
702 len = ntohs(iph->tot_len);
703 if (skb->len < len || len < (iph->ihl * 4))
704 return skb;
705
706 if (ip_is_fragment(ip_hdr(skb))) {
707 skb = skb_share_check(skb, GFP_ATOMIC);
708 if (skb) {
709 if (pskb_trim_rcsum(skb, len))
710 return skb;
711 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
712 if (ip_defrag(skb, user))
713 return NULL;
714 skb->rxhash = 0;
715 }
716 }
717 return skb;
718}
719EXPORT_SYMBOL(ip_check_defrag);
720
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800721#ifdef CONFIG_SYSCTL
722static int zero;
723
Pavel Emelyanov0a64b4b2008-05-19 13:51:29 -0700724static struct ctl_table ip4_frags_ns_ctl_table[] = {
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800725 {
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800726 .procname = "ipfrag_high_thresh",
Pavel Emelyanove31e0bdc72008-01-22 06:10:13 -0800727 .data = &init_net.ipv4.frags.high_thresh,
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800728 .maxlen = sizeof(int),
729 .mode = 0644,
Alexey Dobriyan6d9f2392008-11-03 18:21:05 -0800730 .proc_handler = proc_dointvec
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800731 },
732 {
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800733 .procname = "ipfrag_low_thresh",
Pavel Emelyanove31e0bdc72008-01-22 06:10:13 -0800734 .data = &init_net.ipv4.frags.low_thresh,
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800735 .maxlen = sizeof(int),
736 .mode = 0644,
Alexey Dobriyan6d9f2392008-11-03 18:21:05 -0800737 .proc_handler = proc_dointvec
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800738 },
739 {
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800740 .procname = "ipfrag_time",
Pavel Emelyanovb2fd5322008-01-22 06:09:37 -0800741 .data = &init_net.ipv4.frags.timeout,
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800742 .maxlen = sizeof(int),
743 .mode = 0644,
Alexey Dobriyan6d9f2392008-11-03 18:21:05 -0800744 .proc_handler = proc_dointvec_jiffies,
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800745 },
Pavel Emelyanov7d291eb2008-05-19 13:53:02 -0700746 { }
747};
748
749static struct ctl_table ip4_frags_ctl_table[] = {
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800750 {
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800751 .procname = "ipfrag_secret_interval",
Pavel Emelyanov3b4bc4a2008-01-22 06:11:04 -0800752 .data = &ip4_frags.secret_interval,
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800753 .maxlen = sizeof(int),
754 .mode = 0644,
Alexey Dobriyan6d9f2392008-11-03 18:21:05 -0800755 .proc_handler = proc_dointvec_jiffies,
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800756 },
757 {
758 .procname = "ipfrag_max_dist",
759 .data = &sysctl_ipfrag_max_dist,
760 .maxlen = sizeof(int),
761 .mode = 0644,
Alexey Dobriyan6d9f2392008-11-03 18:21:05 -0800762 .proc_handler = proc_dointvec_minmax,
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800763 .extra1 = &zero
764 },
765 { }
766};
767
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000768static int __net_init ip4_frags_ns_ctl_register(struct net *net)
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800769{
Pavel Emelyanove4a2d5c2008-01-22 06:08:36 -0800770 struct ctl_table *table;
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800771 struct ctl_table_header *hdr;
772
Pavel Emelyanov0a64b4b2008-05-19 13:51:29 -0700773 table = ip4_frags_ns_ctl_table;
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800774 if (!net_eq(net, &init_net)) {
Pavel Emelyanov0a64b4b2008-05-19 13:51:29 -0700775 table = kmemdup(table, sizeof(ip4_frags_ns_ctl_table), GFP_KERNEL);
Pavel Emelyanove4a2d5c2008-01-22 06:08:36 -0800776 if (table == NULL)
777 goto err_alloc;
778
Pavel Emelyanove31e0bdc72008-01-22 06:10:13 -0800779 table[0].data = &net->ipv4.frags.high_thresh;
780 table[1].data = &net->ipv4.frags.low_thresh;
Pavel Emelyanovb2fd5322008-01-22 06:09:37 -0800781 table[2].data = &net->ipv4.frags.timeout;
Pavel Emelyanove4a2d5c2008-01-22 06:08:36 -0800782 }
783
784 hdr = register_net_sysctl_table(net, net_ipv4_ctl_path, table);
785 if (hdr == NULL)
786 goto err_reg;
787
788 net->ipv4.frags_hdr = hdr;
789 return 0;
790
791err_reg:
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800792 if (!net_eq(net, &init_net))
Pavel Emelyanove4a2d5c2008-01-22 06:08:36 -0800793 kfree(table);
794err_alloc:
795 return -ENOMEM;
796}
797
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000798static void __net_exit ip4_frags_ns_ctl_unregister(struct net *net)
Pavel Emelyanove4a2d5c2008-01-22 06:08:36 -0800799{
800 struct ctl_table *table;
801
802 table = net->ipv4.frags_hdr->ctl_table_arg;
803 unregister_net_sysctl_table(net->ipv4.frags_hdr);
804 kfree(table);
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800805}
Pavel Emelyanov7d291eb2008-05-19 13:53:02 -0700806
807static void ip4_frags_ctl_register(void)
808{
809 register_net_sysctl_rotable(net_ipv4_ctl_path, ip4_frags_ctl_table);
810}
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800811#else
Pavel Emelyanov0a64b4b2008-05-19 13:51:29 -0700812static inline int ip4_frags_ns_ctl_register(struct net *net)
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800813{
814 return 0;
815}
Pavel Emelyanove4a2d5c2008-01-22 06:08:36 -0800816
Pavel Emelyanov0a64b4b2008-05-19 13:51:29 -0700817static inline void ip4_frags_ns_ctl_unregister(struct net *net)
Pavel Emelyanove4a2d5c2008-01-22 06:08:36 -0800818{
819}
Pavel Emelyanov7d291eb2008-05-19 13:53:02 -0700820
821static inline void ip4_frags_ctl_register(void)
822{
823}
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800824#endif
825
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000826static int __net_init ipv4_frags_init_net(struct net *net)
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800827{
Pavel Emelyanovb2fd5322008-01-22 06:09:37 -0800828 /*
Pavel Emelyanove31e0bdc72008-01-22 06:10:13 -0800829 * Fragment cache limits. We will commit 256K at one time. Should we
830 * cross that limit we will prune down to 192K. This should cope with
831 * even the most extreme cases without allowing an attacker to
832 * measurably harm machine performance.
833 */
834 net->ipv4.frags.high_thresh = 256 * 1024;
835 net->ipv4.frags.low_thresh = 192 * 1024;
836 /*
Pavel Emelyanovb2fd5322008-01-22 06:09:37 -0800837 * Important NOTE! Fragment queue must be destroyed before MSL expires.
838 * RFC791 is wrong proposing to prolongate timer each fragment arrival
839 * by TTL.
840 */
841 net->ipv4.frags.timeout = IP_FRAG_TIME;
842
Pavel Emelyanove5a2bb82008-01-22 06:06:23 -0800843 inet_frags_init_net(&net->ipv4.frags);
844
Pavel Emelyanov0a64b4b2008-05-19 13:51:29 -0700845 return ip4_frags_ns_ctl_register(net);
Pavel Emelyanov8d8354d2008-01-22 05:58:31 -0800846}
847
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000848static void __net_exit ipv4_frags_exit_net(struct net *net)
Pavel Emelyanov81566e82008-01-22 06:12:39 -0800849{
Pavel Emelyanov0a64b4b2008-05-19 13:51:29 -0700850 ip4_frags_ns_ctl_unregister(net);
Pavel Emelyanov81566e82008-01-22 06:12:39 -0800851 inet_frags_exit_net(&net->ipv4.frags, &ip4_frags);
852}
853
854static struct pernet_operations ip4_frags_ops = {
855 .init = ipv4_frags_init_net,
856 .exit = ipv4_frags_exit_net,
857};
858
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700859void __init ipfrag_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860{
Pavel Emelyanov7d291eb2008-05-19 13:53:02 -0700861 ip4_frags_ctl_register();
Pavel Emelyanov81566e82008-01-22 06:12:39 -0800862 register_pernet_subsys(&ip4_frags_ops);
Pavel Emelyanov321a3a92007-10-15 02:38:08 -0700863 ip4_frags.hashfn = ip4_hashfn;
Pavel Emelyanovc6fda282007-10-17 19:46:47 -0700864 ip4_frags.constructor = ip4_frag_init;
Pavel Emelyanov1e4b8282007-10-15 02:39:14 -0700865 ip4_frags.destructor = ip4_frag_free;
866 ip4_frags.skb_free = NULL;
867 ip4_frags.qsize = sizeof(struct ipq);
Pavel Emelyanovabd65232007-10-17 19:47:21 -0700868 ip4_frags.match = ip4_frag_match;
Pavel Emelyanove521db92007-10-17 19:45:23 -0700869 ip4_frags.frag_expire = ip_expire;
Pavel Emelyanov3b4bc4a2008-01-22 06:11:04 -0800870 ip4_frags.secret_interval = 10 * 60 * HZ;
Pavel Emelyanov7eb95152007-10-15 02:31:52 -0700871 inet_frags_init(&ip4_frags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872}