Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 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 Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 7 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 8 | * Authors: Fred N. van Kempen <waltje@uWalt.NL.Mugnet.ORG> |
Alan Cox | 113aa83 | 2008-10-13 19:01:08 -0700 | [diff] [blame] | 9 | * Alan Cox <alan@lxorguk.ukuu.org.uk> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 10 | * |
| 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 Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 23 | #include <linux/compiler.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #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> |
| 35 | #include <net/sock.h> |
| 36 | #include <net/ip.h> |
| 37 | #include <net/icmp.h> |
| 38 | #include <net/checksum.h> |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 39 | #include <net/inetpeer.h> |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 40 | #include <net/inet_frag.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 41 | #include <linux/tcp.h> |
| 42 | #include <linux/udp.h> |
| 43 | #include <linux/inet.h> |
| 44 | #include <linux/netfilter_ipv4.h> |
| 45 | |
| 46 | /* NOTE. Logic of IP defragmentation is parallel to corresponding IPv6 |
| 47 | * code now. If you change something here, _PLEASE_ update ipv6/reassembly.c |
| 48 | * as well. Or notify me, at least. --ANK |
| 49 | */ |
| 50 | |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 51 | static int sysctl_ipfrag_max_dist __read_mostly = 64; |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 52 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 53 | struct ipfrag_skb_cb |
| 54 | { |
| 55 | struct inet_skb_parm h; |
| 56 | int offset; |
| 57 | }; |
| 58 | |
Jianjun Kong | fd3f8c4 | 2008-11-03 02:47:38 -0800 | [diff] [blame] | 59 | #define FRAG_CB(skb) ((struct ipfrag_skb_cb *)((skb)->cb)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 60 | |
| 61 | /* Describe an entry in the "incomplete datagrams" queue. */ |
| 62 | struct ipq { |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 63 | struct inet_frag_queue q; |
| 64 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 65 | u32 user; |
Al Viro | 1827777 | 2006-09-26 22:19:02 -0700 | [diff] [blame] | 66 | __be32 saddr; |
| 67 | __be32 daddr; |
| 68 | __be16 id; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 69 | u8 protocol; |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 70 | int iif; |
| 71 | unsigned int rid; |
| 72 | struct inet_peer *peer; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 73 | }; |
| 74 | |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 75 | static struct inet_frags ip4_frags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 76 | |
Pavel Emelyanov | e5a2bb8 | 2008-01-22 06:06:23 -0800 | [diff] [blame] | 77 | int ip_frag_nqueues(struct net *net) |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 78 | { |
Pavel Emelyanov | e5a2bb8 | 2008-01-22 06:06:23 -0800 | [diff] [blame] | 79 | return net->ipv4.frags.nqueues; |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 80 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 81 | |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 82 | int ip_frag_mem(struct net *net) |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 83 | { |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 84 | return atomic_read(&net->ipv4.frags.mem); |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 85 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 86 | |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 87 | static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev, |
| 88 | struct net_device *dev); |
| 89 | |
Pavel Emelyanov | c6fda28 | 2007-10-17 19:46:47 -0700 | [diff] [blame] | 90 | struct ip4_create_arg { |
| 91 | struct iphdr *iph; |
| 92 | u32 user; |
| 93 | }; |
| 94 | |
Al Viro | 1827777 | 2006-09-26 22:19:02 -0700 | [diff] [blame] | 95 | static unsigned int ipqhashfn(__be16 id, __be32 saddr, __be32 daddr, u8 prot) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 96 | { |
Al Viro | 1827777 | 2006-09-26 22:19:02 -0700 | [diff] [blame] | 97 | return jhash_3words((__force u32)id << 16 | prot, |
| 98 | (__force u32)saddr, (__force u32)daddr, |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 99 | ip4_frags.rnd) & (INETFRAGS_HASHSZ - 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 100 | } |
| 101 | |
Pavel Emelyanov | 321a3a9 | 2007-10-15 02:38:08 -0700 | [diff] [blame] | 102 | static unsigned int ip4_hashfn(struct inet_frag_queue *q) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 103 | { |
Pavel Emelyanov | 321a3a9 | 2007-10-15 02:38:08 -0700 | [diff] [blame] | 104 | struct ipq *ipq; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 105 | |
Pavel Emelyanov | 321a3a9 | 2007-10-15 02:38:08 -0700 | [diff] [blame] | 106 | ipq = container_of(q, struct ipq, q); |
| 107 | return ipqhashfn(ipq->id, ipq->saddr, ipq->daddr, ipq->protocol); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 108 | } |
| 109 | |
Pavel Emelyanov | abd6523 | 2007-10-17 19:47:21 -0700 | [diff] [blame] | 110 | static int ip4_frag_match(struct inet_frag_queue *q, void *a) |
| 111 | { |
| 112 | struct ipq *qp; |
| 113 | struct ip4_create_arg *arg = a; |
| 114 | |
| 115 | qp = container_of(q, struct ipq, q); |
| 116 | return (qp->id == arg->iph->id && |
| 117 | qp->saddr == arg->iph->saddr && |
| 118 | qp->daddr == arg->iph->daddr && |
| 119 | qp->protocol == arg->iph->protocol && |
| 120 | qp->user == arg->user); |
| 121 | } |
| 122 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 123 | /* Memory Tracking Functions. */ |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 124 | static __inline__ void frag_kfree_skb(struct netns_frags *nf, |
| 125 | struct sk_buff *skb, int *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 126 | { |
| 127 | if (work) |
| 128 | *work -= skb->truesize; |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 129 | atomic_sub(skb->truesize, &nf->mem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 130 | kfree_skb(skb); |
| 131 | } |
| 132 | |
Pavel Emelyanov | c6fda28 | 2007-10-17 19:46:47 -0700 | [diff] [blame] | 133 | static void ip4_frag_init(struct inet_frag_queue *q, void *a) |
| 134 | { |
| 135 | struct ipq *qp = container_of(q, struct ipq, q); |
| 136 | struct ip4_create_arg *arg = a; |
| 137 | |
| 138 | qp->protocol = arg->iph->protocol; |
| 139 | qp->id = arg->iph->id; |
| 140 | qp->saddr = arg->iph->saddr; |
| 141 | qp->daddr = arg->iph->daddr; |
| 142 | qp->user = arg->user; |
| 143 | qp->peer = sysctl_ipfrag_max_dist ? |
| 144 | inet_getpeer(arg->iph->saddr, 1) : NULL; |
| 145 | } |
| 146 | |
Pavel Emelyanov | 1e4b828 | 2007-10-15 02:39:14 -0700 | [diff] [blame] | 147 | static __inline__ void ip4_frag_free(struct inet_frag_queue *q) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 148 | { |
Pavel Emelyanov | 1e4b828 | 2007-10-15 02:39:14 -0700 | [diff] [blame] | 149 | struct ipq *qp; |
| 150 | |
| 151 | qp = container_of(q, struct ipq, q); |
| 152 | if (qp->peer) |
| 153 | inet_putpeer(qp->peer); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 154 | } |
| 155 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 156 | |
| 157 | /* Destruction primitives. */ |
| 158 | |
Pavel Emelyanov | 4b6cb5d | 2007-10-15 02:41:09 -0700 | [diff] [blame] | 159 | static __inline__ void ipq_put(struct ipq *ipq) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 160 | { |
Pavel Emelyanov | 762cc40 | 2007-10-15 02:41:56 -0700 | [diff] [blame] | 161 | inet_frag_put(&ipq->q, &ip4_frags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 162 | } |
| 163 | |
| 164 | /* Kill ipq entry. It is not destroyed immediately, |
| 165 | * because caller (and someone more) holds reference count. |
| 166 | */ |
| 167 | static void ipq_kill(struct ipq *ipq) |
| 168 | { |
Pavel Emelyanov | 277e650 | 2007-10-15 02:37:18 -0700 | [diff] [blame] | 169 | inet_frag_kill(&ipq->q, &ip4_frags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 170 | } |
| 171 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 172 | /* Memory limiting on fragments. Evictor trashes the oldest |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 173 | * fragment queue until we are back under the threshold. |
| 174 | */ |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 175 | static void ip_evictor(struct net *net) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 176 | { |
Pavel Emelyanov | 8e7999c | 2007-10-15 02:40:06 -0700 | [diff] [blame] | 177 | int evicted; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 178 | |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 179 | evicted = inet_frag_evictor(&net->ipv4.frags, &ip4_frags); |
Pavel Emelyanov | 8e7999c | 2007-10-15 02:40:06 -0700 | [diff] [blame] | 180 | if (evicted) |
Pavel Emelyanov | c5346fe | 2008-07-16 20:20:33 -0700 | [diff] [blame] | 181 | IP_ADD_STATS_BH(net, IPSTATS_MIB_REASMFAILS, evicted); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | } |
| 183 | |
| 184 | /* |
| 185 | * Oops, a fragment queue timed out. Kill it and send an ICMP reply. |
| 186 | */ |
| 187 | static void ip_expire(unsigned long arg) |
| 188 | { |
Pavel Emelyanov | e521db9 | 2007-10-17 19:45:23 -0700 | [diff] [blame] | 189 | struct ipq *qp; |
Pavel Emelyanov | 84a3aa0 | 2008-07-16 20:19:08 -0700 | [diff] [blame] | 190 | struct net *net; |
Pavel Emelyanov | e521db9 | 2007-10-17 19:45:23 -0700 | [diff] [blame] | 191 | |
| 192 | qp = container_of((struct inet_frag_queue *) arg, struct ipq, q); |
Pavel Emelyanov | 84a3aa0 | 2008-07-16 20:19:08 -0700 | [diff] [blame] | 193 | net = container_of(qp->q.net, struct net, ipv4.frags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 194 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 195 | spin_lock(&qp->q.lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 196 | |
Joe Perches | bc578a5 | 2008-03-28 16:35:27 -0700 | [diff] [blame] | 197 | if (qp->q.last_in & INET_FRAG_COMPLETE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 198 | goto out; |
| 199 | |
| 200 | ipq_kill(qp); |
| 201 | |
Pavel Emelyanov | 7c73a6f | 2008-07-16 20:20:11 -0700 | [diff] [blame] | 202 | IP_INC_STATS_BH(net, IPSTATS_MIB_REASMTIMEOUT); |
| 203 | IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 204 | |
Joe Perches | bc578a5 | 2008-03-28 16:35:27 -0700 | [diff] [blame] | 205 | if ((qp->q.last_in & INET_FRAG_FIRST_IN) && qp->q.fragments != NULL) { |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 206 | struct sk_buff *head = qp->q.fragments; |
Denis V. Lunev | cb84663 | 2008-03-24 15:31:00 -0700 | [diff] [blame] | 207 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 208 | /* Send an ICMP "Fragment Reassembly Timeout" message. */ |
Denis V. Lunev | cb84663 | 2008-03-24 15:31:00 -0700 | [diff] [blame] | 209 | if ((head->dev = dev_get_by_index(net, qp->iif)) != NULL) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 210 | icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0); |
| 211 | dev_put(head->dev); |
| 212 | } |
| 213 | } |
| 214 | out: |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 215 | spin_unlock(&qp->q.lock); |
Pavel Emelyanov | 4b6cb5d | 2007-10-15 02:41:09 -0700 | [diff] [blame] | 216 | ipq_put(qp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 217 | } |
| 218 | |
Pavel Emelyanov | abd6523 | 2007-10-17 19:47:21 -0700 | [diff] [blame] | 219 | /* Find the correct entry in the "incomplete datagrams" queue for |
| 220 | * this IP datagram, and create new one, if nothing is found. |
| 221 | */ |
Pavel Emelyanov | ac18e75 | 2008-01-22 06:02:14 -0800 | [diff] [blame] | 222 | static inline struct ipq *ip_find(struct net *net, struct iphdr *iph, u32 user) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 223 | { |
Pavel Emelyanov | c6fda28 | 2007-10-17 19:46:47 -0700 | [diff] [blame] | 224 | struct inet_frag_queue *q; |
| 225 | struct ip4_create_arg arg; |
Pavel Emelyanov | abd6523 | 2007-10-17 19:47:21 -0700 | [diff] [blame] | 226 | unsigned int hash; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 227 | |
Pavel Emelyanov | c6fda28 | 2007-10-17 19:46:47 -0700 | [diff] [blame] | 228 | arg.iph = iph; |
| 229 | arg.user = user; |
Pavel Emelyanov | 9a37580 | 2008-06-27 20:06:08 -0700 | [diff] [blame] | 230 | |
| 231 | read_lock(&ip4_frags.lock); |
Pavel Emelyanov | abd6523 | 2007-10-17 19:47:21 -0700 | [diff] [blame] | 232 | hash = ipqhashfn(iph->id, iph->saddr, iph->daddr, iph->protocol); |
Pavel Emelyanov | c6fda28 | 2007-10-17 19:46:47 -0700 | [diff] [blame] | 233 | |
Pavel Emelyanov | ac18e75 | 2008-01-22 06:02:14 -0800 | [diff] [blame] | 234 | q = inet_frag_find(&net->ipv4.frags, &ip4_frags, &arg, hash); |
Pavel Emelyanov | c6fda28 | 2007-10-17 19:46:47 -0700 | [diff] [blame] | 235 | if (q == NULL) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 236 | goto out_nomem; |
| 237 | |
Pavel Emelyanov | c6fda28 | 2007-10-17 19:46:47 -0700 | [diff] [blame] | 238 | return container_of(q, struct ipq, q); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 239 | |
| 240 | out_nomem: |
Patrick McHardy | 64ce207 | 2005-08-09 20:50:53 -0700 | [diff] [blame] | 241 | LIMIT_NETDEBUG(KERN_ERR "ip_frag_create: no memory left !\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 242 | return NULL; |
| 243 | } |
| 244 | |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 245 | /* Is the fragment too far ahead to be part of ipq? */ |
| 246 | static inline int ip_frag_too_far(struct ipq *qp) |
| 247 | { |
| 248 | struct inet_peer *peer = qp->peer; |
| 249 | unsigned int max = sysctl_ipfrag_max_dist; |
| 250 | unsigned int start, end; |
| 251 | |
| 252 | int rc; |
| 253 | |
| 254 | if (!peer || !max) |
| 255 | return 0; |
| 256 | |
| 257 | start = qp->rid; |
| 258 | end = atomic_inc_return(&peer->rid); |
| 259 | qp->rid = end; |
| 260 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 261 | rc = qp->q.fragments && (end - start) > max; |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 262 | |
| 263 | if (rc) { |
Pavel Emelyanov | 7c73a6f | 2008-07-16 20:20:11 -0700 | [diff] [blame] | 264 | struct net *net; |
| 265 | |
| 266 | net = container_of(qp->q.net, struct net, ipv4.frags); |
| 267 | IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS); |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 268 | } |
| 269 | |
| 270 | return rc; |
| 271 | } |
| 272 | |
| 273 | static int ip_frag_reinit(struct ipq *qp) |
| 274 | { |
| 275 | struct sk_buff *fp; |
| 276 | |
Pavel Emelyanov | b2fd532 | 2008-01-22 06:09:37 -0800 | [diff] [blame] | 277 | if (!mod_timer(&qp->q.timer, jiffies + qp->q.net->timeout)) { |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 278 | atomic_inc(&qp->q.refcnt); |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 279 | return -ETIMEDOUT; |
| 280 | } |
| 281 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 282 | fp = qp->q.fragments; |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 283 | do { |
| 284 | struct sk_buff *xp = fp->next; |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 285 | frag_kfree_skb(qp->q.net, fp, NULL); |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 286 | fp = xp; |
| 287 | } while (fp); |
| 288 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 289 | qp->q.last_in = 0; |
| 290 | qp->q.len = 0; |
| 291 | qp->q.meat = 0; |
| 292 | qp->q.fragments = NULL; |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 293 | qp->iif = 0; |
| 294 | |
| 295 | return 0; |
| 296 | } |
| 297 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 298 | /* Add new segment to existing queue. */ |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 299 | static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 300 | { |
| 301 | struct sk_buff *prev, *next; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 302 | struct net_device *dev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 303 | int flags, offset; |
| 304 | int ihl, end; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 305 | int err = -ENOENT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 306 | |
Joe Perches | bc578a5 | 2008-03-28 16:35:27 -0700 | [diff] [blame] | 307 | if (qp->q.last_in & INET_FRAG_COMPLETE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 308 | goto err; |
| 309 | |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 310 | if (!(IPCB(skb)->flags & IPSKB_FRAG_COMPLETE) && |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 311 | unlikely(ip_frag_too_far(qp)) && |
| 312 | unlikely(err = ip_frag_reinit(qp))) { |
Herbert Xu | 89cee8b | 2005-12-13 23:14:27 -0800 | [diff] [blame] | 313 | ipq_kill(qp); |
| 314 | goto err; |
| 315 | } |
| 316 | |
Arnaldo Carvalho de Melo | eddc9ec | 2007-04-20 22:47:35 -0700 | [diff] [blame] | 317 | offset = ntohs(ip_hdr(skb)->frag_off); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 318 | flags = offset & ~IP_OFFSET; |
| 319 | offset &= IP_OFFSET; |
| 320 | offset <<= 3; /* offset is in 8-byte chunks */ |
Arnaldo Carvalho de Melo | c9bdd4b | 2007-03-12 20:09:15 -0300 | [diff] [blame] | 321 | ihl = ip_hdrlen(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 322 | |
| 323 | /* Determine the position of this fragment. */ |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 324 | end = offset + skb->len - ihl; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 325 | err = -EINVAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 326 | |
| 327 | /* Is this the final fragment? */ |
| 328 | if ((flags & IP_MF) == 0) { |
| 329 | /* If we already have some bits beyond end |
| 330 | * or have different end, the segment is corrrupted. |
| 331 | */ |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 332 | if (end < qp->q.len || |
Joe Perches | bc578a5 | 2008-03-28 16:35:27 -0700 | [diff] [blame] | 333 | ((qp->q.last_in & INET_FRAG_LAST_IN) && end != qp->q.len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 334 | goto err; |
Joe Perches | bc578a5 | 2008-03-28 16:35:27 -0700 | [diff] [blame] | 335 | qp->q.last_in |= INET_FRAG_LAST_IN; |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 336 | qp->q.len = end; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 337 | } else { |
| 338 | if (end&7) { |
| 339 | end &= ~7; |
| 340 | if (skb->ip_summed != CHECKSUM_UNNECESSARY) |
| 341 | skb->ip_summed = CHECKSUM_NONE; |
| 342 | } |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 343 | if (end > qp->q.len) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 344 | /* Some bits beyond end -> corruption. */ |
Joe Perches | bc578a5 | 2008-03-28 16:35:27 -0700 | [diff] [blame] | 345 | if (qp->q.last_in & INET_FRAG_LAST_IN) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 346 | goto err; |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 347 | qp->q.len = end; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 348 | } |
| 349 | } |
| 350 | if (end == offset) |
| 351 | goto err; |
| 352 | |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 353 | err = -ENOMEM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 354 | if (pskb_pull(skb, ihl) == NULL) |
| 355 | goto err; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 356 | |
| 357 | err = pskb_trim_rcsum(skb, end - offset); |
| 358 | if (err) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 359 | goto err; |
| 360 | |
| 361 | /* Find out which fragments are in front and at the back of us |
| 362 | * in the chain of fragments so far. We must know where to put |
| 363 | * this fragment, right? |
| 364 | */ |
| 365 | prev = NULL; |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 366 | for (next = qp->q.fragments; next != NULL; next = next->next) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 367 | if (FRAG_CB(next)->offset >= offset) |
| 368 | break; /* bingo! */ |
| 369 | prev = next; |
| 370 | } |
| 371 | |
| 372 | /* We found where to put this one. Check for overlap with |
| 373 | * preceding fragment, and, if needed, align things so that |
| 374 | * any overlaps are eliminated. |
| 375 | */ |
| 376 | if (prev) { |
| 377 | int i = (FRAG_CB(prev)->offset + prev->len) - offset; |
| 378 | |
| 379 | if (i > 0) { |
| 380 | offset += i; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 381 | err = -EINVAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 382 | if (end <= offset) |
| 383 | goto err; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 384 | err = -ENOMEM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 385 | if (!pskb_pull(skb, i)) |
| 386 | goto err; |
| 387 | if (skb->ip_summed != CHECKSUM_UNNECESSARY) |
| 388 | skb->ip_summed = CHECKSUM_NONE; |
| 389 | } |
| 390 | } |
| 391 | |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 392 | err = -ENOMEM; |
| 393 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 394 | while (next && FRAG_CB(next)->offset < end) { |
| 395 | int i = end - FRAG_CB(next)->offset; /* overlap is 'i' bytes */ |
| 396 | |
| 397 | if (i < next->len) { |
| 398 | /* Eat head of the next overlapped fragment |
| 399 | * and leave the loop. The next ones cannot overlap. |
| 400 | */ |
| 401 | if (!pskb_pull(next, i)) |
| 402 | goto err; |
| 403 | FRAG_CB(next)->offset += i; |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 404 | qp->q.meat -= i; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 405 | if (next->ip_summed != CHECKSUM_UNNECESSARY) |
| 406 | next->ip_summed = CHECKSUM_NONE; |
| 407 | break; |
| 408 | } else { |
| 409 | struct sk_buff *free_it = next; |
| 410 | |
Peter Zijlstra | 47c6bf77 | 2006-12-12 19:48:59 +0100 | [diff] [blame] | 411 | /* Old fragment is completely overridden with |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 412 | * new one drop it. |
| 413 | */ |
| 414 | next = next->next; |
| 415 | |
| 416 | if (prev) |
| 417 | prev->next = next; |
| 418 | else |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 419 | qp->q.fragments = next; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 420 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 421 | qp->q.meat -= free_it->len; |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 422 | frag_kfree_skb(qp->q.net, free_it, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 423 | } |
| 424 | } |
| 425 | |
| 426 | FRAG_CB(skb)->offset = offset; |
| 427 | |
| 428 | /* Insert this fragment in the chain of fragments. */ |
| 429 | skb->next = next; |
| 430 | if (prev) |
| 431 | prev->next = skb; |
| 432 | else |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 433 | qp->q.fragments = skb; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 434 | |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 435 | dev = skb->dev; |
| 436 | if (dev) { |
| 437 | qp->iif = dev->ifindex; |
| 438 | skb->dev = NULL; |
| 439 | } |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 440 | qp->q.stamp = skb->tstamp; |
| 441 | qp->q.meat += skb->len; |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 442 | atomic_add(skb->truesize, &qp->q.net->mem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 443 | if (offset == 0) |
Joe Perches | bc578a5 | 2008-03-28 16:35:27 -0700 | [diff] [blame] | 444 | qp->q.last_in |= INET_FRAG_FIRST_IN; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 445 | |
Joe Perches | bc578a5 | 2008-03-28 16:35:27 -0700 | [diff] [blame] | 446 | if (qp->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) && |
| 447 | qp->q.meat == qp->q.len) |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 448 | return ip_frag_reasm(qp, prev, dev); |
| 449 | |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 450 | write_lock(&ip4_frags.lock); |
Pavel Emelyanov | 3140c25 | 2008-01-22 06:11:48 -0800 | [diff] [blame] | 451 | list_move_tail(&qp->q.lru_list, &qp->q.net->lru_list); |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 452 | write_unlock(&ip4_frags.lock); |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 453 | return -EINPROGRESS; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 454 | |
| 455 | err: |
| 456 | kfree_skb(skb); |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 457 | return err; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 458 | } |
| 459 | |
| 460 | |
| 461 | /* Build a new IP datagram from all its fragments. */ |
| 462 | |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 463 | static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev, |
| 464 | struct net_device *dev) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 465 | { |
Jorge Boncompte [DTI2] | 2bad35b | 2009-03-18 23:26:11 -0700 | [diff] [blame] | 466 | struct net *net = container_of(qp->q.net, struct net, ipv4.frags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 467 | struct iphdr *iph; |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 468 | struct sk_buff *fp, *head = qp->q.fragments; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 469 | int len; |
| 470 | int ihlen; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 471 | int err; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 472 | |
| 473 | ipq_kill(qp); |
| 474 | |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 475 | /* Make the one we just received the head. */ |
| 476 | if (prev) { |
| 477 | head = prev->next; |
| 478 | fp = skb_clone(head, GFP_ATOMIC); |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 479 | if (!fp) |
| 480 | goto out_nomem; |
| 481 | |
| 482 | fp->next = head->next; |
| 483 | prev->next = fp; |
| 484 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 485 | skb_morph(head, qp->q.fragments); |
| 486 | head->next = qp->q.fragments->next; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 487 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 488 | kfree_skb(qp->q.fragments); |
| 489 | qp->q.fragments = head; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 490 | } |
| 491 | |
Ilpo Järvinen | 547b792 | 2008-07-25 21:43:18 -0700 | [diff] [blame] | 492 | WARN_ON(head == NULL); |
| 493 | WARN_ON(FRAG_CB(head)->offset != 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 494 | |
| 495 | /* Allocate a new buffer for the datagram. */ |
Arnaldo Carvalho de Melo | c9bdd4b | 2007-03-12 20:09:15 -0300 | [diff] [blame] | 496 | ihlen = ip_hdrlen(head); |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 497 | len = ihlen + qp->q.len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 498 | |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 499 | err = -E2BIG; |
Stephen Hemminger | 132adf5 | 2007-03-08 20:44:43 -0800 | [diff] [blame] | 500 | if (len > 65535) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 501 | goto out_oversize; |
| 502 | |
| 503 | /* Head of list must not be cloned. */ |
| 504 | if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC)) |
| 505 | goto out_nomem; |
| 506 | |
| 507 | /* If the first fragment is fragmented itself, we split |
| 508 | * it to two chunks: the first with data and paged part |
| 509 | * and the second, holding only fragments. */ |
| 510 | if (skb_shinfo(head)->frag_list) { |
| 511 | struct sk_buff *clone; |
| 512 | int i, plen = 0; |
| 513 | |
| 514 | if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL) |
| 515 | goto out_nomem; |
| 516 | clone->next = head->next; |
| 517 | head->next = clone; |
| 518 | skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list; |
| 519 | skb_shinfo(head)->frag_list = NULL; |
| 520 | for (i=0; i<skb_shinfo(head)->nr_frags; i++) |
| 521 | plen += skb_shinfo(head)->frags[i].size; |
| 522 | clone->len = clone->data_len = head->data_len - plen; |
| 523 | head->data_len -= clone->len; |
| 524 | head->len -= clone->len; |
| 525 | clone->csum = 0; |
| 526 | clone->ip_summed = head->ip_summed; |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 527 | atomic_add(clone->truesize, &qp->q.net->mem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 528 | } |
| 529 | |
| 530 | skb_shinfo(head)->frag_list = head->next; |
Arnaldo Carvalho de Melo | d56f90a | 2007-04-10 20:50:43 -0700 | [diff] [blame] | 531 | skb_push(head, head->data - skb_network_header(head)); |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 532 | atomic_sub(head->truesize, &qp->q.net->mem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 533 | |
| 534 | for (fp=head->next; fp; fp = fp->next) { |
| 535 | head->data_len += fp->len; |
| 536 | head->len += fp->len; |
| 537 | if (head->ip_summed != fp->ip_summed) |
| 538 | head->ip_summed = CHECKSUM_NONE; |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 539 | else if (head->ip_summed == CHECKSUM_COMPLETE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 540 | head->csum = csum_add(head->csum, fp->csum); |
| 541 | head->truesize += fp->truesize; |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 542 | atomic_sub(fp->truesize, &qp->q.net->mem); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 543 | } |
| 544 | |
| 545 | head->next = NULL; |
| 546 | head->dev = dev; |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 547 | head->tstamp = qp->q.stamp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 548 | |
Arnaldo Carvalho de Melo | eddc9ec | 2007-04-20 22:47:35 -0700 | [diff] [blame] | 549 | iph = ip_hdr(head); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 550 | iph->frag_off = 0; |
| 551 | iph->tot_len = htons(len); |
Jorge Boncompte [DTI2] | 2bad35b | 2009-03-18 23:26:11 -0700 | [diff] [blame] | 552 | IP_INC_STATS_BH(net, IPSTATS_MIB_REASMOKS); |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 553 | qp->q.fragments = NULL; |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 554 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 555 | |
| 556 | out_nomem: |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 557 | LIMIT_NETDEBUG(KERN_ERR "IP: queue_glue: no memory for gluing " |
Patrick McHardy | 64ce207 | 2005-08-09 20:50:53 -0700 | [diff] [blame] | 558 | "queue %p\n", qp); |
David Howells | 4554247 | 2007-10-17 21:37:22 -0700 | [diff] [blame] | 559 | err = -ENOMEM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 560 | goto out_fail; |
| 561 | out_oversize: |
| 562 | if (net_ratelimit()) |
Harvey Harrison | 673d57e | 2008-10-31 00:53:57 -0700 | [diff] [blame] | 563 | printk(KERN_INFO "Oversized IP packet from %pI4.\n", |
| 564 | &qp->saddr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 565 | out_fail: |
Pavel Emelyanov | 7c73a6f | 2008-07-16 20:20:11 -0700 | [diff] [blame] | 566 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_REASMFAILS); |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 567 | return err; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 568 | } |
| 569 | |
| 570 | /* Process an incoming IP datagram fragment. */ |
Herbert Xu | 776c729 | 2007-10-14 00:38:32 -0700 | [diff] [blame] | 571 | int ip_defrag(struct sk_buff *skb, u32 user) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 572 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 573 | struct ipq *qp; |
Pavel Emelyanov | ac18e75 | 2008-01-22 06:02:14 -0800 | [diff] [blame] | 574 | struct net *net; |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 575 | |
Pavel Emelyanov | 84a3aa0 | 2008-07-16 20:19:08 -0700 | [diff] [blame] | 576 | net = skb->dev ? dev_net(skb->dev) : dev_net(skb->dst->dev); |
Pavel Emelyanov | 7c73a6f | 2008-07-16 20:20:11 -0700 | [diff] [blame] | 577 | IP_INC_STATS_BH(net, IPSTATS_MIB_REASMREQDS); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 578 | |
| 579 | /* Start by cleaning up the memory. */ |
Pavel Emelyanov | e31e0bdc7 | 2008-01-22 06:10:13 -0800 | [diff] [blame] | 580 | if (atomic_read(&net->ipv4.frags.mem) > net->ipv4.frags.high_thresh) |
Pavel Emelyanov | 6ddc082 | 2008-01-22 06:07:25 -0800 | [diff] [blame] | 581 | ip_evictor(net); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 582 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 583 | /* Lookup (or create) queue header */ |
Pavel Emelyanov | ac18e75 | 2008-01-22 06:02:14 -0800 | [diff] [blame] | 584 | if ((qp = ip_find(net, ip_hdr(skb), user)) != NULL) { |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 585 | int ret; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 586 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 587 | spin_lock(&qp->q.lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 588 | |
Herbert Xu | 1706d58 | 2007-10-14 00:38:15 -0700 | [diff] [blame] | 589 | ret = ip_frag_queue(qp, skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 590 | |
Pavel Emelyanov | 5ab11c9 | 2007-10-15 02:24:19 -0700 | [diff] [blame] | 591 | spin_unlock(&qp->q.lock); |
Pavel Emelyanov | 4b6cb5d | 2007-10-15 02:41:09 -0700 | [diff] [blame] | 592 | ipq_put(qp); |
Herbert Xu | 776c729 | 2007-10-14 00:38:32 -0700 | [diff] [blame] | 593 | return ret; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 594 | } |
| 595 | |
Pavel Emelyanov | 7c73a6f | 2008-07-16 20:20:11 -0700 | [diff] [blame] | 596 | IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 597 | kfree_skb(skb); |
Herbert Xu | 776c729 | 2007-10-14 00:38:32 -0700 | [diff] [blame] | 598 | return -ENOMEM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 599 | } |
| 600 | |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 601 | #ifdef CONFIG_SYSCTL |
| 602 | static int zero; |
| 603 | |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 604 | static struct ctl_table ip4_frags_ns_ctl_table[] = { |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 605 | { |
| 606 | .ctl_name = NET_IPV4_IPFRAG_HIGH_THRESH, |
| 607 | .procname = "ipfrag_high_thresh", |
Pavel Emelyanov | e31e0bdc7 | 2008-01-22 06:10:13 -0800 | [diff] [blame] | 608 | .data = &init_net.ipv4.frags.high_thresh, |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 609 | .maxlen = sizeof(int), |
| 610 | .mode = 0644, |
Alexey Dobriyan | 6d9f239 | 2008-11-03 18:21:05 -0800 | [diff] [blame] | 611 | .proc_handler = proc_dointvec |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 612 | }, |
| 613 | { |
| 614 | .ctl_name = NET_IPV4_IPFRAG_LOW_THRESH, |
| 615 | .procname = "ipfrag_low_thresh", |
Pavel Emelyanov | e31e0bdc7 | 2008-01-22 06:10:13 -0800 | [diff] [blame] | 616 | .data = &init_net.ipv4.frags.low_thresh, |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 617 | .maxlen = sizeof(int), |
| 618 | .mode = 0644, |
Alexey Dobriyan | 6d9f239 | 2008-11-03 18:21:05 -0800 | [diff] [blame] | 619 | .proc_handler = proc_dointvec |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 620 | }, |
| 621 | { |
| 622 | .ctl_name = NET_IPV4_IPFRAG_TIME, |
| 623 | .procname = "ipfrag_time", |
Pavel Emelyanov | b2fd532 | 2008-01-22 06:09:37 -0800 | [diff] [blame] | 624 | .data = &init_net.ipv4.frags.timeout, |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 625 | .maxlen = sizeof(int), |
| 626 | .mode = 0644, |
Alexey Dobriyan | 6d9f239 | 2008-11-03 18:21:05 -0800 | [diff] [blame] | 627 | .proc_handler = proc_dointvec_jiffies, |
| 628 | .strategy = sysctl_jiffies |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 629 | }, |
Pavel Emelyanov | 7d291eb | 2008-05-19 13:53:02 -0700 | [diff] [blame] | 630 | { } |
| 631 | }; |
| 632 | |
| 633 | static struct ctl_table ip4_frags_ctl_table[] = { |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 634 | { |
| 635 | .ctl_name = NET_IPV4_IPFRAG_SECRET_INTERVAL, |
| 636 | .procname = "ipfrag_secret_interval", |
Pavel Emelyanov | 3b4bc4a | 2008-01-22 06:11:04 -0800 | [diff] [blame] | 637 | .data = &ip4_frags.secret_interval, |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 638 | .maxlen = sizeof(int), |
| 639 | .mode = 0644, |
Alexey Dobriyan | 6d9f239 | 2008-11-03 18:21:05 -0800 | [diff] [blame] | 640 | .proc_handler = proc_dointvec_jiffies, |
| 641 | .strategy = sysctl_jiffies |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 642 | }, |
| 643 | { |
| 644 | .procname = "ipfrag_max_dist", |
| 645 | .data = &sysctl_ipfrag_max_dist, |
| 646 | .maxlen = sizeof(int), |
| 647 | .mode = 0644, |
Alexey Dobriyan | 6d9f239 | 2008-11-03 18:21:05 -0800 | [diff] [blame] | 648 | .proc_handler = proc_dointvec_minmax, |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 649 | .extra1 = &zero |
| 650 | }, |
| 651 | { } |
| 652 | }; |
| 653 | |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 654 | static int ip4_frags_ns_ctl_register(struct net *net) |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 655 | { |
Pavel Emelyanov | e4a2d5c | 2008-01-22 06:08:36 -0800 | [diff] [blame] | 656 | struct ctl_table *table; |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 657 | struct ctl_table_header *hdr; |
| 658 | |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 659 | table = ip4_frags_ns_ctl_table; |
Pavel Emelyanov | e4a2d5c | 2008-01-22 06:08:36 -0800 | [diff] [blame] | 660 | if (net != &init_net) { |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 661 | table = kmemdup(table, sizeof(ip4_frags_ns_ctl_table), GFP_KERNEL); |
Pavel Emelyanov | e4a2d5c | 2008-01-22 06:08:36 -0800 | [diff] [blame] | 662 | if (table == NULL) |
| 663 | goto err_alloc; |
| 664 | |
Pavel Emelyanov | e31e0bdc7 | 2008-01-22 06:10:13 -0800 | [diff] [blame] | 665 | table[0].data = &net->ipv4.frags.high_thresh; |
| 666 | table[1].data = &net->ipv4.frags.low_thresh; |
Pavel Emelyanov | b2fd532 | 2008-01-22 06:09:37 -0800 | [diff] [blame] | 667 | table[2].data = &net->ipv4.frags.timeout; |
Pavel Emelyanov | e4a2d5c | 2008-01-22 06:08:36 -0800 | [diff] [blame] | 668 | } |
| 669 | |
| 670 | hdr = register_net_sysctl_table(net, net_ipv4_ctl_path, table); |
| 671 | if (hdr == NULL) |
| 672 | goto err_reg; |
| 673 | |
| 674 | net->ipv4.frags_hdr = hdr; |
| 675 | return 0; |
| 676 | |
| 677 | err_reg: |
| 678 | if (net != &init_net) |
| 679 | kfree(table); |
| 680 | err_alloc: |
| 681 | return -ENOMEM; |
| 682 | } |
| 683 | |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 684 | static void ip4_frags_ns_ctl_unregister(struct net *net) |
Pavel Emelyanov | e4a2d5c | 2008-01-22 06:08:36 -0800 | [diff] [blame] | 685 | { |
| 686 | struct ctl_table *table; |
| 687 | |
| 688 | table = net->ipv4.frags_hdr->ctl_table_arg; |
| 689 | unregister_net_sysctl_table(net->ipv4.frags_hdr); |
| 690 | kfree(table); |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 691 | } |
Pavel Emelyanov | 7d291eb | 2008-05-19 13:53:02 -0700 | [diff] [blame] | 692 | |
| 693 | static void ip4_frags_ctl_register(void) |
| 694 | { |
| 695 | register_net_sysctl_rotable(net_ipv4_ctl_path, ip4_frags_ctl_table); |
| 696 | } |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 697 | #else |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 698 | static inline int ip4_frags_ns_ctl_register(struct net *net) |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 699 | { |
| 700 | return 0; |
| 701 | } |
Pavel Emelyanov | e4a2d5c | 2008-01-22 06:08:36 -0800 | [diff] [blame] | 702 | |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 703 | static inline void ip4_frags_ns_ctl_unregister(struct net *net) |
Pavel Emelyanov | e4a2d5c | 2008-01-22 06:08:36 -0800 | [diff] [blame] | 704 | { |
| 705 | } |
Pavel Emelyanov | 7d291eb | 2008-05-19 13:53:02 -0700 | [diff] [blame] | 706 | |
| 707 | static inline void ip4_frags_ctl_register(void) |
| 708 | { |
| 709 | } |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 710 | #endif |
| 711 | |
| 712 | static int ipv4_frags_init_net(struct net *net) |
| 713 | { |
Pavel Emelyanov | b2fd532 | 2008-01-22 06:09:37 -0800 | [diff] [blame] | 714 | /* |
Pavel Emelyanov | e31e0bdc7 | 2008-01-22 06:10:13 -0800 | [diff] [blame] | 715 | * Fragment cache limits. We will commit 256K at one time. Should we |
| 716 | * cross that limit we will prune down to 192K. This should cope with |
| 717 | * even the most extreme cases without allowing an attacker to |
| 718 | * measurably harm machine performance. |
| 719 | */ |
| 720 | net->ipv4.frags.high_thresh = 256 * 1024; |
| 721 | net->ipv4.frags.low_thresh = 192 * 1024; |
| 722 | /* |
Pavel Emelyanov | b2fd532 | 2008-01-22 06:09:37 -0800 | [diff] [blame] | 723 | * Important NOTE! Fragment queue must be destroyed before MSL expires. |
| 724 | * RFC791 is wrong proposing to prolongate timer each fragment arrival |
| 725 | * by TTL. |
| 726 | */ |
| 727 | net->ipv4.frags.timeout = IP_FRAG_TIME; |
| 728 | |
Pavel Emelyanov | e5a2bb8 | 2008-01-22 06:06:23 -0800 | [diff] [blame] | 729 | inet_frags_init_net(&net->ipv4.frags); |
| 730 | |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 731 | return ip4_frags_ns_ctl_register(net); |
Pavel Emelyanov | 8d8354d | 2008-01-22 05:58:31 -0800 | [diff] [blame] | 732 | } |
| 733 | |
Pavel Emelyanov | 81566e8 | 2008-01-22 06:12:39 -0800 | [diff] [blame] | 734 | static void ipv4_frags_exit_net(struct net *net) |
| 735 | { |
Pavel Emelyanov | 0a64b4b | 2008-05-19 13:51:29 -0700 | [diff] [blame] | 736 | ip4_frags_ns_ctl_unregister(net); |
Pavel Emelyanov | 81566e8 | 2008-01-22 06:12:39 -0800 | [diff] [blame] | 737 | inet_frags_exit_net(&net->ipv4.frags, &ip4_frags); |
| 738 | } |
| 739 | |
| 740 | static struct pernet_operations ip4_frags_ops = { |
| 741 | .init = ipv4_frags_init_net, |
| 742 | .exit = ipv4_frags_exit_net, |
| 743 | }; |
| 744 | |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 745 | void __init ipfrag_init(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 746 | { |
Pavel Emelyanov | 7d291eb | 2008-05-19 13:53:02 -0700 | [diff] [blame] | 747 | ip4_frags_ctl_register(); |
Pavel Emelyanov | 81566e8 | 2008-01-22 06:12:39 -0800 | [diff] [blame] | 748 | register_pernet_subsys(&ip4_frags_ops); |
Pavel Emelyanov | 321a3a9 | 2007-10-15 02:38:08 -0700 | [diff] [blame] | 749 | ip4_frags.hashfn = ip4_hashfn; |
Pavel Emelyanov | c6fda28 | 2007-10-17 19:46:47 -0700 | [diff] [blame] | 750 | ip4_frags.constructor = ip4_frag_init; |
Pavel Emelyanov | 1e4b828 | 2007-10-15 02:39:14 -0700 | [diff] [blame] | 751 | ip4_frags.destructor = ip4_frag_free; |
| 752 | ip4_frags.skb_free = NULL; |
| 753 | ip4_frags.qsize = sizeof(struct ipq); |
Pavel Emelyanov | abd6523 | 2007-10-17 19:47:21 -0700 | [diff] [blame] | 754 | ip4_frags.match = ip4_frag_match; |
Pavel Emelyanov | e521db9 | 2007-10-17 19:45:23 -0700 | [diff] [blame] | 755 | ip4_frags.frag_expire = ip_expire; |
Pavel Emelyanov | 3b4bc4a | 2008-01-22 06:11:04 -0800 | [diff] [blame] | 756 | ip4_frags.secret_interval = 10 * 60 * HZ; |
Pavel Emelyanov | 7eb9515 | 2007-10-15 02:31:52 -0700 | [diff] [blame] | 757 | inet_frags_init(&ip4_frags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 758 | } |
| 759 | |
| 760 | EXPORT_SYMBOL(ip_defrag); |