| 1 |
1629 |
jcastillo |
/*
|
| 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 Internet Protocol (IP) module.
|
| 7 |
|
|
*
|
| 8 |
|
|
* Version: @(#)ip.c 1.0.16b 9/1/93
|
| 9 |
|
|
*
|
| 10 |
|
|
* Authors: Ross Biro, <bir7@leland.Stanford.Edu>
|
| 11 |
|
|
* Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
|
| 12 |
|
|
* Donald Becker, <becker@super.org>
|
| 13 |
|
|
* Alan Cox, <Alan.Cox@linux.org>
|
| 14 |
|
|
* Richard Underwood
|
| 15 |
|
|
* Stefan Becker, <stefanb@yello.ping.de>
|
| 16 |
|
|
* Jorge Cwik, <jorge@laser.satlink.net>
|
| 17 |
|
|
* Arnt Gulbrandsen, <agulbra@nvg.unit.no>
|
| 18 |
|
|
*
|
| 19 |
|
|
*
|
| 20 |
|
|
* Fixes:
|
| 21 |
|
|
* Alan Cox : Commented a couple of minor bits of surplus code
|
| 22 |
|
|
* Alan Cox : Undefining IP_FORWARD doesn't include the code
|
| 23 |
|
|
* (just stops a compiler warning).
|
| 24 |
|
|
* Alan Cox : Frames with >=MAX_ROUTE record routes, strict routes or loose routes
|
| 25 |
|
|
* are junked rather than corrupting things.
|
| 26 |
|
|
* Alan Cox : Frames to bad broadcast subnets are dumped
|
| 27 |
|
|
* We used to process them non broadcast and
|
| 28 |
|
|
* boy could that cause havoc.
|
| 29 |
|
|
* Alan Cox : ip_forward sets the free flag on the
|
| 30 |
|
|
* new frame it queues. Still crap because
|
| 31 |
|
|
* it copies the frame but at least it
|
| 32 |
|
|
* doesn't eat memory too.
|
| 33 |
|
|
* Alan Cox : Generic queue code and memory fixes.
|
| 34 |
|
|
* Fred Van Kempen : IP fragment support (borrowed from NET2E)
|
| 35 |
|
|
* Gerhard Koerting: Forward fragmented frames correctly.
|
| 36 |
|
|
* Gerhard Koerting: Fixes to my fix of the above 8-).
|
| 37 |
|
|
* Gerhard Koerting: IP interface addressing fix.
|
| 38 |
|
|
* Linus Torvalds : More robustness checks
|
| 39 |
|
|
* Alan Cox : Even more checks: Still not as robust as it ought to be
|
| 40 |
|
|
* Alan Cox : Save IP header pointer for later
|
| 41 |
|
|
* Alan Cox : ip option setting
|
| 42 |
|
|
* Alan Cox : Use ip_tos/ip_ttl settings
|
| 43 |
|
|
* Alan Cox : Fragmentation bogosity removed
|
| 44 |
|
|
* (Thanks to Mark.Bush@prg.ox.ac.uk)
|
| 45 |
|
|
* Dmitry Gorodchanin : Send of a raw packet crash fix.
|
| 46 |
|
|
* Alan Cox : Silly ip bug when an overlength
|
| 47 |
|
|
* fragment turns up. Now frees the
|
| 48 |
|
|
* queue.
|
| 49 |
|
|
* Linus Torvalds/ : Memory leakage on fragmentation
|
| 50 |
|
|
* Alan Cox : handling.
|
| 51 |
|
|
* Gerhard Koerting: Forwarding uses IP priority hints
|
| 52 |
|
|
* Teemu Rantanen : Fragment problems.
|
| 53 |
|
|
* Alan Cox : General cleanup, comments and reformat
|
| 54 |
|
|
* Alan Cox : SNMP statistics
|
| 55 |
|
|
* Alan Cox : BSD address rule semantics. Also see
|
| 56 |
|
|
* UDP as there is a nasty checksum issue
|
| 57 |
|
|
* if you do things the wrong way.
|
| 58 |
|
|
* Alan Cox : Always defrag, moved IP_FORWARD to the config.in file
|
| 59 |
|
|
* Alan Cox : IP options adjust sk->priority.
|
| 60 |
|
|
* Pedro Roque : Fix mtu/length error in ip_forward.
|
| 61 |
|
|
* Alan Cox : Avoid ip_chk_addr when possible.
|
| 62 |
|
|
* Richard Underwood : IP multicasting.
|
| 63 |
|
|
* Alan Cox : Cleaned up multicast handlers.
|
| 64 |
|
|
* Alan Cox : RAW sockets demultiplex in the BSD style.
|
| 65 |
|
|
* Gunther Mayer : Fix the SNMP reporting typo
|
| 66 |
|
|
* Alan Cox : Always in group 224.0.0.1
|
| 67 |
|
|
* Pauline Middelink : Fast ip_checksum update when forwarding
|
| 68 |
|
|
* Masquerading support.
|
| 69 |
|
|
* Alan Cox : Multicast loopback error for 224.0.0.1
|
| 70 |
|
|
* Alan Cox : IP_MULTICAST_LOOP option.
|
| 71 |
|
|
* Alan Cox : Use notifiers.
|
| 72 |
|
|
* Bjorn Ekwall : Removed ip_csum (from slhc.c too)
|
| 73 |
|
|
* Bjorn Ekwall : Moved ip_fast_csum to ip.h (inline!)
|
| 74 |
|
|
* Stefan Becker : Send out ICMP HOST REDIRECT
|
| 75 |
|
|
* Arnt Gulbrandsen : ip_build_xmit
|
| 76 |
|
|
* Alan Cox : Per socket routing cache
|
| 77 |
|
|
* Alan Cox : Fixed routing cache, added header cache.
|
| 78 |
|
|
* Alan Cox : Loopback didn't work right in original ip_build_xmit - fixed it.
|
| 79 |
|
|
* Alan Cox : Only send ICMP_REDIRECT if src/dest are the same net.
|
| 80 |
|
|
* Alan Cox : Incoming IP option handling.
|
| 81 |
|
|
* Alan Cox : Set saddr on raw output frames as per BSD.
|
| 82 |
|
|
* Alan Cox : Stopped broadcast source route explosions.
|
| 83 |
|
|
* Alan Cox : Can disable source routing
|
| 84 |
|
|
* Takeshi Sone : Masquerading didn't work.
|
| 85 |
|
|
* Dave Bonn,Alan Cox : Faster IP forwarding whenever possible.
|
| 86 |
|
|
* Alan Cox : Memory leaks, tramples, misc debugging.
|
| 87 |
|
|
* Alan Cox : Fixed multicast (by popular demand 8))
|
| 88 |
|
|
* Alan Cox : Fixed forwarding (by even more popular demand 8))
|
| 89 |
|
|
* Alan Cox : Fixed SNMP statistics [I think]
|
| 90 |
|
|
* Gerhard Koerting : IP fragmentation forwarding fix
|
| 91 |
|
|
* Alan Cox : Device lock against page fault.
|
| 92 |
|
|
* Alan Cox : IP_HDRINCL facility.
|
| 93 |
|
|
* Werner Almesberger : Zero fragment bug
|
| 94 |
|
|
* Alan Cox : RAW IP frame length bug
|
| 95 |
|
|
* Alan Cox : Outgoing firewall on build_xmit
|
| 96 |
|
|
* A.N.Kuznetsov : IP_OPTIONS support throughout the kernel
|
| 97 |
|
|
* Alan Cox : Multicast routing hooks
|
| 98 |
|
|
* Jos Vos : Do accounting *before* call_in_firewall
|
| 99 |
|
|
* Willy Konynenberg : Transparent proxying support
|
| 100 |
|
|
*
|
| 101 |
|
|
*
|
| 102 |
|
|
*
|
| 103 |
|
|
* To Fix:
|
| 104 |
|
|
* IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient
|
| 105 |
|
|
* and could be made very efficient with the addition of some virtual memory hacks to permit
|
| 106 |
|
|
* the allocation of a buffer that can then be 'grown' by twiddling page tables.
|
| 107 |
|
|
* Output fragmentation wants updating along with the buffer management to use a single
|
| 108 |
|
|
* interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet
|
| 109 |
|
|
* output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause
|
| 110 |
|
|
* fragmentation anyway.
|
| 111 |
|
|
*
|
| 112 |
|
|
* FIXME: copy frag 0 iph to qp->iph
|
| 113 |
|
|
*
|
| 114 |
|
|
* This program is free software; you can redistribute it and/or
|
| 115 |
|
|
* modify it under the terms of the GNU General Public License
|
| 116 |
|
|
* as published by the Free Software Foundation; either version
|
| 117 |
|
|
* 2 of the License, or (at your option) any later version.
|
| 118 |
|
|
*/
|
| 119 |
|
|
|
| 120 |
|
|
#include <asm/segment.h>
|
| 121 |
|
|
#include <asm/system.h>
|
| 122 |
|
|
#include <linux/types.h>
|
| 123 |
|
|
#include <linux/kernel.h>
|
| 124 |
|
|
#include <linux/sched.h>
|
| 125 |
|
|
#include <linux/mm.h>
|
| 126 |
|
|
#include <linux/string.h>
|
| 127 |
|
|
#include <linux/errno.h>
|
| 128 |
|
|
#include <linux/config.h>
|
| 129 |
|
|
|
| 130 |
|
|
#include <linux/socket.h>
|
| 131 |
|
|
#include <linux/sockios.h>
|
| 132 |
|
|
#include <linux/in.h>
|
| 133 |
|
|
#include <linux/inet.h>
|
| 134 |
|
|
#include <linux/netdevice.h>
|
| 135 |
|
|
#include <linux/etherdevice.h>
|
| 136 |
|
|
#include <linux/proc_fs.h>
|
| 137 |
|
|
#include <linux/stat.h>
|
| 138 |
|
|
|
| 139 |
|
|
#include <net/snmp.h>
|
| 140 |
|
|
#include <net/ip.h>
|
| 141 |
|
|
#include <net/protocol.h>
|
| 142 |
|
|
#include <net/route.h>
|
| 143 |
|
|
#include <net/tcp.h>
|
| 144 |
|
|
#include <net/udp.h>
|
| 145 |
|
|
#include <linux/skbuff.h>
|
| 146 |
|
|
#include <net/sock.h>
|
| 147 |
|
|
#include <net/arp.h>
|
| 148 |
|
|
#include <net/icmp.h>
|
| 149 |
|
|
#include <net/raw.h>
|
| 150 |
|
|
#include <net/checksum.h>
|
| 151 |
|
|
#include <linux/igmp.h>
|
| 152 |
|
|
#include <linux/ip_fw.h>
|
| 153 |
|
|
#ifdef CONFIG_IP_MASQUERADE
|
| 154 |
|
|
#include <net/ip_masq.h>
|
| 155 |
|
|
#endif
|
| 156 |
|
|
#include <linux/firewall.h>
|
| 157 |
|
|
#include <linux/mroute.h>
|
| 158 |
|
|
#include <net/netlink.h>
|
| 159 |
|
|
#ifdef CONFIG_NET_ALIAS
|
| 160 |
|
|
#include <linux/net_alias.h>
|
| 161 |
|
|
#endif
|
| 162 |
|
|
|
| 163 |
|
|
extern int last_retran;
|
| 164 |
|
|
extern void sort_send(struct sock *sk);
|
| 165 |
|
|
|
| 166 |
|
|
#define min(a,b) ((a)<(b)?(a):(b))
|
| 167 |
|
|
|
| 168 |
|
|
/*
|
| 169 |
|
|
* SNMP management statistics
|
| 170 |
|
|
*/
|
| 171 |
|
|
|
| 172 |
|
|
#ifdef CONFIG_IP_FORWARD
|
| 173 |
|
|
struct ip_mib ip_statistics={1,64,}; /* Forwarding=Yes, Default TTL=64 */
|
| 174 |
|
|
#else
|
| 175 |
|
|
struct ip_mib ip_statistics={2,64,}; /* Forwarding=No, Default TTL=64 */
|
| 176 |
|
|
#endif
|
| 177 |
|
|
|
| 178 |
|
|
/*
|
| 179 |
|
|
* Handle the issuing of an ioctl() request
|
| 180 |
|
|
* for the ip device. This is scheduled to
|
| 181 |
|
|
* disappear
|
| 182 |
|
|
*/
|
| 183 |
|
|
|
| 184 |
|
|
int ip_ioctl(struct sock *sk, int cmd, unsigned long arg)
|
| 185 |
|
|
{
|
| 186 |
|
|
switch(cmd)
|
| 187 |
|
|
{
|
| 188 |
|
|
default:
|
| 189 |
|
|
return(-EINVAL);
|
| 190 |
|
|
}
|
| 191 |
|
|
}
|
| 192 |
|
|
|
| 193 |
|
|
#ifdef CONFIG_IP_TRANSPARENT_PROXY
|
| 194 |
|
|
/*
|
| 195 |
|
|
* Check the packet against our socket administration to see
|
| 196 |
|
|
* if it is related to a connection on our system.
|
| 197 |
|
|
* Needed for transparent proxying.
|
| 198 |
|
|
*/
|
| 199 |
|
|
|
| 200 |
|
|
int ip_chksock(struct sk_buff *skb)
|
| 201 |
|
|
{
|
| 202 |
|
|
switch (skb->h.iph->protocol) {
|
| 203 |
|
|
case IPPROTO_ICMP:
|
| 204 |
|
|
return icmp_chkaddr(skb);
|
| 205 |
|
|
case IPPROTO_TCP:
|
| 206 |
|
|
return tcp_chkaddr(skb);
|
| 207 |
|
|
case IPPROTO_UDP:
|
| 208 |
|
|
return udp_chkaddr(skb);
|
| 209 |
|
|
default:
|
| 210 |
|
|
return 0;
|
| 211 |
|
|
}
|
| 212 |
|
|
}
|
| 213 |
|
|
#endif
|
| 214 |
|
|
|
| 215 |
|
|
|
| 216 |
|
|
/*
|
| 217 |
|
|
* This function receives all incoming IP datagrams.
|
| 218 |
|
|
*
|
| 219 |
|
|
* On entry skb->data points to the start of the IP header and
|
| 220 |
|
|
* the MAC header has been removed.
|
| 221 |
|
|
*/
|
| 222 |
|
|
|
| 223 |
|
|
int ip_rcv(struct sk_buff *skb, struct device *dev, struct packet_type *pt)
|
| 224 |
|
|
{
|
| 225 |
|
|
struct iphdr *iph = skb->h.iph;
|
| 226 |
|
|
struct sock *raw_sk=NULL;
|
| 227 |
|
|
unsigned char hash;
|
| 228 |
|
|
unsigned char flag = 0;
|
| 229 |
|
|
struct inet_protocol *ipprot;
|
| 230 |
|
|
int brd=IS_MYADDR;
|
| 231 |
|
|
struct options * opt = NULL;
|
| 232 |
|
|
int is_frag=0;
|
| 233 |
|
|
__u32 daddr;
|
| 234 |
|
|
|
| 235 |
|
|
#ifdef CONFIG_FIREWALL
|
| 236 |
|
|
int fwres;
|
| 237 |
|
|
__u16 rport;
|
| 238 |
|
|
#endif
|
| 239 |
|
|
#ifdef CONFIG_IP_MROUTE
|
| 240 |
|
|
int mroute_pkt=0;
|
| 241 |
|
|
#endif
|
| 242 |
|
|
|
| 243 |
|
|
#ifdef CONFIG_NET_IPV6
|
| 244 |
|
|
/*
|
| 245 |
|
|
* Intercept IPv6 frames. We dump ST-II and invalid types just below..
|
| 246 |
|
|
*/
|
| 247 |
|
|
|
| 248 |
|
|
if(iph->version == 6)
|
| 249 |
|
|
return ipv6_rcv(skb,dev,pt);
|
| 250 |
|
|
#endif
|
| 251 |
|
|
|
| 252 |
|
|
ip_statistics.IpInReceives++;
|
| 253 |
|
|
|
| 254 |
|
|
|
| 255 |
|
|
/*
|
| 256 |
|
|
* Tag the ip header of this packet so we can find it
|
| 257 |
|
|
*/
|
| 258 |
|
|
|
| 259 |
|
|
skb->ip_hdr = iph;
|
| 260 |
|
|
|
| 261 |
|
|
/*
|
| 262 |
|
|
* RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the checksum.
|
| 263 |
|
|
* RFC1122: 3.1.2.3 MUST discard a frame with invalid source address [NEEDS FIXING].
|
| 264 |
|
|
*
|
| 265 |
|
|
* Is the datagram acceptable?
|
| 266 |
|
|
*
|
| 267 |
|
|
* 1. Length at least the size of an ip header
|
| 268 |
|
|
* 2. Version of 4
|
| 269 |
|
|
* 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
|
| 270 |
|
|
* 4. Doesn't have a bogus length
|
| 271 |
|
|
* (5. We ought to check for IP multicast addresses and undefined types.. does this matter ?)
|
| 272 |
|
|
*/
|
| 273 |
|
|
|
| 274 |
|
|
if (skb->len<sizeof(struct iphdr) || iph->ihl<5 || iph->version != 4 || ip_fast_csum((unsigned char *)iph, iph->ihl) !=0
|
| 275 |
|
|
|| skb->len < ntohs(iph->tot_len))
|
| 276 |
|
|
{
|
| 277 |
|
|
ip_statistics.IpInHdrErrors++;
|
| 278 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 279 |
|
|
return(0);
|
| 280 |
|
|
}
|
| 281 |
|
|
|
| 282 |
|
|
/*
|
| 283 |
|
|
* Our transport medium may have padded the buffer out. Now we know it
|
| 284 |
|
|
* is IP we can trim to the true length of the frame.
|
| 285 |
|
|
* Note this now means skb->len holds ntohs(iph->tot_len).
|
| 286 |
|
|
*/
|
| 287 |
|
|
|
| 288 |
|
|
skb_trim(skb,ntohs(iph->tot_len));
|
| 289 |
|
|
|
| 290 |
|
|
if(skb->len < (iph->ihl<<2))
|
| 291 |
|
|
{
|
| 292 |
|
|
ip_statistics.IpInHdrErrors++;
|
| 293 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 294 |
|
|
return 0;
|
| 295 |
|
|
}
|
| 296 |
|
|
|
| 297 |
|
|
/*
|
| 298 |
|
|
* Account for the packet (even if the packet is
|
| 299 |
|
|
* not accepted by the firewall!). We do this after
|
| 300 |
|
|
* the sanity checks and the additional ihl check
|
| 301 |
|
|
* so we dont account garbage as we might do before.
|
| 302 |
|
|
*/
|
| 303 |
|
|
|
| 304 |
|
|
#ifdef CONFIG_IP_ACCT
|
| 305 |
|
|
ip_fw_chk(iph,dev,NULL,ip_acct_chain,0,IP_FW_MODE_ACCT_IN);
|
| 306 |
|
|
#endif
|
| 307 |
|
|
|
| 308 |
|
|
/*
|
| 309 |
|
|
* Try to select closest <src,dst> alias device, if any.
|
| 310 |
|
|
* net_alias_dev_rx32 returns main device if it
|
| 311 |
|
|
* fails to found other.
|
| 312 |
|
|
* If successful, also incr. alias rx count.
|
| 313 |
|
|
*
|
| 314 |
|
|
* Only makes sense for unicasts - Thanks ANK.
|
| 315 |
|
|
*/
|
| 316 |
|
|
|
| 317 |
|
|
#ifdef CONFIG_NET_ALIAS
|
| 318 |
|
|
if (skb->pkt_type == PACKET_HOST && iph->daddr != skb->dev->pa_addr && net_alias_has(skb->dev)) {
|
| 319 |
|
|
skb->dev = dev = net_alias_dev_rx32(skb->dev, AF_INET, iph->saddr, iph->daddr);
|
| 320 |
|
|
}
|
| 321 |
|
|
#endif
|
| 322 |
|
|
|
| 323 |
|
|
if (iph->ihl > 5)
|
| 324 |
|
|
{
|
| 325 |
|
|
skb->ip_summed = 0;
|
| 326 |
|
|
if (ip_options_compile(NULL, skb))
|
| 327 |
|
|
return(0);
|
| 328 |
|
|
opt = (struct options*)skb->proto_priv;
|
| 329 |
|
|
#ifdef CONFIG_IP_NOSR
|
| 330 |
|
|
if (opt->srr)
|
| 331 |
|
|
{
|
| 332 |
|
|
kfree_skb(skb, FREE_READ);
|
| 333 |
|
|
return -EINVAL;
|
| 334 |
|
|
}
|
| 335 |
|
|
#endif
|
| 336 |
|
|
}
|
| 337 |
|
|
|
| 338 |
|
|
#if defined(CONFIG_IP_TRANSPARENT_PROXY) && !defined(CONFIG_IP_ALWAYS_DEFRAG)
|
| 339 |
|
|
#define CONFIG_IP_ALWAYS_DEFRAG 1
|
| 340 |
|
|
#endif
|
| 341 |
|
|
#ifdef CONFIG_IP_ALWAYS_DEFRAG
|
| 342 |
|
|
/*
|
| 343 |
|
|
* Defragment all incoming traffic before even looking at it.
|
| 344 |
|
|
* If you have forwarding enabled, this makes the system a
|
| 345 |
|
|
* defragmenting router. Not a common thing.
|
| 346 |
|
|
* You probably DON'T want to enable this unless you have to.
|
| 347 |
|
|
* You NEED to use this if you want to use transparent proxying,
|
| 348 |
|
|
* otherwise, we can't vouch for your sanity.
|
| 349 |
|
|
*/
|
| 350 |
|
|
|
| 351 |
|
|
/*
|
| 352 |
|
|
* See if the frame is fragmented.
|
| 353 |
|
|
*/
|
| 354 |
|
|
|
| 355 |
|
|
if(iph->frag_off)
|
| 356 |
|
|
{
|
| 357 |
|
|
if (iph->frag_off & htons(IP_MF))
|
| 358 |
|
|
is_frag|=IPFWD_FRAGMENT;
|
| 359 |
|
|
/*
|
| 360 |
|
|
* Last fragment ?
|
| 361 |
|
|
*/
|
| 362 |
|
|
|
| 363 |
|
|
if (iph->frag_off & htons(IP_OFFSET))
|
| 364 |
|
|
is_frag|=IPFWD_LASTFRAG;
|
| 365 |
|
|
|
| 366 |
|
|
/*
|
| 367 |
|
|
* Reassemble IP fragments.
|
| 368 |
|
|
*/
|
| 369 |
|
|
|
| 370 |
|
|
if(is_frag)
|
| 371 |
|
|
{
|
| 372 |
|
|
/* Defragment. Obtain the complete packet if there is one */
|
| 373 |
|
|
skb=ip_defrag(iph,skb,dev);
|
| 374 |
|
|
if(skb==NULL)
|
| 375 |
|
|
return 0;
|
| 376 |
|
|
skb->dev = dev;
|
| 377 |
|
|
iph=skb->h.iph;
|
| 378 |
|
|
is_frag = 0;
|
| 379 |
|
|
/*
|
| 380 |
|
|
* When the reassembled packet gets forwarded, the ip
|
| 381 |
|
|
* header checksum should be correct.
|
| 382 |
|
|
* For better performance, this should actually only
|
| 383 |
|
|
* be done in that particular case, i.e. set a flag
|
| 384 |
|
|
* here and calculate the checksum in ip_forward.
|
| 385 |
|
|
*/
|
| 386 |
|
|
ip_send_check(iph);
|
| 387 |
|
|
}
|
| 388 |
|
|
}
|
| 389 |
|
|
|
| 390 |
|
|
#endif
|
| 391 |
|
|
/*
|
| 392 |
|
|
* See if the firewall wants to dispose of the packet.
|
| 393 |
|
|
*/
|
| 394 |
|
|
|
| 395 |
|
|
#ifdef CONFIG_FIREWALL
|
| 396 |
|
|
|
| 397 |
|
|
if ((fwres=call_in_firewall(PF_INET, skb->dev, iph, &rport))<FW_ACCEPT)
|
| 398 |
|
|
{
|
| 399 |
|
|
if(fwres==FW_REJECT)
|
| 400 |
|
|
icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0, dev);
|
| 401 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 402 |
|
|
return 0;
|
| 403 |
|
|
}
|
| 404 |
|
|
|
| 405 |
|
|
#ifdef CONFIG_IP_TRANSPARENT_PROXY
|
| 406 |
|
|
if (fwres==FW_REDIRECT)
|
| 407 |
|
|
skb->redirport = rport;
|
| 408 |
|
|
else
|
| 409 |
|
|
#endif
|
| 410 |
|
|
skb->redirport = 0;
|
| 411 |
|
|
#endif
|
| 412 |
|
|
|
| 413 |
|
|
#ifndef CONFIG_IP_ALWAYS_DEFRAG
|
| 414 |
|
|
/*
|
| 415 |
|
|
* Remember if the frame is fragmented.
|
| 416 |
|
|
*/
|
| 417 |
|
|
|
| 418 |
|
|
if(iph->frag_off)
|
| 419 |
|
|
{
|
| 420 |
|
|
if (iph->frag_off & htons(IP_MF))
|
| 421 |
|
|
is_frag|=IPFWD_FRAGMENT;
|
| 422 |
|
|
/*
|
| 423 |
|
|
* Last fragment ?
|
| 424 |
|
|
*/
|
| 425 |
|
|
|
| 426 |
|
|
if (iph->frag_off & htons(IP_OFFSET))
|
| 427 |
|
|
is_frag|=IPFWD_LASTFRAG;
|
| 428 |
|
|
}
|
| 429 |
|
|
|
| 430 |
|
|
#endif
|
| 431 |
|
|
/*
|
| 432 |
|
|
* Do any IP forwarding required. chk_addr() is expensive -- avoid it someday.
|
| 433 |
|
|
*
|
| 434 |
|
|
* This is inefficient. While finding out if it is for us we could also compute
|
| 435 |
|
|
* the routing table entry. This is where the great unified cache theory comes
|
| 436 |
|
|
* in as and when someone implements it
|
| 437 |
|
|
*
|
| 438 |
|
|
* For most hosts over 99% of packets match the first conditional
|
| 439 |
|
|
* and don't go via ip_chk_addr. Note: brd is set to IS_MYADDR at
|
| 440 |
|
|
* function entry.
|
| 441 |
|
|
*/
|
| 442 |
|
|
daddr = iph->daddr;
|
| 443 |
|
|
#ifdef CONFIG_IP_TRANSPARENT_PROXY
|
| 444 |
|
|
/*
|
| 445 |
|
|
* ip_chksock adds still more overhead for forwarded traffic...
|
| 446 |
|
|
*/
|
| 447 |
|
|
if ( iph->daddr == skb->dev->pa_addr || skb->redirport || (brd = ip_chk_addr(iph->daddr)) != 0 || ip_chksock(skb))
|
| 448 |
|
|
#else
|
| 449 |
|
|
if ( iph->daddr == skb->dev->pa_addr || (brd = ip_chk_addr(iph->daddr)) != 0)
|
| 450 |
|
|
#endif
|
| 451 |
|
|
{
|
| 452 |
|
|
if (opt && opt->srr)
|
| 453 |
|
|
{
|
| 454 |
|
|
int srrspace, srrptr;
|
| 455 |
|
|
__u32 nexthop;
|
| 456 |
|
|
unsigned char * optptr = ((unsigned char *)iph) + opt->srr;
|
| 457 |
|
|
|
| 458 |
|
|
if (brd != IS_MYADDR || skb->pkt_type != PACKET_HOST)
|
| 459 |
|
|
{
|
| 460 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 461 |
|
|
return 0;
|
| 462 |
|
|
}
|
| 463 |
|
|
|
| 464 |
|
|
for ( srrptr=optptr[2], srrspace = optptr[1];
|
| 465 |
|
|
srrptr <= srrspace;
|
| 466 |
|
|
srrptr += 4
|
| 467 |
|
|
)
|
| 468 |
|
|
{
|
| 469 |
|
|
int brd2;
|
| 470 |
|
|
if (srrptr + 3 > srrspace)
|
| 471 |
|
|
{
|
| 472 |
|
|
icmp_send(skb, ICMP_PARAMETERPROB, 0, opt->srr+2,
|
| 473 |
|
|
skb->dev);
|
| 474 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 475 |
|
|
return 0;
|
| 476 |
|
|
}
|
| 477 |
|
|
memcpy(&nexthop, &optptr[srrptr-1], 4);
|
| 478 |
|
|
if ((brd2 = ip_chk_addr(nexthop)) == 0)
|
| 479 |
|
|
break;
|
| 480 |
|
|
if (brd2 != IS_MYADDR)
|
| 481 |
|
|
{
|
| 482 |
|
|
|
| 483 |
|
|
/*
|
| 484 |
|
|
* ANK: should we implement weak tunneling of multicasts?
|
| 485 |
|
|
* Are they obsolete? DVMRP specs (RFC-1075) is old enough...
|
| 486 |
|
|
* [They are obsolete]
|
| 487 |
|
|
*/
|
| 488 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 489 |
|
|
return -EINVAL;
|
| 490 |
|
|
}
|
| 491 |
|
|
memcpy(&daddr, &optptr[srrptr-1], 4);
|
| 492 |
|
|
}
|
| 493 |
|
|
if (srrptr <= srrspace)
|
| 494 |
|
|
{
|
| 495 |
|
|
opt->srr_is_hit = 1;
|
| 496 |
|
|
opt->is_changed = 1;
|
| 497 |
|
|
if (sysctl_ip_forward) {
|
| 498 |
|
|
if (ip_forward(skb, dev, is_frag, nexthop))
|
| 499 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 500 |
|
|
} else {
|
| 501 |
|
|
ip_statistics.IpInAddrErrors++;
|
| 502 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 503 |
|
|
}
|
| 504 |
|
|
return 0;
|
| 505 |
|
|
}
|
| 506 |
|
|
}
|
| 507 |
|
|
|
| 508 |
|
|
#ifdef CONFIG_IP_MULTICAST
|
| 509 |
|
|
if(!(dev->flags&IFF_ALLMULTI) && brd==IS_MULTICAST && iph->daddr!=IGMP_ALL_HOSTS && !(dev->flags&IFF_LOOPBACK))
|
| 510 |
|
|
{
|
| 511 |
|
|
/*
|
| 512 |
|
|
* Check it is for one of our groups
|
| 513 |
|
|
*/
|
| 514 |
|
|
struct ip_mc_list *ip_mc=dev->ip_mc_list;
|
| 515 |
|
|
do
|
| 516 |
|
|
{
|
| 517 |
|
|
if(ip_mc==NULL)
|
| 518 |
|
|
{
|
| 519 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 520 |
|
|
return 0;
|
| 521 |
|
|
}
|
| 522 |
|
|
if(ip_mc->multiaddr==iph->daddr)
|
| 523 |
|
|
break;
|
| 524 |
|
|
ip_mc=ip_mc->next;
|
| 525 |
|
|
}
|
| 526 |
|
|
while(1);
|
| 527 |
|
|
}
|
| 528 |
|
|
#endif
|
| 529 |
|
|
|
| 530 |
|
|
#ifndef CONFIG_IP_ALWAYS_DEFRAG
|
| 531 |
|
|
/*
|
| 532 |
|
|
* Reassemble IP fragments.
|
| 533 |
|
|
*/
|
| 534 |
|
|
|
| 535 |
|
|
if(is_frag)
|
| 536 |
|
|
{
|
| 537 |
|
|
/* Defragment. Obtain the complete packet if there is one */
|
| 538 |
|
|
skb=ip_defrag(iph,skb,dev);
|
| 539 |
|
|
if(skb==NULL)
|
| 540 |
|
|
return 0;
|
| 541 |
|
|
skb->dev = dev;
|
| 542 |
|
|
iph=skb->h.iph;
|
| 543 |
|
|
}
|
| 544 |
|
|
|
| 545 |
|
|
#endif
|
| 546 |
|
|
|
| 547 |
|
|
#ifdef CONFIG_IP_MASQUERADE
|
| 548 |
|
|
/*
|
| 549 |
|
|
* Do we need to de-masquerade this packet?
|
| 550 |
|
|
*/
|
| 551 |
|
|
{
|
| 552 |
|
|
int ret = ip_fw_demasquerade(&skb,dev);
|
| 553 |
|
|
if (ret < 0) {
|
| 554 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 555 |
|
|
return 0;
|
| 556 |
|
|
}
|
| 557 |
|
|
|
| 558 |
|
|
if (ret)
|
| 559 |
|
|
{
|
| 560 |
|
|
struct iphdr *iph=skb->h.iph;
|
| 561 |
|
|
if (ip_forward(skb, dev, IPFWD_MASQUERADED, iph->daddr))
|
| 562 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 563 |
|
|
return 0;
|
| 564 |
|
|
}
|
| 565 |
|
|
}
|
| 566 |
|
|
#endif
|
| 567 |
|
|
|
| 568 |
|
|
/*
|
| 569 |
|
|
* Point into the IP datagram, just past the header.
|
| 570 |
|
|
*/
|
| 571 |
|
|
|
| 572 |
|
|
skb->ip_hdr = iph;
|
| 573 |
|
|
skb->h.raw += iph->ihl*4;
|
| 574 |
|
|
|
| 575 |
|
|
#ifdef CONFIG_IP_MROUTE
|
| 576 |
|
|
/*
|
| 577 |
|
|
* Check the state on multicast routing (multicast and not 224.0.0.z)
|
| 578 |
|
|
*/
|
| 579 |
|
|
|
| 580 |
|
|
if(brd==IS_MULTICAST && (iph->daddr&htonl(0xFFFFFF00))!=htonl(0xE0000000))
|
| 581 |
|
|
mroute_pkt=1;
|
| 582 |
|
|
|
| 583 |
|
|
#endif
|
| 584 |
|
|
/*
|
| 585 |
|
|
* Deliver to raw sockets. This is fun as to avoid copies we want to make no surplus copies.
|
| 586 |
|
|
*
|
| 587 |
|
|
* RFC 1122: SHOULD pass TOS value up to the transport layer.
|
| 588 |
|
|
*/
|
| 589 |
|
|
|
| 590 |
|
|
/* Note: See raw.c and net/raw.h, RAWV4_HTABLE_SIZE==MAX_INET_PROTOS */
|
| 591 |
|
|
hash = iph->protocol & (MAX_INET_PROTOS - 1);
|
| 592 |
|
|
|
| 593 |
|
|
/*
|
| 594 |
|
|
* If there maybe a raw socket we must check - if not we don't care less
|
| 595 |
|
|
*/
|
| 596 |
|
|
|
| 597 |
|
|
if((raw_sk = raw_v4_htable[hash]) != NULL) {
|
| 598 |
|
|
struct sock *sknext = NULL;
|
| 599 |
|
|
struct sk_buff *skb1;
|
| 600 |
|
|
|
| 601 |
|
|
raw_sk = raw_v4_lookup(raw_sk, iph->protocol,
|
| 602 |
|
|
iph->saddr, iph->daddr);
|
| 603 |
|
|
if(raw_sk) { /* Any raw sockets */
|
| 604 |
|
|
do {
|
| 605 |
|
|
/* Find the next */
|
| 606 |
|
|
sknext = raw_v4_lookup(raw_sk->next,
|
| 607 |
|
|
iph->protocol,
|
| 608 |
|
|
iph->saddr,
|
| 609 |
|
|
iph->daddr);
|
| 610 |
|
|
if(sknext)
|
| 611 |
|
|
skb1 = skb_clone(skb, GFP_ATOMIC);
|
| 612 |
|
|
else
|
| 613 |
|
|
break; /* One pending raw socket left */
|
| 614 |
|
|
if(skb1)
|
| 615 |
|
|
raw_rcv(raw_sk, skb1, dev, iph->saddr,daddr);
|
| 616 |
|
|
raw_sk = sknext;
|
| 617 |
|
|
} while(raw_sk!=NULL);
|
| 618 |
|
|
|
| 619 |
|
|
/*
|
| 620 |
|
|
* Here either raw_sk is the last raw socket, or NULL if none
|
| 621 |
|
|
*/
|
| 622 |
|
|
|
| 623 |
|
|
/*
|
| 624 |
|
|
* We deliver to the last raw socket AFTER the protocol checks as it avoids a surplus copy
|
| 625 |
|
|
*/
|
| 626 |
|
|
}
|
| 627 |
|
|
}
|
| 628 |
|
|
|
| 629 |
|
|
/*
|
| 630 |
|
|
* skb->h.raw now points at the protocol beyond the IP header.
|
| 631 |
|
|
*/
|
| 632 |
|
|
|
| 633 |
|
|
for (ipprot = (struct inet_protocol *)inet_protos[hash];ipprot != NULL;ipprot=(struct inet_protocol *)ipprot->next)
|
| 634 |
|
|
{
|
| 635 |
|
|
struct sk_buff *skb2;
|
| 636 |
|
|
|
| 637 |
|
|
if (ipprot->protocol != iph->protocol)
|
| 638 |
|
|
continue;
|
| 639 |
|
|
/*
|
| 640 |
|
|
* See if we need to make a copy of it. This will
|
| 641 |
|
|
* only be set if more than one protocol wants it.
|
| 642 |
|
|
* and then not for the last one. If there is a pending
|
| 643 |
|
|
* raw delivery wait for that
|
| 644 |
|
|
*/
|
| 645 |
|
|
|
| 646 |
|
|
#ifdef CONFIG_IP_MROUTE
|
| 647 |
|
|
if (ipprot->copy || raw_sk || mroute_pkt)
|
| 648 |
|
|
#else
|
| 649 |
|
|
if (ipprot->copy || raw_sk)
|
| 650 |
|
|
#endif
|
| 651 |
|
|
{
|
| 652 |
|
|
skb2 = skb_clone(skb, GFP_ATOMIC);
|
| 653 |
|
|
if(skb2==NULL)
|
| 654 |
|
|
continue;
|
| 655 |
|
|
}
|
| 656 |
|
|
else
|
| 657 |
|
|
{
|
| 658 |
|
|
skb2 = skb;
|
| 659 |
|
|
}
|
| 660 |
|
|
flag = 1;
|
| 661 |
|
|
|
| 662 |
|
|
/*
|
| 663 |
|
|
* Pass on the datagram to each protocol that wants it,
|
| 664 |
|
|
* based on the datagram protocol. We should really
|
| 665 |
|
|
* check the protocol handler's return values here...
|
| 666 |
|
|
*/
|
| 667 |
|
|
|
| 668 |
|
|
ipprot->handler(skb2, dev, opt, daddr,
|
| 669 |
|
|
(ntohs(iph->tot_len) - (iph->ihl * 4)),
|
| 670 |
|
|
iph->saddr, 0, ipprot);
|
| 671 |
|
|
}
|
| 672 |
|
|
|
| 673 |
|
|
/*
|
| 674 |
|
|
* All protocols checked.
|
| 675 |
|
|
* If this packet was a broadcast, we may *not* reply to it, since that
|
| 676 |
|
|
* causes (proven, grin) ARP storms and a leakage of memory (i.e. all
|
| 677 |
|
|
* ICMP reply messages get queued up for transmission...)
|
| 678 |
|
|
*/
|
| 679 |
|
|
|
| 680 |
|
|
#ifdef CONFIG_IP_MROUTE
|
| 681 |
|
|
/*
|
| 682 |
|
|
* Forward the last copy to the multicast router. If
|
| 683 |
|
|
* there is a pending raw delivery however make a copy
|
| 684 |
|
|
* and forward that.
|
| 685 |
|
|
*/
|
| 686 |
|
|
|
| 687 |
|
|
if(mroute_pkt)
|
| 688 |
|
|
{
|
| 689 |
|
|
flag=1;
|
| 690 |
|
|
if(raw_sk==NULL)
|
| 691 |
|
|
ipmr_forward(skb, is_frag);
|
| 692 |
|
|
else
|
| 693 |
|
|
{
|
| 694 |
|
|
struct sk_buff *skb2=skb_clone(skb, GFP_ATOMIC);
|
| 695 |
|
|
if(skb2)
|
| 696 |
|
|
{
|
| 697 |
|
|
skb2->free=1;
|
| 698 |
|
|
ipmr_forward(skb2, is_frag);
|
| 699 |
|
|
}
|
| 700 |
|
|
}
|
| 701 |
|
|
}
|
| 702 |
|
|
#endif
|
| 703 |
|
|
|
| 704 |
|
|
if(raw_sk!=NULL) /* Shift to last raw user */
|
| 705 |
|
|
raw_rcv(raw_sk, skb, dev, iph->saddr, daddr);
|
| 706 |
|
|
else if (!flag) /* Free and report errors */
|
| 707 |
|
|
{
|
| 708 |
|
|
if (brd != IS_BROADCAST && brd!=IS_MULTICAST)
|
| 709 |
|
|
icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PROT_UNREACH, 0, dev);
|
| 710 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 711 |
|
|
}
|
| 712 |
|
|
|
| 713 |
|
|
return(0);
|
| 714 |
|
|
}
|
| 715 |
|
|
|
| 716 |
|
|
/*
|
| 717 |
|
|
* Do any unicast IP forwarding required.
|
| 718 |
|
|
*/
|
| 719 |
|
|
|
| 720 |
|
|
/*
|
| 721 |
|
|
* Don't forward multicast or broadcast frames.
|
| 722 |
|
|
*/
|
| 723 |
|
|
|
| 724 |
|
|
if(skb->pkt_type!=PACKET_HOST || brd==IS_BROADCAST)
|
| 725 |
|
|
{
|
| 726 |
|
|
kfree_skb(skb,FREE_WRITE);
|
| 727 |
|
|
return 0;
|
| 728 |
|
|
}
|
| 729 |
|
|
|
| 730 |
|
|
/*
|
| 731 |
|
|
* The packet is for another target. Forward the frame
|
| 732 |
|
|
*/
|
| 733 |
|
|
|
| 734 |
|
|
if (sysctl_ip_forward) {
|
| 735 |
|
|
if (opt && opt->is_strictroute)
|
| 736 |
|
|
{
|
| 737 |
|
|
icmp_send(skb, ICMP_PARAMETERPROB, 0, 16, skb->dev);
|
| 738 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 739 |
|
|
return -1;
|
| 740 |
|
|
}
|
| 741 |
|
|
if (ip_forward(skb, dev, is_frag, iph->daddr))
|
| 742 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 743 |
|
|
} else {
|
| 744 |
|
|
/* printk("Machine %lx tried to use us as a forwarder to %lx but we have forwarding disabled!\n",
|
| 745 |
|
|
iph->saddr,iph->daddr);*/
|
| 746 |
|
|
ip_statistics.IpInAddrErrors++;
|
| 747 |
|
|
kfree_skb(skb, FREE_WRITE);
|
| 748 |
|
|
}
|
| 749 |
|
|
return(0);
|
| 750 |
|
|
}
|
| 751 |
|
|
|
| 752 |
|
|
|
