KEIS
Fork of lwip by
Diff: core/ipv4/icmp.c
- Revision:
- 0:51ac1d130fd4
diff -r 000000000000 -r 51ac1d130fd4 core/ipv4/icmp.c --- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/core/ipv4/icmp.c Fri Jun 22 09:25:39 2012 +0000 @@ -0,0 +1,335 @@ +/** + * @file + * ICMP - Internet Control Message Protocol + * + */ + +/* + * Copyright (c) 2001-2004 Swedish Institute of Computer Science. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without modification, + * are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright notice, + * this list of conditions and the following disclaimer in the documentation + * and/or other materials provided with the distribution. + * 3. The name of the author may not be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED + * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT + * SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT + * OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING + * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY + * OF SUCH DAMAGE. + * + * This file is part of the lwIP TCP/IP stack. + * + * Author: Adam Dunkels <adam@sics.se> + * + */ + +/* Some ICMP messages should be passed to the transport protocols. This + is not implemented. */ + +#include "lwip/opt.h" + +#if LWIP_ICMP /* don't build if not configured for use in lwipopts.h */ + +#include "lwip/icmp.h" +#include "lwip/inet_chksum.h" +#include "lwip/ip.h" +#include "lwip/def.h" +#include "lwip/stats.h" +#include "lwip/snmp.h" + +#include <string.h> + +/** Small optimization: set to 0 if incoming PBUF_POOL pbuf always can be + * used to modify and send a response packet (and to 1 if this is not the case, + * e.g. when link header is stripped of when receiving) */ +#ifndef LWIP_ICMP_ECHO_CHECK_INPUT_PBUF_LEN +#define LWIP_ICMP_ECHO_CHECK_INPUT_PBUF_LEN 1 +#endif /* LWIP_ICMP_ECHO_CHECK_INPUT_PBUF_LEN */ + +/* The amount of data from the original packet to return in a dest-unreachable */ +#define ICMP_DEST_UNREACH_DATASIZE 8 + +static void icmp_send_response(struct pbuf *p, u8_t type, u8_t code); + +/** + * Processes ICMP input packets, called from ip_input(). + * + * Currently only processes icmp echo requests and sends + * out the echo response. + * + * @param p the icmp echo request packet, p->payload pointing to the ip header + * @param inp the netif on which this packet was received + */ +void +icmp_input(struct pbuf *p, struct netif *inp) +{ + u8_t type; +#ifdef LWIP_DEBUG + u8_t code; +#endif /* LWIP_DEBUG */ + struct icmp_echo_hdr *iecho; + struct ip_hdr *iphdr; + s16_t hlen; + + ICMP_STATS_INC(icmp.recv); + snmp_inc_icmpinmsgs(); + + + iphdr = (struct ip_hdr *)p->payload; + hlen = IPH_HL(iphdr) * 4; + if (pbuf_header(p, -hlen) || (p->tot_len < sizeof(u16_t)*2)) { + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_input: short ICMP (%"U16_F" bytes) received\n", p->tot_len)); + goto lenerr; + } + + type = *((u8_t *)p->payload); +#ifdef LWIP_DEBUG + code = *(((u8_t *)p->payload)+1); +#endif /* LWIP_DEBUG */ + switch (type) { + case ICMP_ER: + /* This is OK, echo reply might have been parsed by a raw PCB + (as obviously, an echo request has been sent, too). */ + break; + case ICMP_ECHO: +#if !LWIP_MULTICAST_PING || !LWIP_BROADCAST_PING + { + int accepted = 1; +#if !LWIP_MULTICAST_PING + /* multicast destination address? */ + if (ip_addr_ismulticast(¤t_iphdr_dest)) { + accepted = 0; + } +#endif /* LWIP_MULTICAST_PING */ +#if !LWIP_BROADCAST_PING + /* broadcast destination address? */ + if (ip_addr_isbroadcast(¤t_iphdr_dest, inp)) { + accepted = 0; + } +#endif /* LWIP_BROADCAST_PING */ + /* broadcast or multicast destination address not acceptd? */ + if (!accepted) { + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_input: Not echoing to multicast or broadcast pings\n")); + ICMP_STATS_INC(icmp.err); + pbuf_free(p); + return; + } + } +#endif /* !LWIP_MULTICAST_PING || !LWIP_BROADCAST_PING */ + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_input: ping\n")); + if (p->tot_len < sizeof(struct icmp_echo_hdr)) { + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_input: bad ICMP echo received\n")); + goto lenerr; + } + if (inet_chksum_pbuf(p) != 0) { + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_input: checksum failed for received ICMP echo\n")); + pbuf_free(p); + ICMP_STATS_INC(icmp.chkerr); + snmp_inc_icmpinerrors(); + return; + } +#if LWIP_ICMP_ECHO_CHECK_INPUT_PBUF_LEN + if (pbuf_header(p, (PBUF_IP_HLEN + PBUF_LINK_HLEN))) { + /* p is not big enough to contain link headers + * allocate a new one and copy p into it + */ + struct pbuf *r; + /* switch p->payload to ip header */ + if (pbuf_header(p, hlen)) { + LWIP_ASSERT("icmp_input: moving p->payload to ip header failed\n", 0); + goto memerr; + } + /* allocate new packet buffer with space for link headers */ + r = pbuf_alloc(PBUF_LINK, p->tot_len, PBUF_RAM); + if (r == NULL) { + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_input: allocating new pbuf failed\n")); + goto memerr; + } + LWIP_ASSERT("check that first pbuf can hold struct the ICMP header", + (r->len >= hlen + sizeof(struct icmp_echo_hdr))); + /* copy the whole packet including ip header */ + if (pbuf_copy(r, p) != ERR_OK) { + LWIP_ASSERT("icmp_input: copying to new pbuf failed\n", 0); + goto memerr; + } + iphdr = (struct ip_hdr *)r->payload; + /* switch r->payload back to icmp header */ + if (pbuf_header(r, -hlen)) { + LWIP_ASSERT("icmp_input: restoring original p->payload failed\n", 0); + goto memerr; + } + /* free the original p */ + pbuf_free(p); + /* we now have an identical copy of p that has room for link headers */ + p = r; + } else { + /* restore p->payload to point to icmp header */ + if (pbuf_header(p, -(s16_t)(PBUF_IP_HLEN + PBUF_LINK_HLEN))) { + LWIP_ASSERT("icmp_input: restoring original p->payload failed\n", 0); + goto memerr; + } + } +#endif /* LWIP_ICMP_ECHO_CHECK_INPUT_PBUF_LEN */ + /* At this point, all checks are OK. */ + /* We generate an answer by switching the dest and src ip addresses, + * setting the icmp type to ECHO_RESPONSE and updating the checksum. */ + iecho = (struct icmp_echo_hdr *)p->payload; + ip_addr_copy(iphdr->src, *ip_current_dest_addr()); + ip_addr_copy(iphdr->dest, *ip_current_src_addr()); + ICMPH_TYPE_SET(iecho, ICMP_ER); + /* adjust the checksum */ + if (iecho->chksum >= PP_HTONS(0xffffU - (ICMP_ECHO << 8))) { + iecho->chksum += PP_HTONS(ICMP_ECHO << 8) + 1; + } else { + iecho->chksum += PP_HTONS(ICMP_ECHO << 8); + } + + /* Set the correct TTL and recalculate the header checksum. */ + IPH_TTL_SET(iphdr, ICMP_TTL); + IPH_CHKSUM_SET(iphdr, 0); +#if CHECKSUM_GEN_IP + IPH_CHKSUM_SET(iphdr, inet_chksum(iphdr, IP_HLEN)); +#endif /* CHECKSUM_GEN_IP */ + + ICMP_STATS_INC(icmp.xmit); + /* increase number of messages attempted to send */ + snmp_inc_icmpoutmsgs(); + /* increase number of echo replies attempted to send */ + snmp_inc_icmpoutechoreps(); + + if(pbuf_header(p, hlen)) { + LWIP_ASSERT("Can't move over header in packet", 0); + } else { + err_t ret; + /* send an ICMP packet, src addr is the dest addr of the curren packet */ + ret = ip_output_if(p, ip_current_dest_addr(), IP_HDRINCL, + ICMP_TTL, 0, IP_PROTO_ICMP, inp); + if (ret != ERR_OK) { + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_input: ip_output_if returned an error: %c.\n", ret)); + } + } + break; + default: + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_input: ICMP type %"S16_F" code %"S16_F" not supported.\n", + (s16_t)type, (s16_t)code)); + ICMP_STATS_INC(icmp.proterr); + ICMP_STATS_INC(icmp.drop); + } + pbuf_free(p); + return; +lenerr: + pbuf_free(p); + ICMP_STATS_INC(icmp.lenerr); + snmp_inc_icmpinerrors(); + return; +#if LWIP_ICMP_ECHO_CHECK_INPUT_PBUF_LEN +memerr: + pbuf_free(p); + ICMP_STATS_INC(icmp.err); + snmp_inc_icmpinerrors(); + return; +#endif /* LWIP_ICMP_ECHO_CHECK_INPUT_PBUF_LEN */ +} + +/** + * Send an icmp 'destination unreachable' packet, called from ip_input() if + * the transport layer protocol is unknown and from udp_input() if the local + * port is not bound. + * + * @param p the input packet for which the 'unreachable' should be sent, + * p->payload pointing to the IP header + * @param t type of the 'unreachable' packet + */ +void +icmp_dest_unreach(struct pbuf *p, enum icmp_dur_type t) +{ + icmp_send_response(p, ICMP_DUR, t); +} + +#if IP_FORWARD || IP_REASSEMBLY +/** + * Send a 'time exceeded' packet, called from ip_forward() if TTL is 0. + * + * @param p the input packet for which the 'time exceeded' should be sent, + * p->payload pointing to the IP header + * @param t type of the 'time exceeded' packet + */ +void +icmp_time_exceeded(struct pbuf *p, enum icmp_te_type t) +{ + icmp_send_response(p, ICMP_TE, t); +} + +#endif /* IP_FORWARD || IP_REASSEMBLY */ + +/** + * Send an icmp packet in response to an incoming packet. + * + * @param p the input packet for which the 'unreachable' should be sent, + * p->payload pointing to the IP header + * @param type Type of the ICMP header + * @param code Code of the ICMP header + */ +static void +icmp_send_response(struct pbuf *p, u8_t type, u8_t code) +{ + struct pbuf *q; + struct ip_hdr *iphdr; + /* we can use the echo header here */ + struct icmp_echo_hdr *icmphdr; + ip_addr_t iphdr_src; + + /* ICMP header + IP header + 8 bytes of data */ + q = pbuf_alloc(PBUF_IP, sizeof(struct icmp_echo_hdr) + IP_HLEN + ICMP_DEST_UNREACH_DATASIZE, + PBUF_RAM); + if (q == NULL) { + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_time_exceeded: failed to allocate pbuf for ICMP packet.\n")); + return; + } + LWIP_ASSERT("check that first pbuf can hold icmp message", + (q->len >= (sizeof(struct icmp_echo_hdr) + IP_HLEN + ICMP_DEST_UNREACH_DATASIZE))); + + iphdr = (struct ip_hdr *)p->payload; + LWIP_DEBUGF(ICMP_DEBUG, ("icmp_time_exceeded from ")); + ip_addr_debug_print(ICMP_DEBUG, &(iphdr->src)); + LWIP_DEBUGF(ICMP_DEBUG, (" to ")); + ip_addr_debug_print(ICMP_DEBUG, &(iphdr->dest)); + LWIP_DEBUGF(ICMP_DEBUG, ("\n")); + + icmphdr = (struct icmp_echo_hdr *)q->payload; + icmphdr->type = type; + icmphdr->code = code; + icmphdr->id = 0; + icmphdr->seqno = 0; + + /* copy fields from original packet */ + SMEMCPY((u8_t *)q->payload + sizeof(struct icmp_echo_hdr), (u8_t *)p->payload, + IP_HLEN + ICMP_DEST_UNREACH_DATASIZE); + + /* calculate checksum */ + icmphdr->chksum = 0; + icmphdr->chksum = inet_chksum(icmphdr, q->len); + ICMP_STATS_INC(icmp.xmit); + /* increase number of messages attempted to send */ + snmp_inc_icmpoutmsgs(); + /* increase number of destination unreachable messages attempted to send */ + snmp_inc_icmpouttimeexcds(); + ip_addr_copy(iphdr_src, iphdr->src); + ip_output(q, NULL, &iphdr_src, ICMP_TTL, 0, IP_PROTO_ICMP); + pbuf_free(q); +} + +#endif /* LWIP_ICMP */