Ping/ICMP example

I have tried this on my archlinux i686 and works, on arch x86_64 segfaults and I don't no why, I need to work more on this.

/*
 * ping.c
 *
 * An ping example. I have used some iputils code
 * you can get the iputils source here: http://www.skbuff.net/iputils/
 *
 */

#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/socket.h>
#include <sys/types.h>
#include <netinet/in.h>
#include <netinet/ip.h>
#include <netdb.h>
#include <netinet/ip_icmp.h>


#define pexit(s) ({perror(s); exit(EXIT_FAILURE);})

u_short in_cksum(const u_short *addr, register int len, u_short csum);
void print_icmphdr(struct icmphdr *);
void print_iphdr(struct iphdr *);

int main(int argc, char **argv)
{
        int sock;
        int len;
        int bytes;
        int count = -1;
        u_short cksum;
        u_int16_t seq; 

        struct sockaddr_in dst_addr;
        struct sockaddr_in rcv_addr;
        struct hostent *dst_host;

#define BUFLEN 1000000
        char outpack[BUFLEN];
        struct icmphdr *icp; 
        struct iphdr *ip;

        if (argc <= 1) {
                printf("Usage: %s HOST [COUNT]");
                exit(EXIT_FAILURE);
        }



        sock = socket(AF_INET, SOCK_RAW, IPPROTO_ICMP);
        if (sock == -1)
                pexit("socket");
        
        dst_host = gethostbyname(argv[1]);
        if (!dst_host) {
                errno = h_errno;
                pexit("gethostbyname");
        }

        memcpy(&dst_addr.sin_addr.s_addr, dst_host->h_addr_list[0], 
                        sizeof(dst_addr));
        dst_addr.sin_family = PF_INET;
        dst_addr.sin_port = 0;

        if (argc > 2 )
                count = atoi(argv[2]);

        seq = 1;
        while (count--) {
                icp = (struct icmphdr *)outpack;
                icp->type = ICMP_ECHO;
                icp->code = 0;
                icp->un.echo.sequence = seq;
                icp->un.echo.id = getpid(); 
                icp->checksum = 0;
                icp->checksum = in_cksum((u_short *)icp, 
                                sizeof(struct icmphdr), 0);

                bytes = sendto(sock, outpack, sizeof(struct icmphdr),
                                MSG_DONTWAIT, (struct sockaddr *)&dst_addr,
                                sizeof(dst_addr));
                if (bytes < 0)
                        pexit("sendto");

                sleep(1);

                len = sizeof(struct sockaddr_in);
                bytes = recvfrom(sock, outpack, sizeof(struct iphdr) +
                                sizeof(struct icmphdr), MSG_DONTWAIT,
                                (struct sockaddr *)&rcv_addr, &len);
                if (bytes < 0) /* I'm ignoring incoming errors */
                        continue;

                ip = (struct iphdr *)outpack;
                icp = (struct icmphdr *)&outpack[sizeof(struct iphdr)];

                cksum = icp->checksum;
                icp->checksum = 0;
                icp->checksum = in_cksum((u_short *)icp, 
                                sizeof(struct icmphdr), 0);

                if (cksum != icp->checksum) /* and ignoring  */
                        continue;           /* corrupted packets */

                switch(icp->type) {
                case ICMP_ECHOREPLY: /* and repeateds */
                        if (icp->un.echo.sequence < seq)
                                continue;
                        print_iphdr(ip);
                        print_icmphdr(icp);
                        putchar('\n');
                        seq++;
                        break;
                case ICMP_DEST_UNREACH:
                        printf("Destination unreachable\n");
                        break;
                }

        } 

        return 0;
}


/*
 * Taken from iputils/ping.c, at http://www.skbuff.net/iputils/
 */
u_short in_cksum(const u_short *addr, register int len, u_short csum)
{
 register int nleft = len;
 const u_short *w = addr;
 register u_short answer;
 register int sum = csum;

 /*
  *  Our algorithm is simple, using a 32 bit accumulator (sum),
  *  we add sequential 16 bit words to it, and at the end, fold
  *  back all the carry bits from the top 16 bits into the lower
  *  16 bits.
  */
 while (nleft > 1)  {
  sum += *w++;
  nleft -= 2;
 }

 /* mop up an odd byte, if necessary */
 if (nleft == 1)
  sum += htons(*(u_char *)w << 8);

 /*
  * add back carry outs from top 16 bits to low 16 bits
  */
 sum = (sum >> 16) + (sum & 0xffff); /* add hi 16 to low 16 */
 sum += (sum >> 16);   /* add carry */
 answer = ~sum;    /* truncate to 16 bits */
 return (answer);
}

void print_iphdr(struct iphdr *ip)
{
        printf("IP tos=%u id=%u ttl=%u saddr=%s daddr=%s ",
                       ip->tos, ip->id, ip->ttl, inet_ntoa(ip->saddr),
                       inet_ntoa(ip->daddr));
}
void print_icmphdr(struct icmphdr *icp)
{
        printf("ICMP seq=%d ", icp->un.echo.sequence);
}