src/slirp/socket.c

/*
 * Copyright (c) 1995 Danny Gasparovski.
 * 
 * Please read the file COPYRIGHT for the 
 * terms and conditions of the copyright.
 */

#define WANT_SYS_IOCTL_H
#include <slirp.h>
#include "ip_icmp.h"
#include "main.h"

void
so_init()
{
        /* Nothing yet */
}


struct socket *
solookup(head, laddr, lport, faddr, fport)
        struct socket *head;
        struct in_addr laddr;
        u_int lport;
        struct in_addr faddr;
        u_int fport;
{
        struct socket *so;
        
        for (so = head->so_next; so != head; so = so->so_next) {
                if (so->so_lport == lport && 
                    so->so_laddr.s_addr == laddr.s_addr &&
                    so->so_faddr.s_addr == faddr.s_addr &&
                    so->so_fport == fport)
                   break;
        }
        
        if (so == head)
           return (struct socket *)NULL;
        return so;
        
}

/*
 * Create a new socket, initialise the fields
 * It is the responsibility of the caller to
 * insque() it into the correct linked-list
 */
struct socket *
socreate()
{
  struct socket *so;
        
  so = (struct socket *)malloc(sizeof(struct socket));
  if(so) {
    memset(so, 0, sizeof(struct socket));
    so->so_state = SS_NOFDREF;
    so->s = -1;
  }
  return(so);
}

/*
 * remque and free a socket, clobber cache
 */
void
sofree(so)
        struct socket *so;
{
  if (so->so_emu==EMU_RSH && so->extra) {
        sofree(so->extra);
        so->extra=NULL;
  }
  if (so == tcp_last_so)
    tcp_last_so = &tcb;
  else if (so == udp_last_so)
    udp_last_so = &udb;
        
  m_free(so->so_m);
        
  if(so->so_next && so->so_prev) 
    remque(so);  /* crashes if so is not in a queue */

  free(so);
}

/*
 * Read from so's socket into sb_snd, updating all relevant sbuf fields
 * NOTE: This will only be called if it is select()ed for reading, so
 * a read() of 0 (or less) means it's disconnected
 */
int
soread(so)
        struct socket *so;
{
        int n, nn, lss, total;
        struct sbuf *sb = &so->so_snd;
        int len = sb->sb_datalen - sb->sb_cc;
        struct iovec iov[2];
        int mss = so->so_tcpcb->t_maxseg;
        
        DEBUG_CALL("soread");
        DEBUG_ARG("so = %lx", (long )so);
        
        /* 
         * No need to check if there's enough room to read.
         * soread wouldn't have been called if there weren't
         */
        
        len = sb->sb_datalen - sb->sb_cc;
        
        iov[0].iov_base = sb->sb_wptr;
        if (sb->sb_wptr < sb->sb_rptr) {
                iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
                /* Should never succeed, but... */
                if (iov[0].iov_len > len)
                   iov[0].iov_len = len;
                if (iov[0].iov_len > mss)
                   iov[0].iov_len -= iov[0].iov_len%mss;
                n = 1;
        } else {
                iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
                /* Should never succeed, but... */
                if (iov[0].iov_len > len) iov[0].iov_len = len;
                len -= iov[0].iov_len;
                if (len) {
                        iov[1].iov_base = sb->sb_data;
                        iov[1].iov_len = sb->sb_rptr - sb->sb_data;
                        if(iov[1].iov_len > len)
                           iov[1].iov_len = len;
                        total = iov[0].iov_len + iov[1].iov_len;
                        if (total > mss) {
                                lss = total%mss;
                                if (iov[1].iov_len > lss) {
                                        iov[1].iov_len -= lss;
                                        n = 2;
                                } else {
                                        lss -= iov[1].iov_len;
                                        iov[0].iov_len -= lss;
                                        n = 1;
                                }
                        } else
                                n = 2;
                } else {
                        if (iov[0].iov_len > mss)
                           iov[0].iov_len -= iov[0].iov_len%mss;
                        n = 1;
                }
        }
        
#ifdef HAVE_READV
        nn = readv(so->s, (struct iovec *)iov, n);
        DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
#else
        nn = recv(so->s, iov[0].iov_base, iov[0].iov_len,0);
#endif  
        if (nn <= 0) {
                if (nn < 0 && (errno == EINTR || errno == EAGAIN))
                        return 0;
                else {
                        DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n", nn, errno,strerror(errno)));
                        sofcantrcvmore(so);
                        tcp_sockclosed(sototcpcb(so));
                        return -1;
                }
        }
        
#ifndef HAVE_READV
        /*
         * If there was no error, try and read the second time round
         * We read again if n = 2 (ie, there's another part of the buffer)
         * and we read as much as we could in the first read
         * We don't test for <= 0 this time, because there legitimately
         * might not be any more data (since the socket is non-blocking),
         * a close will be detected on next iteration.
         * A return of -1 wont (shouldn't) happen, since it didn't happen above
         */
        if (n == 2 && nn == iov[0].iov_len) {
            int ret;
            ret = recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
            if (ret > 0)
                nn += ret;
        }
        
        DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
#endif
        
        /* Update fields */
        sb->sb_cc += nn;
        sb->sb_wptr += nn;
        if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
                sb->sb_wptr -= sb->sb_datalen;
        return nn;
}
        
/*
 * Get urgent data
 * 
 * When the socket is created, we set it SO_OOBINLINE,
 * so when OOB data arrives, we soread() it and everything
 * in the send buffer is sent as urgent data
 */
void
sorecvoob(so)
        struct socket *so;
{
        struct tcpcb *tp = sototcpcb(so);

        DEBUG_CALL("sorecvoob");
        DEBUG_ARG("so = %lx", (long)so);
        
        /*
         * We take a guess at how much urgent data has arrived.
         * In most situations, when urgent data arrives, the next
         * read() should get all the urgent data.  This guess will
         * be wrong however if more data arrives just after the
         * urgent data, or the read() doesn't return all the 
         * urgent data.
         */
        soread(so);
        tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
        tp->t_force = 1;
        tcp_output(tp);
        tp->t_force = 0;
}

/*
 * Send urgent data
 * There's a lot duplicated code here, but...
 */
int
sosendoob(so)
        struct socket *so;
{
        struct sbuf *sb = &so->so_rcv;
        char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
        
        int n, len;
        
        DEBUG_CALL("sosendoob");
        DEBUG_ARG("so = %lx", (long)so);
        DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
        
        if (so->so_urgc > 2048)
           so->so_urgc = 2048; /* XXXX */
        
        if (sb->sb_rptr < sb->sb_wptr) {
                /* We can send it directly */
                n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* |MSG_DONTWAIT)); */
                so->so_urgc -= n;
                
                DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
        } else {
                /* 
                 * Since there's no sendv or sendtov like writev,
                 * we must copy all data to a linear buffer then
                 * send it all
                 */
                len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
                if (len > so->so_urgc) len = so->so_urgc;
                memcpy(buff, sb->sb_rptr, len);
                so->so_urgc -= len;
                if (so->so_urgc) {
                        n = sb->sb_wptr - sb->sb_data;
                        if (n > so->so_urgc) n = so->so_urgc;
                        memcpy((buff + len), sb->sb_data, n);
                        so->so_urgc -= n;
                        len += n;
                }
                n = send(so->s, buff, len, (MSG_OOB)); /* |MSG_DONTWAIT)); */
#ifdef DEBUG
                if (n != len)
                   DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
#endif          
                DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
        }
        
        sb->sb_cc -= n;
        sb->sb_rptr += n;
        if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
                sb->sb_rptr -= sb->sb_datalen;
        
        return n;
}

/*
 * Write data from so_rcv to so's socket, 
 * updating all sbuf field as necessary
 */
int
sowrite(so)
        struct socket *so;
{
        int  n,nn;
        struct sbuf *sb = &so->so_rcv;
        int len = sb->sb_cc;
        struct iovec iov[2];
        
        DEBUG_CALL("sowrite");
        DEBUG_ARG("so = %lx", (long)so);
        
        if (so->so_urgc) {
                sosendoob(so);
                if (sb->sb_cc == 0)
                        return 0;
        }

        /*
         * No need to check if there's something to write,
         * sowrite wouldn't have been called otherwise
         */
        
        len = sb->sb_cc;
        
        iov[0].iov_base = sb->sb_rptr;
        if (sb->sb_rptr < sb->sb_wptr) {
                iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
                /* Should never succeed, but... */
                if (iov[0].iov_len > len) iov[0].iov_len = len;
                n = 1;
        } else {
                iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
                if (iov[0].iov_len > len) iov[0].iov_len = len;
                len -= iov[0].iov_len;
                if (len) {
                        iov[1].iov_base = sb->sb_data;
                        iov[1].iov_len = sb->sb_wptr - sb->sb_data;
                        if (iov[1].iov_len > len) iov[1].iov_len = len;
                        n = 2;
                } else
                        n = 1;
        }
        /* Check if there's urgent data to send, and if so, send it */

#ifdef HAVE_READV
        nn = writev(so->s, (const struct iovec *)iov, n);
        
        DEBUG_MISC((dfd, "  ... wrote nn = %d bytes\n", nn));
#else
        nn = send(so->s, iov[0].iov_base, iov[0].iov_len,0);
#endif
        /* This should never happen, but people tell me it does *shrug* */
        if (nn < 0 && (errno == EAGAIN || errno == EINTR))
                return 0;
        
        if (nn <= 0) {
                DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
                        so->so_state, errno));
                sofcantsendmore(so);
                tcp_sockclosed(sototcpcb(so));
                return -1;
        }
        
#ifndef HAVE_READV
        if (n == 2 && nn == iov[0].iov_len) {
            int ret;
            ret = send(so->s, iov[1].iov_base, iov[1].iov_len,0);
            if (ret > 0)
                nn += ret;
        }
        DEBUG_MISC((dfd, "  ... wrote nn = %d bytes\n", nn));
#endif
        
        /* Update sbuf */
        sb->sb_cc -= nn;
        sb->sb_rptr += nn;
        if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
                sb->sb_rptr -= sb->sb_datalen;
        
        /*
         * If in DRAIN mode, and there's no more data, set
         * it CANTSENDMORE
         */
        if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
                sofcantsendmore(so);
        
        return nn;
}

/*
 * recvfrom() a UDP socket
 */
void
sorecvfrom(so)
        struct socket *so;
{
        struct sockaddr_in addr;
        int addrlen = sizeof(struct sockaddr_in);
        
        DEBUG_CALL("sorecvfrom");
        DEBUG_ARG("so = %lx", (long)so);
        
        if (so->so_type == IPPROTO_ICMP) {   /* This is a "ping" reply */
          char buff[256];
          int len;
                
          len = recvfrom(so->s, buff, 256, 0, 
                         (struct sockaddr *)&addr, &addrlen);
          /* XXX Check if reply is "correct"? */
          
          if(len == -1 || len == 0) {
            u_char code=ICMP_UNREACH_PORT;

            if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
            else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
            
            DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
                        errno,strerror(errno)));
            icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
          } else {
            icmp_reflect(so->so_m);
            so->so_m = 0; /* Don't m_free() it again! */
          }
          /* No need for this socket anymore, udp_detach it */
          udp_detach(so);
        } else {                                /* A "normal" UDP packet */
          struct mbuf *m;
          int len, n;

          if (!(m = m_get())) return;
          m->m_data += if_maxlinkhdr;
                
          /* 
           * XXX Shouldn't FIONREAD packets destined for port 53,
           * but I don't know the max packet size for DNS lookups
           */
          len = M_FREEROOM(m);
          /* if (so->so_fport != htons(53)) { */
          ioctlsocket(so->s, FIONREAD, &n);
          
          if (n > len) {
            n = (m->m_data - m->m_dat) + m->m_len + n + 1;
            m_inc(m, n);
            len = M_FREEROOM(m);
          }
          /* } */
                
          m->m_len = recvfrom(so->s, m->m_data, len, 0,
                              (struct sockaddr *)&addr, &addrlen);
          DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n", 
                      m->m_len, errno,strerror(errno)));
          if(m->m_len<0) {
            u_char code=ICMP_UNREACH_PORT;

            if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
            else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
            
            DEBUG_MISC((dfd," rx error, tx icmp ICMP_UNREACH:%i\n", code));
            icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
            m_free(m);
          } else {
          /*
           * Hack: domain name lookup will be used the most for UDP,
           * and since they'll only be used once there's no need
           * for the 4 minute (or whatever) timeout... So we time them
           * out much quicker (10 seconds  for now...)
           */
            if (so->so_expire) {
              if (so->so_fport == htons(53))
                so->so_expire = curtime + SO_EXPIREFAST;
              else
                so->so_expire = curtime + SO_EXPIRE;
            }

            /*          if (m->m_len == len) {
             *                  m_inc(m, MINCSIZE);
             *                  m->m_len = 0;
             *          }
             */
            
            /* 
             * If this packet was destined for CTL_ADDR,
             * make it look like that's where it came from, done by udp_output
             */
            udp_output(so, m, &addr);
          } /* rx error */
        } /* if ping packet */
}

/*
 * sendto() a socket
 */
int
sosendto(so, m)
        struct socket *so;
        struct mbuf *m;
{
        int ret;
        struct sockaddr_in addr;

        DEBUG_CALL("sosendto");
        DEBUG_ARG("so = %lx", (long)so);
        DEBUG_ARG("m = %lx", (long)m);
        
        addr.sin_family = AF_INET;
        if ((so->so_faddr.s_addr & htonl(0xffffff00)) == special_addr.s_addr) {
          /* It's an alias */
          switch(ntohl(so->so_faddr.s_addr) & 0xff) {
          case CTL_DNS:
            addr.sin_addr = dns_addr;
            break;
          case CTL_ALIAS:
          default:
            addr.sin_addr = loopback_addr;
            break;
          }
        } else
          addr.sin_addr = so->so_faddr;
        addr.sin_port = so->so_fport;

        DEBUG_MISC((dfd, " sendto()ing, addr.sin_port=%d, addr.sin_addr.s_addr=%.16s\n", ntohs(addr.sin_port), inet_ntoa(addr.sin_addr)));
        
        /* Don't care what port we get */
        ret = sendto(so->s, m->m_data, m->m_len, 0,
                     (struct sockaddr *)&addr, sizeof (struct sockaddr));
        if (ret < 0)
                return -1;
        
        /*
         * Kill the socket if there's no reply in 4 minutes,
         * but only if it's an expirable socket
         */
        if (so->so_expire)
                so->so_expire = curtime + SO_EXPIRE;
        so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
        return 0;
}

/*
 * XXX This should really be tcp_listen
 */
struct socket *
solisten(port, laddr, lport, flags)
        u_int port;
        u_int32_t laddr;
        u_int lport;
        int flags;
{
        struct sockaddr_in addr;
        struct socket *so;
        int s, addrlen = sizeof(addr), opt = 1;

        DEBUG_CALL("solisten");
        DEBUG_ARG("port = %d", port);
        DEBUG_ARG("laddr = %x", laddr);
        DEBUG_ARG("lport = %d", lport);
        DEBUG_ARG("flags = %x", flags);
        
        if ((so = socreate()) == NULL) {
          /* free(so);      Not sofree() ??? free(NULL) == NOP */
          return NULL;
        }
        
        /* Don't tcp_attach... we don't need so_snd nor so_rcv */
        if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL) {
                free(so);
                return NULL;
        }
        insque(so,&tcb);
        
        /* 
         * SS_FACCEPTONCE sockets must time out.
         */
        if (flags & SS_FACCEPTONCE)
           so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
        
        so->so_state = (SS_FACCEPTCONN|flags);
        so->so_lport = lport; /* Kept in network format */
        so->so_laddr.s_addr = laddr; /* Ditto */
        
        addr.sin_family = AF_INET;
        addr.sin_addr.s_addr = INADDR_ANY;
        addr.sin_port = port;
        
        if (((s = socket(AF_INET,SOCK_STREAM,0)) < 0) ||
            (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) ||
            (listen(s,1) < 0)) {
                int tmperrno = errno; /* Don't clobber the real reason we failed */
                
                close(s);
                sofree(so);
                /* Restore the real errno */
#ifdef _WIN32
                WSASetLastError(tmperrno);
#else
                errno = tmperrno;
#endif
                return NULL;
        }
        setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int));
        setsockopt(s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
        
        getsockname(s,(struct sockaddr *)&addr,&addrlen);
        so->so_fport = addr.sin_port;
        if (addr.sin_addr.s_addr == 0 || addr.sin_addr.s_addr == loopback_addr.s_addr)
           so->so_faddr = our_addr;
        else
           so->so_faddr = addr.sin_addr;

        so->s = s;
        return so;
}

/* 
 * Data is available in so_rcv
 * Just write() the data to the socket
 * XXX not yet...
 */
void
sorwakeup(so)
        struct socket *so;
{
/*      sowrite(so); */
/*      FD_CLR(so->s,&writefds); */
}
        
/*
 * Data has been freed in so_snd
 * We have room for a read() if we want to
 * For now, don't read, it'll be done in the main loop
 */
void
sowwakeup(so)
        struct socket *so;
{
        /* Nothing, yet */
}

/*
 * Various session state calls
 * XXX Should be #define's
 * The socket state stuff needs work, these often get call 2 or 3
 * times each when only 1 was needed
 */
void
soisfconnecting(so)
        register struct socket *so;
{
        so->so_state &= ~(SS_NOFDREF|SS_ISFCONNECTED|SS_FCANTRCVMORE|
                          SS_FCANTSENDMORE|SS_FWDRAIN);
        so->so_state |= SS_ISFCONNECTING; /* Clobber other states */
}

void
soisfconnected(so)
        register struct socket *so;
{
        so->so_state &= ~(SS_ISFCONNECTING|SS_FWDRAIN|SS_NOFDREF);
        so->so_state |= SS_ISFCONNECTED; /* Clobber other states */
}

void
sofcantrcvmore(so)
        struct  socket *so;
{
        if ((so->so_state & SS_NOFDREF) == 0) {
                shutdown(so->s,0);
                if(global_writefds) {
                  FD_CLR(so->s,global_writefds);
                }
        }
        so->so_state &= ~(SS_ISFCONNECTING);
        if (so->so_state & SS_FCANTSENDMORE)
           so->so_state = SS_NOFDREF; /* Don't select it */ /* XXX close() here as well? */
        else
           so->so_state |= SS_FCANTRCVMORE;
}

void
sofcantsendmore(so)
        struct socket *so;
{
        if ((so->so_state & SS_NOFDREF) == 0) {
            shutdown(so->s,1);           /* send FIN to fhost */
            if (global_readfds) {
                FD_CLR(so->s,global_readfds);
            }
            if (global_xfds) {
                FD_CLR(so->s,global_xfds);
            }
        }
        so->so_state &= ~(SS_ISFCONNECTING);
        if (so->so_state & SS_FCANTRCVMORE)
           so->so_state = SS_NOFDREF; /* as above */
        else
           so->so_state |= SS_FCANTSENDMORE;
}

void
soisfdisconnected(so)
        struct socket *so;
{
/*      so->so_state &= ~(SS_ISFCONNECTING|SS_ISFCONNECTED); */
/*      close(so->s); */
/*      so->so_state = SS_ISFDISCONNECTED; */
        /*
         * XXX Do nothing ... ?
         */
}

/*
 * Set write drain mode
 * Set CANTSENDMORE once all data has been write()n
 */
void
sofwdrain(so)
        struct socket *so;
{
        if (so->so_rcv.sb_cc)
                so->so_state |= SS_FWDRAIN;
        else
                sofcantsendmore(so);
}

Revision:	1.1
Committed:	2005-05-13T09:00:59Z (19 years, 6 months ago) by gbeauche
Content type:	text/plain
Branch:	MAIN
Log Message:	slirp user mode network emulation code from qemu
#	User	Rev	Content
1	gbeauche	1.1	/*
2			* Copyright (c) 1995 Danny Gasparovski.
3			*
4			* Please read the file COPYRIGHT for the
5			* terms and conditions of the copyright.
6			*/
7
8			#define WANT_SYS_IOCTL_H
9			#include <slirp.h>
10			#include "ip_icmp.h"
11			#include "main.h"
12
13			void
14			so_init()
15			{
16			/* Nothing yet */
17			}
18
19
20			struct socket *
21			solookup(head, laddr, lport, faddr, fport)
22			struct socket *head;
23			struct in_addr laddr;
24			u_int lport;
25			struct in_addr faddr;
26			u_int fport;
27			{
28			struct socket *so;
29
30			for (so = head->so_next; so != head; so = so->so_next) {
31			if (so->so_lport == lport &&
32			so->so_laddr.s_addr == laddr.s_addr &&
33			so->so_faddr.s_addr == faddr.s_addr &&
34			so->so_fport == fport)
35			break;
36			}
37
38			if (so == head)
39			return (struct socket *)NULL;
40			return so;
41
42			}
43
44			/*
45			* Create a new socket, initialise the fields
46			* It is the responsibility of the caller to
47			* insque() it into the correct linked-list
48			*/
49			struct socket *
50			socreate()
51			{
52			struct socket *so;
53
54			so = (struct socket *)malloc(sizeof(struct socket));
55			if(so) {
56			memset(so, 0, sizeof(struct socket));
57			so->so_state = SS_NOFDREF;
58			so->s = -1;
59			}
60			return(so);
61			}
62
63			/*
64			* remque and free a socket, clobber cache
65			*/
66			void
67			sofree(so)
68			struct socket *so;
69			{
70			if (so->so_emu==EMU_RSH && so->extra) {
71			sofree(so->extra);
72			so->extra=NULL;
73			}
74			if (so == tcp_last_so)
75			tcp_last_so = &tcb;
76			else if (so == udp_last_so)
77			udp_last_so = &udb;
78
79			m_free(so->so_m);
80
81			if(so->so_next && so->so_prev)
82			remque(so); /* crashes if so is not in a queue */
83
84			free(so);
85			}
86
87			/*
88			* Read from so's socket into sb_snd, updating all relevant sbuf fields
89			* NOTE: This will only be called if it is select()ed for reading, so
90			* a read() of 0 (or less) means it's disconnected
91			*/
92			int
93			soread(so)
94			struct socket *so;
95			{
96			int n, nn, lss, total;
97			struct sbuf *sb = &so->so_snd;
98			int len = sb->sb_datalen - sb->sb_cc;
99			struct iovec iov[2];
100			int mss = so->so_tcpcb->t_maxseg;
101
102			DEBUG_CALL("soread");
103			DEBUG_ARG("so = %lx", (long )so);
104
105			/*
106			* No need to check if there's enough room to read.
107			* soread wouldn't have been called if there weren't
108			*/
109
110			len = sb->sb_datalen - sb->sb_cc;
111
112			iov[0].iov_base = sb->sb_wptr;
113			if (sb->sb_wptr < sb->sb_rptr) {
114			iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
115			/* Should never succeed, but... */
116			if (iov[0].iov_len > len)
117			iov[0].iov_len = len;
118			if (iov[0].iov_len > mss)
119			iov[0].iov_len -= iov[0].iov_len%mss;
120			n = 1;
121			} else {
122			iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
123			/* Should never succeed, but... */
124			if (iov[0].iov_len > len) iov[0].iov_len = len;
125			len -= iov[0].iov_len;
126			if (len) {
127			iov[1].iov_base = sb->sb_data;
128			iov[1].iov_len = sb->sb_rptr - sb->sb_data;
129			if(iov[1].iov_len > len)
130			iov[1].iov_len = len;
131			total = iov[0].iov_len + iov[1].iov_len;
132			if (total > mss) {
133			lss = total%mss;
134			if (iov[1].iov_len > lss) {
135			iov[1].iov_len -= lss;
136			n = 2;
137			} else {
138			lss -= iov[1].iov_len;
139			iov[0].iov_len -= lss;
140			n = 1;
141			}
142			} else
143			n = 2;
144			} else {
145			if (iov[0].iov_len > mss)
146			iov[0].iov_len -= iov[0].iov_len%mss;
147			n = 1;
148			}
149			}
150
151			#ifdef HAVE_READV
152			nn = readv(so->s, (struct iovec *)iov, n);
153			DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
154			#else
155			nn = recv(so->s, iov[0].iov_base, iov[0].iov_len,0);
156			#endif
157			if (nn <= 0) {
158			if (nn < 0 && (errno == EINTR \|\| errno == EAGAIN))
159			return 0;
160			else {
161			DEBUG_MISC((dfd, " --- soread() disconnected, nn = %d, errno = %d-%s\n", nn, errno,strerror(errno)));
162			sofcantrcvmore(so);
163			tcp_sockclosed(sototcpcb(so));
164			return -1;
165			}
166			}
167
168			#ifndef HAVE_READV
169			/*
170			* If there was no error, try and read the second time round
171			* We read again if n = 2 (ie, there's another part of the buffer)
172			* and we read as much as we could in the first read
173			* We don't test for <= 0 this time, because there legitimately
174			* might not be any more data (since the socket is non-blocking),
175			* a close will be detected on next iteration.
176			* A return of -1 wont (shouldn't) happen, since it didn't happen above
177			*/
178			if (n == 2 && nn == iov[0].iov_len) {
179			int ret;
180			ret = recv(so->s, iov[1].iov_base, iov[1].iov_len,0);
181			if (ret > 0)
182			nn += ret;
183			}
184
185			DEBUG_MISC((dfd, " ... read nn = %d bytes\n", nn));
186			#endif
187
188			/* Update fields */
189			sb->sb_cc += nn;
190			sb->sb_wptr += nn;
191			if (sb->sb_wptr >= (sb->sb_data + sb->sb_datalen))
192			sb->sb_wptr -= sb->sb_datalen;
193			return nn;
194			}
195
196			/*
197			* Get urgent data
198			*
199			* When the socket is created, we set it SO_OOBINLINE,
200			* so when OOB data arrives, we soread() it and everything
201			* in the send buffer is sent as urgent data
202			*/
203			void
204			sorecvoob(so)
205			struct socket *so;
206			{
207			struct tcpcb *tp = sototcpcb(so);
208
209			DEBUG_CALL("sorecvoob");
210			DEBUG_ARG("so = %lx", (long)so);
211
212			/*
213			* We take a guess at how much urgent data has arrived.
214			* In most situations, when urgent data arrives, the next
215			* read() should get all the urgent data. This guess will
216			* be wrong however if more data arrives just after the
217			* urgent data, or the read() doesn't return all the
218			* urgent data.
219			*/
220			soread(so);
221			tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
222			tp->t_force = 1;
223			tcp_output(tp);
224			tp->t_force = 0;
225			}
226
227			/*
228			* Send urgent data
229			* There's a lot duplicated code here, but...
230			*/
231			int
232			sosendoob(so)
233			struct socket *so;
234			{
235			struct sbuf *sb = &so->so_rcv;
236			char buff[2048]; /* XXX Shouldn't be sending more oob data than this */
237
238			int n, len;
239
240			DEBUG_CALL("sosendoob");
241			DEBUG_ARG("so = %lx", (long)so);
242			DEBUG_ARG("sb->sb_cc = %d", sb->sb_cc);
243
244			if (so->so_urgc > 2048)
245			so->so_urgc = 2048; /* XXXX */
246
247			if (sb->sb_rptr < sb->sb_wptr) {
248			/* We can send it directly */
249			n = send(so->s, sb->sb_rptr, so->so_urgc, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
250			so->so_urgc -= n;
251
252			DEBUG_MISC((dfd, " --- sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
253			} else {
254			/*
255			* Since there's no sendv or sendtov like writev,
256			* we must copy all data to a linear buffer then
257			* send it all
258			*/
259			len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
260			if (len > so->so_urgc) len = so->so_urgc;
261			memcpy(buff, sb->sb_rptr, len);
262			so->so_urgc -= len;
263			if (so->so_urgc) {
264			n = sb->sb_wptr - sb->sb_data;
265			if (n > so->so_urgc) n = so->so_urgc;
266			memcpy((buff + len), sb->sb_data, n);
267			so->so_urgc -= n;
268			len += n;
269			}
270			n = send(so->s, buff, len, (MSG_OOB)); /* \|MSG_DONTWAIT)); */
271			#ifdef DEBUG
272			if (n != len)
273			DEBUG_ERROR((dfd, "Didn't send all data urgently XXXXX\n"));
274			#endif
275			DEBUG_MISC((dfd, " ---2 sent %d bytes urgent data, %d urgent bytes left\n", n, so->so_urgc));
276			}
277
278			sb->sb_cc -= n;
279			sb->sb_rptr += n;
280			if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
281			sb->sb_rptr -= sb->sb_datalen;
282
283			return n;
284			}
285
286			/*
287			* Write data from so_rcv to so's socket,
288			* updating all sbuf field as necessary
289			*/
290			int
291			sowrite(so)
292			struct socket *so;
293			{
294			int n,nn;
295			struct sbuf *sb = &so->so_rcv;
296			int len = sb->sb_cc;
297			struct iovec iov[2];
298
299			DEBUG_CALL("sowrite");
300			DEBUG_ARG("so = %lx", (long)so);
301
302			if (so->so_urgc) {
303			sosendoob(so);
304			if (sb->sb_cc == 0)
305			return 0;
306			}
307
308			/*
309			* No need to check if there's something to write,
310			* sowrite wouldn't have been called otherwise
311			*/
312
313			len = sb->sb_cc;
314
315			iov[0].iov_base = sb->sb_rptr;
316			if (sb->sb_rptr < sb->sb_wptr) {
317			iov[0].iov_len = sb->sb_wptr - sb->sb_rptr;
318			/* Should never succeed, but... */
319			if (iov[0].iov_len > len) iov[0].iov_len = len;
320			n = 1;
321			} else {
322			iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_rptr;
323			if (iov[0].iov_len > len) iov[0].iov_len = len;
324			len -= iov[0].iov_len;
325			if (len) {
326			iov[1].iov_base = sb->sb_data;
327			iov[1].iov_len = sb->sb_wptr - sb->sb_data;
328			if (iov[1].iov_len > len) iov[1].iov_len = len;
329			n = 2;
330			} else
331			n = 1;
332			}
333			/* Check if there's urgent data to send, and if so, send it */
334
335			#ifdef HAVE_READV
336			nn = writev(so->s, (const struct iovec *)iov, n);
337
338			DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
339			#else
340			nn = send(so->s, iov[0].iov_base, iov[0].iov_len,0);
341			#endif
342			/* This should never happen, but people tell me it does shrug */
343			if (nn < 0 && (errno == EAGAIN \|\| errno == EINTR))
344			return 0;
345
346			if (nn <= 0) {
347			DEBUG_MISC((dfd, " --- sowrite disconnected, so->so_state = %x, errno = %d\n",
348			so->so_state, errno));
349			sofcantsendmore(so);
350			tcp_sockclosed(sototcpcb(so));
351			return -1;
352			}
353
354			#ifndef HAVE_READV
355			if (n == 2 && nn == iov[0].iov_len) {
356			int ret;
357			ret = send(so->s, iov[1].iov_base, iov[1].iov_len,0);
358			if (ret > 0)
359			nn += ret;
360			}
361			DEBUG_MISC((dfd, " ... wrote nn = %d bytes\n", nn));
362			#endif
363
364			/* Update sbuf */
365			sb->sb_cc -= nn;
366			sb->sb_rptr += nn;
367			if (sb->sb_rptr >= (sb->sb_data + sb->sb_datalen))
368			sb->sb_rptr -= sb->sb_datalen;
369
370			/*
371			* If in DRAIN mode, and there's no more data, set
372			* it CANTSENDMORE
373			*/
374			if ((so->so_state & SS_FWDRAIN) && sb->sb_cc == 0)
375			sofcantsendmore(so);
376
377			return nn;
378			}
379
380			/*
381			* recvfrom() a UDP socket
382			*/
383			void
384			sorecvfrom(so)
385			struct socket *so;
386			{
387			struct sockaddr_in addr;
388			int addrlen = sizeof(struct sockaddr_in);
389
390			DEBUG_CALL("sorecvfrom");
391			DEBUG_ARG("so = %lx", (long)so);
392
393			if (so->so_type == IPPROTO_ICMP) { /* This is a "ping" reply */
394			char buff[256];
395			int len;
396
397			len = recvfrom(so->s, buff, 256, 0,
398			(struct sockaddr *)&addr, &addrlen);
399			/* XXX Check if reply is "correct"? */
400
401			if(len == -1 \|\| len == 0) {
402			u_char code=ICMP_UNREACH_PORT;
403
404			if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
405			else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
406
407			DEBUG_MISC((dfd," udp icmp rx errno = %d-%s\n",
408			errno,strerror(errno)));
409			icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
410			} else {
411			icmp_reflect(so->so_m);
412			so->so_m = 0; /* Don't m_free() it again! */
413			}
414			/* No need for this socket anymore, udp_detach it */
415			udp_detach(so);
416			} else { /* A "normal" UDP packet */
417			struct mbuf *m;
418			int len, n;
419
420			if (!(m = m_get())) return;
421			m->m_data += if_maxlinkhdr;
422
423			/*
424			* XXX Shouldn't FIONREAD packets destined for port 53,
425			* but I don't know the max packet size for DNS lookups
426			*/
427			len = M_FREEROOM(m);
428			/* if (so->so_fport != htons(53)) { */
429			ioctlsocket(so->s, FIONREAD, &n);
430
431			if (n > len) {
432			n = (m->m_data - m->m_dat) + m->m_len + n + 1;
433			m_inc(m, n);
434			len = M_FREEROOM(m);
435			}
436			/* } */
437
438			m->m_len = recvfrom(so->s, m->m_data, len, 0,
439			(struct sockaddr *)&addr, &addrlen);
440			DEBUG_MISC((dfd, " did recvfrom %d, errno = %d-%s\n",
441			m->m_len, errno,strerror(errno)));
442			if(m->m_len<0) {
443			u_char code=ICMP_UNREACH_PORT;
444
445			if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
446			else if(errno == ENETUNREACH) code=ICMP_UNREACH_NET;
447
448			DEBUG_MISC((dfd," rx error, tx icmp ICMP_UNREACH:%i\n", code));
449			icmp_error(so->so_m, ICMP_UNREACH,code, 0,strerror(errno));
450			m_free(m);
451			} else {
452			/*
453			* Hack: domain name lookup will be used the most for UDP,
454			* and since they'll only be used once there's no need
455			* for the 4 minute (or whatever) timeout... So we time them
456			* out much quicker (10 seconds for now...)
457			*/
458			if (so->so_expire) {
459			if (so->so_fport == htons(53))
460			so->so_expire = curtime + SO_EXPIREFAST;
461			else
462			so->so_expire = curtime + SO_EXPIRE;
463			}
464
465			/* if (m->m_len == len) {
466			* m_inc(m, MINCSIZE);
467			* m->m_len = 0;
468			* }
469			*/
470
471			/*
472			* If this packet was destined for CTL_ADDR,
473			* make it look like that's where it came from, done by udp_output
474			*/
475			udp_output(so, m, &addr);
476			} /* rx error */
477			} /* if ping packet */
478			}
479
480			/*
481			* sendto() a socket
482			*/
483			int
484			sosendto(so, m)
485			struct socket *so;
486			struct mbuf *m;
487			{
488			int ret;
489			struct sockaddr_in addr;
490
491			DEBUG_CALL("sosendto");
492			DEBUG_ARG("so = %lx", (long)so);
493			DEBUG_ARG("m = %lx", (long)m);
494
495			addr.sin_family = AF_INET;
496			if ((so->so_faddr.s_addr & htonl(0xffffff00)) == special_addr.s_addr) {
497			/* It's an alias */
498			switch(ntohl(so->so_faddr.s_addr) & 0xff) {
499			case CTL_DNS:
500			addr.sin_addr = dns_addr;
501			break;
502			case CTL_ALIAS:
503			default:
504			addr.sin_addr = loopback_addr;
505			break;
506			}
507			} else
508			addr.sin_addr = so->so_faddr;
509			addr.sin_port = so->so_fport;
510
511			DEBUG_MISC((dfd, " sendto()ing, addr.sin_port=%d, addr.sin_addr.s_addr=%.16s\n", ntohs(addr.sin_port), inet_ntoa(addr.sin_addr)));
512
513			/* Don't care what port we get */
514			ret = sendto(so->s, m->m_data, m->m_len, 0,
515			(struct sockaddr *)&addr, sizeof (struct sockaddr));
516			if (ret < 0)
517			return -1;
518
519			/*
520			* Kill the socket if there's no reply in 4 minutes,
521			* but only if it's an expirable socket
522			*/
523			if (so->so_expire)
524			so->so_expire = curtime + SO_EXPIRE;
525			so->so_state = SS_ISFCONNECTED; /* So that it gets select()ed */
526			return 0;
527			}
528
529			/*
530			* XXX This should really be tcp_listen
531			*/
532			struct socket *
533			solisten(port, laddr, lport, flags)
534			u_int port;
535			u_int32_t laddr;
536			u_int lport;
537			int flags;
538			{
539			struct sockaddr_in addr;
540			struct socket *so;
541			int s, addrlen = sizeof(addr), opt = 1;
542
543			DEBUG_CALL("solisten");
544			DEBUG_ARG("port = %d", port);
545			DEBUG_ARG("laddr = %x", laddr);
546			DEBUG_ARG("lport = %d", lport);
547			DEBUG_ARG("flags = %x", flags);
548
549			if ((so = socreate()) == NULL) {
550			/* free(so); Not sofree() ??? free(NULL) == NOP */
551			return NULL;
552			}
553
554			/* Don't tcp_attach... we don't need so_snd nor so_rcv */
555			if ((so->so_tcpcb = tcp_newtcpcb(so)) == NULL) {
556			free(so);
557			return NULL;
558			}
559			insque(so,&tcb);
560
561			/*
562			* SS_FACCEPTONCE sockets must time out.
563			*/
564			if (flags & SS_FACCEPTONCE)
565			so->so_tcpcb->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT*2;
566
567			so->so_state = (SS_FACCEPTCONN\|flags);
568			so->so_lport = lport; /* Kept in network format */
569			so->so_laddr.s_addr = laddr; /* Ditto */
570
571			addr.sin_family = AF_INET;
572			addr.sin_addr.s_addr = INADDR_ANY;
573			addr.sin_port = port;
574
575			if (((s = socket(AF_INET,SOCK_STREAM,0)) < 0) \|\|
576			(bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) \|\|
577			(listen(s,1) < 0)) {
578			int tmperrno = errno; /* Don't clobber the real reason we failed */
579
580			close(s);
581			sofree(so);
582			/* Restore the real errno */
583			#ifdef _WIN32
584			WSASetLastError(tmperrno);
585			#else
586			errno = tmperrno;
587			#endif
588			return NULL;
589			}
590			setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int));
591			setsockopt(s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
592
593			getsockname(s,(struct sockaddr *)&addr,&addrlen);
594			so->so_fport = addr.sin_port;
595			if (addr.sin_addr.s_addr == 0 \|\| addr.sin_addr.s_addr == loopback_addr.s_addr)
596			so->so_faddr = our_addr;
597			else
598			so->so_faddr = addr.sin_addr;
599
600			so->s = s;
601			return so;
602			}
603
604			/*
605			* Data is available in so_rcv
606			* Just write() the data to the socket
607			* XXX not yet...
608			*/
609			void
610			sorwakeup(so)
611			struct socket *so;
612			{
613			/* sowrite(so); */
614			/* FD_CLR(so->s,&writefds); */
615			}
616
617			/*
618			* Data has been freed in so_snd
619			* We have room for a read() if we want to
620			* For now, don't read, it'll be done in the main loop
621			*/
622			void
623			sowwakeup(so)
624			struct socket *so;
625			{
626			/* Nothing, yet */
627			}
628
629			/*
630			* Various session state calls
631			* XXX Should be #define's
632			* The socket state stuff needs work, these often get call 2 or 3
633			* times each when only 1 was needed
634			*/
635			void
636			soisfconnecting(so)
637			register struct socket *so;
638			{
639			so->so_state &= ~(SS_NOFDREF\|SS_ISFCONNECTED\|SS_FCANTRCVMORE\|
640			SS_FCANTSENDMORE\|SS_FWDRAIN);
641			so->so_state \|= SS_ISFCONNECTING; /* Clobber other states */
642			}
643
644			void
645			soisfconnected(so)
646			register struct socket *so;
647			{
648			so->so_state &= ~(SS_ISFCONNECTING\|SS_FWDRAIN\|SS_NOFDREF);
649			so->so_state \|= SS_ISFCONNECTED; /* Clobber other states */
650			}
651
652			void
653			sofcantrcvmore(so)
654			struct socket *so;
655			{
656			if ((so->so_state & SS_NOFDREF) == 0) {
657			shutdown(so->s,0);
658			if(global_writefds) {
659			FD_CLR(so->s,global_writefds);
660			}
661			}
662			so->so_state &= ~(SS_ISFCONNECTING);
663			if (so->so_state & SS_FCANTSENDMORE)
664			so->so_state = SS_NOFDREF; /* Don't select it / / XXX close() here as well? */
665			else
666			so->so_state \|= SS_FCANTRCVMORE;
667			}
668
669			void
670			sofcantsendmore(so)
671			struct socket *so;
672			{
673			if ((so->so_state & SS_NOFDREF) == 0) {
674			shutdown(so->s,1); /* send FIN to fhost */
675			if (global_readfds) {
676			FD_CLR(so->s,global_readfds);
677			}
678			if (global_xfds) {
679			FD_CLR(so->s,global_xfds);
680			}
681			}
682			so->so_state &= ~(SS_ISFCONNECTING);
683			if (so->so_state & SS_FCANTRCVMORE)
684			so->so_state = SS_NOFDREF; /* as above */
685			else
686			so->so_state \|= SS_FCANTSENDMORE;
687			}
688
689			void
690			soisfdisconnected(so)
691			struct socket *so;
692			{
693			/* so->so_state &= ~(SS_ISFCONNECTING\|SS_ISFCONNECTED); */
694			/* close(so->s); */
695			/* so->so_state = SS_ISFDISCONNECTED; */
696			/*
697			* XXX Do nothing ... ?
698			*/
699			}
700
701			/*
702			* Set write drain mode
703			* Set CANTSENDMORE once all data has been write()n
704			*/
705			void
706			sofwdrain(so)
707			struct socket *so;
708			{
709			if (so->so_rcv.sb_cc)
710			so->so_state \|= SS_FWDRAIN;
711			else
712			sofcantsendmore(so);
713			}
714