2005-01-07 01:45:51 +00:00
|
|
|
/*-
|
2001-03-10 05:24:45 +00:00
|
|
|
* Copyright (c) 1999, 2001 Boris Popov
|
1999-10-02 04:06:24 +00:00
|
|
|
* 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. All advertising materials mentioning features or use of this software
|
|
|
|
* must display the following acknowledgement:
|
|
|
|
* This product includes software developed by Boris Popov.
|
|
|
|
* 4. Neither the name of the author nor the names of any co-contributors
|
|
|
|
* may be used to endorse or promote products derived from this software
|
|
|
|
* without specific prior written permission.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``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 OR CONTRIBUTORS 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.
|
|
|
|
*
|
|
|
|
* Low level socket routines
|
|
|
|
*/
|
|
|
|
|
2003-06-11 05:37:42 +00:00
|
|
|
#include <sys/cdefs.h>
|
|
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
|
1999-10-02 04:06:24 +00:00
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/errno.h>
|
2001-05-15 22:02:46 +00:00
|
|
|
#include <sys/lock.h>
|
1999-10-02 04:06:24 +00:00
|
|
|
#include <sys/malloc.h>
|
2001-05-15 22:02:46 +00:00
|
|
|
#include <sys/mutex.h>
|
1999-10-02 04:06:24 +00:00
|
|
|
#include <sys/systm.h>
|
|
|
|
#include <sys/proc.h>
|
|
|
|
#include <sys/socket.h>
|
|
|
|
#include <sys/socketvar.h>
|
|
|
|
#include <sys/protosw.h>
|
|
|
|
#include <sys/kernel.h>
|
|
|
|
#include <sys/uio.h>
|
|
|
|
#include <sys/syslog.h>
|
|
|
|
#include <sys/mbuf.h>
|
2001-05-14 05:26:48 +00:00
|
|
|
#include <sys/condvar.h>
|
1999-10-02 04:06:24 +00:00
|
|
|
#include <net/route.h>
|
|
|
|
|
|
|
|
#include <netipx/ipx.h>
|
|
|
|
#include <netipx/ipx_pcb.h>
|
|
|
|
|
|
|
|
#include <netncp/ncp.h>
|
|
|
|
#include <netncp/ncp_conn.h>
|
|
|
|
#include <netncp/ncp_sock.h>
|
|
|
|
#include <netncp/ncp_subr.h>
|
|
|
|
#include <netncp/ncp_rq.h>
|
|
|
|
|
|
|
|
#define ipx_setnullnet(x) ((x).x_net.s_net[0]=0); ((x).x_net.s_net[1]=0);
|
|
|
|
#define ipx_setnullhost(x) ((x).x_host.s_host[0] = 0); \
|
|
|
|
((x).x_host.s_host[1] = 0); ((x).x_host.s_host[2] = 0);
|
|
|
|
|
|
|
|
/*static int ncp_getsockname(struct socket *so, caddr_t asa, int *alen);*/
|
2003-02-26 21:25:55 +00:00
|
|
|
static int ncp_soconnect(struct socket *so, struct sockaddr *target,
|
|
|
|
struct thread *td);
|
1999-10-02 04:06:24 +00:00
|
|
|
|
|
|
|
|
|
|
|
/* This will need only if native IP used, or (unlikely) NCP will be
|
|
|
|
* implemented on the socket level
|
|
|
|
*/
|
|
|
|
static int
|
2003-02-26 21:25:55 +00:00
|
|
|
ncp_soconnect(struct socket *so, struct sockaddr *target, struct thread *td)
|
|
|
|
{
|
|
|
|
int error, s;
|
1999-10-02 04:06:24 +00:00
|
|
|
|
2003-02-26 21:25:55 +00:00
|
|
|
error = soconnect(so, (struct sockaddr*)target, td);
|
1999-10-02 04:06:24 +00:00
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
/*
|
|
|
|
* Wait for the connection to complete. Cribbed from the
|
|
|
|
* connect system call but with the wait timing out so
|
|
|
|
* that interruptible mounts don't hang here for a long time.
|
|
|
|
*/
|
|
|
|
error = EIO;
|
|
|
|
s = splnet();
|
|
|
|
while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
|
2002-05-31 11:52:35 +00:00
|
|
|
(void) tsleep((caddr_t)&so->so_timeo, PSOCK, "ncpcon", 2 * hz);
|
1999-10-02 04:06:24 +00:00
|
|
|
if ((so->so_state & SS_ISCONNECTING) &&
|
|
|
|
so->so_error == 0 /*&& rep &&*/) {
|
|
|
|
so->so_state &= ~SS_ISCONNECTING;
|
|
|
|
splx(s);
|
|
|
|
goto bad;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (so->so_error) {
|
|
|
|
error = so->so_error;
|
|
|
|
so->so_error = 0;
|
|
|
|
splx(s);
|
|
|
|
goto bad;
|
|
|
|
}
|
|
|
|
splx(s);
|
|
|
|
error=0;
|
|
|
|
bad:
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
#ifdef notyet
|
|
|
|
static int
|
|
|
|
ncp_getsockname(struct socket *so, caddr_t asa, int *alen) {
|
|
|
|
struct sockaddr *sa;
|
|
|
|
int len=0, error;
|
|
|
|
|
|
|
|
sa = 0;
|
|
|
|
error = (*so->so_proto->pr_usrreqs->pru_sockaddr)(so, &sa);
|
|
|
|
if (error==0) {
|
|
|
|
if (sa) {
|
|
|
|
len = min(len, sa->sa_len);
|
|
|
|
bcopy(sa, (caddr_t)asa, (u_int)len);
|
|
|
|
}
|
|
|
|
*alen=len;
|
|
|
|
}
|
|
|
|
if (sa)
|
|
|
|
FREE(sa, M_SONAME);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
int ncp_sock_recv(struct socket *so, struct mbuf **mp, int *rlen)
|
|
|
|
{
|
|
|
|
struct uio auio;
|
2003-02-26 21:25:55 +00:00
|
|
|
struct thread *td = curthread; /* XXX */
|
1999-10-02 04:06:24 +00:00
|
|
|
int error,flags,len;
|
2003-02-26 21:25:55 +00:00
|
|
|
|
1999-10-02 04:06:24 +00:00
|
|
|
auio.uio_resid = len = 1000000;
|
2003-02-26 21:25:55 +00:00
|
|
|
auio.uio_td = td;
|
1999-10-02 04:06:24 +00:00
|
|
|
flags = MSG_DONTWAIT;
|
|
|
|
|
soreceive_generic(), and sopoll_generic(). Add new functions sosend(),
soreceive(), and sopoll(), which are wrappers for pru_sosend,
pru_soreceive, and pru_sopoll, and are now used univerally by socket
consumers rather than either directly invoking the old so*() functions
or directly invoking the protocol switch method (about an even split
prior to this commit).
This completes an architectural change that was begun in 1996 to permit
protocols to provide substitute implementations, as now used by UDP.
Consumers now uniformly invoke sosend(), soreceive(), and sopoll() to
perform these operations on sockets -- in particular, distributed file
systems and socket system calls.
Architectural head nod: sam, gnn, wollman
2006-07-24 15:20:08 +00:00
|
|
|
/* error = soreceive(so, 0, &auio, (struct mbuf **)0, (struct mbuf **)0,
|
|
|
|
&flags);*/
|
|
|
|
error = soreceive(so, 0, &auio, mp, (struct mbuf **)0, &flags);
|
1999-10-02 04:06:24 +00:00
|
|
|
*rlen = len - auio.uio_resid;
|
|
|
|
/* if (!error) {
|
|
|
|
*rlen=iov.iov_len;
|
|
|
|
} else
|
|
|
|
*rlen=0;*/
|
|
|
|
#ifdef NCP_SOCKET_DEBUG
|
|
|
|
if (error)
|
|
|
|
printf("ncp_recv: err=%d\n", error);
|
|
|
|
#endif
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
ncp_sock_send(struct socket *so, struct mbuf *top, struct ncp_rq *rqp)
|
|
|
|
{
|
2003-02-26 21:25:55 +00:00
|
|
|
struct thread *td = curthread; /* XXX */
|
1999-10-02 04:06:24 +00:00
|
|
|
struct sockaddr *to = 0;
|
2001-03-10 05:24:45 +00:00
|
|
|
struct ncp_conn *conn = rqp->nr_conn;
|
1999-10-02 04:06:24 +00:00
|
|
|
struct mbuf *m;
|
|
|
|
int error, flags=0;
|
|
|
|
|
2003-02-26 21:25:55 +00:00
|
|
|
for (;;) {
|
2003-02-19 05:47:46 +00:00
|
|
|
m = m_copym(top, 0, M_COPYALL, M_TRYWAIT);
|
1999-10-02 04:06:24 +00:00
|
|
|
/* NCPDDEBUG(m);*/
|
soreceive_generic(), and sopoll_generic(). Add new functions sosend(),
soreceive(), and sopoll(), which are wrappers for pru_sosend,
pru_soreceive, and pru_sopoll, and are now used univerally by socket
consumers rather than either directly invoking the old so*() functions
or directly invoking the protocol switch method (about an even split
prior to this commit).
This completes an architectural change that was begun in 1996 to permit
protocols to provide substitute implementations, as now used by UDP.
Consumers now uniformly invoke sosend(), soreceive(), and sopoll() to
perform these operations on sockets -- in particular, distributed file
systems and socket system calls.
Architectural head nod: sam, gnn, wollman
2006-07-24 15:20:08 +00:00
|
|
|
error = sosend(so, to, 0, m, 0, flags, td);
|
1999-10-02 04:06:24 +00:00
|
|
|
if (error == 0 || error == EINTR || error == ENETDOWN)
|
|
|
|
break;
|
|
|
|
if (rqp->rexmit == 0) break;
|
|
|
|
rqp->rexmit--;
|
2007-02-27 17:23:29 +00:00
|
|
|
pause("ncprsn", conn->li.timeout * hz);
|
2003-02-26 21:25:55 +00:00
|
|
|
error = ncp_chkintr(conn, td);
|
1999-10-02 04:06:24 +00:00
|
|
|
if (error == EINTR) break;
|
|
|
|
}
|
|
|
|
if (error) {
|
|
|
|
log(LOG_INFO, "ncp_send: error %d for server %s", error, conn->li.server);
|
|
|
|
}
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2003-02-26 21:25:55 +00:00
|
|
|
/*
|
1999-10-02 04:06:24 +00:00
|
|
|
* Connect to specified server via IPX
|
|
|
|
*/
|
2001-03-10 05:24:45 +00:00
|
|
|
static int
|
|
|
|
ncp_sock_connect_ipx(struct ncp_conn *conn)
|
|
|
|
{
|
1999-10-02 04:06:24 +00:00
|
|
|
struct sockaddr_ipx sipx;
|
|
|
|
struct ipxpcb *npcb;
|
2003-02-26 21:25:55 +00:00
|
|
|
struct thread *td = conn->td;
|
1999-10-02 04:06:24 +00:00
|
|
|
int addrlen, error, count;
|
|
|
|
|
|
|
|
sipx.sipx_port = htons(0);
|
|
|
|
|
|
|
|
for (count = 0;;count++) {
|
|
|
|
if (count > (IPXPORT_WELLKNOWN-IPXPORT_RESERVED)*2) {
|
|
|
|
error = EADDRINUSE;
|
|
|
|
goto bad;
|
|
|
|
}
|
|
|
|
conn->ncp_so = conn->wdg_so = NULL;
|
2003-02-26 21:25:55 +00:00
|
|
|
checkbad(socreate(AF_IPX, &conn->ncp_so, SOCK_DGRAM, 0, td->td_ucred, td));
|
1999-10-02 04:06:24 +00:00
|
|
|
if (conn->li.opt & NCP_OPT_WDOG)
|
2003-02-26 21:25:55 +00:00
|
|
|
checkbad(socreate(AF_IPX, &conn->wdg_so, SOCK_DGRAM, 0, td->td_ucred, td));
|
1999-10-02 04:06:24 +00:00
|
|
|
addrlen = sizeof(sipx);
|
|
|
|
sipx.sipx_family = AF_IPX;
|
|
|
|
ipx_setnullnet(sipx.sipx_addr);
|
|
|
|
ipx_setnullhost(sipx.sipx_addr);
|
|
|
|
sipx.sipx_len = addrlen;
|
2003-02-26 21:25:55 +00:00
|
|
|
error = sobind(conn->ncp_so, (struct sockaddr *)&sipx, td);
|
1999-10-02 04:06:24 +00:00
|
|
|
if (error == 0) {
|
|
|
|
if ((conn->li.opt & NCP_OPT_WDOG) == 0)
|
|
|
|
break;
|
|
|
|
sipx.sipx_addr = sotoipxpcb(conn->ncp_so)->ipxp_laddr;
|
|
|
|
sipx.sipx_port = htons(ntohs(sipx.sipx_port) + 1);
|
|
|
|
ipx_setnullnet(sipx.sipx_addr);
|
|
|
|
ipx_setnullhost(sipx.sipx_addr);
|
2003-02-26 21:25:55 +00:00
|
|
|
error = sobind(conn->wdg_so, (struct sockaddr *)&sipx, td);
|
1999-10-02 04:06:24 +00:00
|
|
|
}
|
|
|
|
if (!error) break;
|
|
|
|
if (error != EADDRINUSE) goto bad;
|
|
|
|
sipx.sipx_port = htons((ntohs(sipx.sipx_port)+4) & 0xfff8);
|
|
|
|
soclose(conn->ncp_so);
|
|
|
|
if (conn->wdg_so)
|
|
|
|
soclose(conn->wdg_so);
|
|
|
|
}
|
|
|
|
npcb = sotoipxpcb(conn->ncp_so);
|
|
|
|
npcb->ipxp_dpt = IPXPROTO_NCP;
|
|
|
|
/* IPXrouted must be running, i.e. route must be presented */
|
|
|
|
conn->li.ipxaddr.sipx_len = sizeof(struct sockaddr_ipx);
|
2003-02-26 21:25:55 +00:00
|
|
|
checkbad(ncp_soconnect(conn->ncp_so, &conn->li.saddr, td));
|
1999-10-02 04:06:24 +00:00
|
|
|
if (conn->wdg_so) {
|
|
|
|
sotoipxpcb(conn->wdg_so)->ipxp_laddr.x_net = npcb->ipxp_laddr.x_net;
|
|
|
|
sotoipxpcb(conn->wdg_so)->ipxp_laddr.x_host= npcb->ipxp_laddr.x_host;
|
|
|
|
}
|
|
|
|
if (!error) {
|
|
|
|
conn->flags |= NCPFL_SOCONN;
|
|
|
|
}
|
|
|
|
#ifdef NCPBURST
|
|
|
|
if (ncp_burst_enabled) {
|
2003-02-26 21:25:55 +00:00
|
|
|
checkbad(socreate(AF_IPX, &conn->bc_so, SOCK_DGRAM, 0, td));
|
1999-10-02 04:06:24 +00:00
|
|
|
bzero(&sipx, sizeof(sipx));
|
|
|
|
sipx.sipx_len = sizeof(sipx);
|
2003-02-26 21:25:55 +00:00
|
|
|
checkbad(sobind(conn->bc_so, (struct sockaddr *)&sipx, td));
|
|
|
|
checkbad(ncp_soconnect(conn->bc_so, &conn->li.saddr, td));
|
1999-10-02 04:06:24 +00:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
if (!error) {
|
|
|
|
conn->flags |= NCPFL_SOCONN;
|
|
|
|
ncp_sock_checksum(conn, 0);
|
|
|
|
}
|
|
|
|
return error;
|
|
|
|
bad:
|
|
|
|
ncp_sock_disconnect(conn);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
2001-03-10 05:24:45 +00:00
|
|
|
ncp_sock_checksum(struct ncp_conn *conn, int enable)
|
|
|
|
{
|
1999-10-02 04:06:24 +00:00
|
|
|
|
|
|
|
if (enable) {
|
|
|
|
sotoipxpcb(conn->ncp_so)->ipxp_flags |= IPXP_CHECKSUM;
|
|
|
|
} else {
|
|
|
|
sotoipxpcb(conn->ncp_so)->ipxp_flags &= ~IPXP_CHECKSUM;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2003-02-26 21:25:55 +00:00
|
|
|
/*
|
1999-10-02 04:06:24 +00:00
|
|
|
* Connect to specified server via IP
|
|
|
|
*/
|
2001-03-10 05:24:45 +00:00
|
|
|
static int
|
|
|
|
ncp_sock_connect_in(struct ncp_conn *conn)
|
|
|
|
{
|
1999-10-02 04:06:24 +00:00
|
|
|
struct sockaddr_in sin;
|
2003-02-26 21:25:55 +00:00
|
|
|
struct thread *td = conn->td;
|
|
|
|
int addrlen = sizeof(sin), error;
|
1999-10-02 04:06:24 +00:00
|
|
|
|
|
|
|
conn->flags = 0;
|
|
|
|
bzero(&sin,addrlen);
|
|
|
|
conn->ncp_so = conn->wdg_so = NULL;
|
2003-02-26 21:25:55 +00:00
|
|
|
checkbad(socreate(AF_INET, &conn->ncp_so, SOCK_DGRAM, IPPROTO_UDP, td->td_ucred, td));
|
1999-10-02 04:06:24 +00:00
|
|
|
sin.sin_family = AF_INET;
|
|
|
|
sin.sin_len = addrlen;
|
2003-02-26 21:25:55 +00:00
|
|
|
checkbad(sobind(conn->ncp_so, (struct sockaddr *)&sin, td));
|
|
|
|
checkbad(ncp_soconnect(conn->ncp_so,(struct sockaddr*)&conn->li.addr, td));
|
1999-10-02 04:06:24 +00:00
|
|
|
if (!error)
|
|
|
|
conn->flags |= NCPFL_SOCONN;
|
|
|
|
return error;
|
|
|
|
bad:
|
|
|
|
ncp_sock_disconnect(conn);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
2001-03-10 05:24:45 +00:00
|
|
|
int
|
|
|
|
ncp_sock_connect(struct ncp_conn *ncp)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
switch (ncp->li.saddr.sa_family) {
|
|
|
|
case AF_IPX:
|
|
|
|
error = ncp_sock_connect_ipx(ncp);
|
|
|
|
break;
|
|
|
|
case AF_INET:
|
|
|
|
error = ncp_sock_connect_in(ncp);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return EPROTONOSUPPORT;
|
|
|
|
}
|
|
|
|
return error;
|
|
|
|
}
|
1999-10-02 04:06:24 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Connection expected to be locked
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
ncp_sock_disconnect(struct ncp_conn *conn) {
|
|
|
|
register struct socket *so;
|
|
|
|
conn->flags &= ~(NCPFL_SOCONN | NCPFL_ATTACHED | NCPFL_LOGGED);
|
|
|
|
if (conn->ncp_so) {
|
|
|
|
so = conn->ncp_so;
|
|
|
|
conn->ncp_so = (struct socket *)0;
|
|
|
|
soshutdown(so, 2);
|
|
|
|
soclose(so);
|
|
|
|
}
|
|
|
|
if (conn->wdg_so) {
|
|
|
|
so = conn->wdg_so;
|
|
|
|
conn->wdg_so = (struct socket *)0;
|
|
|
|
soshutdown(so, 2);
|
|
|
|
soclose(so);
|
|
|
|
}
|
|
|
|
#ifdef NCPBURST
|
|
|
|
if (conn->bc_so) {
|
|
|
|
so = conn->bc_so;
|
|
|
|
conn->bc_so = (struct socket *)NULL;
|
|
|
|
soshutdown(so, 2);
|
|
|
|
soclose(so);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
ncp_watchdog(struct ncp_conn *conn) {
|
|
|
|
char *buf;
|
|
|
|
struct mbuf *m;
|
|
|
|
int error, len, flags;
|
|
|
|
struct socket *so;
|
|
|
|
struct sockaddr *sa;
|
|
|
|
struct uio auio;
|
|
|
|
|
|
|
|
sa = NULL;
|
|
|
|
while (conn->wdg_so) { /* not a loop */
|
|
|
|
so = conn->wdg_so;
|
|
|
|
auio.uio_resid = len = 1000000;
|
2003-02-26 21:25:55 +00:00
|
|
|
auio.uio_td = curthread;
|
1999-10-02 04:06:24 +00:00
|
|
|
flags = MSG_DONTWAIT;
|
soreceive_generic(), and sopoll_generic(). Add new functions sosend(),
soreceive(), and sopoll(), which are wrappers for pru_sosend,
pru_soreceive, and pru_sopoll, and are now used univerally by socket
consumers rather than either directly invoking the old so*() functions
or directly invoking the protocol switch method (about an even split
prior to this commit).
This completes an architectural change that was begun in 1996 to permit
protocols to provide substitute implementations, as now used by UDP.
Consumers now uniformly invoke sosend(), soreceive(), and sopoll() to
perform these operations on sockets -- in particular, distributed file
systems and socket system calls.
Architectural head nod: sam, gnn, wollman
2006-07-24 15:20:08 +00:00
|
|
|
error = soreceive(so, (struct sockaddr**)&sa, &auio, &m,
|
|
|
|
(struct mbuf**)0, &flags);
|
1999-10-02 04:06:24 +00:00
|
|
|
if (error) break;
|
|
|
|
len -= auio.uio_resid;
|
|
|
|
NCPSDEBUG("got watch dog %d\n",len);
|
|
|
|
if (len != 2) break;
|
|
|
|
buf = mtod(m, char*);
|
|
|
|
if (buf[1] != '?') break;
|
|
|
|
buf[1] = 'Y';
|
soreceive_generic(), and sopoll_generic(). Add new functions sosend(),
soreceive(), and sopoll(), which are wrappers for pru_sosend,
pru_soreceive, and pru_sopoll, and are now used univerally by socket
consumers rather than either directly invoking the old so*() functions
or directly invoking the protocol switch method (about an even split
prior to this commit).
This completes an architectural change that was begun in 1996 to permit
protocols to provide substitute implementations, as now used by UDP.
Consumers now uniformly invoke sosend(), soreceive(), and sopoll() to
perform these operations on sockets -- in particular, distributed file
systems and socket system calls.
Architectural head nod: sam, gnn, wollman
2006-07-24 15:20:08 +00:00
|
|
|
error = sosend(so, (struct sockaddr*)sa, 0, m, 0, 0, curthread);
|
1999-10-02 04:06:24 +00:00
|
|
|
NCPSDEBUG("send watch dog %d\n",error);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (sa) FREE(sa, M_SONAME);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
ncp_check_conn(struct ncp_conn *conn) {
|
|
|
|
int s;
|
|
|
|
|
|
|
|
if (conn == NULL || !(conn->flags & NCPFL_ATTACHED))
|
|
|
|
return;
|
|
|
|
s = splnet();
|
|
|
|
ncp_check_rq(conn);
|
|
|
|
splx(s);
|
2001-03-10 05:24:45 +00:00
|
|
|
if (conn->li.saddr.sa_family == AF_IPX)
|
|
|
|
ncp_watchdog(conn);
|
1999-10-02 04:06:24 +00:00
|
|
|
}
|