freebsd-skq/usr.sbin/ppp/fsm.c
Pedro F. Giffuni 1de7b4b805 various: general adoption of SPDX licensing ID tags.
Mainly focus on files that use BSD 2-Clause license, however the tool I
was using misidentified many licenses so this was mostly a manual - error
prone - task.

The Software Package Data Exchange (SPDX) group provides a specification
to make it easier for automated tools to detect and summarize well known
opensource licenses. We are gradually adopting the specification, noting
that the tags are considered only advisory and do not, in any way,
superceed or replace the license texts.

No functional change intended.
2017-11-27 15:37:16 +00:00

1216 lines
31 KiB
C

/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 1996 - 2001 Brian Somers <brian@Awfulhak.org>
* based on work by Toshiharu OHNO <tony-o@iij.ad.jp>
* Internet Initiative Japan, Inc (IIJ)
* 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.
*
* 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.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <string.h>
#include <termios.h>
#include "layer.h"
#include "ua.h"
#include "mbuf.h"
#include "log.h"
#include "defs.h"
#include "timer.h"
#include "fsm.h"
#include "iplist.h"
#include "lqr.h"
#include "hdlc.h"
#include "throughput.h"
#include "slcompress.h"
#include "ncpaddr.h"
#include "ipcp.h"
#include "filter.h"
#include "descriptor.h"
#include "lcp.h"
#include "ccp.h"
#include "link.h"
#include "mp.h"
#ifndef NORADIUS
#include "radius.h"
#endif
#include "ipv6cp.h"
#include "ncp.h"
#include "bundle.h"
#include "async.h"
#include "physical.h"
#include "proto.h"
static void FsmSendConfigReq(struct fsm *);
static void FsmSendTerminateReq(struct fsm *);
static void FsmInitRestartCounter(struct fsm *, int);
typedef void (recvfn)(struct fsm *, struct fsmheader *, struct mbuf *);
static recvfn FsmRecvConfigReq, FsmRecvConfigAck, FsmRecvConfigNak,
FsmRecvConfigRej, FsmRecvTermReq, FsmRecvTermAck,
FsmRecvCodeRej, FsmRecvProtoRej, FsmRecvEchoReq,
FsmRecvEchoRep, FsmRecvDiscReq, FsmRecvIdent,
FsmRecvTimeRemain, FsmRecvResetReq, FsmRecvResetAck;
static const struct fsmcodedesc {
recvfn *recv;
unsigned check_reqid : 1;
unsigned inc_reqid : 1;
const char *name;
} FsmCodes[] = {
{ FsmRecvConfigReq, 0, 0, "ConfigReq" },
{ FsmRecvConfigAck, 1, 1, "ConfigAck" },
{ FsmRecvConfigNak, 1, 1, "ConfigNak" },
{ FsmRecvConfigRej, 1, 1, "ConfigRej" },
{ FsmRecvTermReq, 0, 0, "TerminateReq" },
{ FsmRecvTermAck, 1, 1, "TerminateAck" },
{ FsmRecvCodeRej, 0, 0, "CodeRej" },
{ FsmRecvProtoRej, 0, 0, "ProtocolRej" },
{ FsmRecvEchoReq, 0, 0, "EchoRequest" },
{ FsmRecvEchoRep, 0, 0, "EchoReply" },
{ FsmRecvDiscReq, 0, 0, "DiscardReq" },
{ FsmRecvIdent, 0, 1, "Ident" },
{ FsmRecvTimeRemain,0, 0, "TimeRemain" },
{ FsmRecvResetReq, 0, 0, "ResetReq" },
{ FsmRecvResetAck, 0, 1, "ResetAck" }
};
static const char *
Code2Nam(u_int code)
{
if (code == 0 || code > sizeof FsmCodes / sizeof FsmCodes[0])
return "Unknown";
return FsmCodes[code-1].name;
}
const char *
State2Nam(u_int state)
{
static const char * const StateNames[] = {
"Initial", "Starting", "Closed", "Stopped", "Closing", "Stopping",
"Req-Sent", "Ack-Rcvd", "Ack-Sent", "Opened",
};
if (state >= sizeof StateNames / sizeof StateNames[0])
return "unknown";
return StateNames[state];
}
static void
StoppedTimeout(void *v)
{
struct fsm *fp = (struct fsm *)v;
log_Printf(fp->LogLevel, "%s: Stopped timer expired\n", fp->link->name);
if (fp->OpenTimer.state == TIMER_RUNNING) {
log_Printf(LogWARN, "%s: %s: aborting open delay due to stopped timer\n",
fp->link->name, fp->name);
timer_Stop(&fp->OpenTimer);
}
if (fp->state == ST_STOPPED)
fsm2initial(fp);
}
void
fsm_Init(struct fsm *fp, const char *name, u_short proto, int mincode,
int maxcode, int LogLevel, struct bundle *bundle,
struct link *l, const struct fsm_parent *parent,
struct fsm_callbacks *fn, const char * const timer_names[3])
{
fp->name = name;
fp->proto = proto;
fp->min_code = mincode;
fp->max_code = maxcode;
fp->state = fp->min_code > CODE_TERMACK ? ST_OPENED : ST_INITIAL;
fp->reqid = 1;
fp->restart = 1;
fp->more.reqs = fp->more.naks = fp->more.rejs = 3;
memset(&fp->FsmTimer, '\0', sizeof fp->FsmTimer);
memset(&fp->OpenTimer, '\0', sizeof fp->OpenTimer);
memset(&fp->StoppedTimer, '\0', sizeof fp->StoppedTimer);
fp->LogLevel = LogLevel;
fp->link = l;
fp->bundle = bundle;
fp->parent = parent;
fp->fn = fn;
fp->FsmTimer.name = timer_names[0];
fp->OpenTimer.name = timer_names[1];
fp->StoppedTimer.name = timer_names[2];
}
static void
NewState(struct fsm *fp, int new)
{
log_Printf(fp->LogLevel, "%s: State change %s --> %s\n",
fp->link->name, State2Nam(fp->state), State2Nam(new));
if (fp->state == ST_STOPPED && fp->StoppedTimer.state == TIMER_RUNNING)
timer_Stop(&fp->StoppedTimer);
fp->state = new;
if ((new >= ST_INITIAL && new <= ST_STOPPED) || (new == ST_OPENED)) {
timer_Stop(&fp->FsmTimer);
if (new == ST_STOPPED && fp->StoppedTimer.load) {
timer_Stop(&fp->StoppedTimer);
fp->StoppedTimer.func = StoppedTimeout;
fp->StoppedTimer.arg = (void *) fp;
timer_Start(&fp->StoppedTimer);
}
}
}
void
fsm_Output(struct fsm *fp, u_int code, u_int id, u_char *ptr, unsigned count,
int mtype)
{
int plen;
struct fsmheader lh;
struct mbuf *bp;
if (log_IsKept(fp->LogLevel)) {
log_Printf(fp->LogLevel, "%s: Send%s(%d) state = %s\n",
fp->link->name, Code2Nam(code), id, State2Nam(fp->state));
switch (code) {
case CODE_CONFIGREQ:
case CODE_CONFIGACK:
case CODE_CONFIGREJ:
case CODE_CONFIGNAK:
(*fp->fn->DecodeConfig)(fp, ptr, ptr + count, MODE_NOP, NULL);
if (count < sizeof(struct fsm_opt_hdr))
log_Printf(fp->LogLevel, " [EMPTY]\n");
break;
}
}
plen = sizeof(struct fsmheader) + count;
lh.code = code;
lh.id = id;
lh.length = htons(plen);
bp = m_get(plen, mtype);
memcpy(MBUF_CTOP(bp), &lh, sizeof(struct fsmheader));
if (count)
memcpy(MBUF_CTOP(bp) + sizeof(struct fsmheader), ptr, count);
log_DumpBp(LogDEBUG, "fsm_Output", bp);
link_PushPacket(fp->link, bp, fp->bundle, LINK_QUEUES(fp->link) - 1,
fp->proto);
if (code == CODE_CONFIGREJ)
lcp_SendIdentification(&fp->link->lcp);
}
static void
FsmOpenNow(void *v)
{
struct fsm *fp = (struct fsm *)v;
timer_Stop(&fp->OpenTimer);
if (fp->state <= ST_STOPPED) {
if (fp->state != ST_STARTING) {
/*
* In practice, we're only here in ST_STOPPED (when delaying the
* first config request) or ST_CLOSED (when openmode == 0).
*
* The ST_STOPPED bit is breaking the RFC already :-(
*
* According to the RFC (1661) state transition table, a TLS isn't
* required for an Open event when state == Closed, but the RFC
* must be wrong as TLS hasn't yet been called (since the last TLF)
* ie, Initial gets an `Up' event, Closing gets a RTA etc.
*/
(*fp->fn->LayerStart)(fp);
(*fp->parent->LayerStart)(fp->parent->object, fp);
}
FsmInitRestartCounter(fp, FSM_REQ_TIMER);
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
}
}
void
fsm_Open(struct fsm *fp)
{
switch (fp->state) {
case ST_INITIAL:
NewState(fp, ST_STARTING);
(*fp->fn->LayerStart)(fp);
(*fp->parent->LayerStart)(fp->parent->object, fp);
break;
case ST_CLOSED:
if (fp->open_mode == OPEN_PASSIVE) {
NewState(fp, ST_STOPPED); /* XXX: This is a hack ! */
} else if (fp->open_mode > 0) {
if (fp->open_mode > 1)
log_Printf(LogPHASE, "%s: Entering STOPPED state for %d seconds\n",
fp->link->name, fp->open_mode);
NewState(fp, ST_STOPPED); /* XXX: This is a not-so-bad hack ! */
timer_Stop(&fp->OpenTimer);
fp->OpenTimer.load = fp->open_mode * SECTICKS;
fp->OpenTimer.func = FsmOpenNow;
fp->OpenTimer.arg = (void *)fp;
timer_Start(&fp->OpenTimer);
} else
FsmOpenNow(fp);
break;
case ST_STOPPED: /* XXX: restart option */
case ST_REQSENT:
case ST_ACKRCVD:
case ST_ACKSENT:
case ST_OPENED: /* XXX: restart option */
break;
case ST_CLOSING: /* XXX: restart option */
case ST_STOPPING: /* XXX: restart option */
NewState(fp, ST_STOPPING);
break;
}
}
void
fsm_Up(struct fsm *fp)
{
switch (fp->state) {
case ST_INITIAL:
log_Printf(fp->LogLevel, "FSM: Using \"%s\" as a transport\n",
fp->link->name);
NewState(fp, ST_CLOSED);
break;
case ST_STARTING:
FsmInitRestartCounter(fp, FSM_REQ_TIMER);
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
break;
default:
log_Printf(fp->LogLevel, "%s: Oops, Up at %s\n",
fp->link->name, State2Nam(fp->state));
break;
}
}
void
fsm_Down(struct fsm *fp)
{
switch (fp->state) {
case ST_CLOSED:
NewState(fp, ST_INITIAL);
break;
case ST_CLOSING:
/* This TLF contradicts the RFC (1661), which ``misses it out'' ! */
(*fp->fn->LayerFinish)(fp);
NewState(fp, ST_INITIAL);
(*fp->parent->LayerFinish)(fp->parent->object, fp);
break;
case ST_STOPPED:
NewState(fp, ST_STARTING);
(*fp->fn->LayerStart)(fp);
(*fp->parent->LayerStart)(fp->parent->object, fp);
break;
case ST_STOPPING:
case ST_REQSENT:
case ST_ACKRCVD:
case ST_ACKSENT:
NewState(fp, ST_STARTING);
break;
case ST_OPENED:
(*fp->fn->LayerDown)(fp);
NewState(fp, ST_STARTING);
(*fp->parent->LayerDown)(fp->parent->object, fp);
break;
}
}
void
fsm_Close(struct fsm *fp)
{
switch (fp->state) {
case ST_STARTING:
(*fp->fn->LayerFinish)(fp);
NewState(fp, ST_INITIAL);
(*fp->parent->LayerFinish)(fp->parent->object, fp);
break;
case ST_STOPPED:
NewState(fp, ST_CLOSED);
break;
case ST_STOPPING:
NewState(fp, ST_CLOSING);
break;
case ST_OPENED:
(*fp->fn->LayerDown)(fp);
if (fp->state == ST_OPENED) {
FsmInitRestartCounter(fp, FSM_TRM_TIMER);
FsmSendTerminateReq(fp);
NewState(fp, ST_CLOSING);
(*fp->parent->LayerDown)(fp->parent->object, fp);
}
break;
case ST_REQSENT:
case ST_ACKRCVD:
case ST_ACKSENT:
FsmInitRestartCounter(fp, FSM_TRM_TIMER);
FsmSendTerminateReq(fp);
NewState(fp, ST_CLOSING);
break;
}
}
/*
* Send functions
*/
static void
FsmSendConfigReq(struct fsm *fp)
{
if (fp->more.reqs-- > 0 && fp->restart-- > 0) {
(*fp->fn->SendConfigReq)(fp);
timer_Start(&fp->FsmTimer); /* Start restart timer */
} else {
if (fp->more.reqs < 0)
log_Printf(LogPHASE, "%s: Too many %s REQs sent - abandoning "
"negotiation\n", fp->link->name, fp->name);
lcp_SendIdentification(&fp->link->lcp);
fsm_Close(fp);
}
}
static void
FsmSendTerminateReq(struct fsm *fp)
{
fsm_Output(fp, CODE_TERMREQ, fp->reqid, NULL, 0, MB_UNKNOWN);
(*fp->fn->SentTerminateReq)(fp);
timer_Start(&fp->FsmTimer); /* Start restart timer */
fp->restart--; /* Decrement restart counter */
}
/*
* Timeout actions
*/
static void
FsmTimeout(void *v)
{
struct fsm *fp = (struct fsm *)v;
if (fp->restart) {
switch (fp->state) {
case ST_CLOSING:
case ST_STOPPING:
FsmSendTerminateReq(fp);
break;
case ST_REQSENT:
case ST_ACKSENT:
FsmSendConfigReq(fp);
break;
case ST_ACKRCVD:
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
break;
}
timer_Start(&fp->FsmTimer);
} else {
switch (fp->state) {
case ST_CLOSING:
(*fp->fn->LayerFinish)(fp);
NewState(fp, ST_CLOSED);
(*fp->parent->LayerFinish)(fp->parent->object, fp);
break;
case ST_STOPPING:
(*fp->fn->LayerFinish)(fp);
NewState(fp, ST_STOPPED);
(*fp->parent->LayerFinish)(fp->parent->object, fp);
break;
case ST_REQSENT: /* XXX: 3p */
case ST_ACKSENT:
case ST_ACKRCVD:
(*fp->fn->LayerFinish)(fp);
NewState(fp, ST_STOPPED);
(*fp->parent->LayerFinish)(fp->parent->object, fp);
break;
}
}
}
static void
FsmInitRestartCounter(struct fsm *fp, int what)
{
timer_Stop(&fp->FsmTimer);
fp->FsmTimer.func = FsmTimeout;
fp->FsmTimer.arg = (void *)fp;
(*fp->fn->InitRestartCounter)(fp, what);
}
/*
* Actions when receive packets
*/
static void
FsmRecvConfigReq(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
/* RCR */
{
struct fsm_decode dec;
int plen, flen;
int ackaction = 0;
u_char *cp;
bp = m_pullup(bp);
plen = m_length(bp);
flen = ntohs(lhp->length) - sizeof *lhp;
if (plen < flen) {
log_Printf(LogWARN, "%s: FsmRecvConfigReq: plen (%d) < flen (%d)\n",
fp->link->name, plen, flen);
m_freem(bp);
return;
}
/* Some things must be done before we Decode the packet */
switch (fp->state) {
case ST_OPENED:
(*fp->fn->LayerDown)(fp);
}
dec.ackend = dec.ack;
dec.nakend = dec.nak;
dec.rejend = dec.rej;
cp = MBUF_CTOP(bp);
(*fp->fn->DecodeConfig)(fp, cp, cp + flen, MODE_REQ, &dec);
if (flen < (int)sizeof(struct fsm_opt_hdr))
log_Printf(fp->LogLevel, " [EMPTY]\n");
if (dec.nakend == dec.nak && dec.rejend == dec.rej)
ackaction = 1;
/* Check and process easy case */
switch (fp->state) {
case ST_INITIAL:
if (fp->proto == PROTO_CCP && fp->link->lcp.fsm.state == ST_OPENED) {
/*
* ccp_SetOpenMode() leaves us in initial if we're disabling
* & denying everything.
*/
bp = m_prepend(bp, lhp, sizeof *lhp, 2);
bp = proto_Prepend(bp, fp->proto, 0, 0);
bp = m_pullup(bp);
lcp_SendProtoRej(&fp->link->lcp, MBUF_CTOP(bp), bp->m_len);
m_freem(bp);
return;
}
/* Drop through */
case ST_STARTING:
log_Printf(fp->LogLevel, "%s: Oops, RCR in %s.\n",
fp->link->name, State2Nam(fp->state));
m_freem(bp);
return;
case ST_CLOSED:
(*fp->fn->SendTerminateAck)(fp, lhp->id);
m_freem(bp);
return;
case ST_CLOSING:
log_Printf(fp->LogLevel, "%s: Error: Got ConfigReq while state = %s\n",
fp->link->name, State2Nam(fp->state));
case ST_STOPPING:
m_freem(bp);
return;
case ST_STOPPED:
FsmInitRestartCounter(fp, FSM_REQ_TIMER);
/* Drop through */
case ST_OPENED:
FsmSendConfigReq(fp);
break;
}
if (dec.rejend != dec.rej)
fsm_Output(fp, CODE_CONFIGREJ, lhp->id, dec.rej, dec.rejend - dec.rej,
MB_UNKNOWN);
if (dec.nakend != dec.nak)
fsm_Output(fp, CODE_CONFIGNAK, lhp->id, dec.nak, dec.nakend - dec.nak,
MB_UNKNOWN);
if (ackaction)
fsm_Output(fp, CODE_CONFIGACK, lhp->id, dec.ack, dec.ackend - dec.ack,
MB_UNKNOWN);
switch (fp->state) {
case ST_STOPPED:
/*
* According to the RFC (1661) state transition table, a TLS isn't
* required for a RCR when state == ST_STOPPED, but the RFC
* must be wrong as TLS hasn't yet been called (since the last TLF)
*/
(*fp->fn->LayerStart)(fp);
(*fp->parent->LayerStart)(fp->parent->object, fp);
/* FALLTHROUGH */
case ST_OPENED:
if (ackaction)
NewState(fp, ST_ACKSENT);
else
NewState(fp, ST_REQSENT);
(*fp->parent->LayerDown)(fp->parent->object, fp);
break;
case ST_REQSENT:
if (ackaction)
NewState(fp, ST_ACKSENT);
break;
case ST_ACKRCVD:
if (ackaction) {
NewState(fp, ST_OPENED);
if ((*fp->fn->LayerUp)(fp))
(*fp->parent->LayerUp)(fp->parent->object, fp);
else {
(*fp->fn->LayerDown)(fp);
FsmInitRestartCounter(fp, FSM_TRM_TIMER);
FsmSendTerminateReq(fp);
NewState(fp, ST_CLOSING);
lcp_SendIdentification(&fp->link->lcp);
}
}
break;
case ST_ACKSENT:
if (!ackaction)
NewState(fp, ST_REQSENT);
break;
}
m_freem(bp);
if (dec.rejend != dec.rej && --fp->more.rejs <= 0) {
log_Printf(LogPHASE, "%s: Too many %s REJs sent - abandoning negotiation\n",
fp->link->name, fp->name);
lcp_SendIdentification(&fp->link->lcp);
fsm_Close(fp);
}
if (dec.nakend != dec.nak && --fp->more.naks <= 0) {
log_Printf(LogPHASE, "%s: Too many %s NAKs sent - abandoning negotiation\n",
fp->link->name, fp->name);
lcp_SendIdentification(&fp->link->lcp);
fsm_Close(fp);
}
}
static void
FsmRecvConfigAck(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
/* RCA */
{
struct fsm_decode dec;
int plen, flen;
u_char *cp;
plen = m_length(bp);
flen = ntohs(lhp->length) - sizeof *lhp;
if (plen < flen) {
m_freem(bp);
return;
}
bp = m_pullup(bp);
dec.ackend = dec.ack;
dec.nakend = dec.nak;
dec.rejend = dec.rej;
cp = MBUF_CTOP(bp);
(*fp->fn->DecodeConfig)(fp, cp, cp + flen, MODE_ACK, &dec);
if (flen < (int)sizeof(struct fsm_opt_hdr))
log_Printf(fp->LogLevel, " [EMPTY]\n");
switch (fp->state) {
case ST_CLOSED:
case ST_STOPPED:
(*fp->fn->SendTerminateAck)(fp, lhp->id);
break;
case ST_CLOSING:
case ST_STOPPING:
break;
case ST_REQSENT:
FsmInitRestartCounter(fp, FSM_REQ_TIMER);
NewState(fp, ST_ACKRCVD);
break;
case ST_ACKRCVD:
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
break;
case ST_ACKSENT:
FsmInitRestartCounter(fp, FSM_REQ_TIMER);
NewState(fp, ST_OPENED);
if ((*fp->fn->LayerUp)(fp))
(*fp->parent->LayerUp)(fp->parent->object, fp);
else {
(*fp->fn->LayerDown)(fp);
FsmInitRestartCounter(fp, FSM_TRM_TIMER);
FsmSendTerminateReq(fp);
NewState(fp, ST_CLOSING);
lcp_SendIdentification(&fp->link->lcp);
}
break;
case ST_OPENED:
(*fp->fn->LayerDown)(fp);
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
(*fp->parent->LayerDown)(fp->parent->object, fp);
break;
}
m_freem(bp);
}
static void
FsmRecvConfigNak(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
/* RCN */
{
struct fsm_decode dec;
int plen, flen;
u_char *cp;
plen = m_length(bp);
flen = ntohs(lhp->length) - sizeof *lhp;
if (plen < flen) {
m_freem(bp);
return;
}
/*
* Check and process easy case
*/
switch (fp->state) {
case ST_INITIAL:
case ST_STARTING:
log_Printf(fp->LogLevel, "%s: Oops, RCN in %s.\n",
fp->link->name, State2Nam(fp->state));
m_freem(bp);
return;
case ST_CLOSED:
case ST_STOPPED:
(*fp->fn->SendTerminateAck)(fp, lhp->id);
m_freem(bp);
return;
case ST_CLOSING:
case ST_STOPPING:
m_freem(bp);
return;
}
bp = m_pullup(bp);
dec.ackend = dec.ack;
dec.nakend = dec.nak;
dec.rejend = dec.rej;
cp = MBUF_CTOP(bp);
(*fp->fn->DecodeConfig)(fp, cp, cp + flen, MODE_NAK, &dec);
if (flen < (int)sizeof(struct fsm_opt_hdr))
log_Printf(fp->LogLevel, " [EMPTY]\n");
switch (fp->state) {
case ST_REQSENT:
case ST_ACKSENT:
FsmInitRestartCounter(fp, FSM_REQ_TIMER);
FsmSendConfigReq(fp);
break;
case ST_OPENED:
(*fp->fn->LayerDown)(fp);
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
(*fp->parent->LayerDown)(fp->parent->object, fp);
break;
case ST_ACKRCVD:
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
break;
}
m_freem(bp);
}
static void
FsmRecvTermReq(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
/* RTR */
{
switch (fp->state) {
case ST_INITIAL:
case ST_STARTING:
log_Printf(fp->LogLevel, "%s: Oops, RTR in %s\n",
fp->link->name, State2Nam(fp->state));
break;
case ST_CLOSED:
case ST_STOPPED:
case ST_CLOSING:
case ST_STOPPING:
case ST_REQSENT:
(*fp->fn->SendTerminateAck)(fp, lhp->id);
break;
case ST_ACKRCVD:
case ST_ACKSENT:
(*fp->fn->SendTerminateAck)(fp, lhp->id);
NewState(fp, ST_REQSENT);
break;
case ST_OPENED:
(*fp->fn->LayerDown)(fp);
(*fp->fn->SendTerminateAck)(fp, lhp->id);
FsmInitRestartCounter(fp, FSM_TRM_TIMER);
timer_Start(&fp->FsmTimer); /* Start restart timer */
fp->restart = 0;
NewState(fp, ST_STOPPING);
(*fp->parent->LayerDown)(fp->parent->object, fp);
/* A delayed ST_STOPPED is now scheduled */
break;
}
m_freem(bp);
}
static void
FsmRecvTermAck(struct fsm *fp, struct fsmheader *lhp __unused, struct mbuf *bp)
/* RTA */
{
switch (fp->state) {
case ST_CLOSING:
(*fp->fn->LayerFinish)(fp);
NewState(fp, ST_CLOSED);
(*fp->parent->LayerFinish)(fp->parent->object, fp);
break;
case ST_STOPPING:
(*fp->fn->LayerFinish)(fp);
NewState(fp, ST_STOPPED);
(*fp->parent->LayerFinish)(fp->parent->object, fp);
break;
case ST_ACKRCVD:
NewState(fp, ST_REQSENT);
break;
case ST_OPENED:
(*fp->fn->LayerDown)(fp);
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
(*fp->parent->LayerDown)(fp->parent->object, fp);
break;
}
m_freem(bp);
}
static void
FsmRecvConfigRej(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
/* RCJ */
{
struct fsm_decode dec;
size_t plen;
int flen;
u_char *cp;
plen = m_length(bp);
flen = ntohs(lhp->length) - sizeof *lhp;
if ((int)plen < flen) {
m_freem(bp);
return;
}
lcp_SendIdentification(&fp->link->lcp);
/*
* Check and process easy case
*/
switch (fp->state) {
case ST_INITIAL:
case ST_STARTING:
log_Printf(fp->LogLevel, "%s: Oops, RCJ in %s.\n",
fp->link->name, State2Nam(fp->state));
m_freem(bp);
return;
case ST_CLOSED:
case ST_STOPPED:
(*fp->fn->SendTerminateAck)(fp, lhp->id);
m_freem(bp);
return;
case ST_CLOSING:
case ST_STOPPING:
m_freem(bp);
return;
}
bp = m_pullup(bp);
dec.ackend = dec.ack;
dec.nakend = dec.nak;
dec.rejend = dec.rej;
cp = MBUF_CTOP(bp);
(*fp->fn->DecodeConfig)(fp, cp, cp + flen, MODE_REJ, &dec);
if (flen < (int)sizeof(struct fsm_opt_hdr))
log_Printf(fp->LogLevel, " [EMPTY]\n");
switch (fp->state) {
case ST_REQSENT:
case ST_ACKSENT:
FsmInitRestartCounter(fp, FSM_REQ_TIMER);
FsmSendConfigReq(fp);
break;
case ST_OPENED:
(*fp->fn->LayerDown)(fp);
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
(*fp->parent->LayerDown)(fp->parent->object, fp);
break;
case ST_ACKRCVD:
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
break;
}
m_freem(bp);
}
static void
FsmRecvCodeRej(struct fsm *fp __unused, struct fsmheader *lhp __unused,
struct mbuf *bp)
{
m_freem(bp);
}
static void
FsmRecvProtoRej(struct fsm *fp, struct fsmheader *lhp __unused, struct mbuf *bp)
{
struct physical *p = link2physical(fp->link);
u_short proto;
if (m_length(bp) < 2) {
m_freem(bp);
return;
}
bp = mbuf_Read(bp, &proto, 2);
proto = ntohs(proto);
log_Printf(fp->LogLevel, "%s: -- Protocol 0x%04x (%s) was rejected!\n",
fp->link->name, proto, hdlc_Protocol2Nam(proto));
switch (proto) {
case PROTO_LQR:
if (p)
lqr_Stop(p, LQM_LQR);
else
log_Printf(LogERROR, "%s: FsmRecvProtoRej: Not a physical link !\n",
fp->link->name);
break;
case PROTO_CCP:
if (fp->proto == PROTO_LCP) {
fp = &fp->link->ccp.fsm;
/* Despite the RFC (1661), don't do an out-of-place TLF */
/* (*fp->fn->LayerFinish)(fp); */
switch (fp->state) {
case ST_CLOSED:
case ST_CLOSING:
NewState(fp, ST_CLOSED);
break;
default:
NewState(fp, ST_STOPPED);
break;
}
/* See above */
/* (*fp->parent->LayerFinish)(fp->parent->object, fp); */
}
break;
case PROTO_IPCP:
if (fp->proto == PROTO_LCP) {
log_Printf(LogPHASE, "%s: IPCP protocol reject closes IPCP !\n",
fp->link->name);
fsm_Close(&fp->bundle->ncp.ipcp.fsm);
}
break;
#ifndef NOINET6
case PROTO_IPV6CP:
if (fp->proto == PROTO_LCP) {
log_Printf(LogPHASE, "%s: IPV6CP protocol reject closes IPV6CP !\n",
fp->link->name);
fsm_Close(&fp->bundle->ncp.ipv6cp.fsm);
}
break;
#endif
case PROTO_MP:
if (fp->proto == PROTO_LCP) {
struct lcp *lcp = fsm2lcp(fp);
if (lcp->want_mrru && lcp->his_mrru) {
log_Printf(LogPHASE, "%s: MP protocol reject is fatal !\n",
fp->link->name);
fsm_Close(fp);
}
}
break;
}
m_freem(bp);
}
static void
FsmRecvEchoReq(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
{
struct lcp *lcp = fsm2lcp(fp);
u_char *cp;
u_int32_t magic;
bp = m_pullup(bp);
m_settype(bp, MB_ECHOIN);
if (lcp && ntohs(lhp->length) - sizeof *lhp >= 4) {
cp = MBUF_CTOP(bp);
ua_ntohl(cp, &magic);
if (magic != lcp->his_magic) {
log_Printf(fp->LogLevel, "%s: RecvEchoReq: magic 0x%08lx is wrong,"
" expecting 0x%08lx\n", fp->link->name, (u_long)magic,
(u_long)lcp->his_magic);
/* XXX: We should send terminate request */
}
if (fp->state == ST_OPENED) {
ua_htonl(&lcp->want_magic, cp); /* local magic */
fsm_Output(fp, CODE_ECHOREP, lhp->id, cp,
ntohs(lhp->length) - sizeof *lhp, MB_ECHOOUT);
}
}
m_freem(bp);
}
static void
FsmRecvEchoRep(struct fsm *fp, struct fsmheader *lhp __unused, struct mbuf *bp)
{
if (fsm2lcp(fp))
bp = lqr_RecvEcho(fp, bp);
m_freem(bp);
}
static void
FsmRecvDiscReq(struct fsm *fp __unused, struct fsmheader *lhp __unused,
struct mbuf *bp)
{
m_freem(bp);
}
static void
FsmRecvIdent(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
{
u_int32_t magic;
u_short len;
u_char *cp;
len = ntohs(lhp->length) - sizeof *lhp;
if (len >= 4) {
bp = m_pullup(m_append(bp, "", 1));
cp = MBUF_CTOP(bp);
ua_ntohl(cp, &magic);
if (magic != fp->link->lcp.his_magic)
log_Printf(fp->LogLevel, "%s: RecvIdent: magic 0x%08lx is wrong,"
" expecting 0x%08lx\n", fp->link->name, (u_long)magic,
(u_long)fp->link->lcp.his_magic);
cp[len] = '\0';
lcp_RecvIdentification(&fp->link->lcp, cp + 4);
}
m_freem(bp);
}
static void
FsmRecvTimeRemain(struct fsm *fp __unused, struct fsmheader *lhp __unused,
struct mbuf *bp)
{
m_freem(bp);
}
static void
FsmRecvResetReq(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
{
if ((*fp->fn->RecvResetReq)(fp)) {
/*
* All sendable compressed packets are queued in the first (lowest
* priority) modem output queue.... dump 'em to the priority queue
* so that they arrive at the peer before our ResetAck.
*/
link_SequenceQueue(fp->link);
fsm_Output(fp, CODE_RESETACK, lhp->id, NULL, 0, MB_CCPOUT);
}
m_freem(bp);
}
static void
FsmRecvResetAck(struct fsm *fp, struct fsmheader *lhp, struct mbuf *bp)
{
(*fp->fn->RecvResetAck)(fp, lhp->id);
m_freem(bp);
}
void
fsm_Input(struct fsm *fp, struct mbuf *bp)
{
size_t len;
struct fsmheader lh;
const struct fsmcodedesc *codep;
len = m_length(bp);
if (len < sizeof(struct fsmheader)) {
m_freem(bp);
return;
}
bp = mbuf_Read(bp, &lh, sizeof lh);
if (ntohs(lh.length) > len) {
log_Printf(LogWARN, "%s: Oops: Got %zu bytes but %d byte payload "
"- dropped\n", fp->link->name, len, (int)ntohs(lh.length));
m_freem(bp);
return;
}
if (lh.code < fp->min_code || lh.code > fp->max_code ||
lh.code > sizeof FsmCodes / sizeof *FsmCodes) {
/*
* Use a private id. This is really a response-type packet, but we
* MUST send a unique id for each REQ....
*/
static u_char id;
bp = m_prepend(bp, &lh, sizeof lh, 0);
bp = m_pullup(bp);
fsm_Output(fp, CODE_CODEREJ, id++, MBUF_CTOP(bp), bp->m_len, MB_UNKNOWN);
m_freem(bp);
return;
}
codep = FsmCodes + lh.code - 1;
if (lh.id != fp->reqid && codep->check_reqid &&
Enabled(fp->bundle, OPT_IDCHECK)) {
log_Printf(fp->LogLevel, "%s: Recv%s(%d), dropped (expected %d)\n",
fp->link->name, codep->name, lh.id, fp->reqid);
return;
}
log_Printf(fp->LogLevel, "%s: Recv%s(%d) state = %s\n",
fp->link->name, codep->name, lh.id, State2Nam(fp->state));
if (codep->inc_reqid && (lh.id == fp->reqid ||
(!Enabled(fp->bundle, OPT_IDCHECK) && codep->check_reqid)))
fp->reqid++; /* That's the end of that ``exchange''.... */
(*codep->recv)(fp, &lh, bp);
}
int
fsm_NullRecvResetReq(struct fsm *fp)
{
log_Printf(fp->LogLevel, "%s: Oops - received unexpected reset req\n",
fp->link->name);
return 1;
}
void
fsm_NullRecvResetAck(struct fsm *fp, u_char id __unused)
{
log_Printf(fp->LogLevel, "%s: Oops - received unexpected reset ack\n",
fp->link->name);
}
void
fsm_Reopen(struct fsm *fp)
{
if (fp->state == ST_OPENED) {
(*fp->fn->LayerDown)(fp);
FsmInitRestartCounter(fp, FSM_REQ_TIMER);
FsmSendConfigReq(fp);
NewState(fp, ST_REQSENT);
(*fp->parent->LayerDown)(fp->parent->object, fp);
}
}
void
fsm2initial(struct fsm *fp)
{
timer_Stop(&fp->FsmTimer);
timer_Stop(&fp->OpenTimer);
timer_Stop(&fp->StoppedTimer);
if (fp->state == ST_STOPPED)
fsm_Close(fp);
if (fp->state > ST_INITIAL)
fsm_Down(fp);
if (fp->state > ST_INITIAL)
fsm_Close(fp);
}
struct fsm_opt *
fsm_readopt(u_char **cp)
{
struct fsm_opt *o = (struct fsm_opt *)*cp;
if (o->hdr.len < sizeof(struct fsm_opt_hdr)) {
log_Printf(LogERROR, "Bad option length %d (out of phase?)\n", o->hdr.len);
return NULL;
}
*cp += o->hdr.len;
if (o->hdr.len > sizeof(struct fsm_opt)) {
log_Printf(LogERROR, "Warning: Truncating option length from %d to %d\n",
o->hdr.len, (int)sizeof(struct fsm_opt));
o->hdr.len = sizeof(struct fsm_opt);
}
return o;
}
static int
fsm_opt(u_char *opt, int optlen, const struct fsm_opt *o)
{
unsigned cplen = o->hdr.len;
if (optlen < (int)sizeof(struct fsm_opt_hdr))
optlen = 0;
if ((int)cplen > optlen) {
log_Printf(LogERROR, "Can't REJ length %d - trunating to %d\n",
cplen, optlen);
cplen = optlen;
}
memcpy(opt, o, cplen);
if (cplen)
opt[1] = cplen;
return cplen;
}
void
fsm_rej(struct fsm_decode *dec, const struct fsm_opt *o)
{
if (!dec)
return;
dec->rejend += fsm_opt(dec->rejend, FSM_OPTLEN - (dec->rejend - dec->rej), o);
}
void
fsm_ack(struct fsm_decode *dec, const struct fsm_opt *o)
{
if (!dec)
return;
dec->ackend += fsm_opt(dec->ackend, FSM_OPTLEN - (dec->ackend - dec->ack), o);
}
void
fsm_nak(struct fsm_decode *dec, const struct fsm_opt *o)
{
if (!dec)
return;
dec->nakend += fsm_opt(dec->nakend, FSM_OPTLEN - (dec->nakend - dec->nak), o);
}
void
fsm_opt_normalise(struct fsm_decode *dec)
{
if (dec->rejend != dec->rej) {
/* rejects are preferred */
dec->ackend = dec->ack;
dec->nakend = dec->nak;
} else if (dec->nakend != dec->nak)
/* then NAKs */
dec->ackend = dec->ack;
}