819 lines
23 KiB
C
819 lines
23 KiB
C
/*-
|
|
* 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 <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h> /* memcpy() on some archs */
|
|
#include <termios.h>
|
|
|
|
#include "layer.h"
|
|
#include "defs.h"
|
|
#include "command.h"
|
|
#include "mbuf.h"
|
|
#include "log.h"
|
|
#include "timer.h"
|
|
#include "fsm.h"
|
|
#include "proto.h"
|
|
#include "pred.h"
|
|
#include "deflate.h"
|
|
#include "throughput.h"
|
|
#include "iplist.h"
|
|
#include "slcompress.h"
|
|
#include "lqr.h"
|
|
#include "hdlc.h"
|
|
#include "lcp.h"
|
|
#include "ccp.h"
|
|
#include "ncpaddr.h"
|
|
#include "ipcp.h"
|
|
#include "filter.h"
|
|
#include "descriptor.h"
|
|
#include "prompt.h"
|
|
#include "link.h"
|
|
#include "mp.h"
|
|
#include "async.h"
|
|
#include "physical.h"
|
|
#ifndef NORADIUS
|
|
#include "radius.h"
|
|
#endif
|
|
#ifndef NODES
|
|
#include "mppe.h"
|
|
#endif
|
|
#include "ipv6cp.h"
|
|
#include "ncp.h"
|
|
#include "bundle.h"
|
|
|
|
static void CcpSendConfigReq(struct fsm *);
|
|
static void CcpSentTerminateReq(struct fsm *);
|
|
static void CcpSendTerminateAck(struct fsm *, u_char);
|
|
static void CcpDecodeConfig(struct fsm *, u_char *, u_char *, int,
|
|
struct fsm_decode *);
|
|
static void CcpLayerStart(struct fsm *);
|
|
static void CcpLayerFinish(struct fsm *);
|
|
static int CcpLayerUp(struct fsm *);
|
|
static void CcpLayerDown(struct fsm *);
|
|
static void CcpInitRestartCounter(struct fsm *, int);
|
|
static int CcpRecvResetReq(struct fsm *);
|
|
static void CcpRecvResetAck(struct fsm *, u_char);
|
|
|
|
static struct fsm_callbacks ccp_Callbacks = {
|
|
CcpLayerUp,
|
|
CcpLayerDown,
|
|
CcpLayerStart,
|
|
CcpLayerFinish,
|
|
CcpInitRestartCounter,
|
|
CcpSendConfigReq,
|
|
CcpSentTerminateReq,
|
|
CcpSendTerminateAck,
|
|
CcpDecodeConfig,
|
|
CcpRecvResetReq,
|
|
CcpRecvResetAck
|
|
};
|
|
|
|
static const char * const ccp_TimerNames[] =
|
|
{"CCP restart", "CCP openmode", "CCP stopped"};
|
|
|
|
static const char *
|
|
protoname(int proto)
|
|
{
|
|
static char const * const cftypes[] = {
|
|
/* Check out the latest ``Compression Control Protocol'' rfc (1962) */
|
|
"OUI", /* 0: OUI */
|
|
"PRED1", /* 1: Predictor type 1 */
|
|
"PRED2", /* 2: Predictor type 2 */
|
|
"PUDDLE", /* 3: Puddle Jumber */
|
|
NULL, NULL, NULL, NULL, NULL, NULL,
|
|
NULL, NULL, NULL, NULL, NULL, NULL,
|
|
"HWPPC", /* 16: Hewlett-Packard PPC */
|
|
"STAC", /* 17: Stac Electronics LZS (rfc1974) */
|
|
"MPPE", /* 18: Microsoft PPC (rfc2118) and */
|
|
/* Microsoft PPE (draft-ietf-pppext-mppe) */
|
|
"GAND", /* 19: Gandalf FZA (rfc1993) */
|
|
"V42BIS", /* 20: ARG->DATA.42bis compression */
|
|
"BSD", /* 21: BSD LZW Compress */
|
|
NULL,
|
|
"LZS-DCP", /* 23: LZS-DCP Compression Protocol (rfc1967) */
|
|
"MAGNALINK/DEFLATE",/* 24: Magnalink Variable Resource (rfc1975) */
|
|
/* 24: Deflate (according to pppd-2.3.*) */
|
|
"DCE", /* 25: Data Circuit-Terminating Equip (rfc1976) */
|
|
"DEFLATE", /* 26: Deflate (rfc1979) */
|
|
};
|
|
|
|
if (proto < 0 || proto > sizeof cftypes / sizeof *cftypes ||
|
|
cftypes[proto] == NULL) {
|
|
if (proto == -1)
|
|
return "none";
|
|
return HexStr(proto, NULL, 0);
|
|
}
|
|
|
|
return cftypes[proto];
|
|
}
|
|
|
|
/* We support these algorithms, and Req them in the given order */
|
|
static const struct ccp_algorithm * const algorithm[] = {
|
|
&DeflateAlgorithm,
|
|
&Pred1Algorithm,
|
|
&PppdDeflateAlgorithm
|
|
#ifndef NODES
|
|
, &MPPEAlgorithm
|
|
#endif
|
|
};
|
|
|
|
#define NALGORITHMS (sizeof algorithm/sizeof algorithm[0])
|
|
|
|
int
|
|
ccp_ReportStatus(struct cmdargs const *arg)
|
|
{
|
|
struct ccp_opt **o;
|
|
struct link *l;
|
|
struct ccp *ccp;
|
|
int f;
|
|
|
|
l = command_ChooseLink(arg);
|
|
ccp = &l->ccp;
|
|
|
|
prompt_Printf(arg->prompt, "%s: %s [%s]\n", l->name, ccp->fsm.name,
|
|
State2Nam(ccp->fsm.state));
|
|
if (ccp->fsm.state == ST_OPENED) {
|
|
prompt_Printf(arg->prompt, " My protocol = %s, His protocol = %s\n",
|
|
protoname(ccp->my_proto), protoname(ccp->his_proto));
|
|
prompt_Printf(arg->prompt, " Output: %ld --> %ld, Input: %ld --> %ld\n",
|
|
ccp->uncompout, ccp->compout,
|
|
ccp->compin, ccp->uncompin);
|
|
}
|
|
|
|
if (ccp->in.algorithm != -1)
|
|
prompt_Printf(arg->prompt, "\n Input Options: %s\n",
|
|
(*algorithm[ccp->in.algorithm]->Disp)(&ccp->in.opt));
|
|
|
|
if (ccp->out.algorithm != -1) {
|
|
o = &ccp->out.opt;
|
|
for (f = 0; f < ccp->out.algorithm; f++)
|
|
if (IsEnabled(ccp->cfg.neg[algorithm[f]->Neg]))
|
|
o = &(*o)->next;
|
|
prompt_Printf(arg->prompt, " Output Options: %s\n",
|
|
(*algorithm[ccp->out.algorithm]->Disp)(&(*o)->val));
|
|
}
|
|
|
|
prompt_Printf(arg->prompt, "\n Defaults: ");
|
|
prompt_Printf(arg->prompt, "FSM retry = %us, max %u Config"
|
|
" REQ%s, %u Term REQ%s\n", ccp->cfg.fsm.timeout,
|
|
ccp->cfg.fsm.maxreq, ccp->cfg.fsm.maxreq == 1 ? "" : "s",
|
|
ccp->cfg.fsm.maxtrm, ccp->cfg.fsm.maxtrm == 1 ? "" : "s");
|
|
prompt_Printf(arg->prompt, " deflate windows: ");
|
|
prompt_Printf(arg->prompt, "incoming = %d, ", ccp->cfg.deflate.in.winsize);
|
|
prompt_Printf(arg->prompt, "outgoing = %d\n", ccp->cfg.deflate.out.winsize);
|
|
#ifndef NODES
|
|
prompt_Printf(arg->prompt, " MPPE: ");
|
|
if (ccp->cfg.mppe.keybits)
|
|
prompt_Printf(arg->prompt, "%d bits, ", ccp->cfg.mppe.keybits);
|
|
else
|
|
prompt_Printf(arg->prompt, "any bits, ");
|
|
switch (ccp->cfg.mppe.state) {
|
|
case MPPE_STATEFUL:
|
|
prompt_Printf(arg->prompt, "stateful");
|
|
break;
|
|
case MPPE_STATELESS:
|
|
prompt_Printf(arg->prompt, "stateless");
|
|
break;
|
|
case MPPE_ANYSTATE:
|
|
prompt_Printf(arg->prompt, "any state");
|
|
break;
|
|
}
|
|
prompt_Printf(arg->prompt, "%s\n",
|
|
ccp->cfg.mppe.required ? ", required" : "");
|
|
#endif
|
|
|
|
prompt_Printf(arg->prompt, "\n DEFLATE: %s\n",
|
|
command_ShowNegval(ccp->cfg.neg[CCP_NEG_DEFLATE]));
|
|
prompt_Printf(arg->prompt, " PREDICTOR1: %s\n",
|
|
command_ShowNegval(ccp->cfg.neg[CCP_NEG_PRED1]));
|
|
prompt_Printf(arg->prompt, " DEFLATE24: %s\n",
|
|
command_ShowNegval(ccp->cfg.neg[CCP_NEG_DEFLATE24]));
|
|
#ifndef NODES
|
|
prompt_Printf(arg->prompt, " MPPE: %s\n",
|
|
command_ShowNegval(ccp->cfg.neg[CCP_NEG_MPPE]));
|
|
#endif
|
|
return 0;
|
|
}
|
|
|
|
void
|
|
ccp_SetupCallbacks(struct ccp *ccp)
|
|
{
|
|
ccp->fsm.fn = &ccp_Callbacks;
|
|
ccp->fsm.FsmTimer.name = ccp_TimerNames[0];
|
|
ccp->fsm.OpenTimer.name = ccp_TimerNames[1];
|
|
ccp->fsm.StoppedTimer.name = ccp_TimerNames[2];
|
|
}
|
|
|
|
void
|
|
ccp_Init(struct ccp *ccp, struct bundle *bundle, struct link *l,
|
|
const struct fsm_parent *parent)
|
|
{
|
|
/* Initialise ourselves */
|
|
|
|
fsm_Init(&ccp->fsm, "CCP", PROTO_CCP, 1, CCP_MAXCODE, LogCCP,
|
|
bundle, l, parent, &ccp_Callbacks, ccp_TimerNames);
|
|
|
|
ccp->cfg.deflate.in.winsize = 0;
|
|
ccp->cfg.deflate.out.winsize = 15;
|
|
ccp->cfg.fsm.timeout = DEF_FSMRETRY;
|
|
ccp->cfg.fsm.maxreq = DEF_FSMTRIES;
|
|
ccp->cfg.fsm.maxtrm = DEF_FSMTRIES;
|
|
ccp->cfg.neg[CCP_NEG_DEFLATE] = NEG_ENABLED|NEG_ACCEPTED;
|
|
ccp->cfg.neg[CCP_NEG_PRED1] = NEG_ENABLED|NEG_ACCEPTED;
|
|
ccp->cfg.neg[CCP_NEG_DEFLATE24] = 0;
|
|
#ifndef NODES
|
|
ccp->cfg.mppe.keybits = 0;
|
|
ccp->cfg.mppe.state = MPPE_ANYSTATE;
|
|
ccp->cfg.mppe.required = 0;
|
|
ccp->cfg.neg[CCP_NEG_MPPE] = NEG_ENABLED|NEG_ACCEPTED;
|
|
#endif
|
|
|
|
ccp_Setup(ccp);
|
|
}
|
|
|
|
void
|
|
ccp_Setup(struct ccp *ccp)
|
|
{
|
|
/* Set ourselves up for a startup */
|
|
ccp->fsm.open_mode = 0;
|
|
ccp->his_proto = ccp->my_proto = -1;
|
|
ccp->reset_sent = ccp->last_reset = -1;
|
|
ccp->in.algorithm = ccp->out.algorithm = -1;
|
|
ccp->in.state = ccp->out.state = NULL;
|
|
ccp->in.opt.hdr.id = -1;
|
|
ccp->out.opt = NULL;
|
|
ccp->his_reject = ccp->my_reject = 0;
|
|
ccp->uncompout = ccp->compout = 0;
|
|
ccp->uncompin = ccp->compin = 0;
|
|
}
|
|
|
|
/*
|
|
* Is ccp *REQUIRED* ?
|
|
* We ask each of the configured ccp protocols if they're required and
|
|
* return TRUE if they are.
|
|
*
|
|
* It's not possible for the peer to reject a required ccp protocol
|
|
* without our state machine bringing the supporting lcp layer down.
|
|
*
|
|
* If ccp is required but not open, the NCP layer should not push
|
|
* any data into the link.
|
|
*/
|
|
int
|
|
ccp_Required(struct ccp *ccp)
|
|
{
|
|
int f;
|
|
|
|
for (f = 0; f < NALGORITHMS; f++)
|
|
if (IsEnabled(ccp->cfg.neg[algorithm[f]->Neg]) &&
|
|
(*algorithm[f]->Required)(&ccp->fsm))
|
|
return 1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Report whether it's possible to increase a packet's size after
|
|
* compression (and by how much).
|
|
*/
|
|
int
|
|
ccp_MTUOverhead(struct ccp *ccp)
|
|
{
|
|
if (ccp->fsm.state == ST_OPENED && ccp->out.algorithm >= 0)
|
|
return algorithm[ccp->out.algorithm]->o.MTUOverhead;
|
|
|
|
return 0;
|
|
}
|
|
|
|
static void
|
|
CcpInitRestartCounter(struct fsm *fp, int what)
|
|
{
|
|
/* Set fsm timer load */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
|
|
fp->FsmTimer.load = ccp->cfg.fsm.timeout * SECTICKS;
|
|
switch (what) {
|
|
case FSM_REQ_TIMER:
|
|
fp->restart = ccp->cfg.fsm.maxreq;
|
|
break;
|
|
case FSM_TRM_TIMER:
|
|
fp->restart = ccp->cfg.fsm.maxtrm;
|
|
break;
|
|
default:
|
|
fp->restart = 1;
|
|
break;
|
|
}
|
|
}
|
|
|
|
static void
|
|
CcpSendConfigReq(struct fsm *fp)
|
|
{
|
|
/* Send config REQ please */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
struct ccp_opt **o;
|
|
u_char *cp, buff[100];
|
|
int f, alloc;
|
|
|
|
cp = buff;
|
|
o = &ccp->out.opt;
|
|
alloc = ccp->his_reject == 0 && ccp->out.opt == NULL;
|
|
ccp->my_proto = -1;
|
|
ccp->out.algorithm = -1;
|
|
for (f = 0; f < NALGORITHMS; f++)
|
|
if (IsEnabled(ccp->cfg.neg[algorithm[f]->Neg]) &&
|
|
!REJECTED(ccp, algorithm[f]->id) &&
|
|
(*algorithm[f]->Usable)(fp)) {
|
|
|
|
if (!alloc)
|
|
for (o = &ccp->out.opt; *o != NULL; o = &(*o)->next)
|
|
if ((*o)->val.hdr.id == algorithm[f]->id && (*o)->algorithm == f)
|
|
break;
|
|
|
|
if (alloc || *o == NULL) {
|
|
*o = (struct ccp_opt *)malloc(sizeof(struct ccp_opt));
|
|
(*o)->val.hdr.id = algorithm[f]->id;
|
|
(*o)->val.hdr.len = 2;
|
|
(*o)->next = NULL;
|
|
(*o)->algorithm = f;
|
|
(*algorithm[f]->o.OptInit)(fp->bundle, &(*o)->val, &ccp->cfg);
|
|
}
|
|
|
|
if (cp + (*o)->val.hdr.len > buff + sizeof buff) {
|
|
log_Printf(LogERROR, "%s: CCP REQ buffer overrun !\n", fp->link->name);
|
|
break;
|
|
}
|
|
memcpy(cp, &(*o)->val, (*o)->val.hdr.len);
|
|
cp += (*o)->val.hdr.len;
|
|
|
|
ccp->my_proto = (*o)->val.hdr.id;
|
|
ccp->out.algorithm = f;
|
|
|
|
if (alloc)
|
|
o = &(*o)->next;
|
|
}
|
|
|
|
fsm_Output(fp, CODE_CONFIGREQ, fp->reqid, buff, cp - buff, MB_CCPOUT);
|
|
}
|
|
|
|
void
|
|
ccp_SendResetReq(struct fsm *fp)
|
|
{
|
|
/* We can't read our input - ask peer to reset */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
|
|
ccp->reset_sent = fp->reqid;
|
|
ccp->last_reset = -1;
|
|
fsm_Output(fp, CODE_RESETREQ, fp->reqid, NULL, 0, MB_CCPOUT);
|
|
}
|
|
|
|
static void
|
|
CcpSentTerminateReq(struct fsm *fp)
|
|
{
|
|
/* Term REQ just sent by FSM */
|
|
}
|
|
|
|
static void
|
|
CcpSendTerminateAck(struct fsm *fp, u_char id)
|
|
{
|
|
/* Send Term ACK please */
|
|
fsm_Output(fp, CODE_TERMACK, id, NULL, 0, MB_CCPOUT);
|
|
}
|
|
|
|
static int
|
|
CcpRecvResetReq(struct fsm *fp)
|
|
{
|
|
/* Got a reset REQ, reset outgoing dictionary */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
if (ccp->out.state == NULL)
|
|
return 1;
|
|
return (*algorithm[ccp->out.algorithm]->o.Reset)(ccp->out.state);
|
|
}
|
|
|
|
static void
|
|
CcpLayerStart(struct fsm *fp)
|
|
{
|
|
/* We're about to start up ! */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
|
|
log_Printf(LogCCP, "%s: LayerStart.\n", fp->link->name);
|
|
fp->more.reqs = fp->more.naks = fp->more.rejs = ccp->cfg.fsm.maxreq * 3;
|
|
}
|
|
|
|
static void
|
|
CcpLayerDown(struct fsm *fp)
|
|
{
|
|
/* About to come down */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
struct ccp_opt *next;
|
|
|
|
log_Printf(LogCCP, "%s: LayerDown.\n", fp->link->name);
|
|
if (ccp->in.state != NULL) {
|
|
(*algorithm[ccp->in.algorithm]->i.Term)(ccp->in.state);
|
|
ccp->in.state = NULL;
|
|
ccp->in.algorithm = -1;
|
|
}
|
|
if (ccp->out.state != NULL) {
|
|
(*algorithm[ccp->out.algorithm]->o.Term)(ccp->out.state);
|
|
ccp->out.state = NULL;
|
|
ccp->out.algorithm = -1;
|
|
}
|
|
ccp->his_reject = ccp->my_reject = 0;
|
|
|
|
while (ccp->out.opt) {
|
|
next = ccp->out.opt->next;
|
|
free(ccp->out.opt);
|
|
ccp->out.opt = next;
|
|
}
|
|
ccp_Setup(ccp);
|
|
}
|
|
|
|
static void
|
|
CcpLayerFinish(struct fsm *fp)
|
|
{
|
|
/* We're now down */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
struct ccp_opt *next;
|
|
|
|
log_Printf(LogCCP, "%s: LayerFinish.\n", fp->link->name);
|
|
|
|
/*
|
|
* Nuke options that may be left over from sending a REQ but never
|
|
* coming up.
|
|
*/
|
|
while (ccp->out.opt) {
|
|
next = ccp->out.opt->next;
|
|
free(ccp->out.opt);
|
|
ccp->out.opt = next;
|
|
}
|
|
|
|
if (ccp_Required(ccp)) {
|
|
if (fp->link->lcp.fsm.state == ST_OPENED)
|
|
log_Printf(LogLCP, "%s: Closing due to CCP completion\n", fp->link->name);
|
|
fsm_Close(&fp->link->lcp.fsm);
|
|
}
|
|
}
|
|
|
|
/* Called when CCP has reached the OPEN state */
|
|
static int
|
|
CcpLayerUp(struct fsm *fp)
|
|
{
|
|
/* We're now up */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
struct ccp_opt **o;
|
|
int f, fail;
|
|
|
|
for (f = fail = 0; f < NALGORITHMS; f++)
|
|
if (IsEnabled(ccp->cfg.neg[algorithm[f]->Neg]) &&
|
|
(*algorithm[f]->Required)(&ccp->fsm) &&
|
|
(ccp->in.algorithm != f || ccp->out.algorithm != f)) {
|
|
/* Blow it all away - we haven't negotiated a required algorithm */
|
|
log_Printf(LogWARN, "%s: Failed to negotiate (required) %s\n",
|
|
fp->link->name, protoname(algorithm[f]->id));
|
|
fail = 1;
|
|
}
|
|
|
|
if (fail) {
|
|
ccp->his_proto = ccp->my_proto = -1;
|
|
fsm_Close(fp);
|
|
fsm_Close(&fp->link->lcp.fsm);
|
|
return 0;
|
|
}
|
|
|
|
log_Printf(LogCCP, "%s: LayerUp.\n", fp->link->name);
|
|
|
|
if (ccp->in.state == NULL && ccp->in.algorithm >= 0 &&
|
|
ccp->in.algorithm < NALGORITHMS) {
|
|
ccp->in.state = (*algorithm[ccp->in.algorithm]->i.Init)
|
|
(fp->bundle, &ccp->in.opt);
|
|
if (ccp->in.state == NULL) {
|
|
log_Printf(LogERROR, "%s: %s (in) initialisation failure\n",
|
|
fp->link->name, protoname(ccp->his_proto));
|
|
ccp->his_proto = ccp->my_proto = -1;
|
|
fsm_Close(fp);
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
o = &ccp->out.opt;
|
|
for (f = 0; f < ccp->out.algorithm; f++)
|
|
if (IsEnabled(ccp->cfg.neg[algorithm[f]->Neg]))
|
|
o = &(*o)->next;
|
|
|
|
if (ccp->out.state == NULL && ccp->out.algorithm >= 0 &&
|
|
ccp->out.algorithm < NALGORITHMS) {
|
|
ccp->out.state = (*algorithm[ccp->out.algorithm]->o.Init)
|
|
(fp->bundle, &(*o)->val);
|
|
if (ccp->out.state == NULL) {
|
|
log_Printf(LogERROR, "%s: %s (out) initialisation failure\n",
|
|
fp->link->name, protoname(ccp->my_proto));
|
|
ccp->his_proto = ccp->my_proto = -1;
|
|
fsm_Close(fp);
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
fp->more.reqs = fp->more.naks = fp->more.rejs = ccp->cfg.fsm.maxreq * 3;
|
|
|
|
log_Printf(LogCCP, "%s: Out = %s[%d], In = %s[%d]\n",
|
|
fp->link->name, protoname(ccp->my_proto), ccp->my_proto,
|
|
protoname(ccp->his_proto), ccp->his_proto);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static void
|
|
CcpDecodeConfig(struct fsm *fp, u_char *cp, u_char *end, int mode_type,
|
|
struct fsm_decode *dec)
|
|
{
|
|
/* Deal with incoming data */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
int f;
|
|
const char *disp;
|
|
struct fsm_opt *opt;
|
|
|
|
if (mode_type == MODE_REQ)
|
|
ccp->in.algorithm = -1; /* In case we've received two REQs in a row */
|
|
|
|
while (end - cp >= sizeof(opt->hdr)) {
|
|
if ((opt = fsm_readopt(&cp)) == NULL)
|
|
break;
|
|
|
|
for (f = NALGORITHMS-1; f > -1; f--)
|
|
if (algorithm[f]->id == opt->hdr.id)
|
|
break;
|
|
|
|
disp = f == -1 ? "" : (*algorithm[f]->Disp)(opt);
|
|
if (disp == NULL)
|
|
disp = "";
|
|
|
|
log_Printf(LogCCP, " %s[%d] %s\n", protoname(opt->hdr.id),
|
|
opt->hdr.len, disp);
|
|
|
|
if (f == -1) {
|
|
/* Don't understand that :-( */
|
|
if (mode_type == MODE_REQ) {
|
|
ccp->my_reject |= (1 << opt->hdr.id);
|
|
fsm_rej(dec, opt);
|
|
}
|
|
} else {
|
|
struct ccp_opt *o;
|
|
|
|
switch (mode_type) {
|
|
case MODE_REQ:
|
|
if (IsAccepted(ccp->cfg.neg[algorithm[f]->Neg]) &&
|
|
(*algorithm[f]->Usable)(fp) &&
|
|
ccp->in.algorithm == -1) {
|
|
memcpy(&ccp->in.opt, opt, opt->hdr.len);
|
|
switch ((*algorithm[f]->i.Set)(fp->bundle, &ccp->in.opt, &ccp->cfg)) {
|
|
case MODE_REJ:
|
|
fsm_rej(dec, &ccp->in.opt);
|
|
break;
|
|
case MODE_NAK:
|
|
fsm_nak(dec, &ccp->in.opt);
|
|
break;
|
|
case MODE_ACK:
|
|
fsm_ack(dec, &ccp->in.opt);
|
|
ccp->his_proto = opt->hdr.id;
|
|
ccp->in.algorithm = f; /* This one'll do :-) */
|
|
break;
|
|
}
|
|
} else {
|
|
fsm_rej(dec, opt);
|
|
}
|
|
break;
|
|
case MODE_NAK:
|
|
for (o = ccp->out.opt; o != NULL; o = o->next)
|
|
if (o->val.hdr.id == opt->hdr.id)
|
|
break;
|
|
if (o == NULL)
|
|
log_Printf(LogCCP, "%s: Warning: Ignoring peer NAK of unsent"
|
|
" option\n", fp->link->name);
|
|
else {
|
|
memcpy(&o->val, opt, opt->hdr.len);
|
|
if ((*algorithm[f]->o.Set)(fp->bundle, &o->val, &ccp->cfg) ==
|
|
MODE_ACK)
|
|
ccp->my_proto = algorithm[f]->id;
|
|
else {
|
|
ccp->his_reject |= (1 << opt->hdr.id);
|
|
ccp->my_proto = -1;
|
|
if (algorithm[f]->Required(fp)) {
|
|
log_Printf(LogWARN, "%s: Cannot understand peers (required)"
|
|
" %s negotiation\n", fp->link->name,
|
|
protoname(algorithm[f]->id));
|
|
fsm_Close(&fp->link->lcp.fsm);
|
|
}
|
|
}
|
|
}
|
|
break;
|
|
case MODE_REJ:
|
|
ccp->his_reject |= (1 << opt->hdr.id);
|
|
ccp->my_proto = -1;
|
|
if (algorithm[f]->Required(fp)) {
|
|
log_Printf(LogWARN, "%s: Peer rejected (required) %s negotiation\n",
|
|
fp->link->name, protoname(algorithm[f]->id));
|
|
fsm_Close(&fp->link->lcp.fsm);
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (mode_type != MODE_NOP) {
|
|
fsm_opt_normalise(dec);
|
|
if (dec->rejend != dec->rej || dec->nakend != dec->nak) {
|
|
if (ccp->in.state == NULL) {
|
|
ccp->his_proto = -1;
|
|
ccp->in.algorithm = -1;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
extern struct mbuf *
|
|
ccp_Input(struct bundle *bundle, struct link *l, struct mbuf *bp)
|
|
{
|
|
/* Got PROTO_CCP from link */
|
|
m_settype(bp, MB_CCPIN);
|
|
if (bundle_Phase(bundle) == PHASE_NETWORK)
|
|
fsm_Input(&l->ccp.fsm, bp);
|
|
else {
|
|
if (bundle_Phase(bundle) < PHASE_NETWORK)
|
|
log_Printf(LogCCP, "%s: Error: Unexpected CCP in phase %s (ignored)\n",
|
|
l->ccp.fsm.link->name, bundle_PhaseName(bundle));
|
|
m_freem(bp);
|
|
}
|
|
return NULL;
|
|
}
|
|
|
|
static void
|
|
CcpRecvResetAck(struct fsm *fp, u_char id)
|
|
{
|
|
/* Got a reset ACK, reset incoming dictionary */
|
|
struct ccp *ccp = fsm2ccp(fp);
|
|
|
|
if (ccp->reset_sent != -1) {
|
|
if (id != ccp->reset_sent) {
|
|
log_Printf(LogCCP, "%s: Incorrect ResetAck (id %d, not %d)"
|
|
" ignored\n", fp->link->name, id, ccp->reset_sent);
|
|
return;
|
|
}
|
|
/* Whaddaya know - a correct reset ack */
|
|
} else if (id == ccp->last_reset)
|
|
log_Printf(LogCCP, "%s: Duplicate ResetAck (resetting again)\n",
|
|
fp->link->name);
|
|
else {
|
|
log_Printf(LogCCP, "%s: Unexpected ResetAck (id %d) ignored\n",
|
|
fp->link->name, id);
|
|
return;
|
|
}
|
|
|
|
ccp->last_reset = ccp->reset_sent;
|
|
ccp->reset_sent = -1;
|
|
if (ccp->in.state != NULL)
|
|
(*algorithm[ccp->in.algorithm]->i.Reset)(ccp->in.state);
|
|
}
|
|
|
|
static struct mbuf *
|
|
ccp_LayerPush(struct bundle *b, struct link *l, struct mbuf *bp,
|
|
int pri, u_short *proto)
|
|
{
|
|
if (PROTO_COMPRESSIBLE(*proto)) {
|
|
if (l->ccp.fsm.state != ST_OPENED) {
|
|
if (ccp_Required(&l->ccp)) {
|
|
/* The NCP layer shouldn't have let this happen ! */
|
|
log_Printf(LogERROR, "%s: Unexpected attempt to use an unopened and"
|
|
" required CCP layer\n", l->name);
|
|
m_freem(bp);
|
|
bp = NULL;
|
|
}
|
|
} else if (l->ccp.out.state != NULL) {
|
|
bp = (*algorithm[l->ccp.out.algorithm]->o.Write)
|
|
(l->ccp.out.state, &l->ccp, l, pri, proto, bp);
|
|
switch (*proto) {
|
|
case PROTO_ICOMPD:
|
|
m_settype(bp, MB_ICOMPDOUT);
|
|
break;
|
|
case PROTO_COMPD:
|
|
m_settype(bp, MB_COMPDOUT);
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
return bp;
|
|
}
|
|
|
|
static struct mbuf *
|
|
ccp_LayerPull(struct bundle *b, struct link *l, struct mbuf *bp, u_short *proto)
|
|
{
|
|
/*
|
|
* If proto isn't PROTO_[I]COMPD, we still want to pass it to the
|
|
* decompression routines so that the dictionary's updated
|
|
*/
|
|
if (l->ccp.fsm.state == ST_OPENED) {
|
|
if (*proto == PROTO_COMPD || *proto == PROTO_ICOMPD) {
|
|
/* Decompress incoming data */
|
|
if (l->ccp.reset_sent != -1)
|
|
/* Send another REQ and put the packet in the bit bucket */
|
|
fsm_Output(&l->ccp.fsm, CODE_RESETREQ, l->ccp.reset_sent, NULL, 0,
|
|
MB_CCPOUT);
|
|
else if (l->ccp.in.state != NULL) {
|
|
bp = (*algorithm[l->ccp.in.algorithm]->i.Read)
|
|
(l->ccp.in.state, &l->ccp, proto, bp);
|
|
switch (*proto) {
|
|
case PROTO_ICOMPD:
|
|
m_settype(bp, MB_ICOMPDIN);
|
|
break;
|
|
case PROTO_COMPD:
|
|
m_settype(bp, MB_COMPDIN);
|
|
break;
|
|
}
|
|
return bp;
|
|
}
|
|
m_freem(bp);
|
|
bp = NULL;
|
|
} else if (PROTO_COMPRESSIBLE(*proto) && l->ccp.in.state != NULL) {
|
|
/* Add incoming Network Layer traffic to our dictionary */
|
|
(*algorithm[l->ccp.in.algorithm]->i.DictSetup)
|
|
(l->ccp.in.state, &l->ccp, *proto, bp);
|
|
}
|
|
}
|
|
|
|
return bp;
|
|
}
|
|
|
|
u_short
|
|
ccp_Proto(struct ccp *ccp)
|
|
{
|
|
return !link2physical(ccp->fsm.link) || !ccp->fsm.bundle->ncp.mp.active ?
|
|
PROTO_COMPD : PROTO_ICOMPD;
|
|
}
|
|
|
|
int
|
|
ccp_SetOpenMode(struct ccp *ccp)
|
|
{
|
|
int f;
|
|
|
|
for (f = 0; f < CCP_NEG_TOTAL; f++)
|
|
if (IsEnabled(ccp->cfg.neg[f])) {
|
|
ccp->fsm.open_mode = 0;
|
|
return 1;
|
|
}
|
|
|
|
ccp->fsm.open_mode = OPEN_PASSIVE; /* Go straight to ST_STOPPED ? */
|
|
|
|
for (f = 0; f < CCP_NEG_TOTAL; f++)
|
|
if (IsAccepted(ccp->cfg.neg[f]))
|
|
return 1;
|
|
|
|
return 0; /* No CCP at all */
|
|
}
|
|
|
|
int
|
|
ccp_DefaultUsable(struct fsm *fp)
|
|
{
|
|
return 1;
|
|
}
|
|
|
|
int
|
|
ccp_DefaultRequired(struct fsm *fp)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
struct layer ccplayer = { LAYER_CCP, "ccp", ccp_LayerPush, ccp_LayerPull };
|