changelog

this is a linux pptp software

Mon Feb 13 14:01:34 2006  James Cameron  <[e-mail]>

	* pptp-1.7.1 released.

Fri Feb 10 10:59:31 2006  James Cameron  <[e-mail]>

	* pptp.c: use prctl(2) to set process name for gateway process so
	that killall(1) of pptp will affect control connection call
	manager only.

Mon Aug 29 15:42:03 2005  James Cameron  <[e-mail]>

	* pptp.c: add --version feature.

Mon Aug 22 10:46:38 2005  James Cameron  <[e-mail]>

	* util.c (sigpipe_assign): remove superfluous sigset variable
	and calls; they served no purpose.  Reported by: Pavol Gono.

Sat Aug 20 23:14:25 2005  James Cameron  <[e-mail]>

	* options.pptp: default to not require MPPE, per Red Hat Bugzilla
	#166394.

Thu Aug 11 10:46:32 2005  James Cameron  <[e-mail]>

	* PROTOCOL-SECURITY: add write-up regarding protocol security risks.
	* Makefile (dist): add PROTOCOL-SECURITY write-up.

Wed Jul 27 20:49:30 2005  James Cameron  <[e-mail]>

	* pptp-1.7.0 released.

Mon Jul 11 13:17:37 2005  James Cameron  <[e-mail]>

	* pptp_gre.c (decaps_gre): silently discard packets that are not
	for this call.  Reported by: Alan Jenkins.

Thu Mar 31 17:41:17 2005  James Cameron  <[e-mail]>

	* pptp_ctrl.c (pptp_dispatch): handle signals first, since there's
	a possibility of premature return from the function.
	
	* pptp_callmgr.c (callmgr_main): terminate faster when peer closes
	control connection.

Thu Mar 10 11:07:52 2005  James Cameron  <[e-mail]>

	* pptp_ctrl.c (pptp_handle_timer): when we close connection due to
	loss of echo replies, say so in the log.
	Reported by: Jean Wolter.

	* pptp_ctrl.c: fix signal deadlock on kernel 2.6.x
	Reported by: Jean Wolter.
	
	* util.c, util.h: adopt sigpipe implementation from pptpd.

	* pptp_ctrl.c (pptp_conn_open): use sigpipe implementation instead
	of trying I/O in signal handler (which causes deadlocks).

	* pptp_ctrl.c (pptp_fd_set): include the sigpipe in the file
	descriptor set.
	
	* pptp_ctrl.c (pptp_dispatch): detect queued signals via sigpipe.
	
	* pptp_ctrl.c (pptp_conn_destroy): close the sigpipe.

Fri Feb 18 12:38:18 2005  James Cameron  <[e-mail]>

	* pptp-1.6.0 released.

Fri Feb 18 12:38:02 2005  James Cameron  <[e-mail]>

	* Makefile (install): move options.pptp to this package from the
	RPM spec file.

Tue Dec 14 08:56:07 2004  James Cameron  <[e-mail]>

	* pptp_ctrl.c (ctrlp_disp): fix typo.  From: Avi Kivity
	<[e-mail]>

Tue Dec  7 09:43:00 2004  James Cameron  <[e-mail]>

	* pptp_ctrl.c: add error propagation to the ctrlp_disp callchain,
	so that we don't double-free the connection on a stop control
	connection reply.  From: Avi Kivity <[e-mail]>

Wed Nov 10 09:49:48 2004  James Cameron  <[e-mail]>

	* util.{c,h}: recent compilers spew a packetload of compilation
	warnings.  From: Avi Kivity <[e-mail]>

Tue Nov  9 12:30:57 2004  James Cameron  <[e-mail]>

	* pptp.{8,c}, pptp_ctrl.c, pptp_options.h: add --idle-wait (time
	to wait before sending an echo request on the control connection)
	and --max-echo-wait (time to wait before giving up on lack of
	reply to an echo request).  The latter is not yet implemented.

Tue Nov  9 12:23:48 2004  James Cameron  <[e-mail]>

	* pptp_callmgr.c, pptp_ctrl.{c,h}: pptp_dispatch and its
	subordinates insist on handling fatal errors by calling
	pptp_conn_destroy. unfortunately, as the comments say,
	callmgr_main uses the connection afterwards with the usual
	results.  Added error propagation to the routines mentioned,
	allowing callmgr_main to have the honor of closing the connection.
	From: Avi Kivity <[e-mail]>

Wed Oct 27 21:03:28 2004  James Cameron  <[e-mail]>

	* pptp_callmgr.c (callmgr_main): pptp sometimes crashes in
	pptp_fd_set and leaves a core dump.  Think this is due to a signal
	being caught and longjmp'ed to callmgr_main's shutdown label after
	we are out of that function, causing garbage local variables to be
	used.  From: Avi Kivity <[e-mail]>

Wed Jul 21 16:52:01 CEST 2004 Rein Klazes <[e-mail]>
	
	* pptp_ctrl.c (ctrlp_rep): Do not log sending of echo requests,
        preventing dead-locks observed with Linux 2.6 systems.
        The syslog call is not reentrant and must not be used within a
        signal handler. A better solution would probably be to get rid of
        the SIGALRM timer altogether.

Wed Jul 21 19:00:59 2004  Peter Surda  <[e-mail]>
	
	* pptp_callmgr.c (callmgr_main): prevent kill of init.

Sat Jun 26 14:37:10 2004  James Cameron  <[e-mail]>

	* pptp_gre.c (decaps_hdlc): portability fix, use IPPROTO_IP in
	place of SOL_IP in getsockopt().  From: Thomas Quinot
	<[e-mail]>

Tue Jun 22 19:08:58 2004  James Cameron  <[e-mail]>

	* Makefile, AUTHORS, rename package from pptp-linux to pptp.

Tue Jun 22 18:08:27 2004  James Cameron  <[e-mail]>

	* pptp-linux-1.5.0 released.

Fri Jun 11 09:28:05 2004  Chris Wilson  <[e-mail]>

	* pptp.8, pptp_gre.c: make the statistics work properly when
	buffering is disabled.  Fix the issue with log messages claiming
	buffering when no buffering is being done.

Thu Jun 10 16:58:53 2004  James Cameron  <[e-mail]>

	* pptp.c (main): remove the sleep(3) on termination; it isn't
	really required; and causes the GRE-to-PPP gateway process to hang
	around after the connection has been terminated.

	* pptp.c (open_callmgr): call manager was inheriting the GRE
	socket, so close it before launching the callmgr, a regression
	introduced 18th July 2002, in the patch to bind the GRE socket
	earlier.

Thu Jun 10 08:34:17 CEST 2004 Rein Klazes <[e-mail]>

	* pptp_callmgr.c: fix a case when the call manager does not
	shutdown properly, a regression introduced by the changes on
	2003 Oct 22 and 23.

Wed Jun  9 10:08:02 2004  Chris Wilson  <[e-mail]>

	* pptp.8, pptp.c, pptp_gre.c, pptp_gre.h: add --nobuffer option to
	eliminate all buffering of packets, a "pptp quake patch".

Wed Jun  9 09:37:06 2004  Paul Howarth  <[e-mail]>

	* inststr.c: fix non-setting of the command line.

Tue Jun  8 21:25:27 2004  James Cameron  <[e-mail]>

	* pptp-linux-1.5.0-rc1 released.

Sat Jun  5 22:44:50 2004  James Cameron  <[e-mail]>

	* pptp_ctrl.c (pptp_read_some): fix for CPU loop after pppd killed
	with -9, if read() returns zero, it is because the control
	connection has closed, so destroy the connection.

Wed Mar 24 08:45:12 2004  Mark-Andre Hopf  <[e-mail]>

	* pptp.c: fix compile for ARM architecture.

Mon Mar  8 11:04:00 2004  Chris Wilson <[e-mail]>

	* pptp.8: added documentation for command-line options where
	missing.  Updated to reflect the use of getopt.

Tue Mar  2 09:53:53 2004  Peter McCurdy <[e-mail]>

	* pptp.c (main, do_nothing): do not hang when a connection is
	refused.  When the controlling process (pptp.c) paused while
	waiting for the child to send SIGUSR1, if the child died then the
	controlling process wouldn't ever wake up.  Now if the child dies,
	SIGCHLD gets handled and the controlling process quits.

Tue Feb 10 20:35:18 2004  James Cameron  <[e-mail]>

	* pptp_gre.c (decaps_hdlc): temporarily add code that captures
	additional detail if EMSGSIZE is returned by read().

Fri Jan  2 10:05:17 2004  James Cameron  <[e-mail]>

	* pptp-linux-1.4.0 released.

Mon Dec  1 11:58:24 2003  James Cameron  <[e-mail]>

	* pptp.c (main): remove arguments debugging printf.

Wed Nov 26 12:22:00 2003  Chris Wilson   <[e-mail]>

	* pptp.c: fixed argument handling so that hostname can be supplied
	anywhere on the command line, not just at the start.  pppd options
	starting with "-" but before "--" are no longer supported.
	
	* pptp.c: added some spacing to improve readability, removed
	pointless "Step X" comments.
	
	* pptp.c: fixed usage message: there should not be a "pppd" in the
	command line.

Mon Nov 10 15:39:41 2003  James Cameron  <[e-mail]>

	* pptp-linux-1.4.0-rc1 released.

Thu Oct 23 12:48:54 2003  James Cameron  <[e-mail]>

	* pptp_callmgr.c (callmgr_main): if connection fails, pptp enters
	a CPU loop calling select() with no file descriptors set, and
	being given EBADF.  Introduced by yesterday's modifications.
	Correct code to detect this situation and break the main loop.

Wed Oct 22 13:02:04 2003  James Cameron  <[e-mail]>

	* pptp_ctrl.c (pptp_call_open): add an assertion to prevent a call
	open attempt while the control connection is not established.

	* pptp_callmgr.c (callmgr_main): avoid accepting a UNIX socket
	connection and therefore calling pptp_call_open() if the control
	connection has not yet been established.

	* pptp_ctrl.c (pptp_conn_established): add function for
	pptp_callmgr.c to use to determine if the control connection has
	been established.

	TODO: if a connection reply is not received, what happens?

	Reported by: John BouAntoun

Mon Sep  8 10:33:41 2003  James Cameron  <[e-mail]>

	* pptp.c: fix for compile on OpenBSD.  From: Waldemar Brodkorb

Mon Aug 18 12:12:00 2003  Chris Wilson   <[e-mail]>

	* pptp.c pptp_gre.c util.h: add log level control, to increase
	or reduce verbosity of log messages, for debugging and for people
	who have lossy connections.

Thu Aug  7 12:20:09 2003  James Cameron  <[e-mail]>

	* pptp_gre.c (decaps_gre): fix reporting of packet loss.
	From: Chris Wilson

Wed Jun 25 12:59:28 2003 Rein Klazes <[e-mail]>

        * pptp_callmgr.c: Wait for the replies on our Call-Clear-Request's and
        Stop-Control-Connection-Request's.
        
        * pptp_ctrl.c: small tweak in a message about non existant call:
        report the received call ID's.
        
Sun Jun 22 19:08:14 2003 Rein Klazes <[e-mail]>

        * pptp_ctrl.c: Log the buffering and sending of ALL control
        messages (except Echo messages and Replies of course).
        Change the way the sending of these messages is done. The 
        original way was to first put them into a buffer ( in
        pptp_send_ctrl_packet()) and at some later write the buffered
        bytes to the socket ( in pptp_write_some()). Now
        pptp_send_ctrl_packet first attempts to write the packet directly.
        Only it the write() does not write all of the bytes, the remaining
        ones are buffered. This should help to track bugs in this area (like
        one sending two Start-Control-Request's).

Thu Jun 19 09:09:53 2003 Rein Klazes <[e-mail]>

        * pptp_ctrl.c: Improve logging of received control packets:
        Move duplicated code to report on Result codes and General errors
        to a separate function; Verbosely report error codes embedded in
        start-control-connection replies; Translated some French comment;
        Log when an unexpected Outgoing Call Reply is received; Log the
        receipt of ALL control messages (except in case of echo
        request/reply packets of which only the first 10 are logged);
        Shorten the name of the function "pptp_dispatch_ctrl_packet" to
        "ctrlp_disp", 25 characters is just too much in the log files.

Thu Jun 19 10:11:26 2003  Jan Pieter  <[e-mail]>

	* Makefile (uninstall): include uninstall target, and a minor fix
	to dist.
	
Wed Jun 18 12:19:09 2003 Rein Klazes <[e-mail]>

        * pptp_gre.c: Instead of logging the (a)synchronous mode
        of ppp, only issue a warning when it conflicts with the
        mode of pptp.
        Comparing two sequence numbers for(in)equality is safe even when
        a wrap-around has occurred. This contrasts with comparing for 
        smaller/bigger. Remove a few unneeded tests for wrap-around.

Tue Jun 17 19:40:41 2003 Rein Klazes <[e-mail]>

        * dirutil.c, ppp_fcs.c, pptp.c, pptp_gre.c, util.c, vector_test.c,
        * pptp_ctrl.c, pptp_callmgr.c, vector.c : reformat code with 
        standard 4 spaced indents, uniform function headings and some 
        general tidying up. Patches have been separated in "diff -w"
        invariant patches that should be safe and others that may be 
        not so safe.

Wed Jun 11 14:06:21 2003  Rein Klazes <[e-mail]>

        * pptp_gre.c: When a timeout is specified in the select call
        make it always non-zero. Making this call block makes it likely
        that pppd will run before the select returns. This results in a
        big reduction in transmitted ack-only packets (number down from 
        40% to 0.8% of received packets on my system).

Tue Jun 11 18:42:00 2003  James Cameron  <[e-mail]>

	* pptp-linux-1.3.1 released.

Tue Jun 10 13:53:00 2003  Chris Wilson  <[e-mail]>

	* pptp_callmgr.c: fixed a bug with call manager socket naming,
	introduced since 1.2.0, which caused confusing problems when
	trying to open more than one tunnel from the same machine.

Tue Jun 10 18:43:34 2003  James Cameron  <[e-mail]>

	* pptp-linux-1.3.0 released.

Tue Jun  3 19:16:52 2003  James Cameron  <[e-mail]>

	* pptp.8, pptp.c: change SIGHUP to SIGUSR1 for dumping stats.

Tue May 20 13:20:00 2003  Chris Wilson  <[e-mail]>

	* pptp.c: rewrote command-line usage information, to increase
	clarity and usefulness, and to describe various command-line
	options which have been added recently.

Fri May 23 20:52:05 2003  Rein Klazes  <[e-mail]>

	* pptp_ctrl.c: increase the verbosity of the log message at the
	receipt of a Call Disconnect Notification.

Mon May 12 16:56:14 2003  James Cameron  <[e-mail]>

	* Makefile: don't install the manpage with execute permission,
	swap LDFLAGS and LIBS so that one can say 'make LDFLAGS=-s' to
	create a stripped versions of the executables, introduce the
	optimization flag as a make variable, so that one can say 'make
	DEBUG= OPTIMIZE=-O2 LDFLAGS=-s'.  From: Peter Breitenlohner

Tue Apr 29 19:13:33 2003  James Cameron  <[e-mail]>

	* Makefile (dist): remove CVS directory from distribution.
	Reported by: Ola Lundqvist

Tue Apr 15 10:28:00 2003  Chris Wilson   <[e-mail]>

	* pptp.c, pptp_gre.c, pptp_gre.h: Added GRE statistics counters
	and RTT calculation, which can be dumped to the syslog by sending
	a SIGHUP to the GRE-to-PPP gateway process.

Mon Apr 14 11:57:00 2003  Chris Wilson   <[e-mail]>

	* pptp_gre.c: Fixed one case where an ACK could be immediately
	followed by a data packet, they should be combined into a
	single packet.

Sat Mar 8 2003 15:19:12 2003 Rein Klazes <[e-mail]>

        * pptp_gre.c, pqueue.c: Actually copy the packet into a new entry
        in the packet queue, fixing "unknown protocol" error messages;
        Change the program logic making it obvious that there is not a 
        memory leak;
        Restore the "buffering out-of-order packet" log messages, so we
        have a chance to notice any more side effects.

Mon Feb 17 09:18:30 2003  James Cameron  <[e-mail]>