1 * Mon Sep 19 2005 Brendan O'Dea <bod@optus.net> 2.1.8
2 - Move code from signal handlers into mainloop, avoiding a race
3 condition when forking CLI.
5 * Fri Sep 16 2005 Brendan O'Dea <bod@optus.net> 2.1.7
6 - This time, for sure: really fix Protocol-Reject.
8 * Fri Sep 16 2005 Brendan O'Dea <bod@optus.net> 2.1.6
9 - Any traffic on a tunnel resets lastrec, not just control messages.
10 - Use a unique identifier for LCP.
11 - Fix Code-Reject/Protocol-Reject.
12 - Add l2tp_mtu configuration option, used to define MRU, MSS.
13 - Adjust TCP MSS options in SYN and SYN,ACK packets to avoid
14 fragmentation of tcp packets.
16 * Sat Sep 3 2005 Brendan O'Dea <bod@optus.net> 2.1.5
17 - Avoid Code-Reject loop.
18 - Increase size of PPP buffers to MAXETHER.
19 - Bug fixes for CLI ringbuffer and tunnel HELLO from Yuri.
20 - Restart rather than halt BGP on receipt of CEASE (Dominique Rousseau).
21 - Add cluster_mcast_ttl option to allow a cluster to span multiple
22 subnets (suggested by Tim Devries).
24 * Mon Aug 29 2005 Brendan O'Dea <bod@optus.net> 2.1.4
25 - Drop level of "Unexpected CHAP message" log.
26 - Fix parsing of ProtocolRej (allow 1 or two byte protocols).
27 - Handle rejection of MRU negotiation by peer.
28 - Use local hostname for tunnel in SCCRP (Alex Kiernan).
30 * Wed Aug 17 2005 Brendan O'Dea <bod@optus.net> 2.1.3
31 - Fail IPCP negotiation only on ConfigRej of IP-Address.
33 * Wed Aug 10 2005 Brendan O'Dea <bod@optus.net> 2.1.2
34 - Clear cluster_master on election so that slaves will accept a new master.
35 - Provide more comments/defaults in etc/startup-config.default.
36 - Add DAE support (PoD/CoA) from Vladislav Bjelic.
37 - Clean up new warnings from gcc 4.0.
38 - Replace flags used for LCP/IPCP with state machine.
39 - Include Acct-Session-Time in interim records.
40 - Fix DAE vector, generateload (Alex Kiernan).
41 - Replace RSA MD5 with public domain version.
43 * Tue Jun 14 2005 Brendan O'Dea <bod@optusnet.com.au> 2.1.1
44 - Add missing newline to backtrace macro.
45 - Don't send CDN for each session when shutting down tunnels (this is
47 - Move tunnel shutdown from SIGQUIT signal handler to be run once from
48 still_busy(). Reject new tunnels/sessions while in the process of
50 - Clarify usage of shutdown signals in documentation.
51 - Always initialise PRNG.
52 - Sanity check length of random_vector.
53 - Fix segv in unhide_value.
54 - Ping new master when we get C_MASTER and delay next election to allow
55 the unicast limp-along code to kick in if required.
57 * Sun Jun 5 2005 Brendan O'Dea <bod@optusnet.com.au> 2.1.0
58 - Add IPv6 support from Jonathan McDowell.
59 - Add CHAP support from Jordan Hrycaj.
60 - Add interim accounting support from Vladislav Bjelic.
61 - Negotiate MRU, default 1458 to avoid fragmentation.
62 - Sanity check that cluster_send_session is not called from a child
64 - Use bounds-checking lookup functions for string constants.
65 - Add enum for RADIUS codes.
66 - Make "call_" prefix implict in CSTAT() macro.
67 - Fix some format string problems.
68 - Remove "save_state" option. Not maintained anymore; use clustering
69 to retain state across restarts.
70 - Simplify AVP unhiding code.
71 - Add optional "username" parameter to ungarden control, allowing the
72 username to be reset before going online.
73 - Add result/error codes and message to StopCCN when shutting down tunnels.
74 - Add result/error codes to CDN when shutting down sessions. Sends 2/7
75 (general error, try another LNS) when out of IP addresses, and 3
76 (adminstrative) for everything else (suggestion from Chris Gates).
77 - Use cli_error() for error messages and help.
78 - Don't use LOG() macro in initdata() until the config struct has been
79 allocated (uses config->debug).
80 - Initialise log_stream to stderr to catch errors before the config file
82 - Make "show running-config" a privileged command (contains clear text
84 - Add sessionctl plugin to provide drop/kill via nsctl.
85 - New config option: allow_duplicate_users which determines whether
86 or not to kill older sessions with the same username.
87 - Fix byte counters in accounting records.
88 - Add Acct-Output-Gigawords, Acct-Input-Gigawords attributes to RADIUS
90 - Fix icmp host unreachable to use router address.
91 - Include endpoint address in accounting dump files.
92 - Convert mainloop to use epoll rather than select.
93 - Add note about fragmentation in Docs/manual.html, and a sample
94 iptables rule for MSS clamping.
96 + Show session open time in "show session"/"show user" detailed output.
97 + Have slaves with BGP configured drop BGP on receipt of a shutdown
98 signal, but hang about for an additional 5s to process any remaining
100 + Run regular_cleanups after processing the results of the select,
101 looking at a sufficient slice of each table to ensure that all
102 entries are examined at least once per second.
104 + Cluster changes from Michael, intended to prevent a stray master
105 from trashing a cluster:
106 = Ignore heartbeats from peers claiming to be the master before the
107 timeout on the old master has expired.
108 = A master receiving a stray heartbeat sends a unicast HB back, which
109 should cause the rogue to die due to the tie-breaker code.
110 = Keep probing the master for late heartbeats.
111 = Drop BGP as soon as we become master with the minumum required peers.
112 = Any PING seen from a master forces an election (rather than just
113 where basetime is zero).
114 = A slave which receives a LASTSEEN message (presumably a restarted
115 master) sends back new message type, C_MASTER which indicates the
116 address of the current master.
117 + New config option: cluster_master_min_adv which determines the minimum
118 number of up to date slaves required before the master will drop
121 + Add handling of "throttle=N" RADIUS attributes.
122 + Fix RADIUS indexing (should have 16K entries with 64 sockets).
124 + Fix leak in session freelist when initial RADIUS session allocation
127 + Add a Cisco-Avpair with intercept details to RADIUS Start/Stop
130 + Only send RADIUS stop record in sessionshutdown when there's an ip address.
131 + Reset .die on master takeover (so that dying sessions don't have to
132 hang around until the new master has the same uptime as the old one).
133 + Update .last_packet in cluster_handle_bytes only when there have
134 been bytes received from the modem (dead sessions were having the
135 idle timeout reset by stray packets).
137 + Ensure that sessionkill is not called on an unopened session (borks
139 + Bump MAXSESSION to 60K.
140 + Fix off-by-one errors in session/tunnel initialisation and
141 sessiont <-> sessionidt functions.
142 + Use session[s].opened consistently when checking for in-use sessions
143 (rather than session[s].tunnel).
144 + Use <= cluster_highest_sessionid rather than < MAXSESSION in a
146 + Don't kill a whole tunnel if we're out of sessions.
147 + Change session[s].ip to 0 if set from RADIUS to 255.255.255.254;
148 avoids the possibility that it will be interpreted as a valid IP
150 + Avoid a possible buffer overflow in processpap.
151 + Kill session if authentication was rejected.
153 + More DoS prevention: add packet_limit option to apply a hard limit
154 to downstream packets per session.
155 + Fix "clear counters".
156 + Log "Accepted connection to CLI" at 4 when connection is from localhost
157 to reduce noise in logs.
158 + Show time since last counter reset in "show counters".
160 + Throttle outgoing LASTSEEN packets to at most one per second for a
163 * Fri Dec 17 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.13
164 - Better cluster master collision resolution: keep a counter of state
165 changes, propagated in the heartbeats; the master with the highest #
166 of changes (that has kept in contact with the LAC through the
168 - Skip newlines in ringbuffer messages to CLI.
169 - Drop "Session N is closing" message level to 4; don't process PPPIP
170 packets in this state.
171 - Use gzip --best for man pages, include pid_file in sample
172 startup-config (from Jonathan's Debian package patches).
173 - Read multiple packets off cluster_sockfd as well as udpfd, tunfd in an
174 attempt to avoid losing the cluster in high load (DoS) conditions.
175 - Add counters for select_called, multi_read_used and multi_read_exceeded.
177 - Retain counters of shutdown sessions to dump once per minute.
178 - Use standard uintN_t types for portability.
180 * Wed Dec 1 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.12
181 - The "This time, for sure!" release.
182 - Fix throttlectl plugin argument parsing.
184 * Wed Dec 1 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.11
185 - Don't send a RADIUS start record when ungardening on shutdown.
187 * Wed Dec 1 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.10
188 - Fix byte ordering of LCP header length (thanks Yuri).
189 - Increase ip_conntrack_max due to dropped packets.
191 * Tue Nov 30 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.9
192 - Revise CCP, send ConfigReq once only.
193 - Don't copy the old buffer into Config{Nak,Rej} LCP responses (oops);
194 add length checks when appending.
195 - Do copy the identifier from the request and update length.
196 - Have makeppp print a backtrace on overflow.
197 - Check control serial before clearing window, prevents looping tunnel
198 setup in some instances.
199 - Implement named access-lists which may be applied to a session
200 either via Filter-Id RADIUS responses, or using the CLI.
201 - Drop ip address from LOG.
202 - autothrottle: revise parsing; ignore lcp:interface-config avpairs
203 which don't start with serv[ice-policy].
206 * Sat Nov 20 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.8
207 - Ignore gateway address in Framed-Route (from Jonathan McDowell).
208 - Don't route Framed-IP-Address if contained in a Framed-Route.
209 - Call sessionshutdown() when a tunnel is dropped rather than
210 sessionkill() to ensure that RADIUS stop records are sent.
211 - Cleanup: make a bunch of global functions/variables static.
212 - Remove reference to old -a command line argument.
213 - Add l2tpns(8) and nsctl(8) manpages from Jonathan McDowell.
214 - Add startup-config(5) manpage.
215 - Revise nsctl to allow arbitrary strings/args to be passed to plugins.
216 - Add snoopctl, throttlectl plugins.
217 - Fix deletion from linked list.
218 - Allow LCP re-negotiation after connection completes (thanks Yuri).
220 * Mon Nov 15 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.7
221 - Fix socket creation in host_unreachable() (thanks to Bjørn Augestad)
222 - Don't assume BGP peer sends back negotiated hold time, pick smallest
224 * Thu Nov 11 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.6
225 - Make BGP keepalive/hold time configurable
226 - Revise BGP config to use "router bgp AS" syntax (requires libcli >= 1.8.2)
228 * Tue Nov 9 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.5
229 - Handle routing properly in lone-master case
230 - Fix intercepts: don't double-snoop throttled customers, ensure
231 byte/packet counts are only updated once
232 - Add a callback to allow plugins to fetch values from the running config
234 * Mon Nov 8 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.4
235 - Added setrxspeed plugin
236 - Added peer_address config option
237 - Rename wrapper macros to LOG()/LOG_HEX(), use p->log() in plugins
238 - Replace some PPP{PAP,CHAP} magic numebrs with constants
239 - Nak asyncmap (unless == 0)
240 - Bundle ConfigRej options
241 - Clean up initlcp handling
243 * Wed Nov 3 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.3
244 - Added support for hidden AVPs by Robert Clark
245 - l2tpns-chap-response.patch from Robert Clark
246 - Merge l2tpns-config-hostname.patch from Robert Clark
247 - l2tpns-dont-timeshift-unidirectional-traffic.patch from Robert Clark - Dump accounting data if cin OR cout is non-zero
248 - Don't write accounting files if no accounting dir is set - Yuri
249 - Fix checking for mmap success
250 - Renegotiate MRU - Yuri
251 - Take LCP ConfigReq length from the packet length field - Yuri
252 - Hostname set via command line not config
253 - Make number of throttle buckets configurable
254 - Shared_malloc returns NULL on failure
256 - Unfsck 4->8 indenting change
257 - Use 2 seperate u16 values for throttle rate in/out
258 - Defer adding radius fds to the select loop until become_master
260 * Thu Sep 02 2004 David Parrish <david@dparrish.com> 2.0.2
261 - Combined LCP patches from Iain and Yuri. This should allow Windows 2k/XP
262 clients to connect, as well Linksys DSL modems.
263 - Apply patch to fix -v option from Juergen Kammer.
264 - Makefile fix from Juergen Kammer to not overwrite existing config files on
266 - Configurable radius port patch from Juergen Kammer.
267 - Send my_address if no bind_address when doing IPCP
268 - Write pid file if filename is set
269 - Add startup script and monitor script from Yuri
270 - Some logging correctness fixes from Iain Wade
271 - Add support for LCP Ident and CallBack (rejection only) from Yuri
272 - Initiate LCP if not attempted by the client, or in renegotiation - Yuri
273 - Indentation and style cleanups
274 - Per-user upload and download throttle rates - Yuri
275 - Make autothrottle.so understand cisco lcp:interface-config - Yuri
276 - Show filter stats in show session - Yuri
277 - Cleanup from Michael to change sid to unique_id
278 - Add plugin to remove domain name from auth requests
279 - Add .spec file for RPM generation
281 * Tue Jul 13 2004 Brendan O'Dea <bod@optusnet.com.au> 2.0.1
282 - Update INSTALL, Docs/manual.html documentation.
283 - Add INTERNALS documentation.
284 - Add lock_pages option.
285 - TerminateAck fix from Yuri
286 - Adject cli_loop args for libcli 1.8.0
287 - Allow for backward compatabity in C_PING packets
288 - Don't send RADIUS stop messages from sessionshutdown when called from
291 - Fix for LASTSEEN breakage: don't do anything in the CLI other than
292 flag changes to be made by the parent.
293 - Split out master parts from cluster_check_master() into cluster_check_slaves().
294 - Set hostname in CLI prompt.
295 - Make cluster_hb_interval work; include interval/timeout in heartbeats
296 so that a change on the master is propagated immediately to the slaves.
297 - Use fast heartbeats when there are slaves not up to date.
298 - Ensure basetime of shut down master is set to zero (prevent delayed election).
299 - Fix radius session leak on IPCP timeout.
300 - Fix some off-by-one errors in tunnel/session loops.
301 - Add "limp along" fix for when a slave drops temporarily from the mcast group.
302 - Rename l2tpns.cfg as startup-config to match CONFIGFILE.
303 - Update cli callbacks to work with libcli 1.6.
304 This supports privileged and unprivileged commands, as well as a configuration
306 - Add help for all cli commands.
307 - Add "show version" command.
308 - Fix uptime counter display.
309 - Fix nasty bug where cluster basetime can be set to 0 when sending initial
311 - Don't rmmod ip_conntrack, as this can take a lot of time.
312 - Re-order logging in routeset such that the action is given before any error.
313 - Use the correct gateway address when deleting routes.
314 - Remove any routes when address changes.
315 - Require authentication if telnet from remote ip.
316 - Require enable password always.
317 - Return error if show pool done on slave.
318 - We MUST immediately exit if we're the wrong master!
320 * Wed Jun 23 2004 David Parrish <david@dparrish.com> 2.0.0
322 - Completely replace active/standby clustering with a new peer-to-peer
323 clustering method which allows much greater throughput and is a lot more fault
325 - Add internal tbf implementation for throttling without relying on tc and
327 - Add support for iBGP and eBGP to advertise routes
328 - Add cli commands "show cluster", "show bgp", "show ipcache", "show throttle",
329 "show tbf", "suspend bgp", "restart bgp", "show user"
330 - Interception destination must be set per-user
331 - If SMP machine, allow use of SCHED_FIFO, which should improve performance
332 - Added config option to send GARP at startup
333 - Added plugin_become_master and plugin_new_session_master plugin hooks
334 - Remove useless sessionsendarp(). This isn't needed now that we are using TUN
336 - ICMP rate limiting so not every unreachable packet is replied with an ICMP
338 - mangle table is not required on anything but the cluster master, so slaves
339 will drop the mangle table and attempt to unload the ip_conntrack module
340 - Statically assigned IP addresses (by Radius) work now
341 - Add -d command-line flag to detach and become a daemon
342 - Configuration file is now "/etc/l2tpns/startup-config"
343 - Reduced MIN_IP_SIZE to 0x19 to stop a pile of Short IP warnings
344 - Resend initial IPCP request until it's acknowleged by the client
345 - Better radius session cleanup logic
346 - Many miscellaenous bugfixes and performance enhancements
347 - Thanks to Michael O'Reilly and Brendan O'Dea for most of these new features
349 * Mon May 24 2004 David Parrish <david@dparrish.com> 1.2.0
350 - Fix SEGFAULT in garden module
351 - Use multiple radius sockets to allow more concurrent authentication requests
352 - Add username parameter to "show users" command
353 - Fix counting tunnel rx errors as tunnel tx errors
354 - Add "show throttle" command
355 - Add gcc __attribute__ to logging functions
356 - Fix warnings shown by __attribute__
357 - Make sure regular cleanup happens regularly under high load
358 - Add variable cleanup_interval for changing cleanup interval
359 - Add support for reading more than one packet per fd in each processing loop
360 - This is configurable with the multi_read_count variable
361 - Remove segv handler so core dumps can happen
362 - Use nonblocking sockets
363 - Increase tun queue length
364 - Fix minimum length of IP packets
365 - Remove per-packet plugin hooks (they are slow)
366 - Don't drop session if no free RADIUS
367 - Don't expire more than 1000 sessions per cleanup interval
368 - Remove -a and -c command-line options. They don't work anyway
369 - Don't require file: in log_filename
370 - Bump version to 1.2.0
371 - Check return code when throttling users
373 * Mon Apr 5 2004 David Parrish <david@dparrish.com> 1.1.1
374 - Don't mention configure anymore, it's not used
375 - Added the autosnoop and autothrottle modules
376 - Don't default to using a htb for the class root
378 * Fri Mar 5 2004 David Parrish <david@dparrish.com> 1.1.0
379 - Change all strcpy() calls to strncpy() to avoid buffer overflow potential
380 - Add ICMP host unreachable support
381 - Logging to syslog if log_file = "syslog:facility"
382 - Now requires libcli 1.5
383 - All configuration moves to a config structure
384 - Ability to modify and write config on the fly through command-line interface
385 - Config file support is removed, and now handled by the cli
386 - Show hostname in cli prompt
387 - Keep current state type for tunnels
388 - Add uptime command do CLI, which also shows real-time bandwidth utilisation
389 - Add goodbye command to cluster master, which forces droppping a slave
390 - Cache IP address allocation, so that reconnecting users get the same address
391 - Fix tunnel resend timeouts, so that dead tunnels will be cleaned up
392 - Allocate tunnels and radius without using a linked list which had issues
393 - Fix some off-by-one errors in tunnel and session and radius arrays
394 - Save and reload ip address pool when dieing
395 - Check version and size of reloaded data when restarting
396 - Remove plugin_config support
397 - Remove old support for TBF which didn't work anyway. HTB is required to do throttling now.
398 - Add COPYING and Changes files