d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
b08bf2de64
freebsd-nq/usr.sbin/ppp/bundle.c

2009 lines
56 KiB
C
Raw Normal View History

Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
/*-
* Copyright (c) 1998 Brian Somers <brian@Awfulhak.org>
* 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.
*
$Id$ -> $FreeBSD$
1999-08-28 01:35:59 +00:00
* $FreeBSD$
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
*/
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
#include <sys/param.h>
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include <sys/socket.h>
#include <netinet/in.h>
#include <net/if.h>
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
#include <net/if_tun.h> /* For TUNS* ioctls */
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include <net/route.h>
Move VJ compression state and stats into struct ipcp.
1998-03-16 22:54:35 +00:00
#include <netinet/in_systm.h>
#include <netinet/ip.h>
o Add the link name to modem diagnostics. o Create struct mpserver as part of struct mp. mpserver creates a unix-domain socket based on the peers auth name and endpoint discriminator. If it already exists, ppp will ``pass the link'' over to the owner of the socket, joining it into the bundle of another ppp invocation, otherwise ppp waits for other invocations to pass it links through this socket. The final piece of code will be the code that flattens our datalink info and passes it down this channel (not yet implemented).
1998-04-28 01:25:46 +00:00
#include <sys/un.h>
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include <errno.h>
#include <fcntl.h>
Fix ``set proctitle'' by using setproctitle().
1999-11-17 21:12:35 +00:00
#ifdef __OpenBSD__
#include <util.h>
#else
#include <libutil.h>
#endif
o Use two `cat' processes to connect the modem to an already-running ppp. Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> o Use _PATH_DEVNULL rather than "/dev/null" o Be more paranoid about nuking running timers when transferring links.
1998-05-01 19:20:09 +00:00
#include <paths.h>
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include <stdio.h>
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
#include <stdlib.h>
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include <string.h>
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
#include <sys/uio.h>
o Don't try to transfer tty device descriptors as there's no way of transferring session rights with them. Instead, create two `/bin/cat' processes. A new child is spawned and disassociated from the terminal and the parent, which continues with the rest of the ppp process. Meanwhile, the parent spawns another child, and both the parent and child exec the `/bin/cat' commands with the appropriate descriptors. This way, the session is owned by the parent, and the tty is held open. o Close LCPs that have done a TLF and are now in ST_STOPPED before calling Down. This prevents them from trying to come back up again after the peer has shut them down (it seems a bit strange that the rfc says that a Down in ST_STOPPED will cause a TLS etc). o Don't try to set the physical link name pointer when we're receiving and renaming a datalink. The physical hasn't been created yet, and as it happens, the garbage physical pointer happens to be the value of another physical - so we're pointing that other physical name at ourselves. yeuck. o Re-arrange the order of things in main (DoLoop()). We now handle signals only after the select and not before the UpdateSet. It's possible that either a signal (FSM timeout) or a descriptor_Read() brings a link down, after which we'd better tidy up any dead direct and 1off descriptors before calling UpdateSet() again. o Mention when we detect a PPP packet when we see one before the link is up (then start LCP as before).
1998-05-25 02:22:38 +00:00
#include <sys/wait.h>
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include <termios.h>
#include <unistd.h>
o Redesign the layering mechanism and make the aliasing code part of the layering. We now ``stack'' layers as soon as we open the device (when we figure out what we're dealing with). A static set of `dispatch' routines are also declared for dealing with incoming packets after they've been `pulled' up through the stacked layers. Physical devices are now assigned handlers based on the device type when they're opened. For the moment there are three device types; ttys, execs and tcps. o Increment version number to 2.2 o Make an entry in [uw]tmp for non-tty -direct invocations (after pap/chap authentication). o Make throughput counters quad_t's o Account for the absolute number of mbuf malloc()s and free()s in ``show mem''. o ``show modem'' becomes ``show physical''.
1999-05-08 11:07:56 +00:00
#include "layer.h"
Fix a rather nasty use of `static'. This caused a SEGV when running ``link * load label'' as we ended up recursing back into command_Interpret after nuking our command arg list.
1998-06-15 19:06:25 +00:00
#include "defs.h"
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include "command.h"
#include "mbuf.h"
#include "log.h"
#include "id.h"
#include "timer.h"
#include "fsm.h"
#include "iplist.h"
De-staticise LQR information Increment OutPackets for any packet - not just LQRs MFC: o Fix a few comment typos. o Fix ``set timeout'' usage message and documentation. o Change ifOutPackets, ifOutOctets and ifOutLQRs to `u_int32_t's so that they wrap correctly. o Put the LQR in network byte order using the correct struct size (sizeof u_int32_t, not sizeof u_long). o Wrap LQR ECHO counters correctly. o Don't increment OutLQR count if the last LQR hasn't been replied to. o Initialise last received LQR in StartLqm. o Don't start the LQR timer if we're `disabled' and `accepted'. o Generate LQR responses when both sides are using a timer and we're not going to send our next LQR before the peers max timeout.
1998-03-13 00:44:26 +00:00
#include "lqr.h"
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
#include "hdlc.h"
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include "throughput.h"
Move VJ compression state and stats into struct ipcp.
1998-03-16 22:54:35 +00:00
#include "slcompress.h"
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
#include "ncpaddr.h"
#include "ip.h"
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include "ipcp.h"
Move filter sets into struct bundle.
1998-03-16 22:52:54 +00:00
#include "filter.h"
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
#include "descriptor.h"
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
#include "route.h"
#include "lcp.h"
#include "ccp.h"
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
#include "link.h"
#include "mp.h"
Initial RADIUS support (using libradius). See the man page for details. Compiling with -DNORADIUS (the default for `release') removes support. TODO: The functionality in libradius::rad_send_request() needs to be supplied as a set of routines so that ppp doesn't have to wait indefinitely for the radius server(s). Instead, we need to get a descriptor back, select() on the descriptor, and ask libradius to service it when necessary. For now, ppp blocks SIGALRM while in rad_send_request(), so it misses PAP/CHAP retries & timeouts if they occur. Only PAP is functional. When CHAP is attempted, libradius complains that no User-Password has been specified... rfc2138 says that it *mustn't* be used for CHAP :-( Sponsored by: Internet Business Solutions Ltd., Switzerland
1999-01-28 01:56:34 +00:00
#ifndef NORADIUS
#include "radius.h"
#endif
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
#include "ipv6cp.h"
#include "ncp.h"
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
#include "bundle.h"
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
#include "async.h"
#include "physical.h"
#include "auth.h"
o Redesign the layering mechanism and make the aliasing code part of the layering. We now ``stack'' layers as soon as we open the device (when we figure out what we're dealing with). A static set of `dispatch' routines are also declared for dealing with incoming packets after they've been `pulled' up through the stacked layers. Physical devices are now assigned handlers based on the device type when they're opened. For the moment there are three device types; ttys, execs and tcps. o Increment version number to 2.2 o Make an entry in [uw]tmp for non-tty -direct invocations (after pap/chap authentication). o Make throughput counters quad_t's o Account for the absolute number of mbuf malloc()s and free()s in ``show mem''. o ``show modem'' becomes ``show physical''.
1999-05-08 11:07:56 +00:00
#include "proto.h"
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
#include "chap.h"
#include "tun.h"
Move the terminal/diagnostic socket stuff out of main.c and into the new `prompt.c'. struct prompt is (of course) a `sort' of descriptor.
1998-02-10 03:23:50 +00:00
#include "prompt.h"
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
#include "chat.h"
o Support callback types NONE, E.164, AUTH and CBCP. (see the new ``set callback'' and ``set cbcp'' commands) o Add a ``cbcp'' log level and mbuf type. o Don't dump core when \T is given in ``set login'' or ``set hangup''. o Allow ``*'' and blanks as placeholders in ppp.secret and allow a fifth field for specifying auth/cbcp dialback parameters. o Remove a few extraneous #includes o Define the default number of REQs (restart counter) in defs.h rather than hardcoding ``5'' all over the place. o Fix a few man page inconsistencies.
1998-08-07 18:42:51 +00:00
#include "cbcp.h"
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
#include "datalink.h"
Solve the ``first connection'' problem that occurs on demand-dial links with dynamic IP numbers where the program that causes the dial bind()s to an interface address that is subsequently changed after ppp negotiation. The problem is defeated by adding negotiated addresses to the tun interface as additional alias addresses and providing a set of ``iface'' commands for managing the interface. Libalias is also required (and what a name clash!) - it happily IP-aliases the address so that the source is that of the primary (negotiated) interface and un-IP-aliases it on the way back. An ``enable iface-alias'' is done implicitly by the -alias command line switch. If -alias isn't given, iface-aliasing is disabled by default and can't be enabled 'till an ``alias enable yes'' is done. ``alias enable no'' silently disables iface-alias. So, for dynamic-IP-type-connections, running ``ppp -alias -auto blah'' will work for the first connection, although existing bindings will not survive a disconnect/connect as the TCP peer will be trying to send to the old IP address - the packets won't route. It's now a lot easier to add IPXCP to ppp with minor updates to the new iface.[ch] (if anyone ever gets 'round to it). It's also now possible to manually add interface aliases with something like ``iface add 1.2.3.4/24 5.6.7.8''. This allows multi-homed ppp links :-)
1998-10-22 02:32:50 +00:00
#include "iface.h"
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
#include "server.h"
Probe for the availability of AF_INET6 at startup. If it's not available, default ipv6cp to disabled and refuse to let the user enable it.
2001-08-15 13:53:38 +00:00
#include "probe.h"
Merge the NETGRAPH branch into HEAD. tty devices now use netgraph's line discipline to do the async escaping, but no other benefits are available yet. Change ``ifdef HAVE_DES'' to ``ifndef NODES'' for consistency. Make the Makefile a little more sane WRT RELEASE_CRUNCH.
2002-03-30 12:30:09 +00:00
#ifndef NODES
Make the MPPE MasterKey Invalid messages a bit clearer (it now complains that you can't do MPPE without CHAP81). Reset MasterKeyValid to zero when we hit phase DEAD.
2001-02-04 01:08:24 +00:00
#include "mppe.h"
When we change the interface MTU, run through the routing table and tweak all route MTUs too.
2001-04-05 02:23:48 +00:00
#endif
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
Maintain input and output throughput averages and choose the highest of the two when calculating the MP throughput average for the ``set autoload'' implementation. This makes more sense as all links I know of are full-duplex. This also means that people may need to adjust their autoload settings as 100% bandwidth is now the theoretical maximum rather than 200% (but of course, halfing the current settings is probably not the correct answer either!). This involves a ppp version bump as we need to pass an extra throughput array through the MP local domain socket.
2000-08-15 10:25:42 +00:00
#define SCATTER_SEGMENTS 7 /* version, datalink, name, physical,
throughput, throughput, device */
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
#define SEND_MAXFD 3 /* Max file descriptors passed through
the local domain socket */
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
static int bundle_RemainingIdleTime(struct bundle *);
Add a bunch of `const's and fix a typo. Submitted by: Rich Neswold <rneswold@MCS.Net>
1999-12-27 11:54:57 +00:00
static const char * const PhaseNames[] = {
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
"Dead", "Establish", "Authenticate", "Network", "Terminate"
};
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
const char *
bundle_PhaseName(struct bundle *bundle)
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
{
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
return bundle->phase <= PHASE_TERMINATE ?
PhaseNames[bundle->phase] : "unknown";
}
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
void
Remove the second bundle_NewPhase argument and do the tun config from bundle_LayerUp.
1998-03-13 00:44:33 +00:00
bundle_NewPhase(struct bundle *bundle, u_int new)
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
{
Introduce the ``link'' command for controlling individual links.
1998-02-17 19:28:01 +00:00
if (new == bundle->phase)
return;
Do authentication at the datalink level, not the bundle level. The bundle doesn't get a LayerUp 'till we're authenticated. Introduce DATALINK_LCP and DATALINK_AUTH phases.
1998-03-01 01:07:49 +00:00
if (new <= PHASE_TERMINATE)
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
log_Printf(LogPHASE, "bundle: %s\n", PhaseNames[new]);
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
switch (new) {
case PHASE_DEAD:
bundle->phase = new;
Merge the NETGRAPH branch into HEAD. tty devices now use netgraph's line discipline to do the async escaping, but no other benefits are available yet. Change ``ifdef HAVE_DES'' to ``ifndef NODES'' for consistency. Make the Makefile a little more sane WRT RELEASE_CRUNCH.
2002-03-30 12:30:09 +00:00
#ifndef NODES
Make the MPPE MasterKey Invalid messages a bit clearer (it now complains that you can't do MPPE without CHAP81). Reset MasterKeyValid to zero when we hit phase DEAD.
2001-02-04 01:08:24 +00:00
MPPE_MasterKeyValid = 0;
MPPE_MasterKeyValid is only there if HAVE_DES is defined.
2001-02-04 01:20:33 +00:00
#endif
Make the MPPE MasterKey Invalid messages a bit clearer (it now complains that you can't do MPPE without CHAP81). Reset MasterKeyValid to zero when we hit phase DEAD.
2001-02-04 01:08:24 +00:00
log_DisplayPrompts();
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
break;
case PHASE_ESTABLISH:
bundle->phase = new;
break;
case PHASE_AUTHENTICATE:
Do authentication at the datalink level, not the bundle level. The bundle doesn't get a LayerUp 'till we're authenticated. Introduce DATALINK_LCP and DATALINK_AUTH phases.
1998-03-01 01:07:49 +00:00
bundle->phase = new;
o Move our prompt descriptor list outside of the bundle. It's now dealt with by the `server' object. This simplifies things as we only have one list of prompt descriptors and the log_ routines check prompt::logactive to determine whether it should be used for output. o Include the MP socket UpdateSet() result in bundle::UpdateSet(). o Don't select on the tun device unless we're in NETWORK phase or AUTO mode. o Stop the idle timer when we go to DEAD phase. We may have transferred a link and not had a chance to kill it. o Don't fail when trying to unlink our transferred datalink from our descriptor lists just before the transfer. o Add our link descriptor to the write set if we got a short write the last time (physical::out is set). o Log the connection source address when a connection is closed. o Remove descriptor::next field. Descriptor lists are not required any more.
1998-05-23 22:24:50 +00:00
log_DisplayPrompts();
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
break;
case PHASE_NETWORK:
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
if (ncp_fsmStart(&bundle->ncp, bundle)) {
bundle->phase = new;
log_DisplayPrompts();
} else {
log_Printf(LogPHASE, "bundle: All NCPs are disabled\n");
bundle_Close(bundle, NULL, CLOSE_STAYDOWN);
}
o Initialise MP correctly depending on if we're the first link in mp_Up(). o Bring MP and its CCP down when we enter phase TERMINATE, and ditch everything in the incoming packet queue. o Enable MRRU negotiation. Now, we can multilink mode, but only with one physical link. o Close the link if the peer PROTO REJs PROTO_MP. o Prepend our protocol before passing a packet to struct mp for fragmentation. o Log info messages to DEBUG, not ERROR (oops). o Align `show mp' output (again).
1998-04-23 21:50:13 +00:00
break;
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
case PHASE_TERMINATE:
bundle->phase = new;
o Initialise MP correctly depending on if we're the first link in mp_Up(). o Bring MP and its CCP down when we enter phase TERMINATE, and ditch everything in the incoming packet queue. o Enable MRRU negotiation. Now, we can multilink mode, but only with one physical link. o Close the link if the peer PROTO REJs PROTO_MP. o Prepend our protocol before passing a packet to struct mp for fragmentation. o Log info messages to DEBUG, not ERROR (oops). o Align `show mp' output (again).
1998-04-23 21:50:13 +00:00
mp_Down(&bundle->ncp.mp);
o Move our prompt descriptor list outside of the bundle. It's now dealt with by the `server' object. This simplifies things as we only have one list of prompt descriptors and the log_ routines check prompt::logactive to determine whether it should be used for output. o Include the MP socket UpdateSet() result in bundle::UpdateSet(). o Don't select on the tun device unless we're in NETWORK phase or AUTO mode. o Stop the idle timer when we go to DEAD phase. We may have transferred a link and not had a chance to kill it. o Don't fail when trying to unlink our transferred datalink from our descriptor lists just before the transfer. o Add our link descriptor to the write set if we got a short write the last time (physical::out is set). o Log the connection source address when a connection is closed. o Remove descriptor::next field. Descriptor lists are not required any more.
1998-05-23 22:24:50 +00:00
log_DisplayPrompts();
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
break;
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
}
}
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
static void
bundle_LayerStart(void *v, struct fsm *fp)
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
{
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
/* The given FSM is about to start up ! */
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
}
Tidy up the parent notification stuff for -background mode. Remove extraneous extern decls of tun_{in,out} and netfd.
1998-03-25 18:38:59 +00:00
When running ppp -background, show comfort messages showing when we're redialing/reconnecting. While we're here, log redial, reconnect and phone number announcements to LogCHAT, and reduce some other logging to LogDEBUG.
2000-04-07 23:46:14 +00:00
void
Tidy up the parent notification stuff for -background mode. Remove extraneous extern decls of tun_{in,out} and netfd.
1998-03-25 18:38:59 +00:00
bundle_Notify(struct bundle *bundle, char c)
{
if (bundle->notify.fd != -1) {
When running ppp -background, show comfort messages showing when we're redialing/reconnecting. While we're here, log redial, reconnect and phone number announcements to LogCHAT, and reduce some other logging to LogDEBUG.
2000-04-07 23:46:14 +00:00
int ret;
ret = write(bundle->notify.fd, &c, 1);
if (c != EX_REDIAL && c != EX_RECONNECT) {
if (ret == 1)
log_Printf(LogCHAT, "Parent notified of %s\n",
c == EX_NORMAL ? "success" : "failure");
else
log_Printf(LogERROR, "Failed to notify parent of success\n");
close(bundle->notify.fd);
bundle->notify.fd = -1;
} else if (ret == 1)
log_Printf(LogCHAT, "Parent notified of %s\n", ex_desc(c));
Tidy up the parent notification stuff for -background mode. Remove extraneous extern decls of tun_{in,out} and netfd.
1998-03-25 18:38:59 +00:00
else
When running ppp -background, show comfort messages showing when we're redialing/reconnecting. While we're here, log redial, reconnect and phone number announcements to LogCHAT, and reduce some other logging to LogDEBUG.
2000-04-07 23:46:14 +00:00
log_Printf(LogERROR, "Failed to notify parent of %s\n", ex_desc(c));
Tidy up the parent notification stuff for -background mode. Remove extraneous extern decls of tun_{in,out} and netfd.
1998-03-25 18:38:59 +00:00
}
}
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
If we've got a full output buffer queue and cannot send anything for two mintues (see ``set choked'' and ``show bundle''), nuke the ip, mp and link level buffer queues. This should fix problems where ``ppp -auto'' seems to stop responding after failing to connect to the peer a few times.
1998-08-25 17:48:43 +00:00
static void
bundle_ClearQueues(void *v)
{
struct bundle *bundle = (struct bundle *)v;
struct datalink *dl;
log_Printf(LogPHASE, "Clearing choked output queue\n");
timer_Stop(&bundle->choked.timer);
/*
* Emergency time:
*
* We've had a full queue for PACKET_DEL_SECS seconds without being
* able to get rid of any of the packets. We've probably given up
* on the redials at this point, and the queued data has almost
* definitely been timed out by the layer above. As this is preventing
* us from reading the TUN_NAME device (we don't want to buffer stuff
* indefinitely), we may as well nuke this data and start with a clean
* slate !
*
* Unfortunately, this has the side effect of shafting any compression
* dictionaries in use (causing the relevant RESET_REQ/RESET_ACK).
*/
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
ncp_DeleteQueues(&bundle->ncp);
If we've got a full output buffer queue and cannot send anything for two mintues (see ``set choked'' and ``show bundle''), nuke the ip, mp and link level buffer queues. This should fix problems where ``ppp -auto'' seems to stop responding after failing to connect to the peer a few times.
1998-08-25 17:48:43 +00:00
for (dl = bundle->links; dl; dl = dl->next)
physical_DeleteQueue(dl->physical);
}
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
static void
bundle_LinkAdded(struct bundle *bundle, struct datalink *dl)
{
bundle->phys_type.all |= dl->physical->type;
if (dl->state == DATALINK_OPEN)
bundle->phys_type.open |= dl->physical->type;
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
#ifndef NORADIUS
if ((bundle->phys_type.open & (PHYS_DEDICATED|PHYS_DDIAL))
!= bundle->phys_type.open && bundle->session.timer.state == TIMER_STOPPED)
if (bundle->radius.sessiontime)
bundle_StartSessionTimer(bundle, 0);
#endif
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
if ((bundle->phys_type.open & (PHYS_DEDICATED|PHYS_DDIAL))
!= bundle->phys_type.open && bundle->idle.timer.state == TIMER_STOPPED)
/* We may need to start our idle timer */
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
bundle_StartIdleTimer(bundle, 0);
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
}
o Support callback types NONE, E.164, AUTH and CBCP. (see the new ``set callback'' and ``set cbcp'' commands) o Add a ``cbcp'' log level and mbuf type. o Don't dump core when \T is given in ``set login'' or ``set hangup''. o Allow ``*'' and blanks as placeholders in ppp.secret and allow a fifth field for specifying auth/cbcp dialback parameters. o Remove a few extraneous #includes o Define the default number of REQs (restart counter) in defs.h rather than hardcoding ``5'' all over the place. o Fix a few man page inconsistencies.
1998-08-07 18:42:51 +00:00
void
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
bundle_LinksRemoved(struct bundle *bundle)
{
struct datalink *dl;
bundle->phys_type.all = bundle->phys_type.open = 0;
for (dl = bundle->links; dl; dl = dl->next)
bundle_LinkAdded(bundle, dl);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
bundle_CalculateBandwidth(bundle);
mp_CheckAutoloadTimer(&bundle->ncp.mp);
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
if ((bundle->phys_type.open & (PHYS_DEDICATED|PHYS_DDIAL))
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
== bundle->phys_type.open) {
#ifndef NORADIUS
if (bundle->radius.sessiontime)
bundle_StopSessionTimer(bundle);
#endif
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
bundle_StopIdleTimer(bundle);
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
}
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
}
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
static void
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
bundle_LayerUp(void *v, struct fsm *fp)
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
{
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
/*
* The given fsm is now up
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
* If it's an LCP, adjust our phys_mode.open value and check the
* autoload timer.
* If it's the first NCP, calculate our bandwidth
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
* If it's the first NCP, set our ``upat'' time
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
* If it's the first NCP, start the idle timer.
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
* If it's an NCP, tell our -background parent to go away.
* If it's the first NCP, start the autoload timer
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
*/
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
struct bundle *bundle = (struct bundle *)v;
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
Remove the second bundle_NewPhase argument and do the tun config from bundle_LayerUp.
1998-03-13 00:44:33 +00:00
if (fp->proto == PROTO_LCP) {
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
struct physical *p = link2physical(fp->link);
bundle_LinkAdded(bundle, p->dl);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
mp_CheckAutoloadTimer(&bundle->ncp.mp);
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
} else if (isncp(fp->proto)) {
if (ncp_LayersOpen(&fp->bundle->ncp) == 1) {
bundle_CalculateBandwidth(fp->bundle);
time(&bundle->upat);
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
#ifndef NORADIUS
if (bundle->radius.sessiontime)
bundle_StartSessionTimer(bundle, 0);
#endif
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
bundle_StartIdleTimer(bundle, 0);
mp_CheckAutoloadTimer(&fp->bundle->ncp.mp);
}
Tidy up the parent notification stuff for -background mode. Remove extraneous extern decls of tun_{in,out} and netfd.
1998-03-25 18:38:59 +00:00
bundle_Notify(bundle, EX_NORMAL);
Reduce the interface MTU by 2 when MPPE has been successfully negotiated. This is necessary because MPPE will combine the protocol id with the payload received on the tun interface, encrypt it, then prepend its own protocol id, effectively increasing the payload by two bytes.
2001-07-03 22:20:19 +00:00
} else if (fp->proto == PROTO_CCP)
bundle_CalculateBandwidth(fp->bundle); /* Against ccp_MTUOverhead */
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
}
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
static void
bundle_LayerDown(void *v, struct fsm *fp)
{
/*
* The given FSM has been told to come down.
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
* If it's our last NCP, stop the idle timer.
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
* If it's our last NCP, clear our ``upat'' value.
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
* If it's our last NCP, stop the autoload timer
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
* If it's an LCP, adjust our phys_type.open value and any timers.
o Make sure we adjust our min seq and process any outstanding queued incoming fragments when a link goes down. o Don't use the minimum sequence numbers of links that aren't open. o Understand sequence number wrapping when determining the minimum sequence number. o Add & adjust a few comments.
1998-05-23 17:05:28 +00:00
* If it's an LCP and we're in multilink mode, adjust our tun
I didn't get this right the last time.... When an NCP reaches TLF, *ONLY* datalink_Close() links that are in DATALINK_OPEN. When the last link reaches TLD, DOWN all NCPs (as we used to in the links TLF (which was the wrong place anyway)), as the NCPs aren't now going to datalink_Close() us unexpectedly, we get to continue doing what we were told to do in the first place. The result: When we lose a link, the IPCP layer goes down and we actually call the stuff in ppp.linkdown !
2000-04-06 10:03:48 +00:00
* If it's the last LCP, down all NCPs
o Make sure we adjust our min seq and process any outstanding queued incoming fragments when a link goes down. o Don't use the minimum sequence numbers of links that aren't open. o Understand sequence number wrapping when determining the minimum sequence number. o Add & adjust a few comments.
1998-05-23 17:05:28 +00:00
* speed and make sure our minimum sequence number is adjusted.
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
*/
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
struct bundle *bundle = (struct bundle *)v;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
if (isncp(fp->proto)) {
if (ncp_LayersOpen(&fp->bundle->ncp) == 0) {
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
#ifndef NORADIUS
if (bundle->radius.sessiontime)
bundle_StopSessionTimer(bundle);
#endif
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
bundle_StopIdleTimer(bundle);
bundle->upat = 0;
mp_StopAutoloadTimer(&bundle->ncp.mp);
}
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
} else if (fp->proto == PROTO_LCP) {
I didn't get this right the last time.... When an NCP reaches TLF, *ONLY* datalink_Close() links that are in DATALINK_OPEN. When the last link reaches TLD, DOWN all NCPs (as we used to in the links TLF (which was the wrong place anyway)), as the NCPs aren't now going to datalink_Close() us unexpectedly, we get to continue doing what we were told to do in the first place. The result: When we lose a link, the IPCP layer goes down and we actually call the stuff in ppp.linkdown !
2000-04-06 10:03:48 +00:00
struct datalink *dl;
struct datalink *lost;
int others_active;
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
bundle_LinksRemoved(bundle); /* adjust timers & phys_type values */
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
I didn't get this right the last time.... When an NCP reaches TLF, *ONLY* datalink_Close() links that are in DATALINK_OPEN. When the last link reaches TLD, DOWN all NCPs (as we used to in the links TLF (which was the wrong place anyway)), as the NCPs aren't now going to datalink_Close() us unexpectedly, we get to continue doing what we were told to do in the first place. The result: When we lose a link, the IPCP layer goes down and we actually call the stuff in ppp.linkdown !
2000-04-06 10:03:48 +00:00
lost = NULL;
others_active = 0;
for (dl = bundle->links; dl; dl = dl->next) {
if (fp == &dl->physical->link.lcp.fsm)
lost = dl;
else if (dl->state != DATALINK_CLOSED && dl->state != DATALINK_HANGUP)
others_active++;
}
o Make sure we adjust our min seq and process any outstanding queued incoming fragments when a link goes down. o Don't use the minimum sequence numbers of links that aren't open. o Understand sequence number wrapping when determining the minimum sequence number. o Add & adjust a few comments.
1998-05-23 17:05:28 +00:00
I didn't get this right the last time.... When an NCP reaches TLF, *ONLY* datalink_Close() links that are in DATALINK_OPEN. When the last link reaches TLD, DOWN all NCPs (as we used to in the links TLF (which was the wrong place anyway)), as the NCPs aren't now going to datalink_Close() us unexpectedly, we get to continue doing what we were told to do in the first place. The result: When we lose a link, the IPCP layer goes down and we actually call the stuff in ppp.linkdown !
2000-04-06 10:03:48 +00:00
if (bundle->ncp.mp.active) {
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
bundle_CalculateBandwidth(bundle);
o Make sure we adjust our min seq and process any outstanding queued incoming fragments when a link goes down. o Don't use the minimum sequence numbers of links that aren't open. o Understand sequence number wrapping when determining the minimum sequence number. o Add & adjust a few comments.
1998-05-23 17:05:28 +00:00
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
if (lost)
mp_LinkLost(&bundle->ncp.mp, lost);
else
Change some log levels. ALERTs are only logged when something that can't happen happens or when everyone needs to know. ERRORs are only logged when something unexpected happens.
1998-06-16 19:40:42 +00:00
log_Printf(LogALERT, "Oops, lost an unrecognised datalink (%s) !\n",
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
fp->link->name);
}
I didn't get this right the last time.... When an NCP reaches TLF, *ONLY* datalink_Close() links that are in DATALINK_OPEN. When the last link reaches TLD, DOWN all NCPs (as we used to in the links TLF (which was the wrong place anyway)), as the NCPs aren't now going to datalink_Close() us unexpectedly, we get to continue doing what we were told to do in the first place. The result: When we lose a link, the IPCP layer goes down and we actually call the stuff in ppp.linkdown !
2000-04-06 10:03:48 +00:00
Add a sprinkling of mp_Down() calls to ensure that we don't think that the multi-link NCP is up when it isn't.
2002-03-30 12:52:55 +00:00
if (!others_active) {
I didn't get this right the last time.... When an NCP reaches TLF, *ONLY* datalink_Close() links that are in DATALINK_OPEN. When the last link reaches TLD, DOWN all NCPs (as we used to in the links TLF (which was the wrong place anyway)), as the NCPs aren't now going to datalink_Close() us unexpectedly, we get to continue doing what we were told to do in the first place. The result: When we lose a link, the IPCP layer goes down and we actually call the stuff in ppp.linkdown !
2000-04-06 10:03:48 +00:00
/* Down the NCPs. We don't expect to get fsm_Close()d ourself ! */
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
ncp2initial(&bundle->ncp);
Add a sprinkling of mp_Down() calls to ensure that we don't think that the multi-link NCP is up when it isn't.
2002-03-30 12:52:55 +00:00
mp_Down(&bundle->ncp.mp);
}
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
}
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
}
static void
bundle_LayerFinish(void *v, struct fsm *fp)
{
/* The given fsm is now down (fp cannot be NULL)
*
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
* If it's the last NCP, fsm_Close all LCPs
Add a sprinkling of mp_Down() calls to ensure that we don't think that the multi-link NCP is up when it isn't.
2002-03-30 12:52:55 +00:00
* If it's the last NCP, bring any MP layer down
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
*/
struct bundle *bundle = (struct bundle *)v;
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
struct datalink *dl;
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
if (isncp(fp->proto) && !ncp_LayersUnfinished(&bundle->ncp)) {
o FsmDown() and FsmClose() any NCP that's still around before going into phase DEAD when the last physical link disappears. o Don't go to phase TERMINATE if we're already phase DEAD.
1998-04-16 21:19:05 +00:00
if (bundle_Phase(bundle) != PHASE_DEAD)
bundle_NewPhase(bundle, PHASE_TERMINATE);
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
for (dl = bundle->links; dl; dl = dl->next)
I didn't get this right the last time.... When an NCP reaches TLF, *ONLY* datalink_Close() links that are in DATALINK_OPEN. When the last link reaches TLD, DOWN all NCPs (as we used to in the links TLF (which was the wrong place anyway)), as the NCPs aren't now going to datalink_Close() us unexpectedly, we get to continue doing what we were told to do in the first place. The result: When we lose a link, the IPCP layer goes down and we actually call the stuff in ppp.linkdown !
2000-04-06 10:03:48 +00:00
if (dl->state == DATALINK_OPEN)
datalink_Close(dl, CLOSE_STAYDOWN);
Create & use fsm2initial(), a function to bring a state machine back to ST_INITIAL without going through any unnecessary TLS/TLF pairs.
1998-06-20 00:19:42 +00:00
fsm2initial(fp);
Add a sprinkling of mp_Down() calls to ensure that we don't think that the multi-link NCP is up when it isn't.
2002-03-30 12:52:55 +00:00
mp_Down(&bundle->ncp.mp);
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
}
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
}
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
void
Make `close lcp' just close the LCP layer and not hangup. This is useful for slirp users that wish to get their shell back after the ppp session. `close' with no args still hangs up as expected. Required by: jmz
1998-06-15 19:05:27 +00:00
bundle_Close(struct bundle *bundle, const char *name, int how)
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
{
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
/*
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
* Please close the given datalink.
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
* If name == NULL or name is the last datalink, fsm_Close all NCPs
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
* (except our MP)
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
* If it isn't the last datalink, just Close that datalink.
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
*/
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
struct datalink *dl, *this_dl;
int others_active;
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
others_active = 0;
this_dl = NULL;
for (dl = bundle->links; dl; dl = dl->next) {
if (name && !strcasecmp(name, dl->name))
this_dl = dl;
if (name == NULL || this_dl == dl) {
Make `close lcp' just close the LCP layer and not hangup. This is useful for slirp users that wish to get their shell back after the ppp session. `close' with no args still hangs up as expected. Required by: jmz
1998-06-15 19:05:27 +00:00
switch (how) {
case CLOSE_LCP:
datalink_DontHangup(dl);
Untangle some cunfusion between the CLOSE_STAYDOWN, CLOSE_LCP and CLOSE_NORMAL meanings. CLOSE_NORMAL doesn't change the currently required state, the others do. This should stop ppp from entering DATALINK_READY when LCP shutdown doesn't end up happening cleanly. Bump our version number to reflect this change.
2001-02-04 01:08:22 +00:00
break;
Make `close lcp' just close the LCP layer and not hangup. This is useful for slirp users that wish to get their shell back after the ppp session. `close' with no args still hangs up as expected. Required by: jmz
1998-06-15 19:05:27 +00:00
case CLOSE_STAYDOWN:
datalink_StayDown(dl);
break;
}
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
} else if (dl->state != DATALINK_CLOSED && dl->state != DATALINK_HANGUP)
others_active++;
}
if (name && this_dl == NULL) {
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
log_Printf(LogWARN, "%s: Invalid datalink name\n", name);
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
return;
}
if (!others_active) {
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
#ifndef NORADIUS
if (bundle->radius.sessiontime)
bundle_StopSessionTimer(bundle);
#endif
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
bundle_StopIdleTimer(bundle);
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
if (ncp_LayersUnfinished(&bundle->ncp))
ncp_Close(&bundle->ncp);
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
else {
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
ncp2initial(&bundle->ncp);
Add a sprinkling of mp_Down() calls to ensure that we don't think that the multi-link NCP is up when it isn't.
2002-03-30 12:52:55 +00:00
mp_Down(&bundle->ncp.mp);
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
for (dl = bundle->links; dl; dl = dl->next)
Make `close lcp' just close the LCP layer and not hangup. This is useful for slirp users that wish to get their shell back after the ppp session. `close' with no args still hangs up as expected. Required by: jmz
1998-06-15 19:05:27 +00:00
datalink_Close(dl, how);
Force IPCP back to initial in bundle_Close(). Add a few un-necessary includes to make us build under 2.2.5. These'll go when the code is cleaned up at the end.
1998-03-02 17:25:30 +00:00
}
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
} else if (this_dl && this_dl->state != DATALINK_CLOSED &&
this_dl->state != DATALINK_HANGUP)
Make `close lcp' just close the LCP layer and not hangup. This is useful for slirp users that wish to get their shell back after the ppp session. `close' with no args still hangs up as expected. Required by: jmz
1998-06-15 19:05:27 +00:00
datalink_Close(this_dl, how);
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
}
o Make sure we don't dereference NULL when we've lost all our links. o Use srandomdev() for __FreeBSD__ >= 2, not just >2. o Use srandom((time(NULL)^getpid())+random()), random() when we haven't got srandomdev().
1998-05-06 18:50:12 +00:00
void
Don't bring the modem offline or hangup when ``down lcp'' is done. Instead, behave like ``close lcp''.
1998-06-16 19:40:28 +00:00
bundle_Down(struct bundle *bundle, int how)
o Make sure we don't dereference NULL when we've lost all our links. o Use srandomdev() for __FreeBSD__ >= 2, not just >2. o Use srandom((time(NULL)^getpid())+random()), random() when we haven't got srandomdev().
1998-05-06 18:50:12 +00:00
{
struct datalink *dl;
for (dl = bundle->links; dl; dl = dl->next)
Don't bring the modem offline or hangup when ``down lcp'' is done. Instead, behave like ``close lcp''.
1998-06-16 19:40:28 +00:00
datalink_Down(dl, how);
o Make sure we don't dereference NULL when we've lost all our links. o Use srandomdev() for __FreeBSD__ >= 2, not just >2. o Use srandom((time(NULL)^getpid())+random()), random() when we haven't got srandomdev().
1998-05-06 18:50:12 +00:00
}
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
static int
To avoid namespace polution in NetBSD: ``struct descriptor'' -> ``struct fdescriptor''
2000-03-14 01:46:09 +00:00
bundle_UpdateSet(struct fdescriptor *d, fd_set *r, fd_set *w, fd_set *e, int *n)
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
{
struct bundle *bundle = descriptor2bundle(d);
struct datalink *dl;
Cosmetic: Make struct mbuf more like kernel mbufs.
1999-12-20 20:29:47 +00:00
int result, nlinks;
Add ``set ifqueue'' to control the size of the outgoing packet queue. Doing ``set ifqueue 0'' and ``set urgent none'' will allow full use of luigi's WF2Q code. Requested by: luigi
2000-06-11 02:30:51 +00:00
u_short ifqueue;
Cosmetic: Make struct mbuf more like kernel mbufs.
1999-12-20 20:29:47 +00:00
size_t queued;
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
result = 0;
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
/* If there are aren't many packets queued, look for some more. */
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
for (nlinks = 0, dl = bundle->links; dl; dl = dl->next)
nlinks++;
if (nlinks) {
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
queued = r ? ncp_FillPhysicalQueues(&bundle->ncp, bundle) :
ncp_QueueLen(&bundle->ncp);
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
if (r && (bundle->phase == PHASE_NETWORK ||
bundle->phys_type.all & PHYS_AUTO)) {
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
/* enough surplus so that we can tell if we're getting swamped */
Add ``set ifqueue'' to control the size of the outgoing packet queue. Doing ``set ifqueue 0'' and ``set urgent none'' will allow full use of luigi's WF2Q code. Requested by: luigi
2000-06-11 02:30:51 +00:00
ifqueue = nlinks > bundle->cfg.ifqueue ? nlinks : bundle->cfg.ifqueue;
if (queued < ifqueue) {
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
/* Not enough - select() for more */
If we've got a full output buffer queue and cannot send anything for two mintues (see ``set choked'' and ``show bundle''), nuke the ip, mp and link level buffer queues. This should fix problems where ``ppp -auto'' seems to stop responding after failing to connect to the peer a few times.
1998-08-25 17:48:43 +00:00
if (bundle->choked.timer.state == TIMER_RUNNING)
timer_Stop(&bundle->choked.timer); /* Not needed any more */
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
FD_SET(bundle->dev.fd, r);
if (*n < bundle->dev.fd + 1)
*n = bundle->dev.fd + 1;
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
log_Printf(LogTIMER, "%s: fdset(r) %d\n", TUN_NAME, bundle->dev.fd);
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
result++;
If we've got a full output buffer queue and cannot send anything for two mintues (see ``set choked'' and ``show bundle''), nuke the ip, mp and link level buffer queues. This should fix problems where ``ppp -auto'' seems to stop responding after failing to connect to the peer a few times.
1998-08-25 17:48:43 +00:00
} else if (bundle->choked.timer.state == TIMER_STOPPED) {
bundle->choked.timer.func = bundle_ClearQueues;
bundle->choked.timer.name = "output choke";
bundle->choked.timer.load = bundle->cfg.choked.timeout * SECTICKS;
bundle->choked.timer.arg = bundle;
timer_Start(&bundle->choked.timer);
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
}
}
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
}
Decouple pap & chap output routines from the corresponding input routines and take advantage of the new init/continue interface in libradius. This allows a timely response on other links in an MP setup while RADIUS requests are in progress as well as the ability to handle other data from the peer in parallel. It should also make the future addition of PAM support trivial. While I'm in there, validate pap & chap header IDs if ``idcheck'' is enabled (the default) for other FSM packet types. NOTE: This involved integrating the generation of chap challenges and the validation of chap responses (and commenting what's going on in those routines). I currently have no way of testing ppps ability to respond to M$Chap CHALLENGEs correctly, so if someone could do the honours, it'd be much appreciated (it *looks* ok!). Sponsored by: Internet Business Solutions Ltd., Switzerland
1999-02-06 02:54:47 +00:00
#ifndef NORADIUS
result += descriptor_UpdateSet(&bundle->radius.desc, r, w, e, n);
#endif
Update the write descriptor selection set for our data links *after* shuffling fragments from the IP queue into the individual link queues. This fixes a latency problem pointed out by: Bruce Evans <bde@zeta.org.au>
1998-06-07 00:16:37 +00:00
/* Which links need a select() ? */
for (dl = bundle->links; dl; dl = dl->next)
result += descriptor_UpdateSet(&dl->desc, r, w, e, n);
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
/*
* This *MUST* be called after the datalink UpdateSet()s as it
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
* might be ``holding'' one of the datalinks (death-row) and
Correct some typos introduced in the descriptor -> fdescriptor change.
2000-03-14 01:47:07 +00:00
* wants to be able to de-select() it from the descriptor set.
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
*/
o Move our prompt descriptor list outside of the bundle. It's now dealt with by the `server' object. This simplifies things as we only have one list of prompt descriptors and the log_ routines check prompt::logactive to determine whether it should be used for output. o Include the MP socket UpdateSet() result in bundle::UpdateSet(). o Don't select on the tun device unless we're in NETWORK phase or AUTO mode. o Stop the idle timer when we go to DEAD phase. We may have transferred a link and not had a chance to kill it. o Don't fail when trying to unlink our transferred datalink from our descriptor lists just before the transfer. o Add our link descriptor to the write set if we got a short write the last time (physical::out is set). o Log the connection source address when a connection is closed. o Remove descriptor::next field. Descriptor lists are not required any more.
1998-05-23 22:24:50 +00:00
result += descriptor_UpdateSet(&bundle->ncp.mp.server.desc, r, w, e, n);
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
return result;
}
static int
To avoid namespace polution in NetBSD: ``struct descriptor'' -> ``struct fdescriptor''
2000-03-14 01:46:09 +00:00
bundle_IsSet(struct fdescriptor *d, const fd_set *fdset)
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
{
struct bundle *bundle = descriptor2bundle(d);
struct datalink *dl;
for (dl = bundle->links; dl; dl = dl->next)
if (descriptor_IsSet(&dl->desc, fdset))
return 1;
Decouple pap & chap output routines from the corresponding input routines and take advantage of the new init/continue interface in libradius. This allows a timely response on other links in an MP setup while RADIUS requests are in progress as well as the ability to handle other data from the peer in parallel. It should also make the future addition of PAM support trivial. While I'm in there, validate pap & chap header IDs if ``idcheck'' is enabled (the default) for other FSM packet types. NOTE: This involved integrating the generation of chap challenges and the validation of chap responses (and commenting what's going on in those routines). I currently have no way of testing ppps ability to respond to M$Chap CHALLENGEs correctly, so if someone could do the honours, it'd be much appreciated (it *looks* ok!). Sponsored by: Internet Business Solutions Ltd., Switzerland
1999-02-06 02:54:47 +00:00
#ifndef NORADIUS
if (descriptor_IsSet(&bundle->radius.desc, fdset))
return 1;
#endif
Check the MP server in bundle_IsSet().
1998-05-17 10:16:14 +00:00
if (descriptor_IsSet(&bundle->ncp.mp.server.desc, fdset))
return 1;
Categorize the fields in struct bundle, and make the tunnel device speed available in ``show bundle''.
1998-05-06 23:50:25 +00:00
return FD_ISSET(bundle->dev.fd, fdset);
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
}
static void
To avoid namespace polution in NetBSD: ``struct descriptor'' -> ``struct fdescriptor''
2000-03-14 01:46:09 +00:00
bundle_DescriptorRead(struct fdescriptor *d, struct bundle *bundle,
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
const fd_set *fdset)
{
struct datalink *dl;
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
unsigned secs;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
u_int32_t af;
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
if (descriptor_IsSet(&bundle->ncp.mp.server.desc, fdset))
descriptor_Read(&bundle->ncp.mp.server.desc, bundle, fdset);
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
for (dl = bundle->links; dl; dl = dl->next)
if (descriptor_IsSet(&dl->desc, fdset))
descriptor_Read(&dl->desc, bundle, fdset);
Deglobalise `struct prompt': o Our diagnostic socket has its password set in the `set socket' line only (not in ppp.secret). o Passwords are per server socket (*VarAuthKey are gone) o Authority is per prompt (VarLocalAuth is gone). o Local logging is per prompt. o Add a `show who' command to see who's connected. No identd routine - just a `where the connection came from' display. o SIGUSR1 is disabled for now - we have no way of choosing a password for the socket created :-( Prompts are attached as a list of `struct descriptor's in struct bundle, and serviced under the bundles descriptor service routines. Ultimately, everything should be done like this. Cosmetic: o alphabeticalise SRCS in Makefile. o Add a few comments in command.h TODO: Start checking that we don't overflow the descriptor sets in select() now that we can have any number of descriptors.
1998-04-03 19:26:02 +00:00
Decouple pap & chap output routines from the corresponding input routines and take advantage of the new init/continue interface in libradius. This allows a timely response on other links in an MP setup while RADIUS requests are in progress as well as the ability to handle other data from the peer in parallel. It should also make the future addition of PAM support trivial. While I'm in there, validate pap & chap header IDs if ``idcheck'' is enabled (the default) for other FSM packet types. NOTE: This involved integrating the generation of chap challenges and the validation of chap responses (and commenting what's going on in those routines). I currently have no way of testing ppps ability to respond to M$Chap CHALLENGEs correctly, so if someone could do the honours, it'd be much appreciated (it *looks* ok!). Sponsored by: Internet Business Solutions Ltd., Switzerland
1999-02-06 02:54:47 +00:00
#ifndef NORADIUS
if (descriptor_IsSet(&bundle->radius.desc, fdset))
descriptor_Read(&bundle->radius.desc, bundle, fdset);
#endif
Categorize the fields in struct bundle, and make the tunnel device speed available in ``show bundle''.
1998-05-06 23:50:25 +00:00
if (FD_ISSET(bundle->dev.fd, fdset)) {
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
struct tun_data tun;
int n, pri;
Tweak a data type from char * to u_char *
2002-05-04 21:47:43 +00:00
u_char *data;
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
size_t sz;
if (bundle->dev.header) {
Tweak a data type from char * to u_char *
2002-05-04 21:47:43 +00:00
data = (u_char *)&tun;
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
sz = sizeof tun;
} else {
data = tun.data;
sz = sizeof tun.data;
}
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
/* something to read from tun */
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
n = read(bundle->dev.fd, data, sz);
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
if (n < 0) {
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
log_Printf(LogWARN, "%s: read: %s\n", bundle->dev.Name, strerror(errno));
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
return;
}
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
if (bundle->dev.header) {
n -= sz - sizeof tun.data;
if (n <= 0) {
log_Printf(LogERROR, "%s: read: Got only %d bytes of data !\n",
bundle->dev.Name, n);
return;
}
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
af = ntohl(tun.header.family);
#ifndef NOINET6
Back out the previous fix to deal with kernels that don't support IPv6, and implement a far more subtle and correct fix. The reason behind the infinite loop was that ppp was trying to make up initial IPv6 numbers and wasn't giving up when it failed unexpectedly to assign the addresses it just fabricated to it's interface (thinking that the reason was because another interface was using the same address). It now attempts this up to 100 times before just failing and trying to muddle along (in reality, this should never happen more than a couple of times unless our random number generator doesn't work). Also, when IPv6 is not available, don't even try to assign the IPv6 interface address in the first place...
2001-08-18 19:07:13 +00:00
if (af != AF_INET && af != AF_INET6)
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
#else
if (af != AF_INET)
#endif
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
/* XXX: Should be maintaining drop/family counts ! */
return;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
} else
af = AF_INET;
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
if (af == AF_INET && ((struct ip *)tun.data)->ip_dst.s_addr ==
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
bundle->ncp.ipcp.my_ip.s_addr) {
/* we've been asked to send something addressed *to* us :( */
if (Enabled(bundle, OPT_LOOPBACK)) {
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
pri = PacketCheck(bundle, af, tun.data, n, &bundle->filter.in,
NULL, NULL);
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
if (pri >= 0) {
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
n += sz - sizeof tun.data;
write(bundle->dev.fd, data, n);
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
log_Printf(LogDEBUG, "Looped back packet addressed to myself\n");
}
return;
} else
log_Printf(LogDEBUG, "Oops - forwarding packet addressed to myself\n");
}
/*
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
* Process on-demand dialup. Output packets are queued within the tunnel
* device until the appropriate NCP is opened.
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
*/
if (bundle_Phase(bundle) == PHASE_DEAD) {
/*
* Note, we must be in AUTO mode :-/ otherwise our interface should
* *not* be UP and we can't receive data
*/
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
pri = PacketCheck(bundle, af, tun.data, n, &bundle->filter.dial,
NULL, NULL);
Fix -auto breakage introduced with the last commit.
2000-07-11 09:26:02 +00:00
if (pri >= 0)
Allow an optional ``!'' in the open, dial & call commands. When used, the redial timer is ignored and the modem is opened immediately.
1998-07-29 18:21:17 +00:00
bundle_Open(bundle, NULL, PHYS_AUTO, 0);
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
else
/*
* Drop the packet. If we were to queue it, we'd just end up with
* a pile of timed-out data in our output queue by the time we get
* around to actually dialing. We'd also prematurely reach the
* threshold at which we stop select()ing to read() the tun
* device - breaking auto-dial.
*/
return;
}
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
secs = 0;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
pri = PacketCheck(bundle, af, tun.data, n, &bundle->filter.out,
NULL, &secs);
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
if (pri >= 0) {
/* Prepend the number of seconds timeout given in the filter */
tun.header.timeout = secs;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
ncp_Enqueue(&bundle->ncp, af, pri, (char *)&tun, n + sizeof tun.header);
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
}
o Don't enter phase TERMINATE in the LCP TLD, leave it exclusively for the last NCP TLF. o Move tun reading from the main loop into the bundle descriptor handling routines. o Cosmetic: Add a few `const's and make some diagnostics clearer.
1998-05-06 18:49:45 +00:00
}
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
}
o If we come out of select() with only write descriptors that end up writing zero bytes, sleep for 1/10 of a second so that we don't end up using up too much cpu. This should only ever happen on systems that wrongly report a descriptor as writable despite the tty buffer being full. Discussed with: Jeff Evarts o Do an initial run-time check to see if select() alters the passed timeval. This knowledge isn't yet used, but will be soon.
1998-06-24 19:33:36 +00:00
static int
To avoid namespace polution in NetBSD: ``struct descriptor'' -> ``struct fdescriptor''
2000-03-14 01:46:09 +00:00
bundle_DescriptorWrite(struct fdescriptor *d, struct bundle *bundle,
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
const fd_set *fdset)
{
struct datalink *dl;
o If we come out of select() with only write descriptors that end up writing zero bytes, sleep for 1/10 of a second so that we don't end up using up too much cpu. This should only ever happen on systems that wrongly report a descriptor as writable despite the tty buffer being full. Discussed with: Jeff Evarts o Do an initial run-time check to see if select() alters the passed timeval. This knowledge isn't yet used, but will be soon.
1998-06-24 19:33:36 +00:00
int result = 0;
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
/* This is not actually necessary as struct mpserver doesn't Write() */
if (descriptor_IsSet(&bundle->ncp.mp.server.desc, fdset))
Merge the NETGRAPH branch into HEAD. tty devices now use netgraph's line discipline to do the async escaping, but no other benefits are available yet. Change ``ifdef HAVE_DES'' to ``ifndef NODES'' for consistency. Make the Makefile a little more sane WRT RELEASE_CRUNCH.
2002-03-30 12:30:09 +00:00
if (descriptor_Write(&bundle->ncp.mp.server.desc, bundle, fdset) == 1)
result++;
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
for (dl = bundle->links; dl; dl = dl->next)
if (descriptor_IsSet(&dl->desc, fdset))
Merge the NETGRAPH branch into HEAD. tty devices now use netgraph's line discipline to do the async escaping, but no other benefits are available yet. Change ``ifdef HAVE_DES'' to ``ifndef NODES'' for consistency. Make the Makefile a little more sane WRT RELEASE_CRUNCH.
2002-03-30 12:30:09 +00:00
switch (descriptor_Write(&dl->desc, bundle, fdset)) {
case -1:
datalink_ComeDown(dl, CLOSE_NORMAL);
break;
case 1:
result++;
}
o If we come out of select() with only write descriptors that end up writing zero bytes, sleep for 1/10 of a second so that we don't end up using up too much cpu. This should only ever happen on systems that wrongly report a descriptor as writable despite the tty buffer being full. Discussed with: Jeff Evarts o Do an initial run-time check to see if select() alters the passed timeval. This knowledge isn't yet used, but will be soon.
1998-06-24 19:33:36 +00:00
return result;
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
}
Put the correct pid in /var/run/tunX.pid Suggested by: many
1998-06-06 20:50:57 +00:00
void
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
bundle_LockTun(struct bundle *bundle)
{
FILE *lockfile;
MAXPATHLEN -> PATH_MAX Don't assume MAXHOSTNAMELEN includes the NUL Correct a diagnostic Use "localhost" in our prompt instead of ""
2001-03-08 23:51:50 +00:00
char pidfile[PATH_MAX];
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
snprintf(pidfile, sizeof pidfile, "%stun%d.pid", _PATH_VARRUN, bundle->unit);
lockfile = ID0fopen(pidfile, "w");
if (lockfile != NULL) {
fprintf(lockfile, "%d\n", (int)getpid());
fclose(lockfile);
}
#ifndef RELEASE_CRUNCH
else
log_Printf(LogERROR, "Warning: Can't create %s: %s\n",
pidfile, strerror(errno));
#endif
}
static void
bundle_UnlockTun(struct bundle *bundle)
{
MAXPATHLEN -> PATH_MAX Don't assume MAXHOSTNAMELEN includes the NUL Correct a diagnostic Use "localhost" in our prompt instead of ""
2001-03-08 23:51:50 +00:00
char pidfile[PATH_MAX];
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
snprintf(pidfile, sizeof pidfile, "%stun%d.pid", _PATH_VARRUN, bundle->unit);
ID0unlink(pidfile);
}
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
struct bundle *
Fix ``set proctitle'' by using setproctitle().
1999-11-17 21:12:35 +00:00
bundle_Create(const char *prefix, int type, int unit)
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
{
static struct bundle bundle; /* there can be only one */
Add the -unit command line switch for specifying the tun device. Warn about -alias being depricated (but still allow it). Don't moan twice about failing to open any tun device. Fix a diagnostic and add the -quiet switch to the usage message.
1999-10-19 15:21:09 +00:00
int enoentcount, err, minunit, maxunit;
Move the code for tweaking interface flags into one function.
1999-05-27 08:42:17 +00:00
const char *ifname;
Use modfind() to check if a kld is already loaded. Submitted mostly by: green
1999-11-16 21:57:34 +00:00
#if defined(__FreeBSD__) && !defined(NOKLDLOAD)
Do a kldload() if we get ENXIO trying to open /dev/tun0 Originally submitted by: green
1999-09-22 00:40:47 +00:00
int kldtried;
#endif
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
#if defined(TUNSIFMODE) || defined(TUNSLMODE) || defined(TUNSIFHEAD)
Undo possible damage done by the new TUNSIFMODE ioctl in FreeBSD-current.
1999-03-25 11:37:51 +00:00
int iff;
#endif
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
Solve the ``first connection'' problem that occurs on demand-dial links with dynamic IP numbers where the program that causes the dial bind()s to an interface address that is subsequently changed after ppp negotiation. The problem is defeated by adding negotiated addresses to the tun interface as additional alias addresses and providing a set of ``iface'' commands for managing the interface. Libalias is also required (and what a name clash!) - it happily IP-aliases the address so that the source is that of the primary (negotiated) interface and un-IP-aliases it on the way back. An ``enable iface-alias'' is done implicitly by the -alias command line switch. If -alias isn't given, iface-aliasing is disabled by default and can't be enabled 'till an ``alias enable yes'' is done. ``alias enable no'' silently disables iface-alias. So, for dynamic-IP-type-connections, running ``ppp -alias -auto blah'' will work for the first connection, although existing bindings will not survive a disconnect/connect as the TCP peer will be trying to send to the old IP address - the packets won't route. It's now a lot easier to add IPXCP to ppp with minor updates to the new iface.[ch] (if anyone ever gets 'round to it). It's also now possible to manually add interface aliases with something like ``iface add 1.2.3.4/24 5.6.7.8''. This allows multi-homed ppp links :-)
1998-10-22 02:32:50 +00:00
if (bundle.iface != NULL) { /* Already allocated ! */
Change some log levels. ALERTs are only logged when something that can't happen happens or when everyone needs to know. ERRORs are only logged when something unexpected happens.
1998-06-16 19:40:42 +00:00
log_Printf(LogALERT, "bundle_Create: There's only one BUNDLE !\n");
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
return NULL;
}
Add the -unit command line switch for specifying the tun device. Warn about -alias being depricated (but still allow it). Don't moan twice about failing to open any tun device. Fix a diagnostic and add the -quiet switch to the usage message.
1999-10-19 15:21:09 +00:00
if (unit == -1) {
minunit = 0;
maxunit = -1;
} else {
minunit = unit;
maxunit = unit + 1;
}
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
err = ENOENT;
enoentcount = 0;
Use modfind() to check if a kld is already loaded. Submitted mostly by: green
1999-11-16 21:57:34 +00:00
#if defined(__FreeBSD__) && !defined(NOKLDLOAD)
Do a kldload() if we get ENXIO trying to open /dev/tun0 Originally submitted by: green
1999-09-22 00:40:47 +00:00
kldtried = 0;
#endif
Add the -unit command line switch for specifying the tun device. Warn about -alias being depricated (but still allow it). Don't moan twice about failing to open any tun device. Fix a diagnostic and add the -quiet switch to the usage message.
1999-10-19 15:21:09 +00:00
for (bundle.unit = minunit; bundle.unit != maxunit; bundle.unit++) {
Categorize the fields in struct bundle, and make the tunnel device speed available in ``show bundle''.
1998-05-06 23:50:25 +00:00
snprintf(bundle.dev.Name, sizeof bundle.dev.Name, "%s%d",
prefix, bundle.unit);
bundle.dev.fd = ID0open(bundle.dev.Name, O_RDWR);
if (bundle.dev.fd >= 0)
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
break;
Try to kldload if_tun if we get ENOENT from opening /dev/tunN - not just if we get ENXIO. This makes ppp work with DEVFS when if_tun isn't built into the kernel (without needing to manually kldload it).
2001-02-02 01:41:26 +00:00
else if (errno == ENXIO || errno == ENOENT) {
Use modfind() to check if a kld is already loaded. Submitted mostly by: green
1999-11-16 21:57:34 +00:00
#if defined(__FreeBSD__) && !defined(NOKLDLOAD)
Add the -unit command line switch for specifying the tun device. Warn about -alias being depricated (but still allow it). Don't moan twice about failing to open any tun device. Fix a diagnostic and add the -quiet switch to the usage message.
1999-10-19 15:21:09 +00:00
if (bundle.unit == minunit && !kldtried++) {
Do a kldload() if we get ENXIO trying to open /dev/tun0 Originally submitted by: green
1999-09-22 00:40:47 +00:00
/*
When it's necessary to kldload tun(4), don't forget to re-try to open tun0. Submitted by: qhwt@myrealbox.com
2002-04-15 00:14:40 +00:00
* Attempt to load the tunnel interface KLD if it isn't loaded
* already.
Do a kldload() if we get ENXIO trying to open /dev/tun0 Originally submitted by: green
1999-09-22 00:40:47 +00:00
*/
When it's necessary to kldload tun(4), don't forget to re-try to open tun0. Submitted by: qhwt@myrealbox.com
2002-04-15 00:14:40 +00:00
if (loadmodules(LOAD_VERBOSLY, "if_tun", NULL))
bundle.unit--;
Merge the NETGRAPH branch into HEAD. tty devices now use netgraph's line discipline to do the async escaping, but no other benefits are available yet. Change ``ifdef HAVE_DES'' to ``ifndef NODES'' for consistency. Make the Makefile a little more sane WRT RELEASE_CRUNCH.
2002-03-30 12:30:09 +00:00
continue;
Do a kldload() if we get ENXIO trying to open /dev/tun0 Originally submitted by: green
1999-09-22 00:40:47 +00:00
}
#endif
Deal with ENOENT properly on non-devfs machines Broken in revision 1.109
2001-03-12 12:53:53 +00:00
if (errno != ENOENT || ++enoentcount > 2) {
err = errno;
o Don't limit our tun device number to 256. As long as there's another /dev/ entry, keep trying to open them. o Don't allow ``open ccp'' if lcp isn't open.
1998-04-27 01:40:38 +00:00
break;
Deal with ENOENT properly on non-devfs machines Broken in revision 1.109
2001-03-12 12:53:53 +00:00
}
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
} else
err = errno;
}
Categorize the fields in struct bundle, and make the tunnel device speed available in ``show bundle''.
1998-05-06 23:50:25 +00:00
if (bundle.dev.fd < 0) {
Add the -unit command line switch for specifying the tun device. Warn about -alias being depricated (but still allow it). Don't moan twice about failing to open any tun device. Fix a diagnostic and add the -quiet switch to the usage message.
1999-10-19 15:21:09 +00:00
if (unit == -1)
log_Printf(LogWARN, "No available tunnel devices found (%s)\n",
strerror(err));
else
log_Printf(LogWARN, "%s%d: %s\n", prefix, unit, strerror(err));
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
return NULL;
}
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
log_SetTun(bundle.unit);
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
Solve the ``first connection'' problem that occurs on demand-dial links with dynamic IP numbers where the program that causes the dial bind()s to an interface address that is subsequently changed after ppp negotiation. The problem is defeated by adding negotiated addresses to the tun interface as additional alias addresses and providing a set of ``iface'' commands for managing the interface. Libalias is also required (and what a name clash!) - it happily IP-aliases the address so that the source is that of the primary (negotiated) interface and un-IP-aliases it on the way back. An ``enable iface-alias'' is done implicitly by the -alias command line switch. If -alias isn't given, iface-aliasing is disabled by default and can't be enabled 'till an ``alias enable yes'' is done. ``alias enable no'' silently disables iface-alias. So, for dynamic-IP-type-connections, running ``ppp -alias -auto blah'' will work for the first connection, although existing bindings will not survive a disconnect/connect as the TCP peer will be trying to send to the old IP address - the packets won't route. It's now a lot easier to add IPXCP to ppp with minor updates to the new iface.[ch] (if anyone ever gets 'round to it). It's also now possible to manually add interface aliases with something like ``iface add 1.2.3.4/24 5.6.7.8''. This allows multi-homed ppp links :-)
1998-10-22 02:32:50 +00:00
ifname = strrchr(bundle.dev.Name, '/');
if (ifname == NULL)
ifname = bundle.dev.Name;
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
else
Solve the ``first connection'' problem that occurs on demand-dial links with dynamic IP numbers where the program that causes the dial bind()s to an interface address that is subsequently changed after ppp negotiation. The problem is defeated by adding negotiated addresses to the tun interface as additional alias addresses and providing a set of ``iface'' commands for managing the interface. Libalias is also required (and what a name clash!) - it happily IP-aliases the address so that the source is that of the primary (negotiated) interface and un-IP-aliases it on the way back. An ``enable iface-alias'' is done implicitly by the -alias command line switch. If -alias isn't given, iface-aliasing is disabled by default and can't be enabled 'till an ``alias enable yes'' is done. ``alias enable no'' silently disables iface-alias. So, for dynamic-IP-type-connections, running ``ppp -alias -auto blah'' will work for the first connection, although existing bindings will not survive a disconnect/connect as the TCP peer will be trying to send to the old IP address - the packets won't route. It's now a lot easier to add IPXCP to ppp with minor updates to the new iface.[ch] (if anyone ever gets 'round to it). It's also now possible to manually add interface aliases with something like ``iface add 1.2.3.4/24 5.6.7.8''. This allows multi-homed ppp links :-)
1998-10-22 02:32:50 +00:00
ifname++;
bundle.iface = iface_Create(ifname);
if (bundle.iface == NULL) {
close(bundle.dev.fd);
return NULL;
}
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
Undo possible damage done by the new TUNSIFMODE ioctl in FreeBSD-current.
1999-03-25 11:37:51 +00:00
#ifdef TUNSIFMODE
o Enable IFF_MULTICAST when first opening the tun device (and keep the flag when we ioctl(TUNSIFINFO) under OpenBSD) o Don't bring the interface up immediately o Don't complain about unrecognised interface flags in ``show iface''.
2001-08-21 11:07:58 +00:00
/* Make sure we're POINTOPOINT & IFF_MULTICAST */
iff = IFF_POINTOPOINT | IFF_MULTICAST;
Undo possible damage done by the new TUNSIFMODE ioctl in FreeBSD-current.
1999-03-25 11:37:51 +00:00
if (ID0ioctl(bundle.dev.fd, TUNSIFMODE, &iff) < 0)
log_Printf(LogERROR, "bundle_Create: ioctl(TUNSIFMODE): %s\n",
strerror(errno));
#endif
Disable TUNSLMODE when we first open the tun device. Submitted by: Ian West <ian@niw.com.au>
1999-06-22 11:31:42 +00:00
#ifdef TUNSLMODE
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
/* Make sure we're not prepending sockaddrs */
Disable TUNSLMODE when we first open the tun device. Submitted by: Ian West <ian@niw.com.au>
1999-06-22 11:31:42 +00:00
iff = 0;
if (ID0ioctl(bundle.dev.fd, TUNSLMODE, &iff) < 0)
log_Printf(LogERROR, "bundle_Create: ioctl(TUNSLMODE): %s\n",
strerror(errno));
#endif
Handle the availability of TUNSIFHEAD. If it's there, use it. For the moment this is just overhead, but it'll be used for INET6 support later.
2000-01-23 01:48:19 +00:00
#ifdef TUNSIFHEAD
/* We want the address family please ! */
iff = 1;
if (ID0ioctl(bundle.dev.fd, TUNSIFHEAD, &iff) < 0) {
log_Printf(LogERROR, "bundle_Create: ioctl(TUNSIFHEAD): %s\n",
strerror(errno));
bundle.dev.header = 0;
} else
bundle.dev.header = 1;
#else
#ifdef __OpenBSD__
/* Always present for OpenBSD */
bundle.dev.header = 1;
#else
/*
* If TUNSIFHEAD isn't available and we're not OpenBSD, assume
* everything's AF_INET (hopefully the tun device won't pass us
* anything else !).
*/
bundle.dev.header = 0;
#endif
#endif
Solve the ``first connection'' problem that occurs on demand-dial links with dynamic IP numbers where the program that causes the dial bind()s to an interface address that is subsequently changed after ppp negotiation. The problem is defeated by adding negotiated addresses to the tun interface as additional alias addresses and providing a set of ``iface'' commands for managing the interface. Libalias is also required (and what a name clash!) - it happily IP-aliases the address so that the source is that of the primary (negotiated) interface and un-IP-aliases it on the way back. An ``enable iface-alias'' is done implicitly by the -alias command line switch. If -alias isn't given, iface-aliasing is disabled by default and can't be enabled 'till an ``alias enable yes'' is done. ``alias enable no'' silently disables iface-alias. So, for dynamic-IP-type-connections, running ``ppp -alias -auto blah'' will work for the first connection, although existing bindings will not survive a disconnect/connect as the TCP peer will be trying to send to the old IP address - the packets won't route. It's now a lot easier to add IPXCP to ppp with minor updates to the new iface.[ch] (if anyone ever gets 'round to it). It's also now possible to manually add interface aliases with something like ``iface add 1.2.3.4/24 5.6.7.8''. This allows multi-homed ppp links :-)
1998-10-22 02:32:50 +00:00
log_Printf(LogPHASE, "Using interface: %s\n", ifname);
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
bundle.bandwidth = 0;
Bring the routing seqno into struct bundle.
1998-02-02 19:33:02 +00:00
bundle.routing_seq = 0;
Move `CleaningUp' into struct bundle.
1998-03-16 22:53:15 +00:00
bundle.phase = PHASE_DEAD;
bundle.CleaningUp = 0;
o Add the -foreground switch. This switch behaves like -background except that ppp stays in the foreground. o Add the -quiet switch to quieten ppps startup o Add the -nat flag and discourage the use of the -alias flag. Both do the same thing. o Correct some nat usage strings. o Change the internal ``alias'' command to ``nat''.
1999-08-19 18:15:52 +00:00
bundle.NatEnabled = 0;
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
bundle.fsm.LayerStart = bundle_LayerStart;
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
bundle.fsm.LayerUp = bundle_LayerUp;
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
bundle.fsm.LayerDown = bundle_LayerDown;
bundle.fsm.LayerFinish = bundle_LayerFinish;
bundle.fsm.object = &bundle;
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
bundle.cfg.idle.timeout = NCP_IDLE_TIMEOUT;
bundle.cfg.idle.min_timeout = 0;
o Move the accept/deny/disable/enable globals either to the bundle (non-negotiated vars) or to their respective IPCP, LCP or CCP. o Enable rolling throughput statistics by default. o Remove the `display' command. These values now appear in `show bundle', `show ipcp', `show ccp' and `show lcp'. o Initialise auth name & key at bundle create time (oops). o Rename pppd-deflate (the id-24 hack) to deflate24. o Don't send both a REJ and a NAK to an IPCP or LCP REQ. Favour the REJ (already done at the CCP level). o Recurse in datalink_UpdateSet() when we change state, otherwise we end up setting no descriptors and getting jammed in the imminent select() instead of doing the dial/login/hangup. o Display our CHAP encryption method despite being built with DES. o Display VJ as not negotiated in ``show ipcp'' when necessary.
1998-04-16 00:26:21 +00:00
*bundle.cfg.auth.name = '\0';
*bundle.cfg.auth.key = '\0';
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
bundle.cfg.opt = OPT_IDCHECK | OPT_LOOPBACK | OPT_SROUTES | OPT_TCPMSSFIXUP |
Probe for the availability of AF_INET6 at startup. If it's not available, default ipv6cp to disabled and refuse to let the user enable it.
2001-08-15 13:53:38 +00:00
OPT_THROUGHPUT | OPT_UTMP;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
#ifndef NOINET6
Probe for the availability of AF_INET6 at startup. If it's not available, default ipv6cp to disabled and refuse to let the user enable it.
2001-08-15 13:53:38 +00:00
bundle.cfg.opt |= OPT_IPCP;
if (probe.ipv6_available)
bundle.cfg.opt |= OPT_IPV6CP;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
#endif
Negotiate MRRU, SHORTSEQ and ENDDISC. ENDDISC doesn't imply multilink ('cos I've seen my ISP REQ it without multilink). Setting MRRU is ifdef'd out until it's debugged and we can merge -direct links with other running programs. Fix MTU setting.
1998-04-23 03:23:03 +00:00
*bundle.cfg.label = '\0';
Add ``set ifqueue'' to control the size of the outgoing packet queue. Doing ``set ifqueue 0'' and ``set urgent none'' will allow full use of luigi's WF2Q code. Requested by: luigi
2000-06-11 02:30:51 +00:00
bundle.cfg.ifqueue = DEF_IFQUEUE;
If we've got a full output buffer queue and cannot send anything for two mintues (see ``set choked'' and ``show bundle''), nuke the ip, mp and link level buffer queues. This should fix problems where ``ppp -auto'' seems to stop responding after failing to connect to the peer a few times.
1998-08-25 17:48:43 +00:00
bundle.cfg.choked.timeout = CHOKED_TIMEOUT;
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
bundle.phys_type.all = type;
bundle.phys_type.open = 0;
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
bundle.upat = 0;
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
bundle.links = datalink_Create("deflink", &bundle, type);
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
if (bundle.links == NULL) {
Change some log levels. ALERTs are only logged when something that can't happen happens or when everyone needs to know. ERRORs are only logged when something unexpected happens.
1998-06-16 19:40:42 +00:00
log_Printf(LogALERT, "Cannot create data link: %s\n", strerror(errno));
Solve the ``first connection'' problem that occurs on demand-dial links with dynamic IP numbers where the program that causes the dial bind()s to an interface address that is subsequently changed after ppp negotiation. The problem is defeated by adding negotiated addresses to the tun interface as additional alias addresses and providing a set of ``iface'' commands for managing the interface. Libalias is also required (and what a name clash!) - it happily IP-aliases the address so that the source is that of the primary (negotiated) interface and un-IP-aliases it on the way back. An ``enable iface-alias'' is done implicitly by the -alias command line switch. If -alias isn't given, iface-aliasing is disabled by default and can't be enabled 'till an ``alias enable yes'' is done. ``alias enable no'' silently disables iface-alias. So, for dynamic-IP-type-connections, running ``ppp -alias -auto blah'' will work for the first connection, although existing bindings will not survive a disconnect/connect as the TCP peer will be trying to send to the old IP address - the packets won't route. It's now a lot easier to add IPXCP to ppp with minor updates to the new iface.[ch] (if anyone ever gets 'round to it). It's also now possible to manually add interface aliases with something like ``iface add 1.2.3.4/24 5.6.7.8''. This allows multi-homed ppp links :-)
1998-10-22 02:32:50 +00:00
iface_Destroy(bundle.iface);
bundle.iface = NULL;
Categorize the fields in struct bundle, and make the tunnel device speed available in ``show bundle''.
1998-05-06 23:50:25 +00:00
close(bundle.dev.fd);
Move pppVars.physical into the bundle (for the moment).
1998-02-06 02:23:48 +00:00
return NULL;
}
Make struct bundle into a sort of `struct descriptor'. It does the fdsets/reads/writes for each of it's datalinks.
1998-03-20 19:48:28 +00:00
bundle.desc.type = BUNDLE_DESCRIPTOR;
bundle.desc.UpdateSet = bundle_UpdateSet;
bundle.desc.IsSet = bundle_IsSet;
bundle.desc.Read = bundle_DescriptorRead;
bundle.desc.Write = bundle_DescriptorWrite;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
ncp_Init(&bundle.ncp, &bundle);
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
Move filter sets into struct bundle.
1998-03-16 22:52:54 +00:00
memset(&bundle.filter, '\0', sizeof bundle.filter);
bundle.filter.in.fragok = bundle.filter.in.logok = 1;
bundle.filter.in.name = "IN";
bundle.filter.out.fragok = bundle.filter.out.logok = 1;
bundle.filter.out.name = "OUT";
bundle.filter.dial.name = "DIAL";
o Drop packets that fail the dial filter when we're in phase DEAD. They'll almost definitely have timed out by the time we dial anyway. o Log dial filters again (LogTCPIP). o Make DEBUG diagnostics for filter checking actually mean something to the common observer. o Do our best to keep any already-configured IP numbers at IPCP negotiation time. We always first request our configured IP, and if the peer asks for an invalid IP, we NAK with HISADDR Cosmetic: o Add a linefeed to the `set timeout' arg count error message. o Log unacceptable address errors to LogPHASE if LogIPCP is switched off. o Fix ``destination system not found'' error message. o Get out immediately if we get a fatal error before entering the main loop.
1998-04-03 19:25:07 +00:00
bundle.filter.dial.logok = 1;
Move filter sets into struct bundle.
1998-03-16 22:52:54 +00:00
bundle.filter.alive.name = "ALIVE";
bundle.filter.alive.logok = 1;
o Overhaul filtering, adding facilities to jump over rules and to negate the sense of rules. o Remove the redundant (and undocumented) ``host'' and ``port'' words (README.changes updated). o Don't permit (and ignore) garbage instead of the protocol. Mostly submitted by: Peter Jeremy <jeremyp@gsmx07.alcatel.com.au>
1999-07-27 23:44:00 +00:00
{
int i;
for (i = 0; i < MAXFILTERS; i++) {
bundle.filter.in.rule[i].f_action = A_NONE;
bundle.filter.out.rule[i].f_action = A_NONE;
bundle.filter.dial.rule[i].f_action = A_NONE;
bundle.filter.alive.rule[i].f_action = A_NONE;
}
}
Display the correct amount of remaining time in `show time'.
1998-03-18 23:15:31 +00:00
memset(&bundle.idle.timer, '\0', sizeof bundle.idle.timer);
bundle.idle.done = 0;
Tidy up the parent notification stuff for -background mode. Remove extraneous extern decls of tun_{in,out} and netfd.
1998-03-25 18:38:59 +00:00
bundle.notify.fd = -1;
If we've got a full output buffer queue and cannot send anything for two mintues (see ``set choked'' and ``show bundle''), nuke the ip, mp and link level buffer queues. This should fix problems where ``ppp -auto'' seems to stop responding after failing to connect to the peer a few times.
1998-08-25 17:48:43 +00:00
memset(&bundle.choked.timer, '\0', sizeof bundle.choked.timer);
Initial RADIUS support (using libradius). See the man page for details. Compiling with -DNORADIUS (the default for `release') removes support. TODO: The functionality in libradius::rad_send_request() needs to be supplied as a set of routines so that ppp doesn't have to wait indefinitely for the radius server(s). Instead, we need to get a descriptor back, select() on the descriptor, and ask libradius to service it when necessary. For now, ppp blocks SIGALRM while in rad_send_request(), so it misses PAP/CHAP retries & timeouts if they occur. Only PAP is functional. When CHAP is attempted, libradius complains that no User-Password has been specified... rfc2138 says that it *mustn't* be used for CHAP :-( Sponsored by: Internet Business Solutions Ltd., Switzerland
1999-01-28 01:56:34 +00:00
#ifndef NORADIUS
radius_Init(&bundle.radius);
#endif
Display the correct amount of remaining time in `show time'.
1998-03-18 23:15:31 +00:00
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
/* Clean out any leftover crud */
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
iface_Clear(bundle.iface, &bundle.ncp, 0, IFACE_CLEAR_ALL);
Create struct fsm_parent. FSMs are created with one of these, and the FSM passes subsequent events to them. The datalink now hides its CCP from the bundle layer.
1998-02-27 01:22:39 +00:00
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
bundle_LockTun(&bundle);
Create `struct bundle' - the top level control structure. This structure will eventually contain a list of NCPs (currently only IPCP is supported) and a list of physical `struct link's. It will also derive from a struct link itself. Make ModemTimeout() static - it's way to dangerous to be called from outside ! Bump version to 1.9. Our first MP release should be 2.0.
1998-02-02 19:32:16 +00:00
return &bundle;
}
o Add bundle_Destroy(). o Don't remove routes and DOWN the interface when we're closed in auto mode. o Initialize the FSMs in bundle_Create. o Initialize ipcp::if_mine & ipcp::if_peer only once (so that we don't forget that we've SIOCAIFADDR'd the interface). o Do a SIOCDIFADDR on the specific address to avoid hurting any other (still non-existent) NCPs. o Fix some error/diagnostic messages.
1998-02-08 19:29:45 +00:00
static void
bundle_DownInterface(struct bundle *bundle)
{
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
route_IfDelete(bundle, 1);
Bring the PPPoE interface UP if required Suggested by: archie
2001-03-28 09:45:27 +00:00
iface_ClearFlags(bundle->iface->name, IFF_UP);
o Add bundle_Destroy(). o Don't remove routes and DOWN the interface when we're closed in auto mode. o Initialize the FSMs in bundle_Create. o Initialize ipcp::if_mine & ipcp::if_peer only once (so that we don't forget that we've SIOCAIFADDR'd the interface). o Do a SIOCDIFADDR on the specific address to avoid hurting any other (still non-existent) NCPs. o Fix some error/diagnostic messages.
1998-02-08 19:29:45 +00:00
}
void
bundle_Destroy(struct bundle *bundle)
{
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
struct datalink *dl;
Deglobalise `struct prompt': o Our diagnostic socket has its password set in the `set socket' line only (not in ppp.secret). o Passwords are per server socket (*VarAuthKey are gone) o Authority is per prompt (VarLocalAuth is gone). o Local logging is per prompt. o Add a `show who' command to see who's connected. No identd routine - just a `where the connection came from' display. o SIGUSR1 is disabled for now - we have no way of choosing a password for the socket created :-( Prompts are attached as a list of `struct descriptor's in struct bundle, and serviced under the bundles descriptor service routines. Ultimately, everything should be done like this. Cosmetic: o alphabeticalise SRCS in Makefile. o Add a few comments in command.h TODO: Start checking that we don't overflow the descriptor sets in select() now that we can have any number of descriptors.
1998-04-03 19:26:02 +00:00
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
/*
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
* Clean up the interface. We don't really need to do the timer_Stop()s,
* mp_Down(), iface_Clear() and bundle_DownInterface() unless we're getting
Correct some typos introduced in the descriptor -> fdescriptor change.
2000-03-14 01:47:07 +00:00
* out under exceptional conditions such as a descriptor exception.
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
*/
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
timer_Stop(&bundle->idle.timer);
If we've got a full output buffer queue and cannot send anything for two mintues (see ``set choked'' and ``show bundle''), nuke the ip, mp and link level buffer queues. This should fix problems where ``ppp -auto'' seems to stop responding after failing to connect to the peer a few times.
1998-08-25 17:48:43 +00:00
timer_Stop(&bundle->choked.timer);
o Do an mp_Down() when aborting ppp. o Re-assign the link name when strdup()ing the datalink name. o Add the link name to some more hdlc log messages.
1998-05-11 23:39:30 +00:00
mp_Down(&bundle->ncp.mp);
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
iface_Clear(bundle->iface, &bundle->ncp, 0, IFACE_CLEAR_ALL);
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
bundle_DownInterface(bundle);
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
Initial RADIUS support (using libradius). See the man page for details. Compiling with -DNORADIUS (the default for `release') removes support. TODO: The functionality in libradius::rad_send_request() needs to be supplied as a set of routines so that ppp doesn't have to wait indefinitely for the radius server(s). Instead, we need to get a descriptor back, select() on the descriptor, and ask libradius to service it when necessary. For now, ppp blocks SIGALRM while in rad_send_request(), so it misses PAP/CHAP retries & timeouts if they occur. Only PAP is functional. When CHAP is attempted, libradius complains that no User-Password has been specified... rfc2138 says that it *mustn't* be used for CHAP :-( Sponsored by: Internet Business Solutions Ltd., Switzerland
1999-01-28 01:56:34 +00:00
#ifndef NORADIUS
/* Tell the radius server the bad news */
radius_Destroy(&bundle->radius);
#endif
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
/* Again, these are all DATALINK_CLOSED unless we're abending */
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
dl = bundle->links;
while (dl)
dl = datalink_Destroy(dl);
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
ncp_Destroy(&bundle->ncp);
o Split the two IPCP queues into three - one for FSM data (LCP/CCP/IPCP), one for urgent IP traffic and one for everything else. o Add the ``set urgent'' command for adjusting the list of urgent port numbers. The default urgent ports are 21, 22, 23, 513, 514, 543 and 544 (Ports 80 and 81 have been removed from the default priority list). o Increase the buffered packet threshold from 20 to 30. o Report the number of packets in the IP output queue and the list of urgent ports under ``show ipcp''.
1999-09-04 00:00:21 +00:00
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
close(bundle->dev.fd);
bundle_UnlockTun(bundle);
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
/* In case we never made PHASE_NETWORK */
Tidy up the parent notification stuff for -background mode. Remove extraneous extern decls of tun_{in,out} and netfd.
1998-03-25 18:38:59 +00:00
bundle_Notify(bundle, EX_ERRDEAD);
Deglobalise `struct prompt': o Our diagnostic socket has its password set in the `set socket' line only (not in ppp.secret). o Passwords are per server socket (*VarAuthKey are gone) o Authority is per prompt (VarLocalAuth is gone). o Local logging is per prompt. o Add a `show who' command to see who's connected. No identd routine - just a `where the connection came from' display. o SIGUSR1 is disabled for now - we have no way of choosing a password for the socket created :-( Prompts are attached as a list of `struct descriptor's in struct bundle, and serviced under the bundles descriptor service routines. Ultimately, everything should be done like this. Cosmetic: o alphabeticalise SRCS in Makefile. o Add a few comments in command.h TODO: Start checking that we don't overflow the descriptor sets in select() now that we can have any number of descriptors.
1998-04-03 19:26:02 +00:00
Solve the ``first connection'' problem that occurs on demand-dial links with dynamic IP numbers where the program that causes the dial bind()s to an interface address that is subsequently changed after ppp negotiation. The problem is defeated by adding negotiated addresses to the tun interface as additional alias addresses and providing a set of ``iface'' commands for managing the interface. Libalias is also required (and what a name clash!) - it happily IP-aliases the address so that the source is that of the primary (negotiated) interface and un-IP-aliases it on the way back. An ``enable iface-alias'' is done implicitly by the -alias command line switch. If -alias isn't given, iface-aliasing is disabled by default and can't be enabled 'till an ``alias enable yes'' is done. ``alias enable no'' silently disables iface-alias. So, for dynamic-IP-type-connections, running ``ppp -alias -auto blah'' will work for the first connection, although existing bindings will not survive a disconnect/connect as the TCP peer will be trying to send to the old IP address - the packets won't route. It's now a lot easier to add IPXCP to ppp with minor updates to the new iface.[ch] (if anyone ever gets 'round to it). It's also now possible to manually add interface aliases with something like ``iface add 1.2.3.4/24 5.6.7.8''. This allows multi-homed ppp links :-)
1998-10-22 02:32:50 +00:00
iface_Destroy(bundle->iface);
bundle->iface = NULL;
o Add bundle_Destroy(). o Don't remove routes and DOWN the interface when we're closed in auto mode. o Initialize the FSMs in bundle_Create. o Initialize ipcp::if_mine & ipcp::if_peer only once (so that we don't forget that we've SIOCAIFADDR'd the interface). o Do a SIOCDIFADDR on the specific address to avoid hurting any other (still non-existent) NCPs. o Fix some error/diagnostic messages.
1998-02-08 19:29:45 +00:00
}
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
void
bundle_LinkClosed(struct bundle *bundle, struct datalink *dl)
{
/*
* Our datalink has closed.
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
* CleanDatalinks() (called from DoLoop()) will remove closed
Make -foreground a proper option (allowing ``allow mode foreground'', ``set mode foreground'' etc.
1999-11-28 15:50:08 +00:00
* BACKGROUND, FOREGROUND and DIRECT links.
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
* If it's the last data link, enter phase DEAD.
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
*
* NOTE: dl may not be in our list (bundle_SendDatalink()) !
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
*/
Move the dial, login and hangup scripts into struct datalink. Don't set these scripts in -direct mode. Always set reconnect_tries to zero for -direct mode.
1998-02-16 19:10:03 +00:00
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
struct datalink *odl;
int other_links;
Move the dial, login and hangup scripts into struct datalink. Don't set these scripts in -direct mode. Always set reconnect_tries to zero for -direct mode.
1998-02-16 19:10:03 +00:00
When entering ``term'' mode, don't output any messages until the device is successfully opened. If we fail to open it, mention the fact. Also go back into command mode as soon as the device is closed rather than waiting for the user to type something before noticing.
1998-08-09 15:34:11 +00:00
log_SetTtyCommandMode(dl);
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
other_links = 0;
for (odl = bundle->links; odl; odl = odl->next)
if (odl != dl && odl->state != DATALINK_CLOSED)
other_links++;
Sort out (fix) the `term' command. datalink_Up() can now be told to skip the dial/login/hangup scripts and can be told whether to enter packet mode when entering the DATALINK_OPENED state.
1998-02-17 01:05:47 +00:00
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
if (!other_links) {
o Make modes consistent throughout ppp. The same strings are used in `set mode', `allow modes', on the command line and when outputting mode names. The strings are matched so that only enough characters to uniquely identify the string are required, so you can now ppp -a mylabel (for auto mode) ppp -b mylabel (for background mode) ppp -dd mylabel (for direct dial mode) etc. o Make -ddial dial when specified on the command line (oops). Pointed out by: Alex <garbanzo@hooked.net>
1998-05-29 18:32:11 +00:00
if (dl->physical->type != PHYS_AUTO) /* Not in -auto mode */
Don't bring the interface down with the last link when we're in -auto mode (broken a few commits ago).
1998-05-19 23:05:10 +00:00
bundle_DownInterface(bundle);
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
ncp2initial(&bundle->ncp);
Add a sprinkling of mp_Down() calls to ensure that we don't think that the multi-link NCP is up when it isn't.
2002-03-30 12:52:55 +00:00
mp_Down(&bundle->ncp.mp);
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
bundle_NewPhase(bundle, PHASE_DEAD);
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
#ifndef NORADIUS
if (bundle->radius.sessiontime)
bundle_StopSessionTimer(bundle);
#endif
o Move our prompt descriptor list outside of the bundle. It's now dealt with by the `server' object. This simplifies things as we only have one list of prompt descriptors and the log_ routines check prompt::logactive to determine whether it should be used for output. o Include the MP socket UpdateSet() result in bundle::UpdateSet(). o Don't select on the tun device unless we're in NETWORK phase or AUTO mode. o Stop the idle timer when we go to DEAD phase. We may have transferred a link and not had a chance to kill it. o Don't fail when trying to unlink our transferred datalink from our descriptor lists just before the transfer. o Add our link descriptor to the write set if we got a short write the last time (physical::out is set). o Log the connection source address when a connection is closed. o Remove descriptor::next field. Descriptor lists are not required any more.
1998-05-23 22:24:50 +00:00
bundle_StopIdleTimer(bundle);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
}
o Hook the FSMs into our bundle. o The FSM layering is now more sane. o Move a lot of the NCP stuff into our ipcpstate rather than having it in the bundle, including control of the configured IP addresses. We don't need hacks like the global `linkup' variable any more as the FSM decides when our ppp.link* files get run. This is going to eventually be configurable based on FSM events anyway. o Fix a few inconsistencies when both sides require authentication. o We now have "Ppp..." and "PPp" prompts, reflecting authentication and network phase. We don't print loads of spurious prompts as we change phases any more. o Our phase is part of the bundle now. o Fix a bug where the FSM wasn't calling LayerFinish. o Close the FSM down correctly with a signal rather than slamming it down as if the line was dropped (the undocumented ``down'' command is still available though). o Remove the forgotten `tunno' variable and fix references to it.
1998-02-07 20:50:08 +00:00
}
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
void
Allow an optional ``!'' in the open, dial & call commands. When used, the redial timer is ignored and the modem is opened immediately.
1998-07-29 18:21:17 +00:00
bundle_Open(struct bundle *bundle, const char *name, int mask, int force)
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
{
/*
* Please open the given datalink, or all if name == NULL
*/
struct datalink *dl;
for (dl = bundle->links; dl; dl = dl->next)
if (name == NULL || !strcasecmp(dl->name, name)) {
Allow an optional ``!'' in the open, dial & call commands. When used, the redial timer is ignored and the modem is opened immediately.
1998-07-29 18:21:17 +00:00
if ((mask & dl->physical->type) &&
(dl->state == DATALINK_CLOSED ||
(force && dl->state == DATALINK_OPENING &&
When we do a ``dial'' or ``open'' from DATALINK_READY, go to DATALINK_CARRIER and turn off scripting. This should fix instances where ``term'' is used followed by ~. and then ``dial''/``open'' (it currently just sits there looking at you). Reported by: Tim Vanderhoek <vanderh@ecf.utoronto.ca>
2000-05-26 08:26:56 +00:00
dl->dial.timer.state == TIMER_RUNNING) ||
dl->state == DATALINK_READY)) {
timer_Stop(&dl->dial.timer); /* We're finished with this */
o Remove the `mode' global - it's now per physical device. o Shuffle things that live at the datalink level into ``show link'' rather than ``show modem''. o Make both ``show'' commands prettier and more consistent, and display carrier status, link type and our name in ``show modem''. o Show redial and reconnect information in ``show link'' and remove ``show redial'' and ``show reconnect''. o Down the correct link in bundle_LinkLost(). o Remove stale -direct and -background links at the end of our main loop, not when we know they're going. This prevents unexpected pointer-invalidations... o If we ``set server'' with the same values twice, notice and don't moan about failure. o Record dial script despite our link mode. The mode may be changed later (next mod) :-) We never run scripts in -direct and -dedicated modes. o Make ``set server none'' functional again. o Correct datalink state array so that we don't report an ``unknown'' state. o Pass struct ipcp to IpcpCleanInterface, not struct fsm. o Create TUN_PREFIX define rather than hard-coding in main.c o prompt_TtyInit now handles a NULL prompt for -direct mode rather than having to create one then destroy it uncleanly. o Mention our mode in the "PPP Started" LogPHASE message. o Bring all auto links up when we have something to send. o Remove some redundant Physical_*() functions. o Show which connection is running a command when logging commands. o Initialise throughput uptime correctly.
1998-04-10 13:19:23 +00:00
datalink_Up(dl, 1, 1);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
if (mask & PHYS_AUTO)
When we do a ``dial'' or ``open'' from DATALINK_READY, go to DATALINK_CARRIER and turn off scripting. This should fix instances where ``term'' is used followed by ~. and then ``dial''/``open'' (it currently just sits there looking at you). Reported by: Tim Vanderhoek <vanderh@ecf.utoronto.ca>
2000-05-26 08:26:56 +00:00
break; /* Only one AUTO link at a time */
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
}
Create struct datalink. This is a type of physical link that can chat and talk LCP & CCP. A bundle contains a list of these (only one in the list for the moment). The datalink is a type of descriptor, and dials, enters LCP (& does CCP), kicks the bundle when its FSMs do something interesting and does the hangup chat script on the way down. It also handles redials and reconnects. There are lots of loose ends, and probably lots of bugs, but the data structures are getting there !
1998-02-16 00:01:12 +00:00
if (name != NULL)
break;
}
}
struct datalink *
bundle2datalink(struct bundle *bundle, const char *name)
{
struct datalink *dl;
if (name != NULL) {
for (dl = bundle->links; dl; dl = dl->next)
if (!strcasecmp(dl->name, name))
return dl;
} else if (bundle->links && !bundle->links->next)
return bundle->links;
return NULL;
}
Introduce the ``link'' command for controlling individual links.
1998-02-17 19:28:01 +00:00
int
bundle_ShowLinks(struct cmdargs const *arg)
{
o Defer setting up pap/chap based IP numbers & labels until after we've determined if we're going to join another ppp invocation. o Make ``show link'' show all link details, and ``show links'' just give a list of links and their current status. o Show our current label in ``show bundle''. o Allow link cloning and removal as soon as our MRRU is set. o Make ``show lcp'' require context as nothing will ever change in our MP LCP (it's auto-configured as per rfc1990). o Initialise our LQM owner in hdlc_Init(). o Store our endpoint discriminator and authentication name at both the datalink and multilink level and compare them when we've finished AUTHENTICATE phase and before entering NETWORK phase. If they don't match, close the link again. Display the information in the appropriate ``show'' command. o Initialise datalink::phone and datalink::fsmp.object properly when we're cloning the link. o Show which link we're passing LQRs on in our diagnostics. o Reject endpoint discriminator REQs at the logical multilink level. o Remove the rest of our CARRIER and LINK logging setup.
1998-04-24 19:16:15 +00:00
struct datalink *dl;
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
struct pppThroughput *t;
Maintain input and output throughput averages and choose the highest of the two when calculating the MP throughput average for the ``set autoload'' implementation. This makes more sense as all links I know of are full-duplex. This also means that people may need to adjust their autoload settings as 100% bandwidth is now the theoretical maximum rather than 200% (but of course, halfing the current settings is probably not the correct answer either!). This involves a ppp version bump as we need to pass an extra throughput array through the MP local domain socket.
2000-08-15 10:25:42 +00:00
unsigned long long octets;
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
int secs;
Introduce the ``link'' command for controlling individual links.
1998-02-17 19:28:01 +00:00
o Display current link throughput in `show links' (assuming throughput measurement is enabled). o Load balance the links based on weight *and* on a round-robin basis. This makes things fairly even on an output basis. We don't try to allow for the peer sending all his data down one link (and try to send ours back up the other). o Show the number of pending input buffers that can't be processed in ``show mp''. o Fix a typo in the man page.
1998-05-03 22:13:14 +00:00
for (dl = arg->bundle->links; dl; dl = dl->next) {
Maintain input and output throughput averages and choose the highest of the two when calculating the MP throughput average for the ``set autoload'' implementation. This makes more sense as all links I know of are full-duplex. This also means that people may need to adjust their autoload settings as 100% bandwidth is now the theoretical maximum rather than 200% (but of course, halfing the current settings is probably not the correct answer either!). This involves a ppp version bump as we need to pass an extra throughput array through the MP local domain socket.
2000-08-15 10:25:42 +00:00
octets = MAX(dl->physical->link.stats.total.in.OctetsPerSecond,
dl->physical->link.stats.total.out.OctetsPerSecond);
Show the link mode in `show links'.
1998-05-23 22:28:19 +00:00
prompt_Printf(arg->prompt, "Name: %s [%s, %s]",
dl->name, mode2Nam(dl->physical->type), datalink_State(dl));
Calculate the average link throughput using a counter based on the cumulative total of all active links rather than basing it on the total of PROTO_MP traffic. This fixes a problem whereby Cisco routers send PROTO_IP packets only when there's only one link (hmm, what a good idea!).
2000-08-15 00:59:21 +00:00
if (dl->physical->link.stats.total.rolling && dl->state == DATALINK_OPEN)
Change printf formats %q[du] -> %ll[du]
1999-08-09 22:56:18 +00:00
prompt_Printf(arg->prompt, " bandwidth %d, %llu bps (%llu bytes/sec)",
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
dl->mp.bandwidth ? dl->mp.bandwidth :
physical_GetSpeed(dl->physical),
Maintain input and output throughput averages and choose the highest of the two when calculating the MP throughput average for the ``set autoload'' implementation. This makes more sense as all links I know of are full-duplex. This also means that people may need to adjust their autoload settings as 100% bandwidth is now the theoretical maximum rather than 200% (but of course, halfing the current settings is probably not the correct answer either!). This involves a ppp version bump as we need to pass an extra throughput array through the MP local domain socket.
2000-08-15 10:25:42 +00:00
octets * 8, octets);
o Display current link throughput in `show links' (assuming throughput measurement is enabled). o Load balance the links based on weight *and* on a round-robin basis. This makes things fairly even on an output basis. We don't try to allow for the peer sending all his data down one link (and try to send ours back up the other). o Show the number of pending input buffers that can't be processed in ``show mp''. o Fix a typo in the man page.
1998-05-03 22:13:14 +00:00
prompt_Printf(arg->prompt, "\n");
}
Introduce the ``link'' command for controlling individual links.
1998-02-17 19:28:01 +00:00
Calculate the average link throughput using a counter based on the cumulative total of all active links rather than basing it on the total of PROTO_MP traffic. This fixes a problem whereby Cisco routers send PROTO_IP packets only when there's only one link (hmm, what a good idea!).
2000-08-15 00:59:21 +00:00
t = &arg->bundle->ncp.mp.link.stats.total;
Maintain input and output throughput averages and choose the highest of the two when calculating the MP throughput average for the ``set autoload'' implementation. This makes more sense as all links I know of are full-duplex. This also means that people may need to adjust their autoload settings as 100% bandwidth is now the theoretical maximum rather than 200% (but of course, halfing the current settings is probably not the correct answer either!). This involves a ppp version bump as we need to pass an extra throughput array through the MP local domain socket.
2000-08-15 10:25:42 +00:00
octets = MAX(t->in.OctetsPerSecond, t->out.OctetsPerSecond);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
secs = t->downtime ? 0 : throughput_uptime(t);
if (secs > t->SamplePeriod)
secs = t->SamplePeriod;
if (secs)
Change printf formats %q[du] -> %ll[du]
1999-08-09 22:56:18 +00:00
prompt_Printf(arg->prompt, "Currently averaging %llu bps (%llu bytes/sec)"
Maintain input and output throughput averages and choose the highest of the two when calculating the MP throughput average for the ``set autoload'' implementation. This makes more sense as all links I know of are full-duplex. This also means that people may need to adjust their autoload settings as 100% bandwidth is now the theoretical maximum rather than 200% (but of course, halfing the current settings is probably not the correct answer either!). This involves a ppp version bump as we need to pass an extra throughput array through the MP local domain socket.
2000-08-15 10:25:42 +00:00
" over the last %d secs\n", octets * 8, octets, secs);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
Introduce the ``link'' command for controlling individual links.
1998-02-17 19:28:01 +00:00
return 0;
}
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
o Move the accept/deny/disable/enable globals either to the bundle (non-negotiated vars) or to their respective IPCP, LCP or CCP. o Enable rolling throughput statistics by default. o Remove the `display' command. These values now appear in `show bundle', `show ipcp', `show ccp' and `show lcp'. o Initialise auth name & key at bundle create time (oops). o Rename pppd-deflate (the id-24 hack) to deflate24. o Don't send both a REJ and a NAK to an IPCP or LCP REQ. Favour the REJ (already done at the CCP level). o Recurse in datalink_UpdateSet() when we change state, otherwise we end up setting no descriptors and getting jammed in the imminent select() instead of doing the dial/login/hangup. o Display our CHAP encryption method despite being built with DES. o Display VJ as not negotiated in ``show ipcp'' when necessary.
1998-04-16 00:26:21 +00:00
static const char *
optval(struct bundle *bundle, int bit)
{
return (bundle->cfg.opt & bit) ? "enabled" : "disabled";
}
Remove ``show timeout'' and ``show auth'' and add ``show bundle''. Mention ``show timer'' in the man page.
1998-04-11 21:50:47 +00:00
int
bundle_ShowStatus(struct cmdargs const *arg)
{
int remaining;
prompt_Printf(arg->prompt, "Phase %s\n", bundle_PhaseName(arg->bundle));
Categorize the fields in struct bundle, and make the tunnel device speed available in ``show bundle''.
1998-05-06 23:50:25 +00:00
prompt_Printf(arg->prompt, " Device: %s\n", arg->bundle->dev.Name);
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
prompt_Printf(arg->prompt, " Interface: %s @ %lubps",
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
arg->bundle->iface->name, arg->bundle->bandwidth);
Remove ``show timeout'' and ``show auth'' and add ``show bundle''. Mention ``show timer'' in the man page.
1998-04-11 21:50:47 +00:00
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
if (arg->bundle->upat) {
Add a ``log'' command for logging specific information. Add an ``UPTIME'' variable to indicate the bundle uptime. It's now possible to put something like this in ppp.linkdown for a server setup: MYADDR: log Session closing: User USER, address HISADDR, up UPTIME Fixed some memory leakage with commands that expand words. Made some functions static. Fixed a diagnostic bug (iface add .... SIOCDIFADDR)
2001-11-03 21:45:32 +00:00
int secs = bundle_Uptime(arg->bundle);
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
prompt_Printf(arg->prompt, ", up time %d:%02d:%02d", secs / 3600,
(secs / 60) % 60, secs % 60);
}
Fix a printf-style format error
2000-06-18 10:07:53 +00:00
prompt_Printf(arg->prompt, "\n Queued: %lu of %u\n",
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
(unsigned long)ncp_QueueLen(&arg->bundle->ncp),
Fix a printf-style arg cast (again)
2000-06-23 09:48:26 +00:00
arg->bundle->cfg.ifqueue);
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
Add ``set ifqueue'' to control the size of the outgoing packet queue. Doing ``set ifqueue 0'' and ``set urgent none'' will allow full use of luigi's WF2Q code. Requested by: luigi
2000-06-11 02:30:51 +00:00
prompt_Printf(arg->prompt, "\nDefaults:\n");
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " Label: %s\n",
arg->bundle->cfg.label);
prompt_Printf(arg->prompt, " Auth name: %s\n",
Create `struct sticky_route'. Any `add' or `delete' command that uses MYADDR or HISADDR will be added to the sticky route list (show ipcp). When MYADDR or HISADDR change due to IPCP negotiations, and if `sroutes' is enabled (the default), all sticky route entries are updated in the routing table. The end result is that `add default hisaddr' will ``stick'', as will ``add myaddr 255.255.255.255 127.0.0.1'' and ``add 1.2.3.4 255.255.255.0 hisaddr''.
1998-05-05 23:30:13 +00:00
arg->bundle->cfg.auth.name);
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " Diagnostic socket: ");
Only remove socket files with ``set server open''. Only show the mask in ``show bundle'' when it's been specified. Complain about unexpected arguments after ``set server {none,open,closed}'' Log re-open failures as warnings rather than phase messages. Fix some markup for the ``set server'' man page description.
2001-01-29 01:35:06 +00:00
if (*server.cfg.sockname != '\0') {
prompt_Printf(arg->prompt, "%s", server.cfg.sockname);
if (server.cfg.mask != (mode_t)-1)
prompt_Printf(arg->prompt, ", mask 0%03o", (int)server.cfg.mask);
prompt_Printf(arg->prompt, "%s\n", server.fd == -1 ? " (not open)" : "");
} else if (server.cfg.port != 0)
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, "TCP port %d%s\n", server.cfg.port,
server.fd == -1 ? " (not open)" : "");
else
prompt_Printf(arg->prompt, "none\n");
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " Choked Timer: %ds\n",
If we've got a full output buffer queue and cannot send anything for two mintues (see ``set choked'' and ``show bundle''), nuke the ip, mp and link level buffer queues. This should fix problems where ``ppp -auto'' seems to stop responding after failing to connect to the peer a few times.
1998-08-25 17:48:43 +00:00
arg->bundle->cfg.choked.timeout);
Initial RADIUS support (using libradius). See the man page for details. Compiling with -DNORADIUS (the default for `release') removes support. TODO: The functionality in libradius::rad_send_request() needs to be supplied as a set of routines so that ppp doesn't have to wait indefinitely for the radius server(s). Instead, we need to get a descriptor back, select() on the descriptor, and ask libradius to service it when necessary. For now, ppp blocks SIGALRM while in rad_send_request(), so it misses PAP/CHAP retries & timeouts if they occur. Only PAP is functional. When CHAP is attempted, libradius complains that no User-Password has been specified... rfc2138 says that it *mustn't* be used for CHAP :-( Sponsored by: Internet Business Solutions Ltd., Switzerland
1999-01-28 01:56:34 +00:00
#ifndef NORADIUS
radius_Show(&arg->bundle->radius, arg->prompt);
#endif
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " Idle Timer: ");
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
if (arg->bundle->cfg.idle.timeout) {
prompt_Printf(arg->prompt, "%ds", arg->bundle->cfg.idle.timeout);
if (arg->bundle->cfg.idle.min_timeout)
prompt_Printf(arg->prompt, ", min %ds",
arg->bundle->cfg.idle.min_timeout);
Remove ``show timeout'' and ``show auth'' and add ``show bundle''. Mention ``show timer'' in the man page.
1998-04-11 21:50:47 +00:00
remaining = bundle_RemainingIdleTime(arg->bundle);
if (remaining != -1)
prompt_Printf(arg->prompt, " (%ds remaining)", remaining);
prompt_Printf(arg->prompt, "\n");
} else
prompt_Printf(arg->prompt, "disabled\n");
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
prompt_Printf(arg->prompt, " Filter Decap: %-20.20s",
o Log the (payload/size) of all packet types, not just TCP packets o If the new ``filter-decapsulation'' is enabled, delve into UDP packets that contain 0xff 0x03 as the first two bytes, and if we recognise it as PROTO_IP, decapsulate it for the purpose of filter checking. If we recognise it as PROTO_<anything else> mention this for logging purposes only. This change is aimed at people running PPPoUDP where the UDP traffic is being sent over another PPP link. It's desireable to have the top level link connected all the time, but to have the bottom level link capable of decapsulating the traffic and comparing the payload against the filters, thus allowing ``set filter dial ...'' to work in tunnelled environments. The caveat here is that the top ppp cannot employ any compression layers without making the data unreadable for the bottom ppp. ``disable deflate pred1 vj'' and ``deny deflate pred1 vj'' is suggested.
2000-07-07 14:22:08 +00:00
optval(arg->bundle, OPT_FILTERDECAP));
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
prompt_Printf(arg->prompt, " ID check: %s\n",
o Move the accept/deny/disable/enable globals either to the bundle (non-negotiated vars) or to their respective IPCP, LCP or CCP. o Enable rolling throughput statistics by default. o Remove the `display' command. These values now appear in `show bundle', `show ipcp', `show ccp' and `show lcp'. o Initialise auth name & key at bundle create time (oops). o Rename pppd-deflate (the id-24 hack) to deflate24. o Don't send both a REJ and a NAK to an IPCP or LCP REQ. Favour the REJ (already done at the CCP level). o Recurse in datalink_UpdateSet() when we change state, otherwise we end up setting no descriptors and getting jammed in the imminent select() instead of doing the dial/login/hangup. o Display our CHAP encryption method despite being built with DES. o Display VJ as not negotiated in ``show ipcp'' when necessary.
1998-04-16 00:26:21 +00:00
optval(arg->bundle, OPT_IDCHECK));
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
prompt_Printf(arg->prompt, " Iface-Alias: %-20.20s",
optval(arg->bundle, OPT_IFACEALIAS));
#ifndef NOINET6
prompt_Printf(arg->prompt, " IPCP: %s\n",
optval(arg->bundle, OPT_IPCP));
prompt_Printf(arg->prompt, " IPV6CP: %-20.20s",
optval(arg->bundle, OPT_IPV6CP));
#endif
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " Keep-Session: %s\n",
Introduce the ``keep-session'' option. Refer to the man page for details. This allows MP over non-tty devices where the original ppp process must not exit (such as sshd-spawned ppp sessions).
1999-06-02 00:46:55 +00:00
optval(arg->bundle, OPT_KEEPSESSION));
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " Loopback: %-20.20s",
o Move the accept/deny/disable/enable globals either to the bundle (non-negotiated vars) or to their respective IPCP, LCP or CCP. o Enable rolling throughput statistics by default. o Remove the `display' command. These values now appear in `show bundle', `show ipcp', `show ccp' and `show lcp'. o Initialise auth name & key at bundle create time (oops). o Rename pppd-deflate (the id-24 hack) to deflate24. o Don't send both a REJ and a NAK to an IPCP or LCP REQ. Favour the REJ (already done at the CCP level). o Recurse in datalink_UpdateSet() when we change state, otherwise we end up setting no descriptors and getting jammed in the imminent select() instead of doing the dial/login/hangup. o Display our CHAP encryption method despite being built with DES. o Display VJ as not negotiated in ``show ipcp'' when necessary.
1998-04-16 00:26:21 +00:00
optval(arg->bundle, OPT_LOOPBACK));
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " PasswdAuth: %s\n",
o Move the accept/deny/disable/enable globals either to the bundle (non-negotiated vars) or to their respective IPCP, LCP or CCP. o Enable rolling throughput statistics by default. o Remove the `display' command. These values now appear in `show bundle', `show ipcp', `show ccp' and `show lcp'. o Initialise auth name & key at bundle create time (oops). o Rename pppd-deflate (the id-24 hack) to deflate24. o Don't send both a REJ and a NAK to an IPCP or LCP REQ. Favour the REJ (already done at the CCP level). o Recurse in datalink_UpdateSet() when we change state, otherwise we end up setting no descriptors and getting jammed in the imminent select() instead of doing the dial/login/hangup. o Display our CHAP encryption method despite being built with DES. o Display VJ as not negotiated in ``show ipcp'' when necessary.
1998-04-16 00:26:21 +00:00
optval(arg->bundle, OPT_PASSWDAUTH));
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " Proxy: %-20.20s",
o Move the accept/deny/disable/enable globals either to the bundle (non-negotiated vars) or to their respective IPCP, LCP or CCP. o Enable rolling throughput statistics by default. o Remove the `display' command. These values now appear in `show bundle', `show ipcp', `show ccp' and `show lcp'. o Initialise auth name & key at bundle create time (oops). o Rename pppd-deflate (the id-24 hack) to deflate24. o Don't send both a REJ and a NAK to an IPCP or LCP REQ. Favour the REJ (already done at the CCP level). o Recurse in datalink_UpdateSet() when we change state, otherwise we end up setting no descriptors and getting jammed in the imminent select() instead of doing the dial/login/hangup. o Display our CHAP encryption method despite being built with DES. o Display VJ as not negotiated in ``show ipcp'' when necessary.
1998-04-16 00:26:21 +00:00
optval(arg->bundle, OPT_PROXY));
Allow ``set server closed'' to close the diagnostic socket. Allow ``set server open'' to re-open the diagnostic socket. Handle SIGUSR1 by re-opening the diagnostic socket When receiving SIGUSR2 (and in ``set server none''), don't forget the socket details so that ``set server open'' and SIGUSR1 open it again. Don't create the diagnostic socket as uid 0 ! It's far to dangerous.
2001-01-26 01:41:34 +00:00
prompt_Printf(arg->prompt, " Proxyall: %s\n",
Add ``enable proxyall'' support. This adds proxy ARP entries for every machine on every class C or smaller subnet that we route to. Add ``set {send,recv}pipe'' for controlling our socket buffer sizes. Mention the IP number with the problem in a few error messages. All submitted by: Craig Leres <leres@ee.lbl.gov> Modified slightly by: me
1998-10-26 19:07:39 +00:00
optval(arg->bundle, OPT_PROXYALL));
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
prompt_Printf(arg->prompt, " Sticky Routes: %-20.20s",
optval(arg->bundle, OPT_SROUTES));
prompt_Printf(arg->prompt, " TCPMSS Fixup: %s\n",
Add ``enable/disable tcpmssfixup'', defaulting to enabled. Suggested by: julian Hijacked from: ru (ports/net/tcpmssd)
2000-11-28 13:18:35 +00:00
optval(arg->bundle, OPT_TCPMSSFIXUP));
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
prompt_Printf(arg->prompt, " Throughput: %-20.20s",
o Move the accept/deny/disable/enable globals either to the bundle (non-negotiated vars) or to their respective IPCP, LCP or CCP. o Enable rolling throughput statistics by default. o Remove the `display' command. These values now appear in `show bundle', `show ipcp', `show ccp' and `show lcp'. o Initialise auth name & key at bundle create time (oops). o Rename pppd-deflate (the id-24 hack) to deflate24. o Don't send both a REJ and a NAK to an IPCP or LCP REQ. Favour the REJ (already done at the CCP level). o Recurse in datalink_UpdateSet() when we change state, otherwise we end up setting no descriptors and getting jammed in the imminent select() instead of doing the dial/login/hangup. o Display our CHAP encryption method despite being built with DES. o Display VJ as not negotiated in ``show ipcp'' when necessary.
1998-04-16 00:26:21 +00:00
optval(arg->bundle, OPT_THROUGHPUT));
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
prompt_Printf(arg->prompt, " Utmp Logging: %s\n",
o Move the accept/deny/disable/enable globals either to the bundle (non-negotiated vars) or to their respective IPCP, LCP or CCP. o Enable rolling throughput statistics by default. o Remove the `display' command. These values now appear in `show bundle', `show ipcp', `show ccp' and `show lcp'. o Initialise auth name & key at bundle create time (oops). o Rename pppd-deflate (the id-24 hack) to deflate24. o Don't send both a REJ and a NAK to an IPCP or LCP REQ. Favour the REJ (already done at the CCP level). o Recurse in datalink_UpdateSet() when we change state, otherwise we end up setting no descriptors and getting jammed in the imminent select() instead of doing the dial/login/hangup. o Display our CHAP encryption method despite being built with DES. o Display VJ as not negotiated in ``show ipcp'' when necessary.
1998-04-16 00:26:21 +00:00
optval(arg->bundle, OPT_UTMP));
Remove ``show timeout'' and ``show auth'' and add ``show bundle''. Mention ``show timer'' in the man page.
1998-04-11 21:50:47 +00:00
return 0;
}
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
static void
bundle_IdleTimeout(void *v)
{
struct bundle *bundle = (struct bundle *)v;
When running ppp -background, show comfort messages showing when we're redialing/reconnecting. While we're here, log redial, reconnect and phone number announcements to LogCHAT, and reduce some other logging to LogDEBUG.
2000-04-07 23:46:14 +00:00
log_Printf(LogPHASE, "Idle timer expired\n");
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
bundle_StopIdleTimer(bundle);
Make `close lcp' just close the LCP layer and not hangup. This is useful for slirp users that wish to get their shell back after the ppp session. `close' with no args still hangs up as expected. Required by: jmz
1998-06-15 19:05:27 +00:00
bundle_Close(bundle, NULL, CLOSE_STAYDOWN);
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
}
/*
* Start Idle timer. If timeout is reached, we call bundle_Close() to
* close LCP and link.
*/
void
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
bundle_StartIdleTimer(struct bundle *bundle, unsigned secs)
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
{
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
timer_Stop(&bundle->idle.timer);
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
if ((bundle->phys_type.open & (PHYS_DEDICATED|PHYS_DDIAL)) !=
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
bundle->phys_type.open && bundle->cfg.idle.timeout) {
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
time_t now = time(NULL);
if (secs == 0)
secs = bundle->cfg.idle.timeout;
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
/* We want at least `secs' */
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
if (bundle->cfg.idle.min_timeout > secs && bundle->upat) {
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
int up = now - bundle->upat;
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
if ((long long)bundle->cfg.idle.min_timeout - up > (long long)secs)
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
/* Only increase from the current `remaining' value */
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
secs = bundle->cfg.idle.min_timeout - up;
}
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
bundle->idle.timer.func = bundle_IdleTimeout;
bundle->idle.timer.name = "idle";
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
bundle->idle.timer.load = secs * SECTICKS;
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
bundle->idle.timer.arg = bundle;
timer_Start(&bundle->idle.timer);
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
bundle->idle.done = now + secs;
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
}
}
void
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
bundle_SetIdleTimer(struct bundle *bundle, int timeout, int min_timeout)
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
{
Implement a minimum idle time value as an optional second argument to ``set timeout''. This is useful for situations where your minimum call charge is (say) 5 minutes (like mine is)
1999-08-17 17:22:46 +00:00
bundle->cfg.idle.timeout = timeout;
if (min_timeout >= 0)
bundle->cfg.idle.min_timeout = min_timeout;
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
if (ncp_LayersOpen(&bundle->ncp))
Allow a ``timeout secs'' filter option to let specific packet types effect the idle timer in different ways. Submitted by: Stefan Esser <se@freebsd.org> With adjustments by me to document the option in the man page and to give the same semantics for outgoing traffic as incoming. I made the style more consistent in ip.c - this should really have been done as a separate commit.
2000-07-11 22:11:36 +00:00
bundle_StartIdleTimer(bundle, 0);
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
}
void
bundle_StopIdleTimer(struct bundle *bundle)
{
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
timer_Stop(&bundle->idle.timer);
Initialise bundle::idle::done when the timer's stopped.
1998-04-07 01:49:24 +00:00
bundle->idle.done = 0;
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
}
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
static int
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
bundle_RemainingIdleTime(struct bundle *bundle)
{
Display the correct amount of remaining time in `show time'.
1998-03-18 23:15:31 +00:00
if (bundle->idle.done)
return bundle->idle.done - time(NULL);
return -1;
Move the idle timer into struct bundle. There's no link-level idle timer for the moment.
1998-03-09 19:26:41 +00:00
}
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
#ifndef NORADIUS
static void
bundle_SessionTimeout(void *v)
{
struct bundle *bundle = (struct bundle *)v;
log_Printf(LogPHASE, "Session-Timeout timer expired\n");
bundle_StopSessionTimer(bundle);
bundle_Close(bundle, NULL, CLOSE_STAYDOWN);
}
void
bundle_StartSessionTimer(struct bundle *bundle, unsigned secs)
{
timer_Stop(&bundle->session.timer);
if ((bundle->phys_type.open & (PHYS_DEDICATED|PHYS_DDIAL)) !=
bundle->phys_type.open && bundle->radius.sessiontime) {
time_t now = time(NULL);
if (secs == 0)
secs = bundle->radius.sessiontime;
bundle->session.timer.func = bundle_SessionTimeout;
bundle->session.timer.name = "session";
bundle->session.timer.load = secs * SECTICKS;
bundle->session.timer.arg = bundle;
timer_Start(&bundle->session.timer);
bundle->session.done = now + secs;
}
}
void
bundle_StopSessionTimer(struct bundle *bundle)
{
timer_Stop(&bundle->session.timer);
bundle->session.done = 0;
}
#endif
o Move struct lcp and struct ccp into struct link. o Remove bundle2lcp(), bundle2ccp() and bundle2link(). They're too resource-hungry and we have `owner pointers' to do their job. o Make our FSM understand LCPs that are always ST_OPENED (with a minimum code that != 1). o Send FSM code rejects for invalid codes. o Make our bundle fsm_parent deal with multiple links. o Make timer diagnostics pretty and allow access via ~t in `term' mode (not just when logging debug) and `show timers'. Only show timers every second in debug mode, otherwise we get too many diagnostics to be useful (we probably still do). Also, don't restrict ~m in term mode to depend on debug logging. o Rationalise our bundles' phases. o Create struct mp (multilink protocol). This is both an NCP and a type of struct link. It feeds off other NCPs for output, passing fragmented packets into the queues of available datalinks. It also gets PROTO_MP input, reassembles the fragments into ppp frames, and passes them back to the HDLC layer that the fragments were passed from. ** It's not yet possible to enter multilink mode :-( ** o Add `set weight' (requires context) for deciding on a links weighting in multilink mode. Weighting is simplistic (and probably badly implemented) for now. o Remove the function pointers in struct link. They ended up only applying to physical links. o Configure our tun device with an MTU equal to the MRU from struct mp's LCP and a speed equal to the sum of our link speeds. o `show {lcp,ccp,proto}' and `set deflate' now have optional context and use ChooseLink() to decide on which `struct link' to use. This allows behaviour as before when in non-multilink mode, and allows access to the MP logical link in multilink mode. o Ignore reconnect and redial values when in -direct mode and when cleaning up. Always redial when in -ddial or -dedicated mode (unless cleaning up). o Tell our links to `staydown' when we close them due to a signal. o Remove remaining `#ifdef SIGALRM's (ppp doesn't function without alarms). o Don't bother strdup()ing our physical link name. o Various other cosmetic changes.
1998-04-03 19:21:56 +00:00
int
bundle_IsDead(struct bundle *bundle)
{
return !bundle->links || (bundle->phase == PHASE_DEAD && bundle->CleaningUp);
}
Deglobalise `struct prompt': o Our diagnostic socket has its password set in the `set socket' line only (not in ppp.secret). o Passwords are per server socket (*VarAuthKey are gone) o Authority is per prompt (VarLocalAuth is gone). o Local logging is per prompt. o Add a `show who' command to see who's connected. No identd routine - just a `where the connection came from' display. o SIGUSR1 is disabled for now - we have no way of choosing a password for the socket created :-( Prompts are attached as a list of `struct descriptor's in struct bundle, and serviced under the bundles descriptor service routines. Ultimately, everything should be done like this. Cosmetic: o alphabeticalise SRCS in Makefile. o Add a few comments in command.h TODO: Start checking that we don't overflow the descriptor sets in select() now that we can have any number of descriptors.
1998-04-03 19:26:02 +00:00
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
static struct datalink *
bundle_DatalinkLinkout(struct bundle *bundle, struct datalink *dl)
Add the `clone' and `remove' commands for creating and destroying links.
1998-04-05 22:48:25 +00:00
{
struct datalink **dlp;
o Move our prompt descriptor list outside of the bundle. It's now dealt with by the `server' object. This simplifies things as we only have one list of prompt descriptors and the log_ routines check prompt::logactive to determine whether it should be used for output. o Include the MP socket UpdateSet() result in bundle::UpdateSet(). o Don't select on the tun device unless we're in NETWORK phase or AUTO mode. o Stop the idle timer when we go to DEAD phase. We may have transferred a link and not had a chance to kill it. o Don't fail when trying to unlink our transferred datalink from our descriptor lists just before the transfer. o Add our link descriptor to the write set if we got a short write the last time (physical::out is set). o Log the connection source address when a connection is closed. o Remove descriptor::next field. Descriptor lists are not required any more.
1998-05-23 22:24:50 +00:00
for (dlp = &bundle->links; *dlp; dlp = &(*dlp)->next)
if (*dlp == dl) {
*dlp = dl->next;
dl->next = NULL;
bundle_LinksRemoved(bundle);
return dl;
}
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
return NULL;
}
static void
bundle_DatalinkLinkin(struct bundle *bundle, struct datalink *dl)
{
struct datalink **dlp = &bundle->links;
while (*dlp)
dlp = &(*dlp)->next;
*dlp = dl;
dl->next = NULL;
bundle_LinkAdded(bundle, dl);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
mp_CheckAutoloadTimer(&bundle->ncp.mp);
o Remove the `mode' global - it's now per physical device. o Shuffle things that live at the datalink level into ``show link'' rather than ``show modem''. o Make both ``show'' commands prettier and more consistent, and display carrier status, link type and our name in ``show modem''. o Show redial and reconnect information in ``show link'' and remove ``show redial'' and ``show reconnect''. o Down the correct link in bundle_LinkLost(). o Remove stale -direct and -background links at the end of our main loop, not when we know they're going. This prevents unexpected pointer-invalidations... o If we ``set server'' with the same values twice, notice and don't moan about failure. o Record dial script despite our link mode. The mode may be changed later (next mod) :-) We never run scripts in -direct and -dedicated modes. o Make ``set server none'' functional again. o Correct datalink state array so that we don't report an ``unknown'' state. o Pass struct ipcp to IpcpCleanInterface, not struct fsm. o Create TUN_PREFIX define rather than hard-coding in main.c o prompt_TtyInit now handles a NULL prompt for -direct mode rather than having to create one then destroy it uncleanly. o Mention our mode in the "PPP Started" LogPHASE message. o Bring all auto links up when we have something to send. o Remove some redundant Physical_*() functions. o Show which connection is running a command when logging commands. o Initialise throughput uptime correctly.
1998-04-10 13:19:23 +00:00
}
void
bundle_CleanDatalinks(struct bundle *bundle)
{
struct datalink **dlp = &bundle->links;
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
int found = 0;
o Remove the `mode' global - it's now per physical device. o Shuffle things that live at the datalink level into ``show link'' rather than ``show modem''. o Make both ``show'' commands prettier and more consistent, and display carrier status, link type and our name in ``show modem''. o Show redial and reconnect information in ``show link'' and remove ``show redial'' and ``show reconnect''. o Down the correct link in bundle_LinkLost(). o Remove stale -direct and -background links at the end of our main loop, not when we know they're going. This prevents unexpected pointer-invalidations... o If we ``set server'' with the same values twice, notice and don't moan about failure. o Record dial script despite our link mode. The mode may be changed later (next mod) :-) We never run scripts in -direct and -dedicated modes. o Make ``set server none'' functional again. o Correct datalink state array so that we don't report an ``unknown'' state. o Pass struct ipcp to IpcpCleanInterface, not struct fsm. o Create TUN_PREFIX define rather than hard-coding in main.c o prompt_TtyInit now handles a NULL prompt for -direct mode rather than having to create one then destroy it uncleanly. o Mention our mode in the "PPP Started" LogPHASE message. o Bring all auto links up when we have something to send. o Remove some redundant Physical_*() functions. o Show which connection is running a command when logging commands. o Initialise throughput uptime correctly.
1998-04-10 13:19:23 +00:00
while (*dlp)
if ((*dlp)->state == DATALINK_CLOSED &&
Make -foreground a proper option (allowing ``allow mode foreground'', ``set mode foreground'' etc.
1999-11-28 15:50:08 +00:00
(*dlp)->physical->type &
(PHYS_DIRECT|PHYS_BACKGROUND|PHYS_FOREGROUND)) {
o Remove the `mode' global - it's now per physical device. o Shuffle things that live at the datalink level into ``show link'' rather than ``show modem''. o Make both ``show'' commands prettier and more consistent, and display carrier status, link type and our name in ``show modem''. o Show redial and reconnect information in ``show link'' and remove ``show redial'' and ``show reconnect''. o Down the correct link in bundle_LinkLost(). o Remove stale -direct and -background links at the end of our main loop, not when we know they're going. This prevents unexpected pointer-invalidations... o If we ``set server'' with the same values twice, notice and don't moan about failure. o Record dial script despite our link mode. The mode may be changed later (next mod) :-) We never run scripts in -direct and -dedicated modes. o Make ``set server none'' functional again. o Correct datalink state array so that we don't report an ``unknown'' state. o Pass struct ipcp to IpcpCleanInterface, not struct fsm. o Create TUN_PREFIX define rather than hard-coding in main.c o prompt_TtyInit now handles a NULL prompt for -direct mode rather than having to create one then destroy it uncleanly. o Mention our mode in the "PPP Started" LogPHASE message. o Bring all auto links up when we have something to send. o Remove some redundant Physical_*() functions. o Show which connection is running a command when logging commands. o Initialise throughput uptime correctly.
1998-04-10 13:19:23 +00:00
*dlp = datalink_Destroy(*dlp);
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
found++;
} else
o Remove the `mode' global - it's now per physical device. o Shuffle things that live at the datalink level into ``show link'' rather than ``show modem''. o Make both ``show'' commands prettier and more consistent, and display carrier status, link type and our name in ``show modem''. o Show redial and reconnect information in ``show link'' and remove ``show redial'' and ``show reconnect''. o Down the correct link in bundle_LinkLost(). o Remove stale -direct and -background links at the end of our main loop, not when we know they're going. This prevents unexpected pointer-invalidations... o If we ``set server'' with the same values twice, notice and don't moan about failure. o Record dial script despite our link mode. The mode may be changed later (next mod) :-) We never run scripts in -direct and -dedicated modes. o Make ``set server none'' functional again. o Correct datalink state array so that we don't report an ``unknown'' state. o Pass struct ipcp to IpcpCleanInterface, not struct fsm. o Create TUN_PREFIX define rather than hard-coding in main.c o prompt_TtyInit now handles a NULL prompt for -direct mode rather than having to create one then destroy it uncleanly. o Mention our mode in the "PPP Started" LogPHASE message. o Bring all auto links up when we have something to send. o Remove some redundant Physical_*() functions. o Show which connection is running a command when logging commands. o Initialise throughput uptime correctly.
1998-04-10 13:19:23 +00:00
dlp = &(*dlp)->next;
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
if (found)
bundle_LinksRemoved(bundle);
}
int
bundle_DatalinkClone(struct bundle *bundle, struct datalink *dl,
const char *name)
{
if (bundle2datalink(bundle, name)) {
log_Printf(LogWARN, "Clone: %s: name already exists\n", name);
return 0;
}
bundle_DatalinkLinkin(bundle, datalink_Clone(dl, name));
return 1;
}
void
bundle_DatalinkRemove(struct bundle *bundle, struct datalink *dl)
{
dl = bundle_DatalinkLinkout(bundle, dl);
if (dl)
datalink_Destroy(dl);
Add the `clone' and `remove' commands for creating and destroying links.
1998-04-05 22:48:25 +00:00
}
Negotiate MRRU, SHORTSEQ and ENDDISC. ENDDISC doesn't imply multilink ('cos I've seen my ISP REQ it without multilink). Setting MRRU is ifdef'd out until it's debugged and we can merge -direct links with other running programs. Fix MTU setting.
1998-04-23 03:23:03 +00:00
void
bundle_SetLabel(struct bundle *bundle, const char *label)
{
if (label)
strncpy(bundle->cfg.label, label, sizeof bundle->cfg.label - 1);
else
*bundle->cfg.label = '\0';
}
const char *
bundle_GetLabel(struct bundle *bundle)
{
return *bundle->cfg.label ? bundle->cfg.label : NULL;
}
o Add the link name to modem diagnostics. o Create struct mpserver as part of struct mp. mpserver creates a unix-domain socket based on the peers auth name and endpoint discriminator. If it already exists, ppp will ``pass the link'' over to the owner of the socket, joining it into the bundle of another ppp invocation, otherwise ppp waits for other invocations to pass it links through this socket. The final piece of code will be the code that flattens our datalink info and passes it down this channel (not yet implemented).
1998-04-28 01:25:46 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
int
bundle_LinkSize()
{
struct iovec iov[SCATTER_SEGMENTS];
int niov, expect, f;
iov[0].iov_len = strlen(Version) + 1;
iov[0].iov_base = NULL;
niov = 1;
if (datalink2iov(NULL, iov, &niov, SCATTER_SEGMENTS, NULL, NULL) == -1) {
log_Printf(LogERROR, "Cannot determine space required for link\n");
return 0;
}
for (f = expect = 0; f < niov; f++)
expect += iov[f].iov_len;
return expect;
}
o Add the link name to modem diagnostics. o Create struct mpserver as part of struct mp. mpserver creates a unix-domain socket based on the peers auth name and endpoint discriminator. If it already exists, ppp will ``pass the link'' over to the owner of the socket, joining it into the bundle of another ppp invocation, otherwise ppp waits for other invocations to pass it links through this socket. The final piece of code will be the code that flattens our datalink info and passes it down this channel (not yet implemented).
1998-04-28 01:25:46 +00:00
void
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
bundle_ReceiveDatalink(struct bundle *bundle, int s)
o Add the link name to modem diagnostics. o Create struct mpserver as part of struct mp. mpserver creates a unix-domain socket based on the peers auth name and endpoint discriminator. If it already exists, ppp will ``pass the link'' over to the owner of the socket, joining it into the bundle of another ppp invocation, otherwise ppp waits for other invocations to pass it links through this socket. The final piece of code will be the code that flattens our datalink info and passes it down this channel (not yet implemented).
1998-04-28 01:25:46 +00:00
{
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
char cmsgbuf[sizeof(struct cmsghdr) + sizeof(int) * SEND_MAXFD];
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
int niov, expect, f, *fd, nfd, onfd, got;
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
struct iovec iov[SCATTER_SEGMENTS];
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
struct cmsghdr *cmsg;
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
struct msghdr msg;
o Rename datalinks as soon as the name has been received so that LQM and HDLC timer diagnostics come out with the correct name. o Don't send an LQR immediately upon reviving a datalink. Leave it 'till the next timeout. o Add the link name to some more LQR diagnostics. o Break out of the main loop when a descriptor exception is seen in select(). o Remove the evil nointr_[u]sleep() functions. Timers should be (and are) used instead. o Treat a read() of 0 bytes as an error that's fatal to the link on which the read() is done. We should never read() 0 after select() says there's something there - not unless the link has been closed by the other side. o Write the data seen before a HDLC header to the terminal in `term' mode, *not* back to the modem :-/ o Initialise our transmitted file descriptor before starting any timers. o Only send data links that have *no* pending output data. This means that our final ACK will be written rather than being nuked with the datalink transmission.
1998-05-08 01:15:19 +00:00
struct datalink *dl;
o Don't forget to close our transfer socket if we cannot generate the data to transfer. o Transfer uucp lock ownership for the transferred device. o Don't assume we know the correct values of dev_is_modem and mbits after the transfer.
1998-05-28 23:15:40 +00:00
pid_t pid;
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
log_Printf(LogPHASE, "Receiving datalink\n");
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
/*
* Create our scatter/gather array - passing NULL gets the space
* allocation requirement rather than actually flattening the
* structures.
*/
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
iov[0].iov_len = strlen(Version) + 1;
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
iov[0].iov_base = NULL;
niov = 1;
if (datalink2iov(NULL, iov, &niov, SCATTER_SEGMENTS, NULL, NULL) == -1) {
log_Printf(LogERROR, "Cannot determine space required for link\n");
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
return;
o Don't try to transfer tty device descriptors as there's no way of transferring session rights with them. Instead, create two `/bin/cat' processes. A new child is spawned and disassociated from the terminal and the parent, which continues with the rest of the ppp process. Meanwhile, the parent spawns another child, and both the parent and child exec the `/bin/cat' commands with the appropriate descriptors. This way, the session is owned by the parent, and the tty is held open. o Close LCPs that have done a TLF and are now in ST_STOPPED before calling Down. This prevents them from trying to come back up again after the peer has shut them down (it seems a bit strange that the rfc says that a Down in ST_STOPPED will cause a TLS etc). o Don't try to set the physical link name pointer when we're receiving and renaming a datalink. The physical hasn't been created yet, and as it happens, the garbage physical pointer happens to be the value of another physical - so we're pointing that other physical name at ourselves. yeuck. o Re-arrange the order of things in main (DoLoop()). We now handle signals only after the select and not before the UpdateSet. It's possible that either a signal (FSM timeout) or a descriptor_Read() brings a link down, after which we'd better tidy up any dead direct and 1off descriptors before calling UpdateSet() again. o Mention when we detect a PPP packet when we see one before the link is up (then start LCP as before).
1998-05-25 02:22:38 +00:00
}
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
/* Allocate the scatter/gather array for recvmsg() */
for (f = expect = 0; f < niov; f++) {
if ((iov[f].iov_base = malloc(iov[f].iov_len)) == NULL) {
log_Printf(LogERROR, "Cannot allocate space to receive link\n");
return;
}
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
if (f)
expect += iov[f].iov_len;
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
}
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
/* Set up our message */
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
cmsg = (struct cmsghdr *)cmsgbuf;
cmsg->cmsg_len = sizeof cmsgbuf;
cmsg->cmsg_level = SOL_SOCKET;
cmsg->cmsg_type = 0;
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
memset(&msg, '\0', sizeof msg);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
msg.msg_name = NULL;
msg.msg_namelen = 0;
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
msg.msg_iov = iov;
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
msg.msg_iovlen = 1; /* Only send the version at the first pass */
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
msg.msg_control = cmsgbuf;
msg.msg_controllen = sizeof cmsgbuf;
Fix some printf-style argument bugs
2000-03-14 01:47:19 +00:00
log_Printf(LogDEBUG, "Expecting %u scatter/gather bytes\n",
(unsigned)iov[0].iov_len);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
if ((got = recvmsg(s, &msg, MSG_WAITALL)) != iov[0].iov_len) {
if (got == -1)
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
log_Printf(LogERROR, "Failed recvmsg: %s\n", strerror(errno));
else
Fix some printf-style argument bugs
2000-03-14 01:47:19 +00:00
log_Printf(LogERROR, "Failed recvmsg: Got %d, not %u\n",
got, (unsigned)iov[0].iov_len);
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
while (niov--)
free(iov[niov].iov_base);
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
return;
}
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
if (cmsg->cmsg_level != SOL_SOCKET || cmsg->cmsg_type != SCM_RIGHTS) {
log_Printf(LogERROR, "Recvmsg: no descriptors received !\n");
while (niov--)
free(iov[niov].iov_base);
return;
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
}
Hopefully improve control message passing over Unix domain sockets. 1) Allow the sending of more than one control message at a time over a unix domain socket. This should cover the PR 29499. 2) This requires that unp_{ex,in}ternalize and unp_scan understand mbufs with more than one control message at a time. 3) Internalize and externalize used to work on the mbuf in-place. This made life quite complicated and the code for sizeof(int) < sizeof(file *) could end up doing the wrong thing. The patch always create a new mbuf/cluster now. This resulted in the change of the prototype for the domain externalise function. 4) You can now send SCM_TIMESTAMP messages. 5) Always use CMSG_DATA(cm) to determine the start where the data in unp_{ex,in}ternalize. It was using ((struct cmsghdr *)cm + 1) in some places, which gives the wrong alignment on the alpha. (NetBSD made this fix some time ago). This results in an ABI change for discriptor passing and creds passing on the alpha. (Probably on the IA64 and Spare ports too). 6) Fix userland programs to use CMSG_* macros too. 7) Be more careful about freeing mbufs containing (file *)s. This is made possible by the prototype change of externalise. PR: 29499 MFC after: 6 weeks
2001-10-04 13:11:48 +00:00
fd = (int *)CMSG_DATA(cmsg);
nfd = ((caddr_t)cmsg + cmsg->cmsg_len - (caddr_t)fd) / sizeof(int);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
if (nfd < 2) {
Correct some typos introduced in the descriptor -> fdescriptor change.
2000-03-14 01:47:07 +00:00
log_Printf(LogERROR, "Recvmsg: %d descriptor%s received (too few) !\n",
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
nfd, nfd == 1 ? "" : "s");
while (nfd--)
close(fd[nfd]);
Make sure our recvmsg() is a SCM_RIGHTS message when we receive a link from another ppp.
1998-06-18 23:38:04 +00:00
while (niov--)
free(iov[niov].iov_base);
return;
o Don't try to transfer tty device descriptors as there's no way of transferring session rights with them. Instead, create two `/bin/cat' processes. A new child is spawned and disassociated from the terminal and the parent, which continues with the rest of the ppp process. Meanwhile, the parent spawns another child, and both the parent and child exec the `/bin/cat' commands with the appropriate descriptors. This way, the session is owned by the parent, and the tty is held open. o Close LCPs that have done a TLF and are now in ST_STOPPED before calling Down. This prevents them from trying to come back up again after the peer has shut them down (it seems a bit strange that the rfc says that a Down in ST_STOPPED will cause a TLS etc). o Don't try to set the physical link name pointer when we're receiving and renaming a datalink. The physical hasn't been created yet, and as it happens, the garbage physical pointer happens to be the value of another physical - so we're pointing that other physical name at ourselves. yeuck. o Re-arrange the order of things in main (DoLoop()). We now handle signals only after the select and not before the UpdateSet. It's possible that either a signal (FSM timeout) or a descriptor_Read() brings a link down, after which we'd better tidy up any dead direct and 1off descriptors before calling UpdateSet() again. o Mention when we detect a PPP packet when we see one before the link is up (then start LCP as before).
1998-05-25 02:22:38 +00:00
}
When transmitting link file descriptors in multilink mode, ACK our received file descriptor, and don't close a sent descriptor 'till the ack has been received.
1998-05-09 14:44:11 +00:00
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
/*
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
* We've successfully received two or more open file descriptors
* through our socket, plus a version string. Make sure it's the
* correct version, and drop the connection if it's not.
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
*/
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
if (strncmp(Version, iov[0].iov_base, iov[0].iov_len)) {
log_Printf(LogWARN, "Cannot receive datalink, incorrect version"
" (\"%.*s\", not \"%s\")\n", (int)iov[0].iov_len,
Cast void * to char * for log_Printf().
1998-06-27 12:03:46 +00:00
(char *)iov[0].iov_base, Version);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
while (nfd--)
close(fd[nfd]);
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
while (niov--)
free(iov[niov].iov_base);
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
return;
}
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
/*
* Everything looks good. Send the other side our process id so that
* they can transfer lock ownership, and wait for them to send the
* actual link data.
*/
pid = getpid();
if ((got = write(fd[1], &pid, sizeof pid)) != sizeof pid) {
if (got == -1)
log_Printf(LogERROR, "Failed write: %s\n", strerror(errno));
else
log_Printf(LogERROR, "Failed write: Got %d, not %d\n", got,
(int)(sizeof pid));
while (nfd--)
close(fd[nfd]);
while (niov--)
free(iov[niov].iov_base);
return;
}
if ((got = readv(fd[1], iov + 1, niov - 1)) != expect) {
if (got == -1)
log_Printf(LogERROR, "Failed write: %s\n", strerror(errno));
else
log_Printf(LogERROR, "Failed write: Got %d, not %d\n", got, expect);
while (nfd--)
close(fd[nfd]);
while (niov--)
free(iov[niov].iov_base);
return;
}
close(fd[1]);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
onfd = nfd; /* We've got this many in our array */
Correct some typos introduced in the descriptor -> fdescriptor change.
2000-03-14 01:47:07 +00:00
nfd -= 2; /* Don't include p->fd and our reply descriptor */
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
niov = 1; /* Skip the version id */
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
dl = iov2datalink(bundle, iov, &niov, sizeof iov / sizeof *iov, fd[0],
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
fd + 2, &nfd);
o Rename datalinks as soon as the name has been received so that LQM and HDLC timer diagnostics come out with the correct name. o Don't send an LQR immediately upon reviving a datalink. Leave it 'till the next timeout. o Add the link name to some more LQR diagnostics. o Break out of the main loop when a descriptor exception is seen in select(). o Remove the evil nointr_[u]sleep() functions. Timers should be (and are) used instead. o Treat a read() of 0 bytes as an error that's fatal to the link on which the read() is done. We should never read() 0 after select() says there's something there - not unless the link has been closed by the other side. o Write the data seen before a HDLC header to the terminal in `term' mode, *not* back to the modem :-/ o Initialise our transmitted file descriptor before starting any timers. o Only send data links that have *no* pending output data. This means that our final ACK will be written rather than being nuked with the datalink transmission.
1998-05-08 01:15:19 +00:00
if (dl) {
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
if (nfd) {
log_Printf(LogERROR, "bundle_ReceiveDatalink: Failed to handle %d "
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
"auxiliary file descriptors (%d remain)\n", onfd, nfd);
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
datalink_Destroy(dl);
while (nfd--)
close(fd[onfd--]);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
close(fd[0]);
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
} else {
bundle_DatalinkLinkin(bundle, dl);
datalink_AuthOk(dl);
bundle_CalculateBandwidth(dl->bundle);
}
} else {
while (nfd--)
close(fd[onfd--]);
close(fd[0]);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
close(fd[1]);
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
}
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
free(iov[0].iov_base);
o Add the link name to modem diagnostics. o Create struct mpserver as part of struct mp. mpserver creates a unix-domain socket based on the peers auth name and endpoint discriminator. If it already exists, ppp will ``pass the link'' over to the owner of the socket, joining it into the bundle of another ppp invocation, otherwise ppp waits for other invocations to pass it links through this socket. The final piece of code will be the code that flattens our datalink info and passes it down this channel (not yet implemented).
1998-04-28 01:25:46 +00:00
}
void
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
bundle_SendDatalink(struct datalink *dl, int s, struct sockaddr_un *sun)
o Add the link name to modem diagnostics. o Create struct mpserver as part of struct mp. mpserver creates a unix-domain socket based on the peers auth name and endpoint discriminator. If it already exists, ppp will ``pass the link'' over to the owner of the socket, joining it into the bundle of another ppp invocation, otherwise ppp waits for other invocations to pass it links through this socket. The final piece of code will be the code that flattens our datalink info and passes it down this channel (not yet implemented).
1998-04-28 01:25:46 +00:00
{
Hopefully improve control message passing over Unix domain sockets. 1) Allow the sending of more than one control message at a time over a unix domain socket. This should cover the PR 29499. 2) This requires that unp_{ex,in}ternalize and unp_scan understand mbufs with more than one control message at a time. 3) Internalize and externalize used to work on the mbuf in-place. This made life quite complicated and the code for sizeof(int) < sizeof(file *) could end up doing the wrong thing. The patch always create a new mbuf/cluster now. This resulted in the change of the prototype for the domain externalise function. 4) You can now send SCM_TIMESTAMP messages. 5) Always use CMSG_DATA(cm) to determine the start where the data in unp_{ex,in}ternalize. It was using ((struct cmsghdr *)cm + 1) in some places, which gives the wrong alignment on the alpha. (NetBSD made this fix some time ago). This results in an ABI change for discriptor passing and creds passing on the alpha. (Probably on the IA64 and Spare ports too). 6) Fix userland programs to use CMSG_* macros too. 7) Be more careful about freeing mbufs containing (file *)s. This is made possible by the prototype change of externalise. PR: 29499 MFC after: 6 weeks
2001-10-04 13:11:48 +00:00
char cmsgbuf[CMSG_SPACE(sizeof(int) * SEND_MAXFD)];
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
const char *constlock;
char *lock;
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
struct cmsghdr *cmsg;
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
struct msghdr msg;
struct iovec iov[SCATTER_SEGMENTS];
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
int niov, f, expect, newsid, fd[SEND_MAXFD], nfd, reply[2], got;
o Don't forget to close our transfer socket if we cannot generate the data to transfer. o Transfer uucp lock ownership for the transferred device. o Don't assume we know the correct values of dev_is_modem and mbits after the transfer.
1998-05-28 23:15:40 +00:00
pid_t newpid;
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
Cosmetic: Make our external function names consistent.
1998-05-01 19:26:12 +00:00
log_Printf(LogPHASE, "Transmitting datalink %s\n", dl->name);
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
/* Record the base device name for a lock transfer later */
constlock = physical_LockedDevice(dl->physical);
if (constlock) {
lock = alloca(strlen(constlock) + 1);
strcpy(lock, constlock);
} else
lock = NULL;
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
bundle_LinkClosed(dl->bundle, dl);
o Move our prompt descriptor list outside of the bundle. It's now dealt with by the `server' object. This simplifies things as we only have one list of prompt descriptors and the log_ routines check prompt::logactive to determine whether it should be used for output. o Include the MP socket UpdateSet() result in bundle::UpdateSet(). o Don't select on the tun device unless we're in NETWORK phase or AUTO mode. o Stop the idle timer when we go to DEAD phase. We may have transferred a link and not had a chance to kill it. o Don't fail when trying to unlink our transferred datalink from our descriptor lists just before the transfer. o Add our link descriptor to the write set if we got a short write the last time (physical::out is set). o Log the connection source address when a connection is closed. o Remove descriptor::next field. Descriptor lists are not required any more.
1998-05-23 22:24:50 +00:00
bundle_DatalinkLinkout(dl->bundle, dl);
o Call bundle_LinkClosed() when transferring a datalink so that the bundle has the opportunity to go PHASE_DEAD and cleanup the interface (if it's the last link). o Regnerate our phys_type value when we transfer the link. o Always clean up the interface when destroying our bundle in case we're abending. o Always clean up our interface when the last link is gone rather than delaying things 'till exit time in the -direct case (the interface is useless anyway). Do this *after* slamming down our NCPs (if they're still around). o Our MP server descriptor now clears the relevent device descriptor from our descriptor [fd]sets when a datalink is on death-row (to be transferred to another running ppp), thus avoiding the possibility of passing a bum descriptor to select() and having ppp abend. o Handle the MP socket descriptor functions from within the bundle descriptor functions. Now we ensure that the MP socket descriptor functions see the descriptor sets *after* they've been seen by our datalinks. o Add/fix a few more comments.
1998-05-15 18:21:45 +00:00
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
/* Build our scatter/gather array */
iov[0].iov_len = strlen(Version) + 1;
iov[0].iov_base = strdup(Version);
niov = 1;
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
nfd = 0;
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
fd[0] = datalink2iov(dl, iov, &niov, SCATTER_SEGMENTS, fd + 2, &nfd);
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
if (fd[0] != -1 && socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, reply) != -1) {
/*
* fd[1] is used to get the peer process id back, then to confirm that
* we've transferred any device locks to that process id.
*/
fd[1] = reply[1];
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
nfd += 2; /* Include fd[0] and fd[1] */
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
memset(&msg, '\0', sizeof msg);
o Don't try to transfer tty device descriptors as there's no way of transferring session rights with them. Instead, create two `/bin/cat' processes. A new child is spawned and disassociated from the terminal and the parent, which continues with the rest of the ppp process. Meanwhile, the parent spawns another child, and both the parent and child exec the `/bin/cat' commands with the appropriate descriptors. This way, the session is owned by the parent, and the tty is held open. o Close LCPs that have done a TLF and are now in ST_STOPPED before calling Down. This prevents them from trying to come back up again after the peer has shut them down (it seems a bit strange that the rfc says that a Down in ST_STOPPED will cause a TLS etc). o Don't try to set the physical link name pointer when we're receiving and renaming a datalink. The physical hasn't been created yet, and as it happens, the garbage physical pointer happens to be the value of another physical - so we're pointing that other physical name at ourselves. yeuck. o Re-arrange the order of things in main (DoLoop()). We now handle signals only after the select and not before the UpdateSet. It's possible that either a signal (FSM timeout) or a descriptor_Read() brings a link down, after which we'd better tidy up any dead direct and 1off descriptors before calling UpdateSet() again. o Mention when we detect a PPP packet when we see one before the link is up (then start LCP as before).
1998-05-25 02:22:38 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
msg.msg_name = NULL;
msg.msg_namelen = 0;
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
/*
* Only send the version to start... We used to send the whole lot, but
* this caused problems with our RECVBUF size as a single link is about
* 22k ! This way, we should bump into no limits.
*/
msg.msg_iovlen = 1;
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
msg.msg_iov = iov;
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
msg.msg_control = cmsgbuf;
Hopefully improve control message passing over Unix domain sockets. 1) Allow the sending of more than one control message at a time over a unix domain socket. This should cover the PR 29499. 2) This requires that unp_{ex,in}ternalize and unp_scan understand mbufs with more than one control message at a time. 3) Internalize and externalize used to work on the mbuf in-place. This made life quite complicated and the code for sizeof(int) < sizeof(file *) could end up doing the wrong thing. The patch always create a new mbuf/cluster now. This resulted in the change of the prototype for the domain externalise function. 4) You can now send SCM_TIMESTAMP messages. 5) Always use CMSG_DATA(cm) to determine the start where the data in unp_{ex,in}ternalize. It was using ((struct cmsghdr *)cm + 1) in some places, which gives the wrong alignment on the alpha. (NetBSD made this fix some time ago). This results in an ABI change for discriptor passing and creds passing on the alpha. (Probably on the IA64 and Spare ports too). 6) Fix userland programs to use CMSG_* macros too. 7) Be more careful about freeing mbufs containing (file *)s. This is made possible by the prototype change of externalise. PR: 29499 MFC after: 6 weeks
2001-10-04 13:11:48 +00:00
msg.msg_controllen = CMSG_SPACE(sizeof(int) * nfd);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
msg.msg_flags = 0;
o Don't try to transfer tty device descriptors as there's no way of transferring session rights with them. Instead, create two `/bin/cat' processes. A new child is spawned and disassociated from the terminal and the parent, which continues with the rest of the ppp process. Meanwhile, the parent spawns another child, and both the parent and child exec the `/bin/cat' commands with the appropriate descriptors. This way, the session is owned by the parent, and the tty is held open. o Close LCPs that have done a TLF and are now in ST_STOPPED before calling Down. This prevents them from trying to come back up again after the peer has shut them down (it seems a bit strange that the rfc says that a Down in ST_STOPPED will cause a TLS etc). o Don't try to set the physical link name pointer when we're receiving and renaming a datalink. The physical hasn't been created yet, and as it happens, the garbage physical pointer happens to be the value of another physical - so we're pointing that other physical name at ourselves. yeuck. o Re-arrange the order of things in main (DoLoop()). We now handle signals only after the select and not before the UpdateSet. It's possible that either a signal (FSM timeout) or a descriptor_Read() brings a link down, after which we'd better tidy up any dead direct and 1off descriptors before calling UpdateSet() again. o Mention when we detect a PPP packet when we see one before the link is up (then start LCP as before).
1998-05-25 02:22:38 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
cmsg = (struct cmsghdr *)cmsgbuf;
cmsg->cmsg_len = msg.msg_controllen;
cmsg->cmsg_level = SOL_SOCKET;
cmsg->cmsg_type = SCM_RIGHTS;
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
for (f = 0; f < nfd; f++)
Hopefully improve control message passing over Unix domain sockets. 1) Allow the sending of more than one control message at a time over a unix domain socket. This should cover the PR 29499. 2) This requires that unp_{ex,in}ternalize and unp_scan understand mbufs with more than one control message at a time. 3) Internalize and externalize used to work on the mbuf in-place. This made life quite complicated and the code for sizeof(int) < sizeof(file *) could end up doing the wrong thing. The patch always create a new mbuf/cluster now. This resulted in the change of the prototype for the domain externalise function. 4) You can now send SCM_TIMESTAMP messages. 5) Always use CMSG_DATA(cm) to determine the start where the data in unp_{ex,in}ternalize. It was using ((struct cmsghdr *)cm + 1) in some places, which gives the wrong alignment on the alpha. (NetBSD made this fix some time ago). This results in an ABI change for discriptor passing and creds passing on the alpha. (Probably on the IA64 and Spare ports too). 6) Fix userland programs to use CMSG_* macros too. 7) Be more careful about freeing mbufs containing (file *)s. This is made possible by the prototype change of externalise. PR: 29499 MFC after: 6 weeks
2001-10-04 13:11:48 +00:00
*((int *)CMSG_DATA(cmsg) + f) = fd[f];
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
for (f = 1, expect = 0; f < niov; f++)
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
expect += iov[f].iov_len;
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
if (setsockopt(reply[0], SOL_SOCKET, SO_SNDBUF, &expect, sizeof(int)) == -1)
log_Printf(LogERROR, "setsockopt(SO_RCVBUF, %d): %s\n", expect,
strerror(errno));
if (setsockopt(reply[1], SOL_SOCKET, SO_RCVBUF, &expect, sizeof(int)) == -1)
log_Printf(LogERROR, "setsockopt(SO_RCVBUF, %d): %s\n", expect,
strerror(errno));
Fix some printf-style argument bugs
2000-03-14 01:47:19 +00:00
log_Printf(LogDEBUG, "Sending %d descriptor%s and %u bytes in scatter"
"/gather array\n", nfd, nfd == 1 ? "" : "s",
(unsigned)iov[0].iov_len);
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
if ((got = sendmsg(s, &msg, 0)) == -1)
log_Printf(LogERROR, "Failed sendmsg: %s: %s\n",
sun->sun_path, strerror(errno));
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
else if (got != iov[0].iov_len)
Fix some printf-style argument bugs
2000-03-14 01:47:19 +00:00
log_Printf(LogERROR, "%s: Failed initial sendmsg: Only sent %d of %u\n",
sun->sun_path, got, (unsigned)iov[0].iov_len);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
else {
Correct some typos introduced in the descriptor -> fdescriptor change.
2000-03-14 01:47:07 +00:00
/* We must get the ACK before closing the descriptor ! */
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
int res;
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
if ((got = read(reply[0], &newpid, sizeof newpid)) == sizeof newpid) {
log_Printf(LogDEBUG, "Received confirmation from pid %d\n",
(int)newpid);
if (lock && (res = ID0uu_lock_txfr(lock, newpid)) != UU_LOCK_OK)
When running ppp -background, show comfort messages showing when we're redialing/reconnecting. While we're here, log redial, reconnect and phone number announcements to LogCHAT, and reduce some other logging to LogDEBUG.
2000-04-07 23:46:14 +00:00
log_Printf(LogERROR, "uu_lock_txfr: %s\n", uu_lockerr(res));
Change the way we transfer links (again). The previous method avoided all race conditions, but suffered from sometimes running out of buffer space if enough clients were piled up at the same time. Now, the client pushes the link descriptor, one end of a socketpair() and the ppp version via sendmsg() at the server. The server replies with a pid. The client then transfers any link lock with uu_lock_txfr() and writev()s the actual link contents. The socketpair is now the only place we need to have large socket buffers and the bind()ed socket can keep the default 4k buffer while still handling around 90 racing clients.
1999-11-30 23:52:37 +00:00
log_Printf(LogDEBUG, "Transmitting link (%d bytes)\n", expect);
if ((got = writev(reply[0], iov + 1, niov - 1)) != expect) {
if (got == -1)
log_Printf(LogERROR, "%s: Failed writev: %s\n",
sun->sun_path, strerror(errno));
else
log_Printf(LogERROR, "%s: Failed writev: Wrote %d of %d\n",
sun->sun_path, got, expect);
}
} else if (got == -1)
log_Printf(LogERROR, "%s: Failed socketpair read: %s\n",
sun->sun_path, strerror(errno));
else
log_Printf(LogERROR, "%s: Failed socketpair read: Got %d of %d\n",
sun->sun_path, got, (int)(sizeof newpid));
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
}
close(reply[0]);
close(reply[1]);
o Don't try to transfer tty device descriptors as there's no way of transferring session rights with them. Instead, create two `/bin/cat' processes. A new child is spawned and disassociated from the terminal and the parent, which continues with the rest of the ppp process. Meanwhile, the parent spawns another child, and both the parent and child exec the `/bin/cat' commands with the appropriate descriptors. This way, the session is owned by the parent, and the tty is held open. o Close LCPs that have done a TLF and are now in ST_STOPPED before calling Down. This prevents them from trying to come back up again after the peer has shut them down (it seems a bit strange that the rfc says that a Down in ST_STOPPED will cause a TLS etc). o Don't try to set the physical link name pointer when we're receiving and renaming a datalink. The physical hasn't been created yet, and as it happens, the garbage physical pointer happens to be the value of another physical - so we're pointing that other physical name at ourselves. yeuck. o Re-arrange the order of things in main (DoLoop()). We now handle signals only after the select and not before the UpdateSet. It's possible that either a signal (FSM timeout) or a descriptor_Read() brings a link down, after which we'd better tidy up any dead direct and 1off descriptors before calling UpdateSet() again. o Mention when we detect a PPP packet when we see one before the link is up (then start LCP as before).
1998-05-25 02:22:38 +00:00
Introduce the ``keep-session'' option. Refer to the man page for details. This allows MP over non-tty devices where the original ppp process must not exit (such as sshd-spawned ppp sessions).
1999-06-02 00:46:55 +00:00
newsid = Enabled(dl->bundle, OPT_KEEPSESSION) ||
Support PPPoE Help (lots) from: julian, archie Facilities from: ahebert@pubnix.net
1999-11-06 22:50:59 +00:00
tcgetpgrp(fd[0]) == getpgrp();
while (nfd)
close(fd[--nfd]);
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
if (newsid)
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
bundle_setsid(dl->bundle, got != -1);
o Create a new ``timer'' log level. This lets us ``set log debug'' without filling our filesystem/screen with junk that we don't really want to see. o change PHYS_STDIN to PHYS_DIRECT - we can handle incoming connections that aren't on STDIN_FILENO now. o Allow return values from our FSM LayerUp functions. If LayerUp() fails, the FSM does an immediate FsmDown() without calling the fsm_parent's Layer{Up,Down} functions. o Clear the close-on-exec flag of file descriptor 3 when executing chat programs so that our documented ability to communicate with /dev/tty via that descriptor works. Also document it as descriptor 3, not 4 :-O o Allow a ``rm'' command as an alias for ``remove''. o Fix the bind()/connect()/accept() calls made by the MP server. o Create bundle_SendDatalink() and bundle_ReceiveDatalink(). This allows `struct datalink's to flatten themselves, pass through a pipe (read: the eye of a needle !) and come alive at the other end. The donator then fork()s & exec()s pppmpipe, ``passing'' the connection to another ppp instance. *** PPP NOW TALKS MULTILINK :-))) *** Our link utilization is hideous, and lots of code needs tidying still. It's also probably riddled with bugs ! It's been tested against itself only, and has hung once, so confidence isn't high....
1998-04-30 23:53:56 +00:00
}
o Don't forget to close our transfer socket if we cannot generate the data to transfer. o Transfer uucp lock ownership for the transferred device. o Don't assume we know the correct values of dev_is_modem and mbits after the transfer.
1998-05-28 23:15:40 +00:00
close(s);
o Transfer file descriptors using {send,recv}msg & SCM_RIGHTS, using the scatter/gather array to transfer the link information. The whole link is now passed in one message. This is far better than the two `/bin/cat' processes per additional link :-) I remember years ago thinking that file descriptor transferral would be a really amazing facility ! Suggested by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> and: Eivind Eklund <eivind@yes.no>
1998-05-02 21:57:50 +00:00
while (niov--)
free(iov[niov].iov_base);
o Add the link name to modem diagnostics. o Create struct mpserver as part of struct mp. mpserver creates a unix-domain socket based on the peers auth name and endpoint discriminator. If it already exists, ppp will ``pass the link'' over to the owner of the socket, joining it into the bundle of another ppp invocation, otherwise ppp waits for other invocations to pass it links through this socket. The final piece of code will be the code that flattens our datalink info and passes it down this channel (not yet implemented).
1998-04-28 01:25:46 +00:00
}
o Add the `set mode' command for change a links current mode. It is not possible to switch to or from dedicated or direct mode, but all other combinations are ok (eg. -auto -> -ddial). o Cope with the fact that commands with optional context may not be able to obtain a link with command_ChooseLink() (if all links have been deleted for example). o Allow `clone'ing in non-multilink mode. We may for example want to configure two links in unilink mode and dial them both, using the one that comes up first. It's also possible to rename ``deflink'' by cloning it, deleting the original, then setting the mode of the new link.
1998-05-15 23:58:30 +00:00
Add the `rename' command for renaming links.
1998-05-16 23:47:28 +00:00
int
bundle_RenameDatalink(struct bundle *bundle, struct datalink *ndl,
const char *name)
{
struct datalink *dl;
if (!strcasecmp(ndl->name, name))
return 1;
for (dl = bundle->links; dl; dl = dl->next)
if (!strcasecmp(dl->name, name))
return 0;
datalink_Rename(ndl, name);
return 1;
}
o Add the `set mode' command for change a links current mode. It is not possible to switch to or from dedicated or direct mode, but all other combinations are ok (eg. -auto -> -ddial). o Cope with the fact that commands with optional context may not be able to obtain a link with command_ChooseLink() (if all links have been deleted for example). o Allow `clone'ing in non-multilink mode. We may for example want to configure two links in unilink mode and dial them both, using the one that comes up first. It's also possible to rename ``deflink'' by cloning it, deleting the original, then setting the mode of the new link.
1998-05-15 23:58:30 +00:00
int
bundle_SetMode(struct bundle *bundle, struct datalink *dl, int mode)
{
int omode;
omode = dl->physical->type;
if (omode == mode)
return 1;
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
if (mode == PHYS_AUTO && !(bundle->phys_type.all & PHYS_AUTO))
/* First auto link */
o Add the `set mode' command for change a links current mode. It is not possible to switch to or from dedicated or direct mode, but all other combinations are ok (eg. -auto -> -ddial). o Cope with the fact that commands with optional context may not be able to obtain a link with command_ChooseLink() (if all links have been deleted for example). o Allow `clone'ing in non-multilink mode. We may for example want to configure two links in unilink mode and dial them both, using the one that comes up first. It's also possible to rename ``deflink'' by cloning it, deleting the original, then setting the mode of the new link.
1998-05-15 23:58:30 +00:00
if (bundle->ncp.ipcp.peer_ip.s_addr == INADDR_ANY) {
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
log_Printf(LogWARN, "You must `set ifaddr' or `open' before"
" changing mode to %s\n", mode2Nam(mode));
o Add the `set mode' command for change a links current mode. It is not possible to switch to or from dedicated or direct mode, but all other combinations are ok (eg. -auto -> -ddial). o Cope with the fact that commands with optional context may not be able to obtain a link with command_ChooseLink() (if all links have been deleted for example). o Allow `clone'ing in non-multilink mode. We may for example want to configure two links in unilink mode and dial them both, using the one that comes up first. It's also possible to rename ``deflink'' by cloning it, deleting the original, then setting the mode of the new link.
1998-05-15 23:58:30 +00:00
return 0;
}
if (!datalink_SetMode(dl, mode))
return 0;
o Maintain a link-type mask for open datalinks as well as for all datalinks in a bundle. Ppp now deals correctly with link types that are changed while open o When changing the type of the last AUTO link, only clear the interface if we're not in PHASE_NETWORK. This allows us to switch to -ddial mode while we have a connection without suddenly unexpectedly throttling ourselves by clearing the interface configuration. Problem area noted by: Aaron Jeremias Luz <aaron@csh.rit.edu>
1998-06-12 17:45:10 +00:00
if (mode == PHYS_AUTO && !(bundle->phys_type.all & PHYS_AUTO) &&
bundle->phase != PHASE_NETWORK)
/* First auto link, we need an interface */
o Add the `set mode' command for change a links current mode. It is not possible to switch to or from dedicated or direct mode, but all other combinations are ok (eg. -auto -> -ddial). o Cope with the fact that commands with optional context may not be able to obtain a link with command_ChooseLink() (if all links have been deleted for example). o Allow `clone'ing in non-multilink mode. We may for example want to configure two links in unilink mode and dial them both, using the one that comes up first. It's also possible to rename ``deflink'' by cloning it, deleting the original, then setting the mode of the new link.
1998-05-15 23:58:30 +00:00
ipcp_InterfaceUp(&bundle->ncp.ipcp);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
/* Regenerate phys_type and adjust idle timer */
o Add `set autoload'. You can now set the minimum and maximum thresholds (in terms of queued packets for a period of time) where -auto links will be brought up and down. By default, all auto links come up when we reach NETWORK phase and never go down. o Display current autoload state in `show bundle'. o Disable the idle timer as soon as it's called. o Disable the idle and autoload timers when exiting (in case we're abending).
1998-05-21 01:13:32 +00:00
bundle_LinksRemoved(bundle);
o Add the `set mode' command for change a links current mode. It is not possible to switch to or from dedicated or direct mode, but all other combinations are ok (eg. -auto -> -ddial). o Cope with the fact that commands with optional context may not be able to obtain a link with command_ChooseLink() (if all links have been deleted for example). o Allow `clone'ing in non-multilink mode. We may for example want to configure two links in unilink mode and dial them both, using the one that comes up first. It's also possible to rename ``deflink'' by cloning it, deleting the original, then setting the mode of the new link.
1998-05-15 23:58:30 +00:00
return 1;
}
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
void
bundle_setsid(struct bundle *bundle, int holdsession)
{
/*
* Lose the current session. This means getting rid of our pid
* too so that the tty device will really go away, and any getty
* etc will be allowed to restart.
*/
pid_t pid, orig;
int fds[2];
char done;
struct datalink *dl;
Don't bother fork()ing after closing a ctty if ppp is about to terminate anyway.
1999-12-23 21:43:25 +00:00
if (!holdsession && bundle_IsDead(bundle)) {
/*
* No need to lose our session after all... we're going away anyway
*
* We should really stop the timer and pause if holdsession is set and
* the bundle's dead, but that leaves other resources lying about :-(
*/
return;
}
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
orig = getpid();
if (pipe(fds) == -1) {
log_Printf(LogERROR, "pipe: %s\n", strerror(errno));
return;
}
switch ((pid = fork())) {
case -1:
log_Printf(LogERROR, "fork: %s\n", strerror(errno));
close(fds[0]);
close(fds[1]);
return;
case 0:
close(fds[1]);
Read from fd[0] and write to fd[1] where `fd' is the result of a pipe(). This matters under OpenBSD.
1999-03-07 01:41:40 +00:00
read(fds[0], &done, 1); /* uu_locks are mine ! */
close(fds[0]);
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
if (pipe(fds) == -1) {
log_Printf(LogERROR, "pipe(2): %s\n", strerror(errno));
return;
}
switch ((pid = fork())) {
case -1:
Change some log levels. ALERTs are only logged when something that can't happen happens or when everyone needs to know. ERRORs are only logged when something unexpected happens.
1998-06-16 19:40:42 +00:00
log_Printf(LogERROR, "fork(2): %s\n", strerror(errno));
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
close(fds[0]);
close(fds[1]);
return;
case 0:
close(fds[1]);
Read from fd[0] and write to fd[1] where `fd' is the result of a pipe(). This matters under OpenBSD.
1999-03-07 01:41:40 +00:00
bundle_LockTun(bundle); /* update pid */
read(fds[0], &done, 1); /* uu_locks are mine ! */
close(fds[0]);
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
setsid();
Use TUNSIFPID when we change our PID (if it exists).
2000-01-21 00:33:27 +00:00
bundle_ChangedPID(bundle);
When running ppp -background, show comfort messages showing when we're redialing/reconnecting. While we're here, log redial, reconnect and phone number announcements to LogCHAT, and reduce some other logging to LogDEBUG.
2000-04-07 23:46:14 +00:00
log_Printf(LogDEBUG, "%d -> %d: %s session control\n",
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
(int)orig, (int)getpid(),
holdsession ? "Passed" : "Dropped");
Rather than interrupting 10 times per second then checking to see if there's anything to do, schedule the next alarm based on the next required timeout. This decreases the load when there are lots of relatively idle ppp processes. While I'm in there, handle the possibility that a timeout makes the timer element go out of scope by grabbing the enext pointer before executing the timer function.
1998-12-14 19:24:30 +00:00
timer_InitService(0); /* Start the Timer Service */
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
break;
default:
Read from fd[0] and write to fd[1] where `fd' is the result of a pipe(). This matters under OpenBSD.
1999-03-07 01:41:40 +00:00
close(fds[0]);
o Redesign the layering mechanism and make the aliasing code part of the layering. We now ``stack'' layers as soon as we open the device (when we figure out what we're dealing with). A static set of `dispatch' routines are also declared for dealing with incoming packets after they've been `pulled' up through the stacked layers. Physical devices are now assigned handlers based on the device type when they're opened. For the moment there are three device types; ttys, execs and tcps. o Increment version number to 2.2 o Make an entry in [uw]tmp for non-tty -direct invocations (after pap/chap authentication). o Make throughput counters quad_t's o Account for the absolute number of mbuf malloc()s and free()s in ``show mem''. o ``show modem'' becomes ``show physical''.
1999-05-08 11:07:56 +00:00
/* Give away all our physical locks (to the final process) */
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
for (dl = bundle->links; dl; dl = dl->next)
if (dl->state != DATALINK_CLOSED)
o Redesign the layering mechanism and make the aliasing code part of the layering. We now ``stack'' layers as soon as we open the device (when we figure out what we're dealing with). A static set of `dispatch' routines are also declared for dealing with incoming packets after they've been `pulled' up through the stacked layers. Physical devices are now assigned handlers based on the device type when they're opened. For the moment there are three device types; ttys, execs and tcps. o Increment version number to 2.2 o Make an entry in [uw]tmp for non-tty -direct invocations (after pap/chap authentication). o Make throughput counters quad_t's o Account for the absolute number of mbuf malloc()s and free()s in ``show mem''. o ``show modem'' becomes ``show physical''.
1999-05-08 11:07:56 +00:00
physical_ChangedPid(dl->physical, pid);
Read from fd[0] and write to fd[1] where `fd' is the result of a pipe(). This matters under OpenBSD.
1999-03-07 01:41:40 +00:00
write(fds[1], "!", 1); /* done */
close(fds[1]);
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
_exit(0);
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
break;
}
break;
default:
Read from fd[0] and write to fd[1] where `fd' is the result of a pipe(). This matters under OpenBSD.
1999-03-07 01:41:40 +00:00
close(fds[0]);
o Redesign the layering mechanism and make the aliasing code part of the layering. We now ``stack'' layers as soon as we open the device (when we figure out what we're dealing with). A static set of `dispatch' routines are also declared for dealing with incoming packets after they've been `pulled' up through the stacked layers. Physical devices are now assigned handlers based on the device type when they're opened. For the moment there are three device types; ttys, execs and tcps. o Increment version number to 2.2 o Make an entry in [uw]tmp for non-tty -direct invocations (after pap/chap authentication). o Make throughput counters quad_t's o Account for the absolute number of mbuf malloc()s and free()s in ``show mem''. o ``show modem'' becomes ``show physical''.
1999-05-08 11:07:56 +00:00
/* Give away all our physical locks (to the intermediate process) */
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
for (dl = bundle->links; dl; dl = dl->next)
if (dl->state != DATALINK_CLOSED)
o Redesign the layering mechanism and make the aliasing code part of the layering. We now ``stack'' layers as soon as we open the device (when we figure out what we're dealing with). A static set of `dispatch' routines are also declared for dealing with incoming packets after they've been `pulled' up through the stacked layers. Physical devices are now assigned handlers based on the device type when they're opened. For the moment there are three device types; ttys, execs and tcps. o Increment version number to 2.2 o Make an entry in [uw]tmp for non-tty -direct invocations (after pap/chap authentication). o Make throughput counters quad_t's o Account for the absolute number of mbuf malloc()s and free()s in ``show mem''. o ``show modem'' becomes ``show physical''.
1999-05-08 11:07:56 +00:00
physical_ChangedPid(dl->physical, pid);
Read from fd[0] and write to fd[1] where `fd' is the result of a pipe(). This matters under OpenBSD.
1999-03-07 01:41:40 +00:00
write(fds[1], "!", 1); /* done */
close(fds[1]);
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
if (holdsession) {
int fd, status;
timer_TermService();
signal(SIGPIPE, SIG_DFL);
signal(SIGALRM, SIG_DFL);
signal(SIGHUP, SIG_DFL);
signal(SIGTERM, SIG_DFL);
signal(SIGINT, SIG_DFL);
signal(SIGQUIT, SIG_DFL);
for (fd = getdtablesize(); fd >= 0; fd--)
close(fd);
/*
* Reap the intermediate process. As we're not exiting but the
* intermediate is, we don't want it to become defunct.
*/
waitpid(pid, &status, 0);
o If there's a session leader left running for a descriptor that we're now closing, manually HUP that session leader so that the tty is fully released. o Always restart our carrier detect timer in the receiving process if it was running in the sending process (as we now *always* pass the descriptor). o Tweak argv when we go into pause() mode to keep our session so that ps can see what's going on (without checking for a `pause' state in `ps -l').
1998-05-29 18:33:10 +00:00
/* Tweak our process arguments.... */
setproctitle() doesn't need to be called with root privs, so move it from id.c into defs.c
2000-08-16 09:07:23 +00:00
SetTitle("session owner");
Make -DNOSUID (or -DPPP_NOSUID) possible to build ppp without SUID capabilities.
2000-08-18 00:01:44 +00:00
#ifndef NOSUID
Correct usages of getuid() and geteuid() Pointed out by: billf
1999-12-30 03:36:11 +00:00
setuid(ID0realuid());
Make -DNOSUID (or -DPPP_NOSUID) possible to build ppp without SUID capabilities.
2000-08-18 00:01:44 +00:00
#endif
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
/*
* Hang around for a HUP. This should happen as soon as the
Correct some typos introduced in the descriptor -> fdescriptor change.
2000-03-14 01:47:07 +00:00
* ppp that we passed our ctty descriptor to closes it.
* NOTE: If this process dies, the passed descriptor becomes
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
* invalid and will give a select() error by setting one
* of the error fds, aborting the other ppp. We don't
* want that to happen !
*/
pause();
}
Rewrite the link descriptor transfer code in MP mode. Previously, ppp attempted to bind() to a local domain tcp socket based on the peer authname & enddisc. If it succeeded, it listen()ed and became MP server. If it failed, it connect()ed and became MP client. The server then select()ed on the descriptor, accept()ed it and wrote its pid to it then read the link data & link file descriptor, and finally sent an ack (``!''). The client would read() the server pid, transfer the link lock to that pid, send the link data & descriptor and read the ack. It would then close the descriptor and clean up. There was a race between the bind() and listen() where someone could attempt to connect() and fail. This change removes the race. Now ppp makes the RCVBUF big enough on a socket descriptor and attempts to bind() to a local domain *udp* socket (same name as before). If it succeeds, it becomes MP server. If it fails, it sets the SNDBUF and connect()s, becoming MP client. The server select()s on the descriptor and recvmsg()s the message, insisting on at least two descriptors (plus the link data). It uses the second descriptor to write() its pid then read()s an ack (``!''). The client creates a socketpair() and sendmsg()s the link data, link descriptor and one of the socketpair descriptors. It then read()s the server pid from the other socketpair descriptor, transfers any locks and write()s an ack. Now, there can be no race, and a connect() failure indicates a stale socket file. This also fixes MP ppp over ethernet, where the struct msghdr was being misconstructed when transferring the control socket descriptor. Also, if we fail to send the link, don't hang around in a ``session owner'' state, just do the setsid() and fork() if it's required to disown a tty. UDP idea suggested by: Chris Bennet from Mindspring at FreeBSDCon
1999-11-25 02:47:04 +00:00
_exit(0);
o #define the name "tun" in defs.h against the future possibility of supporting architectures with different device names. o Close /dev/tunX when destroying the bundle. o Don't forget to close the parent end of the pipe in the child process when exec'ing a program from a chat script. o If we close our controlling terminal, ditch the current session with it, allowing getty(8) (or whatever) to regain control. o After transferring our controlling terminal descriptor to another ppp instance, we now fork a new ppp to continue where we left off, transferring ownership of all uucp locks and the /var/run/tunX.pid file. Meanwhile the parent closes all file descriptors, defaults all signals and does a pause() to wait for a HUP after the transferred descriptor is finally closed. We don't run /bin/cat any more (again!). Suggested by: bde TODO: It seems clocal devices need their pause()d session leader to be given a manual HUP, as closing the last open descriptor doesn't do the job.
1998-05-28 23:17:51 +00:00
break;
}
}
Loosen our restrictions on setting enddisc, mrru, shortseq, authname and authkey. o Auth{name,key} may additionally be set in PHASE_ESTABLISH. o The others may be set in PHASE_ESTABLISH as long as no links have yet reached DATALINK_LCP.
1998-10-24 01:08:45 +00:00
int
bundle_HighestState(struct bundle *bundle)
{
struct datalink *dl;
int result = DATALINK_CLOSED;
for (dl = bundle->links; dl; dl = dl->next)
if (result < dl->state)
result = dl->state;
return result;
}
When using ``set device !someprogram'', when ``someprogram'' exits, it causes a select() exception. Handle these select() exceptions on link descriptors in pretty much the same way as loss of carrier rather than dropping out in confusion.
1998-12-10 18:36:30 +00:00
int
bundle_Exception(struct bundle *bundle, int fd)
{
struct datalink *dl;
for (dl = bundle->links; dl; dl = dl->next)
if (dl->physical->fd == fd) {
datalink_Down(dl, CLOSE_NORMAL);
return 1;
}
return 0;
}
Remember if MYADDR or HISADDR is used in a filter add tweak all filters any time either value changes.
1999-05-31 23:57:40 +00:00
void
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
bundle_AdjustFilters(struct bundle *bundle, struct ncpaddr *local,
struct ncpaddr *remote)
Remember if MYADDR or HISADDR is used in a filter add tweak all filters any time either value changes.
1999-05-31 23:57:40 +00:00
{
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
filter_AdjustAddr(&bundle->filter.in, local, remote, NULL);
filter_AdjustAddr(&bundle->filter.out, local, remote, NULL);
filter_AdjustAddr(&bundle->filter.dial, local, remote, NULL);
filter_AdjustAddr(&bundle->filter.alive, local, remote, NULL);
Add the ``resolv'' command for telling ppp how to deal with resolv.conf. You can now ``resolv restore'' in ppp.linkdown ! Add DNS0 and DNS1 macros.
2000-03-14 01:47:27 +00:00
}
void
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
bundle_AdjustDNS(struct bundle *bundle)
Add the ``resolv'' command for telling ppp how to deal with resolv.conf. You can now ``resolv restore'' in ppp.linkdown ! Add DNS0 and DNS1 macros.
2000-03-14 01:47:27 +00:00
{
o Add ipv6 support, abstracting most NCP addresses into opaque structures (well, they're treated as opaque). It's now possible to manage IPv6 interface addresses and routing table entries and to filter IPV6 traffic whether encapsulated or not. IPV6CP support is crude for now, and hasn't been tested against any other implementations. RADIUS and IPv6 are independent of eachother for now. ppp.linkup/ppp.linkdown aren't currently used by IPV6CP o Understand all protocols(5) in filter rules rather than only a select few. o Allow a mask specification for the ``delete'' command. It's now possible to specifically delete one of two conflicting routes. o When creating and deleting proxy arp entries, do it for all IPv4 interface addresses rather than doing it just for the ``current'' peer address. o When iface-alias isn't in effect, don't blow away manually (via ``iface add'') added interface addresses. o When listening on a tcp server (diagnostic) socket, bind so that a tcp46 socket is created -- allowing both IPv4 and IPv6 connections. o When displaying ICMP traffic, don't display the icmp type twice. When display traffic, display at least some information about unrecognised traffic. o Bump version Inspired after filtering work by: Makoto MATSUSHITA <matusita@jp.FreeBSD.org>
2001-08-14 16:05:52 +00:00
struct in_addr *dns = bundle->ncp.ipcp.ns.dns;
Add the ``resolv'' command for telling ppp how to deal with resolv.conf. You can now ``resolv restore'' in ppp.linkdown ! Add DNS0 and DNS1 macros.
2000-03-14 01:47:27 +00:00
filter_AdjustAddr(&bundle->filter.in, NULL, NULL, dns);
filter_AdjustAddr(&bundle->filter.out, NULL, NULL, dns);
filter_AdjustAddr(&bundle->filter.dial, NULL, NULL, dns);
filter_AdjustAddr(&bundle->filter.alive, NULL, NULL, dns);
Remember if MYADDR or HISADDR is used in a filter add tweak all filters any time either value changes.
1999-05-31 23:57:40 +00:00
}
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
void
bundle_CalculateBandwidth(struct bundle *bundle)
{
struct datalink *dl;
Reduce the interface MTU by 2 when MPPE has been successfully negotiated. This is necessary because MPPE will combine the protocol id with the payload received on the tun interface, encrypt it, then prepend its own protocol id, effectively increasing the payload by two bytes.
2001-07-03 22:20:19 +00:00
int sp, overhead, maxoverhead;
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
bundle->bandwidth = 0;
Handle hardware-imposed MTU/MRU limitations. PPPoE will no longer allow MRU/MTU negotiations to exceed 1492. Add an optional ``max'' specifier to ``set m[rt]u'', ie. set mtu max 1480 Bump the ppp version number. Sponsored by: Monzoon Networks AG and FreeBSD Services Limited
2001-06-18 14:59:36 +00:00
bundle->iface->mtu = 0;
Reduce the interface MTU by 2 when MPPE has been successfully negotiated. This is necessary because MPPE will combine the protocol id with the payload received on the tun interface, encrypt it, then prepend its own protocol id, effectively increasing the payload by two bytes.
2001-07-03 22:20:19 +00:00
maxoverhead = 0;
for (dl = bundle->links; dl; dl = dl->next) {
overhead = ccp_MTUOverhead(&dl->physical->link.ccp);
if (maxoverhead < overhead)
maxoverhead = overhead;
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
if (dl->state == DATALINK_OPEN) {
if ((sp = dl->mp.bandwidth) == 0 &&
(sp = physical_GetSpeed(dl->physical)) == 0)
log_Printf(LogDEBUG, "%s: %s: Cannot determine bandwidth\n",
dl->name, dl->physical->name.full);
else
bundle->bandwidth += sp;
if (!bundle->ncp.mp.active) {
Handle hardware-imposed MTU/MRU limitations. PPPoE will no longer allow MRU/MTU negotiations to exceed 1492. Add an optional ``max'' specifier to ``set m[rt]u'', ie. set mtu max 1480 Bump the ppp version number. Sponsored by: Monzoon Networks AG and FreeBSD Services Limited
2001-06-18 14:59:36 +00:00
bundle->iface->mtu = dl->physical->link.lcp.his_mru;
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
break;
}
}
Reduce the interface MTU by 2 when MPPE has been successfully negotiated. This is necessary because MPPE will combine the protocol id with the payload received on the tun interface, encrypt it, then prepend its own protocol id, effectively increasing the payload by two bytes.
2001-07-03 22:20:19 +00:00
}
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
Understand the Session-Timeout RADIUS attribute Store the Filter-Id attribute (we don't do anything with it yet) Submitted mostly by: andrew pavlov <and@kremenchug.net>
2002-05-07 10:06:54 +00:00
if (bundle->bandwidth == 0)
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
bundle->bandwidth = 115200; /* Shrug */
Reduce the interface MTU by 2 when MPPE has been successfully negotiated. This is necessary because MPPE will combine the protocol id with the payload received on the tun interface, encrypt it, then prepend its own protocol id, effectively increasing the payload by two bytes.
2001-07-03 22:20:19 +00:00
if (bundle->ncp.mp.active) {
Handle hardware-imposed MTU/MRU limitations. PPPoE will no longer allow MRU/MTU negotiations to exceed 1492. Add an optional ``max'' specifier to ``set m[rt]u'', ie. set mtu max 1480 Bump the ppp version number. Sponsored by: Monzoon Networks AG and FreeBSD Services Limited
2001-06-18 14:59:36 +00:00
bundle->iface->mtu = bundle->ncp.mp.peer_mrru;
Reduce the interface MTU by 2 when MPPE has been successfully negotiated. This is necessary because MPPE will combine the protocol id with the payload received on the tun interface, encrypt it, then prepend its own protocol id, effectively increasing the payload by two bytes.
2001-07-03 22:20:19 +00:00
overhead = ccp_MTUOverhead(&bundle->ncp.mp.link.ccp);
if (maxoverhead < overhead)
maxoverhead = overhead;
} else if (!bundle->iface->mtu)
Handle hardware-imposed MTU/MRU limitations. PPPoE will no longer allow MRU/MTU negotiations to exceed 1492. Add an optional ``max'' specifier to ``set m[rt]u'', ie. set mtu max 1480 Bump the ppp version number. Sponsored by: Monzoon Networks AG and FreeBSD Services Limited
2001-06-18 14:59:36 +00:00
bundle->iface->mtu = DEF_MRU;
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
#ifndef NORADIUS
Add ``enable/disable tcpmssfixup'', defaulting to enabled. Suggested by: julian Hijacked from: ru (ports/net/tcpmssd)
2000-11-28 13:18:35 +00:00
if (bundle->radius.valid && bundle->radius.mtu &&
Handle hardware-imposed MTU/MRU limitations. PPPoE will no longer allow MRU/MTU negotiations to exceed 1492. Add an optional ``max'' specifier to ``set m[rt]u'', ie. set mtu max 1480 Bump the ppp version number. Sponsored by: Monzoon Networks AG and FreeBSD Services Limited
2001-06-18 14:59:36 +00:00
bundle->radius.mtu < bundle->iface->mtu) {
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
log_Printf(LogLCP, "Reducing MTU to radius value %lu\n",
bundle->radius.mtu);
Handle hardware-imposed MTU/MRU limitations. PPPoE will no longer allow MRU/MTU negotiations to exceed 1492. Add an optional ``max'' specifier to ``set m[rt]u'', ie. set mtu max 1480 Bump the ppp version number. Sponsored by: Monzoon Networks AG and FreeBSD Services Limited
2001-06-18 14:59:36 +00:00
bundle->iface->mtu = bundle->radius.mtu;
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
}
#endif
Reduce the interface MTU by 2 when MPPE has been successfully negotiated. This is necessary because MPPE will combine the protocol id with the payload received on the tun interface, encrypt it, then prepend its own protocol id, effectively increasing the payload by two bytes.
2001-07-03 22:20:19 +00:00
if (maxoverhead) {
log_Printf(LogLCP, "Reducing MTU from %d to %d (CCP requirement)\n",
bundle->iface->mtu, bundle->iface->mtu - maxoverhead);
bundle->iface->mtu -= maxoverhead;
}
Add ``enable/disable tcpmssfixup'', defaulting to enabled. Suggested by: julian Hijacked from: ru (ports/net/tcpmssd)
2000-11-28 13:18:35 +00:00
tun_configure(bundle);
When we change the interface MTU, run through the routing table and tweak all route MTUs too.
2001-04-05 02:23:48 +00:00
route_UpdateMTU(bundle);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
}
void
bundle_AutoAdjust(struct bundle *bundle, int percent, int what)
{
struct datalink *dl, *choice, *otherlinkup;
choice = otherlinkup = NULL;
for (dl = bundle->links; dl; dl = dl->next)
if (dl->physical->type == PHYS_AUTO) {
if (dl->state == DATALINK_OPEN) {
if (what == AUTO_DOWN) {
if (choice)
otherlinkup = choice;
choice = dl;
}
} else if (dl->state == DATALINK_CLOSED) {
if (what == AUTO_UP) {
choice = dl;
break;
}
} else {
/* An auto link in an intermediate state - forget it for the moment */
choice = NULL;
break;
}
} else if (dl->state == DATALINK_OPEN && what == AUTO_DOWN)
otherlinkup = dl;
if (choice) {
if (what == AUTO_UP) {
log_Printf(LogPHASE, "%d%% saturation -> Opening link ``%s''\n",
percent, choice->name);
datalink_Up(choice, 1, 1);
Use mp_CheckAutoloadTimer() rather than mp_StopAutoloadTimer() when opening or closing an auto link due to the autoload setting. Spotted by: David Hedley <david@inty.co.uk>
2000-05-31 15:35:29 +00:00
mp_CheckAutoloadTimer(&bundle->ncp.mp);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
} else if (otherlinkup) { /* Only bring the second-last link down */
log_Printf(LogPHASE, "%d%% saturation -> Closing link ``%s''\n",
percent, choice->name);
Bring links down gently when their throughput isn't enough for the ``set autoload'' value. Don't just ``down'' the link. Don't get stuck in ``ready'' the first time we ``close'' a link.
1999-10-05 05:01:44 +00:00
datalink_Close(choice, CLOSE_STAYDOWN);
Use mp_CheckAutoloadTimer() rather than mp_StopAutoloadTimer() when opening or closing an auto link due to the autoload setting. Spotted by: David Hedley <david@inty.co.uk>
2000-05-31 15:35:29 +00:00
mp_CheckAutoloadTimer(&bundle->ncp.mp);
o Obsolete the undocumented ``set weight'' command. o If we're using RADIUS and the RADIUS mtu is less than our peers mru/mrru, reduce our mtu to this value for NetBSD too. o Make struct throughput's sample period dynamic and tweak the ppp version number to reflect the extra stuff being passed through the local domain socket as a result (MP mode). o Measure the current throughput based on the number of samples actually taken rather than on the full sample period. o Keep the throughput statisics persistent while being passed to another ppp invocation through the local domain socket. o When showing throughput statistics after the timer has stopped, use the stopped time for overall calculations, not the current time. Also show the stopped time and how long the current throughput has been sampled for. o Use time() consistently in throughput.c o Tighten up the ``show bundle'' output. o Introduce the ``set bandwidth'' command. o Rewrite the ``set autoload'' command. It now takes three arguments and works based on a rolling bundle throughput average compared against the theoretical bundle bandwidth over a given period (read: it's now functional).
1999-08-05 10:32:16 +00:00
}
}
}
int
bundle_WantAutoloadTimer(struct bundle *bundle)
{
struct datalink *dl;
int autolink, opened;
if (bundle->phase == PHASE_NETWORK) {
for (autolink = opened = 0, dl = bundle->links; dl; dl = dl->next)
if (dl->physical->type == PHYS_AUTO) {
if (++autolink == 2 || (autolink == 1 && opened))
/* Two auto links or one auto and one open in NETWORK phase */
return 1;
} else if (dl->state == DATALINK_OPEN) {
opened++;
if (autolink)
/* One auto and one open link in NETWORK phase */
return 1;
}
}
return 0;
}
Use TUNSIFPID when we change our PID (if it exists).
2000-01-21 00:33:27 +00:00
void
bundle_ChangedPID(struct bundle *bundle)
{
#ifdef TUNSIFPID
ioctl(bundle->dev.fd, TUNSIFPID, 0);
#endif
}
Add a ``log'' command for logging specific information. Add an ``UPTIME'' variable to indicate the bundle uptime. It's now possible to put something like this in ppp.linkdown for a server setup: MYADDR: log Session closing: User USER, address HISADDR, up UPTIME Fixed some memory leakage with commands that expand words. Made some functions static. Fixed a diagnostic bug (iface add .... SIOCDIFADDR)
2001-11-03 21:45:32 +00:00
int
bundle_Uptime(struct bundle *bundle)
{
if (bundle->upat)
return time(NULL) - bundle->upat;
return 0;
}
Reference in New Issue Copy Permalink