| Age | Commit message (Collapse) | Author |
|---|
|
Add the "l2tp create tunnel" command to initiate new tunnels.
This command accepts three parameters:
* peer-addr (mandatory): IPv4 address of the remote tunnel end.
* host-addr (optional): IPv4 address of the local tunnel end.
By default the "l2tp bind" configuration option is used (if
defined).
* mode (optional): tunnel operating mode: lac (default) or lns.
Example with the telnet interface:
accel-ppp# l2tp create tunnel peer-addr 192.0.2.1 host-addr 192.0.2.2
The above command tells accel-ppp to establish a new tunnel with the
peer at 192.0.2.1. The local address will be 192.0.2.2 and the tunnel
will operate as lac.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
When peer's port isn't known (i.e. right after a tunnel establishment
request has been sent), use the source port of the first peer's
response as the destination port of any future L2TP message sent over
that tunnel.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Send L2TP messages to the host set in the addr field of l2tp_packet_t,
instead of relying on the file descriptor's connection state. This
will allow the file descriptor to be disconnected (allowing it to
wait for a message from an arbitrary source port) while still being
able to retransmit previous L2TP messages.
This possibility is necessary for initiating tunnel establishment: the
initial SCCRQ is sent to the peer on port 1701, but the peer may reply
with an arbitrary source port. The socket thus needs to be disconnected
to be able to receive the peer's SCCRP. But in the mean time, the
initial SCCRQ may need to be retransmitted (if no answer is received
from the peer after a given amount of time). Hence the need for sending
L2TP messages over unconnected sockets.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Rename field lns_addr, as it identifies the host address, whether it be
a LAC or a LNS.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Rename field lac_addr, as it identifies the peer address, whether it be
a LAC or a LNS.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Define the lns_mode flag for tunnels and sessions to keep track
of their operating state. A session's mode is inherited from its
parent tunnel.
The session's mode is used to set the PPPOL2TP_SO_LNSMODE socket
option to its PPP socket.
For tunnels, the mode is used to define how new sessions may be
created:
-In LNS mode, new sessions may be created by accepting incomming
calls (ICRQ) or by placing outgoing calls (OCRQ).
-In LAC mode, tunnels will create sessions by accepting outgoing
calls or by placing incomming calls.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Define STATE_WAIT_SCCRP which will be required for creating new tunnels
on behalf of the host. This state will be entered right after sending
an SCCRQ message.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Change the meaning of the Nr field in struct l2tp_conn_t: instead of
setting it to the last received message sequence, set it with the
value of the next expected message sequence number (i.e. one above
the previous value). This simplifies processing and avoids the
l2tp_send() function to increment the L2TP packet Nr field (it's now
directly set to the correct value).
Having Nr correctly set before calling l2tp_send() is necessary for
implementing tunnel establishment requests (SCCRQ messages) since the
Nr field of this message must be 0.
This patch also replaces the formula used to check if a message Ns is
lower than a tunnel Nr.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
When allocating tunnels, set the lns_addr field of tunnels using
getsockname() instead of using the "host" parameter. This will allow
the caller to pass an unspecified port in "host" (for binding to
any source port) and to set the lns_addr field with the address really
bound to.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Remove the "serv" parameter from l2tp_tunnel_alloc() prototype.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Send the Challenge AVP in SCCRP messages and check for the Challenge
Response AVP in SCCCN.
This patch adds two functions:
-l2tp_tunnel_genchall() generates the Challenge, stores it in the
tunnel structure and adds the AVP in the packet to be sent.
-l2tp_tunnel_checkchallresp() reads a Challenge Response, computes
the expected response using the Challenge stored in the tunnel
structure and finally checks if both responses match.
These two functions are meant to be used in the upcoming implementation
of SCCRQ message sending and SCCRP message reception.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Create the l2tp_tunnel_genchallresp() function for building and
adding the Challenge Response AVP to messages that need it.
Support for this AVP is needed for sending SCCRP and SCCCN messages.
This patch replaces the previous Challenge Response computation by
calling the new function and prepares implementation of SCCCN message
sending.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Move challenge computation out of tunnel allocation by using a
generic CHAP-MD5 calculation function. This will allow to initiate
tunnels on the host side (i.e. sending SCCRQ) and to compute the
challenge response later on, when handling the peer response (i.e.
during SCCRP handling).
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Use a "struct sockaddr_in" pointer for passing the peer address to the
tunnel allocation function (instead of a "struct l2tp_packet_t *"). This
makes l2tp_tunnel_alloc() usable in situations where tunnel creation
doesn't happen in response to a peer request.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
The "recv <SCCRQ packet>" message printing should logically belong to
the SCCRQ message reception function. This patch moves this message
from l2tp_tunnel_alloc() to l2tp_recv_SCCRQ() in order to make
l2tp_tunnel_alloc() usable in situations different than SCCRQ
message handling.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Pass the host address as a directly usable sockaddr_in in
l2tp_tunnel_alloc(), instead of an in_pktinfo. This will allow to
allocate tunnels without having to artificially create in_pktinfo
structures in situations where the tunnel is not created in response
to an SCCRQ.
As a side effect, this patch makes accel-ppp reply to SCCRQ using
an arbitrary source port, as specified in RFC 2661.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Directly pass the framing capabilities argument of l2tp_tunnel_alloc()
as an integer. The previous type (struct l2tp_attr_t *) won't
make make sense when creating a tunnel in situations different from
SCCRQ message handling.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Peer TID assignment is specific to SCCRQ message processing. When
creating a tunnel in different conditions (e.g. upon user request),
the peer TID will be unknown.
This patch moves peer TID assignment from l2tp_tunnel_alloc() to
l2tp_recv_SCCRQ().
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Register and start triton contexts/handlers in l2tp_tunnel_start()
instead of l2tp_tunnel_alloc(). The objective is to make
l2tp_tunnel_alloc() more generic by removing some specific
initialisation code. At a later stage, these would be done by the
caller between the calls to l2tp_tunnel_alloc() and l2tp_tunnel_start().
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Avoid unregistering uninitialised tunnel data when freeing tunnel
structures. This makes l2tp_tunnel_free() suitable for freeing
partially initialised tunnels.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Update the prototype of l2tp_tunnel_alloc() so that it returns the
allocated tunnel structure. This will allow the caller to act on the
newly created tunnel in different ways depending on the creation
context.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Verify return values of functions called by l2tp_tunnel_alloc().
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Don't send the EV_CTRL_FINISHED event when closing sessions which
haven't been established, because no corresponding EV_CTRL_STARTED
events have been sent by such sessions.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Call regexp command handler without command line's leading spaces.
This makes leading spaces handling consistent between the exec and help
command handlers (leading spaces are already skipped before calling
help handlers).
This also gives the possibility to register an anchored regexp command
without having to manage the leading spaces in that regexp.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Match PPPoE "exec" and "help" commands. This makes the PPPoE module
behave nicely with the new help command handling.
For example:
* "help pppoe set" won't print help messages for the "pppoe show"
set of commands
* "help pppoe mac-filter add" will only print help message for this
specific command (i.e. without the other "pppoe mac-filter" ones)
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
When arguments are supplied to the "help" command, only display help
messages corresponding to the command these arguments refer to (e.g.
"help pppoe" will only show help messages for commands starting
with "pppoe").
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Pass a full string, instead of an array of words, to the "help" callback
of regexp command handlers. Also register these command handlers using
a regexp to apply on "help" commands, just like for exec commands. The
"help" callback will then be called only if the command matches the
"help" regexp.
As a side effect, the "help" word and its following spaces are skipped
before calling the "help" callback for both simple and regexp commands.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
* skip_word(): ensure the for() loop won't pass over a '\0' character
* split(): ensure the for() loop won't try to fill buf[MAX_CMD_ITEMS]
* split(): skip leading spaces.
* split(): fix exit value if no item is found (return 0 instead of 1)
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Split regexp and simple command handling from the generic command
processing function and handle regexp commands before simple ones.
This avoids splitting the original command line before processing
regexp command handlers, so that these handlers get the full command
line instead of only the first word.
The regexp command handlers are now only called when the command
matches the regexp (regexp handlers were called unconditionaly
before this patch). The rest of the processing properties should
remain unchanged.
Command processing summary:
-If the command starts with the "help" keyword, then all help
commands are called, starting with regexp help commands.
-Otherwise, regexp command handlers whose regexp matches the
command line are called. Then, simple command handlers whose
header matches the command line are called.
Any command handler that returns a value different from
CLI_CMD_OK stops the process, thus preventing the following
handlers from being called.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Process "help" commands in a separated function with no side effect
when the command doesn't match. This avoids splitting the input
command line if not necessary (regexp command handlers need to
receive the entire command line for being useful).
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
* Call pcre_compile2() with all mandatory arguments and
improve error message.
* Register regexp commands in regexp_cmd_list instead of
simple_cmd_list.
* Fix "help" function call.
* Interpret regexp command return codes (CLI_CMD_*) in the
same way as for simple commands.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Since it's not possible to mix incomming and outgoing states in
a session, the "state2" field isn't necessary.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Add the "l2tp create session" command to place an outgoing call.
For example, "l2tp create session tid 3" will send an OCRQ within
tunnel 3.
This allows the LNS to request establishment of a new session inside
an existing tunnel.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
This patch removes the automatic sending of a CDN message when
cancelling creation of new sessions. This is necessary in order
to give the possibility for a tunnel to create (and possibly cancel)
new sessions on its own (e.g. for placing outgoing calls). In such
cases, sending a CDN message would not make sense if the session gets
cancelled before any creation message was sent.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Define the u_readlong() function as a wrapper for strtol(). It performs
stricter checks on the input value and let the caller specify an interval
of acceptable values.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
Move definition of ACCEL_PPP_VERSION from /accel-pppd/CMakeLists.txt
to /CMakeLists.txt so that it gets visible by any cmake file.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
|
|
|
|
ipoe: fixed DHCP Discover handling when session is in active state
|
|
|
|
|
|
|
|
|
|
This reverts commit cde7ee4d629b32c810ceac836b70a4289b4f00b3.
|
|
|
|
|
|
|
|
This reverts commit 2b9c6812f4558b74843c6e3740ee5319dc1df48d.
|
|
|
|
|
|
|
