Tomislav Čohar [Tue, 26 Aug 2014 22:25:12 +0000 (00:25 +0200)]
Configure minimum reconnect timeouts.
Enable the configuration of minimum reconnect timeout via a
configuration directive "MinTimeout". This functionality is missing in
the default tinc stable distribution. The minimum timeout is, in code,
set to 0 seconds. This patch makes it configurable.
You might ask yourself why is that needed at all ?
Well, we've been using tinc with success for quite some time in a cross DC
setup. Tinc is used to create a virtual network switch and to connect our
distributed database nodes into a virtual local network. Our database nodes
exchange information, synchronize and do failover over the created
tinc-backed network.
Every now and then, when a node has a physical networking issue and is
unreachable by some or all neighboring nodes, tinc will relay traffic over
reachable neighboring nodes and thus save our cluster. But, sometimes,
especially when BGP route changes take place, minor outages of physical
connectivity towards some nodes may cause tinc to become as reliable as
packet-loss is :).
Tinc is fast, it can and does re-establish a lost connection in a jiffy,
but it cannot detect the reason for the loss of the connection. A
re-established connection might last for a few seconds (ping timeout) to
get lost again just because the packet loss is huge at that time. Then it
reconnects again and the story repeats itself.
This process keeps repeating until the physical network stabilizes. Packet
loss on a physical link means disaster in a database replication scenario.
In such cases it is better for tinc to remain disconnected from the
unreachable/destabilized nodes for some time and relay traffic over the
reachable (unaffected) nodes then to use an unreliable route.
This patch enables us to slow down the re-connection process and eliminate
application level issues we had.
Guus Sliepen [Sun, 11 May 2014 14:35:37 +0000 (16:35 +0200)]
Drop h and hh length modifiers from printf format strings.
C already guarantees that chars and shorts get passed as int. The few uses in tinc are mainly
to print fields of struct addrinfo, and fields like ai_family have different sizes on different
platforms, which actually caused some warnings to be generated.
Guus Sliepen [Tue, 6 May 2014 19:34:26 +0000 (21:34 +0200)]
Fix issues found by Coverity.
Most of the problems found were resource leaks in error paths, some NULL
pointer dereferences that do not happen in practice, and a few other issues.
They have all been fixed now anyway.
Guus Sliepen [Tue, 6 May 2014 10:39:59 +0000 (12:39 +0200)]
Nexthop calculation should always use the shortest path.
When tinc runs the graph algorithms and updates the nexthop and via pointers,
it uses a breadth-first search, but it can sometimes revisit nodes that have
already been visited if the previous path is marked as being indirect, and
there is a longer path that is "direct". The via pointer should be updated in
this case, because this points to the closest hop to the destination that can
be reached directly. However, the nexthop pointer should not be updated.
This fixes a bug where there could potentially be a routing loop if a node in
the graph has an edge with the indirect flag set, and some other edge without
that flag, the indirect edge is part of the minimum spanning tree, and a
broadcast packet is being sent.
Check RAND_bytes() return value, fail when getting random fails.
When RAND_bytes() does not return success, the buffer contents cannot be
used. This patch makes sure the return code is checked, and the connection
fails when keys or challenges cannot be trusted.
Use constant time memcmp() when comparing packet HMACs.
This eliminates a timing side channel vulnerability, which could
potentially allow an attacker to compute a valid HMAC, and insert arbitrary
ciphertext data into the connection. If an attacker also identifies packets
with a plaintext it can guess (e.g. small packets every 1s are probably
pings), the attacker can xor the ciphertext to mangle the packet to
arbitrary plaintext. Although this type of attack is rarely seen in the
wild, it is generally considered technically viable.
Guus Sliepen [Fri, 7 Feb 2014 15:34:08 +0000 (16:34 +0100)]
Handle errors from TAP-Win32/64 adapter in a better way.
Before, the tapreader thread would just exit immediately after encountering the
first error, without notifying the main thread. Now, the tapreader thead never
exits itself, but tells the main thread to stop when more than ten errors are
encountered in a row.
Guus Sliepen [Thu, 5 Dec 2013 13:30:41 +0000 (14:30 +0100)]
If no Port is specified, set myport to actual port of first listening socket.
If the Port statement is not used, there are two other ways to let tinc listen
on a non-default port: either by specifying one or more BindToAddress
statements including port numbers, or by starting it from systemd with socket
activation. Tinc announces its own port to other nodes, but before it only
announced what was set using the Port statement.
Recent versions of autoconf and automake were giving a lot of warnings. This is
sort of a backport of similar build system changes from the 1.1 banch:
- Fix all autoconf/automake warnings.
- Merge lib/ into src/.
- Don't use symlinks for device.c any more, but use automake conditionals.
- Use explicit paths to local header files.
- Get rid of alloca.c/malloc.c/memcmp.c/realloc.c.
Windows doesn't actually support it, but MinGW provides it. However, with some versions of
MinGW it doesn't work correctly. Instead, we vsnprintf() to a local buffer and xstrdup() the
results.
Fix combination of Mode = router and DeviceType = tap on Linux.
I believe I have found a bug in tinc on Linux when it is used with
Mode = router and DeviceType = tap. This combination is useful because
it allows global broadcast packets to be used in router mode. However,
when tinc receives a packet in this situation, it needs to make sure its
destination MAC address matches the address of the TAP adapter, which is
typically not the case since the sending node doesn't know the MAC
address of the recipient. Unfortunately, this is not the case on Linux,
which breaks connectivity.
Guus Sliepen [Thu, 30 May 2013 14:43:20 +0000 (16:43 +0200)]
Better optional argument handling.
Some options can take an optional argument. However, in this case GNU getopt
requires that the optional argument is right next to the option without
whitespace inbetween. If there is whitespace, getopt will treat it as a
non-option argument, but tinc ignored those without a warning. Now tinc will
allow optional arguments with whitespace inbetween, and will give an error when
it encounters any other non-option arguments.
Drop packets forwarded via TCP if they are too big (CVE-2013-1428).
Normally all requests sent via the meta connections are checked so that they
cannot be larger than the input buffer. However, when packets are forwarded via
meta connections, they are copied into a packet buffer without checking whether
it fits into it. Since the packet buffer is allocated on the stack, this in
effect allows an authenticated remote node to cause a stack overflow.
Guus Sliepen [Thu, 31 Jan 2013 14:58:33 +0000 (15:58 +0100)]
Detect increases in PMTU.
Tinc never restarts PMTU discovery unless a node becomes unreachable. However,
it can be that the PMTU was very low during the initial discovery, but has
increased later. To detect this, tinc now tries to send an extra packet every
PingInterval, with a size slightly higher than the currently known PMTU. If
this packet is succesfully received back, we partially restart PMTU discovery
to find out the new maximum.
Guus Sliepen [Tue, 13 Nov 2012 14:05:41 +0000 (15:05 +0100)]
Send broadcast packets using a random socket, and properly support IPv6.
Before it would always use the first socket, and always send an IPv4 broadcast packet. That
works fine in a lot of situations, but it is better to try all sockets, and to send IPv6 packets
on IPv6 sockets. This is especially important for users that are on IPv6-only networks or that
have multiple physical network interfaces, although in the latter case it probably requires
them to use the ListenAddress variable to create a separate socket for each interface.
Guus Sliepen [Sat, 10 Nov 2012 22:45:22 +0000 (23:45 +0100)]
Make sure PMTU discovery works in switch mode with VLAN tags.
Before, when tinc saw a packet larger than the PMTU with a VLAN tag, it would
not know what to do with it, and would just forward it via TCP. Now, tinc
handles 802.1q packets correctly, as long as there is only one tag.
Guus Sliepen [Sat, 6 Oct 2012 19:05:02 +0000 (21:05 +0200)]
Clear connection options and status fields in free_connection_partially().
Most fields should be zero when reusing a connection. In particular, when an
outgoing connection to a node which is reachable on more than one address is
made, the second connection to that node will have status.encryptout set but
outctx will be NULL, causing a NULL pointer dereference when
EVP_EncryptUpdate() is called in send_meta() when it shouldn't.
Vilbrekin [Sat, 25 Aug 2012 17:14:00 +0000 (19:14 +0200)]
Replace hard-code with new ScriptsInterpreter configuration property.
This new setting allows choosing a custom script interpreter used for the various tinc callbacks.
If none is specified, the script itself is called as executable (as before).
This is particularly useful when storing tinc configuration and script on a mount point with no-exec attribute.
Vilbrekin [Wed, 22 Aug 2012 08:46:24 +0000 (10:46 +0200)]
Basic patch for android cross-compilation.
Commented non-existing functions in android NDK.
Prefix scripts execution with shell binary to allow execution on no-exec mount points.
Everyything is currently hard coded, while it should use pre-compiler variables...