aboutsummaryrefslogtreecommitdiffstats
path: root/Source/charon/doc/Architecture.txt
blob: 3e8855fa9ee2062676931b2a46eecf5cfa95dabb (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
 strongSwans overall design
============================

IKEv1 and IKEv2 is handled in different keying daemons. The ole IKEv1 stuff is
completely handled in pluto, as it was all the times. IKEv2 is handled in the
new keying daemon, which is called charon. 
Daemon control is done over unix sockets. Pluto uses whack, as it did for years.
Charon uses another socket interface, called stroke. Stroke uses another
format as whack and therefore is not compatible to whack. The starter utility,
wich does fast configuration parsing, speaks both the protocols, whack and
stroke. It also handles daemon startup and termination. 
Pluto uses starter for some commands, for other it uses the whack utility. To be
as close to pluto as possible, charon has the same split up of commands to
starter and stroke. All commands are wrapped together in the ipsec script, which
allows transparent control of both daemons.

         +-----------------------------------------+
         |                  ipsec                  |
         +-----+--------------+---------------+----+
               |              |               |
               |              |               |
               |        +-----+-----+         |
         +-----+----+   |           |   +-----+----+
         |          |   |  starter  |   |          |
         |  stroke  |   |           |   |   whack  |
         |          |   +---+--+----+   |          |
         +------+---+       |  |        +--+-------+
                |           |  |           |
            +---+------+    |  |    +------+--+
            |          |    |  |    |         |
            |  charon  +----+  +----+  pluto  |
            |          |            |         |
            +-----+----+            +----+----+
                  |                      |
            +-----+----+                 |
            |    LSF   |                 |
            +-----+----+                 |
                  |                      |
            +-----+----+            +----+----+
            | RAW Sock |            | UDP/500 |
            +----------+            +---------+

Since IKEv2 uses the same port as IKEv1, both daemons must listen to UDP port
500. Under Linux, there is no clean way to set up two sockets at the same port.
To reslove this problem, charon uses a RAW socket, as they are used in network
sniffers. An installed Linux Socket Filter (LSF) filters out all none-IKEv2
traffic. Pluto receives any IKE message, independant of charons behavior.
Therefore plutos behavior is changed to discard any IKEv2 traffic silently.


 IKEv2 keying daemon: charon
=============================

 Threading modell
------------------

All IKEv2 stuff is handled in charon. It uses a newer and more flexible
architecture than pluto. Charon uses a thread-pool, which allows parallel
execution SA-management. Beside the thread-pool, there are some special purpose
threads which do their job for the common health of the daemon.

                       +------+
                       | E  Q |
                       | v  u |---+                   +------+  +------+
                       | e  e |   |                   |      |  | IKE- |
                       | n  u |  +-----------+        |      |--| SA   |
                       | t  e |  |           |        | I  M |  +------+
     +------------+    | -    |  | Scheduler |        | K  a |
     |  receiver  |    +------+  |           |        | E  n |  +------+
     +----+-------+              +-----------+        | -  a |  | IKE- |
          |      |     +------+   |                   | S  g |--| SA   |
  +-------+--+   +-----| J  Q |---+  +------------+   | A  e |  +------+
 -|  socket  |         | o  u |      |            |   | -  r |
  +-------+--+         | b  e |      |   Thread-  |   |      |
          |            | -  u |      |   Pool     |   |      |
     +----+-------+    |    e |------|            |---|      |
     |   sender   |    +------+      +------------+   +------+
     +----+-------+
          |            +------+
          |            | S  Q |
          |            | e  u |
          |            | n  e |
          +------------| d  u |
                       | -  e |
                       +--+---+

The thread-pool is the heart of the architecture. It processes jobs from a
(fully synchronized) job-queue. Mostly, a job is associated with a specific
IKE SA. These IKE SAs are synchronized, only one thread can work one an IKE SA.
This makes it unnecesary to use further synchronisation methods once a IKE SA
is checked out. The (rather complex) synchronization of IKE SAs is completely
done in the IKE SA manager.
The sceduler is responsible for event firing. It waits until a event in the
(fully synchronized) event-queue is ready for processing and pushes the event
down to the job-queue. A thread form the pool will pick it up as quick as
possible. Every thread can queue events or jobs. Furter, an event can place a
packet in the send-queue. The sender thread waits for those packets and sends
them over the wire, via the socket. The receiver does exactly the opposite of
the sender. It waits on the socket, reads in packets an places them on the
job-queue for further processing by a thread from the pool.
There are even more threads, not drawn in the upper scheme. The stroke thread
is responsible for reading and processessing commands from another process. The
kernel interface thread handles communication from and to the kernel via a
netlink socket. It waits for kernel events and processes them appropriately.


 configuration backends
------------------------

The configuration architecture for charon is complex, but is flexible and
extensible. All configuration stuff is split up in multiple parts:

connection      Defines a connection between two hosts. Proposals define with
                wich algorithms a IKE SA should be set up.
policy          Defines the rules to apply ontop of a connection. A policy is
                defined between two IDs. Proposals and traffic selectors allow
                fine grained configuration of the CHILD SAs (AH and ESP) to set
                up.
credential      A credential is something used for authentication, such as a
                preshared key, a RSA private or public key, certificate, ...
configuration   The configuration itself handles daemon related configuration
                stuff, such as interface binding or logging settings.

These configuration types are defined as interfaces, and are currently
implemented only in the stroke class. Through the modular design, parts could be
replaced with more powerful backends, such as a RADIUS server for the
credentials, a SQL database for the connections, policy definitions on an LDAP
server, and so on...