mitmproxy/pathod/templates/docs_pathoc.html
2016-02-18 13:03:40 +01:00

212 lines
7.4 KiB
HTML

{% extends "docframe.html" %} {% block body %}
<div class="page-header">
<h1>
pathoc
<small>A perverse HTTP client.</small>
</h1>
</div>
<p>
Pathoc is a perverse HTTP daemon designed to let you craft almost any conceivable
HTTP request, including ones that creatively violate the standards. HTTP requests
are specified using a
<a href="/docs/language">small, terse language</a>, which pathod shares with
its server-side twin <a href="/docs/pathod">pathod</a>. To view pathoc's complete
range of options, use the command-line help:
</p>
<pre class="terminal">pathoc --help</pre>
<section>
<div class="page-header">
<h1>Getting Started</h1>
</div>
<p>The basic pattern for pathoc commands is as follows: </p>
<pre class="terminal">pathoc hostname request [request ...]</pre>
<p>
That is, we specify the hostname to connect to, followed by one or more requests.
Lets start with a simple example:
</p>
<pre class="terminal">
&gt; pathoc google.com get:/ &lt;&lt; 301 Moved Permanently: 219 bytes
</pre>
<p>
Here, we make a GET request to the path / on port 80 of google.com. Pathoc's output
tells us that the server responded with a 301. We can tell pathoc to connect
using SSL, in which case the default port is changed to 443 (you can over-ride
the default port with the <b>-p</b> command-line option):
</p>
<pre class="terminal">
&gt; pathoc -s google.com get:/ &lt;&lt; 301 Moved Permanently: 219 bytes
</pre>
</section>
<section>
<div class="page-header">
<h1>Multiple Requests</h1>
</div>
<p>
There are two ways to tell pathoc to issue multiple requests. The first is to specify
them on the command-line, like so:
</p>
<pre class="terminal">
&gt; pathoc google.com get:/ get:/ &lt;&lt; 301 Moved Permanently: 219 bytes &lt;&lt;
301 Moved Permanently: 219 bytes
</pre>
<p>
In this case, pathoc issues the specified requests over the same TCP connection -
so in the above example only one connection is made to google.com
</p>
<p>The other way to issue multiple requets is to use the <b>-n</b> flag:</p>
<pre class="terminal">
&gt; pathoc -n 2 google.com get:/ &lt;&lt; 301 Moved Permanently: 219 bytes &lt;&lt; 301
Moved Permanently: 219 bytes
</pre>
<p>
The output is identical, but two separate TCP connections are made to the upstream
server. These two specification styles can be combined:
</p>
<pre class="terminal">
&gt; pathoc -n 2 google.com get:/ get:/ &lt;&lt; 301 Moved Permanently: 219 bytes &lt;&lt;
301 Moved Permanently: 219 bytes &lt;&lt; 301 Moved Permanently: 219 bytes &lt;&lt;
301 Moved Permanently: 219 bytes
</pre>
<p>Here, two distinct TCP connections are made, with two requests issued over each.</p>
</section>
<section>
<div class="page-header">
<h1>Basic Fuzzing</h1>
</div>
<p>
The combination of pathoc's powerful request specification language and a few of
its command-line options makes for quite a powerful basic fuzzer. Here's
an example:
</p>
<pre class="terminal">
&gt; pathoc -e -I 200 -t 2 -n 1000 localhost get:/:b@10:ir,@1
</pre>
<p>
The request specified here is a valid GET with a body consisting of 10 random bytes,
but with 1 random byte inserted in a random place. This could be in the headers,
in the initial request line, or in the body itself. There are a few things
to note here:
</p>
<ul>
<li>
Corrupting the request in this way will often make the server enter a state where
it's awaiting more input from the client. This is where the
<b>-t</b> option comes in, which sets a timeout that causes pathoc to
disconnect after two seconds.
</li>
<li>
The <b>-n</b> option tells pathoc to repeat the request 1000 times.
</li>
<li>
The <b>-I</b> option tells pathoc to ignore HTTP 200 response codes.
You can use this to fine-tune what pathoc considers to be an exceptional
condition, and therefore log-worthy.
</li>
<li>
The <b>-e</b> option tells pathoc to print an explanation of each logged
request, in the form of an expanded pathoc specification with all random
portions and automatic header additions resolved. This lets you precisely
replay a request that triggered an error.
</li>
</ul>
</section>
<section>
<div class="page-header">
<h1>Interacting with Proxies</h1>
</div>
<p>
Pathoc has a reasonably sophisticated suite of features for interacting with proxies.
The proxy request syntax very closely mirrors that of straight HTTP, which
means that it is possible to make proxy-style requests using pathoc without
any additional syntax, by simply specifying a full URL instead of a simple
path:
</p>
<pre class="terminal">&gt; pathoc -p 8080 localhost "get:'http://google.com'"</pre>
<p>
Another common use case is to use an HTTP CONNECT request to probe remote servers
via a proxy. This is done with the <b>-c</b> command-line option,
which allows you to specify a remote host and port pair:
</p>
<pre class="terminal">&gt; pathoc -c google.com:80 -p 8080 localhost get:/</pre>
<p>
Note that pathoc does <b>not</b> negotiate SSL without being explictly instructed
to do so. If you're making a CONNECT request to an SSL-protected resource,
you must also pass the <b>-s</b> flag:
</p>
<pre class="terminal">&gt; pathoc -sc google.com:443 -p 8080 localhost get:/</pre>
</section>
<section>
<div class="page-header">
<h1>Embedded response specification</h1>
</div>
<p>
One interesting feature of the Request sppecification language is that you can embed
a response specifcation in it, which is then added to the request path. Here's
an example:
</p>
<pre class="terminal">&gt; pathoc localhost:9999 "get:/p/:s'401:ir,@1'"</pre>
<p>
This crafts a request that connects to the pathod server, and which then crafts a
response that generates a 401, with one random byte embedded at a random
point. The response specification is parsed and expanded by pathoc, so you
see syntax errors immediately. This really becomes handy when combined with
the <b>-e</b> flag to show the expanded request:
</p>
<pre class="terminal">
&gt; > pathoc -e localhost:9999 "get:/p/:s'401:ir,@1'" >> Spec: get:/p/:s'401:i15,\'o\':h\'Content-Length\'=\'0\'':h'Content-Length'='0'
<< 401 Unoauthorized: 0 bytes </pre>
<p>
Note that the embedded response has been resolved <i>before</i> being sent
to the server, so that "ir,@1" (embed a random byte at a random location)
has become "i15,\'o\'" (embed the character "o" at offset 15). You now have
a pathoc request specification that is precisely reproducable, even with
random components. This feature comes in terribly handy when testing a proxy,
since you can now drive the server repsonse completely from the client, and
have a complete log of reproducible requests to analyse afterwards.
</p>
</section>
{% endblock %}