Files
kcgi/index.xml
T

375 lines
17 KiB
XML

<!DOCTYPE html>
<html>
<head>
<meta name="viewport" content="initial-scale=1.0, user-scalable=no" />
<meta charset='utf-8' />
<title>kcgi | minimal CGI library for C</title>
<script src="https://google-code-prettify.googlecode.com/svn/loader/run_prettify.js?skin=sons-of-obsidian"></script>
<link href="https://maxcdn.bootstrapcdn.com/font-awesome/4.2.0/css/font-awesome.min.css" rel="stylesheet" />
<link rel="stylesheet" href="index.css" type="text/css" />
<script>
function showBits(name)
{
var show, nshow;
show = document.getElementById(name);
noshow = document.getElementById('no' + name);
console.log('maxheight = ' + show.style.maxHeight);
if (show.style.maxHeight == '' || show.style.maxHeight == '0px') {
show.style.maxHeight = '500em';
noshow.style.maxHeight = 0;
} else {
show.style.maxHeight = 0;
noshow.style.maxHeight = '500em';
}
}
</script>
</head>
<body itemscope="itemscope" itemtype="http://schema.org/SoftwareApplication">
<header>
<section id="header">
<h1>
<span itemprop="name" class="nm">kcgi</span>
&#8211;
<span itemprop="description" class="nd">minimal CGI library in C</span>
</h1>
<nav>
<a itemprop="downloadURL" href="snapshots/kcgi.tgz"><i class="fa fa-fw fa-gears"></i></a>
<a href="snapshots/kcgi.tgz.sha512"><i class="fa fa-fw fa-lock"></i></a>
<a href="snapshots"><i class="fa fa-fw fa-archive"></i></a>
</nav>
</section>
<section id="version">
<nav data-sblg-nav="1" data-sblg-navtag="version" data-sblg-navsz="1" data-sblg-navcontent="1" id="nolongnews">
<div>Version <span class="version">${sblg-title}</span> (<time datetime="${sblg-date}">${sblg-date}</time>)</div>
<div>${sblg-aside}</div>
</nav>
<nav data-sblg-nav="1" data-sblg-navtag="version" data-sblg-navcontent="1" id="longnews">
<div>Version <span class="version">${sblg-title}</span> (<time datetime="${sblg-date}">${sblg-date}</time>)</div>
<div>${sblg-aside}</div>
</nav>
<a href="#version" onclick="showBits('longnews');">Toggle All News</a>
</section>
</header>
<article>
<section id="intro">
<p>
<span class="nm">kcgi</span> is an <a href="http://opensource.org/licenses/ISC" rel="license">open
source</a> CGI and FastCGI library for C web applications.
It is minimal, secure, and auditable.
To start, <a href="#install">install</a> the library then read the <a href="#usage">usage guide</a>.
Contact <a href="http://kristaps.bsd.lv">Kristaps</a> with questions or comments.
<span class="nm">kcgi</span> is a <a href="http://www.bsd.lv">BSD.lv</a> project.
</p>
</section>
<figure id="sample" class="sample">
<figcaption>
The following simple example implements a server that just echoes <q>Hello, World!</q> as an HTTP
response to a CGI request.
Click on any italicitised fields to link to the documentation.
</figcaption>
<pre id="nocomments" class="prettyprint linenums">#include &lt;stdint.h&gt;
#include &lt;stdlib.h&gt;
#include &lt;<a href="kcgi.3.html">kcgi.h</a>&gt;
int main(void) {
struct kreq r;
const char *page = "index";
if (KCGI_OK != <a href="khttp_parse.3.html">khttp_parse</a>(&amp;r, NULL, 0, &amp;page, 1, 0))
return(EXIT_FAILURE);
<a href="khttp_head.3.html">khttp_head</a>(&amp;r, kresps[KRESP_STATUS], "%s", khttps[KHTTP_200]);
khttp_head(&amp;r, kresps[KRESP_CONTENT_TYPE], "%s", kmimetypes[r.mime]);
<a href="khttp_body.3.html">khttp_body</a>(&amp;r);
<a href="khttp_write.3.html">khttp_puts</a>(&amp;r, &quot;Hello, world!&quot;);
<a href="khttp_free.3.html">khttp_free</a>(&amp;r);
return(EXIT_SUCCESS);
}</pre>
<pre id="comments" class="prettyprint linenums">#include &lt;stdint.h&gt;
#include &lt;stdlib.h&gt;
#include &lt;<a href="kcgi.3.html">kcgi.h</a>&gt;
int main(void) {
struct kreq r;
const char *page = "index";
/*
* Parse the HTTP environment.
* We only know a single page, "index", which is also
* the default page if none is supplied.
* (We don't validate any input fields.)
*/
if (KCGI_OK != <a href="khttp_parse.3.html">khttp_parse</a>(&amp;r, NULL, 0, &amp;page, 1, 0))
return(EXIT_FAILURE);
/*
* Ordinarily, here I'd switch on the method (OPTIONS, etc.,
* defined in the <q>method</q> variable) then switch on which
* page was requested (<q>page</q> variable).
* But for the same of example, just output a response.
*/
/* Emit the HTTP status 200 header: everything's ok. */
<a href="khttp_head.3.html">khttp_head</a>(&amp;r, kresps[KRESP_STATUS], "%s", khttps[KHTTP_200]);
/* Echo our content-type, defaulting to HTML if none was specified. */
khttp_head(&amp;r, kresps[KRESP_CONTENT_TYPE], "%s", kmimetypes[r.mime]);
/* No more HTTP headers: start the HTTP document body. */
<a href="khttp_body.3.html">khttp_body</a>(&amp;r);
/*
* We can put any content below here: JSON, HTML, etc.
* Usually we'd switch on our MIME type.
* However, we're just going to put the literal string as noted...
*/
<a href="khttp_write.3.html">khttp_puts</a>(&amp;r, &quot;Hello, world!&quot;);
/* Flush the document and free resources. */
<a href="khttp_free.3.html">khttp_free</a>(&amp;r);
return(EXIT_SUCCESS);
}</pre>
<nav>
<a href="#sample" onclick="showBits('comments');"><i class="fa fa-comments"></i> Toggle comments</a>
</nav>
</figure>
<section id="install">
<h2>
Installation
</h2>
<p>
First, check if <span class="nm">kcgi</span> isn't already a third-part port for your system, such as
for <a href="http://cvsweb.openbsd.org/cgi-bin/cvsweb/ports/www/kcgi/">OpenBSD</a> or <a
href="https://svnweb.freebsd.org/ports/head/www/kcgi/">FreeBSD</a>.
If so, install using that system.
</p>
<p>
If not, you'll need a modern UNIX system.
To date, <span class="nm">kcgi</span> has been built and run on GNU/<a
href="http://www.linux.org">Linux</a> machines, BSD (<a
href="http://www.openbsd.org">OpenBSD</a>, <a href="http://www.freebsd.org">FreeBSD</a>), and <a
href="http://www.apple.com/osx">Mac OSX</a> (Snow Leopard, Lion) on i386 and AMD64.
It has been deployed under <a href="https://httpd.apache.org/">Apache</a>, <a
href="http://nginx.org">nginx</a>, and OpenBSD's <a
href="http://www.openbsd.org/cgi-bin/man.cgi/OpenBSD-current/man8/httpd.8?query=httpd">httpd(8)</a>
(the latter two natively over FastCGI and via the <code>slowcgi</code> wrapper).
Begin by downloading <a href="snapshots/kcgi.tgz">kcgi.tgz</a> and verify the archive with <a
href="snapshots/kcgi.tgz.sha512">kcgi.tgz.sha512</a>.
Once downloaded, compile the software with <code>make</code>, which will automatically run a
configuration script to conditionally deploy portability glue.
Finally, install the software using <code>make install</code>, optionally specifying the
<code>PREFIX</code> if you don't intend to use <span class="file">/usr/local</span>.
</p>
<p>
If <span class="nm">kcgi</span> doesn't compile, please send me the <span class="file">config.log</span>
file and the output of the failed compilation.
If you're running on an operating system with an unsupported sandbox, let me know and we can work
together to fit it into the configuration and portability layer.
Lastly, I'd love to compile with <a href="http://www.mingw.org/">mingw</a> for Microsoft machines:
please contact me if you can do the small amount of work (I think?) to port the <code>poll(2)</code> and
other non-Microsoft functions.
</p>
</section>
<section id="usage">
<h2>
Usage
</h2>
<p>
The <span class="nm">kcgi</span> manpages, starting with <a href="kcgi.3.html">kcgi(3)</a>, are the
canonical source of documentation.
The following are introductory materials to the system.
</p>
<nav data-sblg-nav="1" data-sblg-navtag="tutorial" data-sblg-navcontent="1">
<h3><a href="${sblg-base}.html">${sblg-titletext}</a></h3>
${sblg-aside}
</nav>
</section>
<section id="features">
<h2>
Deploying
</h2>
<p>
Applications using <span class="nm">kcgi</span> behave just like any other application.
To compile <span class="nm">kcgi</span> applications, just include the <span class="nm">kcgi.h</span>
header file and make sure it appears in the compiler inclusion path.
(According to C99, you'll need to include <span class="nm">stdint.h</span> before it for the
<code>int64_t</code> type used for parsing integers.)
Linking is similarly normative: link to <span class="file">libkcgi</span> and, if your system has
compression support, <span class="file">libz</span>.
</p>
<p>
Well-deployed web servers, such as the default <a href="http://www.openbsd.org">OpenBSD</a> server, by
default are deployed within a <a
href="http://www.openbsd.org/cgi-bin/man.cgi/OpenBSD-current/man2/chroot.2">chroot(2)</a>. If
this is the case, you'll need to statically link your binary.
If running within a <a
href="http://www.openbsd.org/cgi-bin/man.cgi/OpenBSD-current/man2/chroot.2">chroot(2)</a> and on
OpenBSD, be aware that the sandbox method requires <span class="file">/dev/systrace</span> within the
server root.
By default, this file does not exist in the web server root.
Moreover, the default web server root mount-point, <span class="file">/var</span>, is mounted
<code>nodev</code>.
This complication does not exist for the other sandboxes.
</p>
<p>
FastCGI applications may either be started directly by the web server (which is popular with <a
href="https://httpd.apache.org/">Apache</a>) or <q>externally</q> given a socket and <a
href="kfcgi.8.html">kfcgi(8)</a> (this method is normative for OpenBSD's <a
href="http://www.openbsd.org/cgi-bin/man.cgi/OpenBSD-current/man8/httpd.8">httpd(8)</a> and
suggested for the security precautions taken by the wrapper).
</p>
</section>
<section id="features">
<h2>
Implementation Details
</h2>
<p>
The bulk of <span class="nm">kcgi</span>'s CGI handling lies in <a
href="khttp_parse.3.html">khttp_parse(3)</a>, which fully parses the HTTP request.
Application developers must invoke this function before all others.
For FastCGI, it's split between <a href="khttp_fcgi_init.3.html">khttp_fcgi_init(3)</a>, which
initialises context; and <a href="khttp_fcgi_parse.3.html">khttp_fcgi_parse(3)</a>, which receives new
parsed requests.
In either case, requests must be freed by an <a href="khttp_free.3.html">khttp_free(3)</a>.
</p>
<p>
All functions isolate the parsing and validation of untrusted network data within a <i>sandboxed</i>
child process.
Sandboxes limit the environment available to a process, so exploitable errors in the parsing process (or
validation with third-party libraries) cannot touch the system environment.
This parsed data is returned to the parent process over a socket.
In the following, the <i>HTTP parser</i> and <i>input validator</i> manage a single HTTP request, while
<i>connection delegator</i> accepts new HTTP requests and passes them along.
</p>
<figure>
<img src="figure1.svg" alt="Implementation Details" />
<img src="figure4.svg" alt="Implementation Details" />
</figure>
<p>
This method of sandboxing the untrusted parsing process follows <a
href="http://www.openssh.org">OpenSSH</a>, and requires special handling for each operating
system:
</p>
<dl>
<dt>
<a href="http://man7.org/linux/man-pages/man2/seccomp.2.html">seccomp(2)</a>
(Linux)
</dt>
<dd>
This requires a fairly new kernel (&#x2265;Linux 3.5).
It is supplemented by <code>setrlimit(2)</code> limiting.
For the time being, this feature is only available for x86, x86_64, and arm architectures.
If you're using another one, please send me your <code>uname -m</code> and, if you know if it,
the correct <code>AUDIT_ARCH_xxx</code> found in <code>/usr/include/linux/audit.h</code>.
</dd>
<dt>
<a href="http://www.openbsd.org/cgi-bin/man.cgi/OpenBSD-current/man4/systrace.4">systrace(4)</a>
(<a href="http://www.openbsd.org">OpenBSD</a>)
</dt>
<dd>
This requires the existence of <i>/dev/systrace</i> if running in a
<code>chroot(2)</code>, which is strongly suggested.
If you're using a stock <a href="http://www.openbsd.org">OpenBSD</a>, make sure that the
mount-point of <i>/dev/systrace</i> isn't mounted <code>nodev</code>!
</dd>
<dt>
<a href="https://developer.apple.com/library/mac/documentation/Darwin/Reference/ManPages/man3/sandbox_init.3.html">sandbox_init(3)</a>
(<a href="https://www.apple.com/osx/">Apple OSX</a>)
</dt>
<dd>
This uses the sandboxing profile for <q>pure computation</q> as provided in Mac OS X Leopard and
later.
This is supplemented by resource limiting via <code>setrlimit(2)</code>.
</dd>
<dt>
<a href="https://www.freebsd.org/cgi/man.cgi?query=capsicum&amp;sektion=4">capsicum(4)</a>
(<a href="https://www.freebsd.org">FreeBSD</a>)
</dt>
<dd>
Uses the capabilities facility on <a href="http://www.freebsd.org">FreeBSD</a> 10 and later.
This is supplemented by resource limiting with <code>setrlimit(2)</code>.
</dd>
</dl>
<p>
Since validation occurs within the sandbox, special care must be taken that validation routines don't
access the environment (e.g., by opening files, network connections, etc.), as the child
<strong>might</strong> be abruptly killed by the sandbox facility.
(Not all sandboxes do this.)
If required, this kind of validation can take place after the parse validation sequence.
</p>
<p>
The connection delegator is similar, but has different sandboxing rules, as it must manage an open
socket connection and respond to new requests.
</p>
</section>
<section>
<h2 id="testing">
Testing
</h2>
<p>
<span class="nm">kcgi</span> is shipped with a fully automated testing framework executed with
<code>make regress</code>.
Interfacing systems can also make use of this by working with the <a
href="kcgiregress.3.html">kcgiregress(3)</a> library.
This framework acts as a mini-webserver, listening on a local port, translating an HTTP document into a
minimal CGI request, and passing the request to a <span class="nm">kcgi</span> CGI client.
For internal tests, test requests are constructed with <a
href="http://curl.haxx.se/libcurl/">libcurl</a>.
</p>
<p>
The automated test framework, at the moment, only has a few tests for basic functionality and
sandboxing.
The binding local port is fixed, too; so if you plan on running the regression suite, you may need to
tweak its access port.
</p>
<p>
Another testing framework exists for use with the <a href="http://lcamtuf.coredump.cx/afl/">American
fuzzy lop</a>.
To use this, you'll need to compile the <code>make afl</code> target with your compiler of choice, e.g.,
<code>make clean</code>, then <code>make afl CC=afl-gcc</code>.
Then run the <code>afl-fuzz</code> tool on the <code>afl-multipart</code>, <code>afl-plain</code>, and
<code>afl-urlencoded</code> binaries using the test cases (and dictionaries, for the first) provided.
</p>
</section>
<section>
<h2>
Performance
</h2>
<p>
Security comes at a price&#8212;but not a stiff price.
By design, <span class="nm">kcgi</span> incurs overhead in three ways: first, spawning a child to
process the untrusted network data; second, enacting the sandbox framework; and third, passing parsed
pairs back to the parent context.
In the case of running CGI scripts, <span class="nm">kcgi</span> performance is bound to the operating
system's ability to spawn and reap processes.
For FastCGI, the bottleneck becomes the transfer of data.
In the following graph, I graph the responsiveness of <span class="nm">kcgi</span> against the baseline
web-server performance.
</p>
<figure>
<img src="figure2.png" />
</figure>
<p>
This shows the empirical cumulative distribution of a statisically-significant number of page requests
as measured by <a href="https://httpd.apache.org/docs/2.2/programs/ab.html">ab(1)</a> with 10 concurrent
requests.
The <strong>CGI</strong> line is the CGI sample included in the source;
the <strong>FastCGI</strong> line is the FastCGI sample;
the <strong>CGI (simple)</strong> simply emits a 200 HTTP status and <q>Hello, World</q>; and
the <strong>static</strong> is a small static file on the web server.
The operating system is Mac OS X 10.7.5 <q>Air</q> laptop (1.86 GHz Intel Core 2 Duo, 2 GB RAM) with the
stock Apache.
The FastCGI server was started using the <a href="kfcgi.8.html">kfcgi(8)</a> defaults.
</p>
</section>
</article>
<footer>
<div>&#169; 2014&#8211;2015 <a rel="author" href="http://kristaps.bsd.lv">Kristaps Dzonsons</a></div>
<div>
<a href="snapshots/kcgi.tgz"><i class="fa fa-fw fa-gears"></i></a>
<a href="snapshots/kcgi.tgz.sha512"><i class="fa fa-fw fa-lock"></i></a>
<a href="snapshots"><i class="fa fa-fw fa-archive"></i></a>
<span>Built with <a href="http://kristaps.bsd.lv/sblg">sblg</a> on <a href="http://www.openbsd.org">OpenBSD</a></span>
</div>
</footer>
</body>
</html>