Skip to content
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
.travis.yml Update .travis.yml Sep 18, 2016 Update May 25, 2016
LICENSE Update LICENSE Oct 2, 2016 fixe apache 2.4 unixd_config change. Sep 11, 2016 Update Nov 2, 2018
_config.yml Update _config.yml Mar 29, 2017
circle.yml Create circle.yml May 11, 2017
configure v1.0rc May 25, 2016
mod_chroot.c Update mod_chroot.c Oct 9, 2016
mod_chroot.h ap 2.2 regression Feb 12, 2017

Build Status CircleCI License: GPL v3


What is it?

mod_chroot makes running Apache in a secure chroot environment easy. You don't need to create a special directory hierarchy containing /dev, /lib, /etc...

This project are major rewrite of mod_chroot by hobbit at (dead project, wayback machine link) focused on apache 2 only (removed Apache 1 support).

Note: From 2.2.10 of apache they include ChrootDir that are similar but not so simple to use as this version of mod_chroot.

Why chroot?

For security.

chroot(2) changes the root directory of a process to a directory other than "/". It means the process is locked inside a virtual filesystem root. If you configure your chroot jail properly, Apache and its child processes (think CGI scripts) won't be able to access anything except the jail.

A non-root process is not able to leave a chroot jail. Still it's not wise to put device files, suid binaries or hardlinks inside the jail.

Chroot - the hard way

There are many documents about running programs inside a chroot jail. Some daemons (tinydns, dnscache, vsftpd) support it out of the box. For others (like Apache) you need to carefully build a "virtual root", containing every file the program may need. This usually includes:

  • C library
  • various other libraries (libssl? libm? libmysqlclient?)
  • resolver configuration files (/etc/nsswitch.conf, /etc/resolv.conf)
  • user files (/etc/passwd, /etc/group)
  • separate directory for log files
  • additional modules needed by the program (for Apache: mod_php and other modules)

Creating this structure is great fun. Run the program, read the error message, copy the missing file, start over. Now think about upgrading - you have to keep your "virtual root" current - if there is a bug in libssl, you need to put a new version in two places. Scared enough? Read on.

Chroot - the mod_chroot way

mod_chroot allows you to run Apache in a chroot jail with no additional files. The chroot() system call is performed at the end of startup procedure - when all libraries are loaded and log files open.

Major change between 0.x and 1.x version:

Starting from version 0.3 mod_chroot supports apache 2.0. While most problems with Apache 1.3 are solved in 2.0 (no more module ordering hassle, no need to apply EAPI patches), architecture changes that appeared in 2.0 created one new problem: multi-processing modules (MPMs). MPMs are core Apache modules responsible for handling requests and dispatching them to child processes/threads.

Unfortunately, MPMs are initialized after all "normal" Apache modules. This basically means that with mod_chroot, MPM initialization is done after a chroot(2) call; when control is handed to MPM, Apache is already inside a jail. And MPMs need to create some files during startup (at least one, a pidfile) - these have to be placed inside the jail.

Once chrooted, Apache cannot access anything located above ChrootDir. For that reason restarting Apache with apachectl reload, apachectl graceful or kill -HUP apache_pid will not work as expected. Apache will not be able to read its config file, open logs or load modules.

Starting with the 1.0 of mod_chroot solve all this probleme.

  • In first problem are resolved by placing the chroot in child_init phase, after the mpm initialisation, in that way they don't need to place pid file, scoreboard, Lock File inside the jail.

  • And the seconds probleme are command like DocumentRoot and CoreDumpDirectory that are tested at start of the server, that make oblige to point to real directory outside the chroot and in the chroot...

    This proleme is resolved by faking all map to storage transparantly like mod_alias (by setting ChrootFixRoot option). This also fake the CoreDumpDirectory directory.

in that way using mod_chroot is simple like to load the module and set the chroot dir... the only constraint are to set document root and coredump directory in the path of the chroot dir...

Ex. this are ok:

ChrootDir         /srv/www1
DocumentRoot      /srv/www1/htdocs
CoreDumpDirectory /srv/www1/dump

Ex. this are not ok:

ChrootDir         /srv/www1
DocumentRoot      /htdocs
CoreDumpDirectory /dump



in order to build this module you need:

  • apxs (shipped as apxs2 by some distributors) and Apache headers. If you compiled Apache from source you already have these headers. If you use Debian, you need to install apache2-prefork-dev or apache2-threaded-dev.
  • Make
  • C compiler (ex:gcc).

Warning: starting 1.0 they support only Apache 2.x.

Building mod_chroot

1 - Go to mod_chroot source directory and type:

# ./configure --with-apxs=/path/to/apxs
# make
# make install

2 - add in your httpd.conf the loading of the module:

<IfModule !mod_chroot.c>
  LoadModule chroot_module modules/


mod_chroot provides two configuration directive:


It can only be used in main server configuration. You can't put ChrootDir inside a <Directory>, <Files>, <Location>, <VirtualHost> sections or .htaccess files.

They define the chroot virtual path base.

Example: if you store your www files in /var/www and you want to make it as virtual root, set ChrootDir this path.

You can use apache server root relative path.


ChrootDir chroot

if the server root is /srv/www1 and ChrootDir is set to "chroot" while point to /srv/www1/chroot directory

you can use some hints to set the chroot directory to value with apache internal value:

  • DOCUMENT_ROOT => set the chroot virtual path base to the globale apache document root value. (to use this you must set document root before ChrootDir cmd).

  • SERVER_ROOT => set the chroot virtual path base to the apache server root. (-d args of httpd, or ServerRoot apache directive)



to set the chroot to the apache document root.


It can only be used in main server configuration. You can't put ChrootDir inside a <Directory>, <Files>, <Location>, <VirtualHost> sections or .htaccess files.

The DocumentRoot (global config context) apache commande verify if directory exist before the chroot is activated, in that way you cannot use DocumentRoot (global config context) relative to chroot (/www in place /var/chroot/www if chroot is /var/chroot).

Note: if you not set global config context DocumentRoot apache set it to default value (define at compile time) and check it in the same way.

Exemple: If you had DocumentRoot pointing to /var/chroot/www and ChootDir pointing to /var/chroot Apache will be looking for /var/chroot/www inside of your chroot (/var/chroot/www/var/chroot/www)

And if you set ChrootFixRoot to yes (are off by default), you dont need to take care about that.

This options fake all apache map to file operation transparantly (like apache mod_alias), and fake also the CoreDumpDirectory directory.

In that way you can use the real (with DocumentRoot/Alias/UserDir...) path in apache configuration file, and this option while translate all real path to relative path to chroot directory (/var/chroot/www is translated to /www if chroot dir is /var/chroot).

Example 1: if you setup your apache like that.

ServerRoot /srv/www1
DocumentRoot /srv/www1/chroot/htdocs
ChrootFixRoot on
# path relative to server root ==> chrootdir is set to /srv/www1/chroot
ChrootDir chroot 

all request to the server like http://mysserver/path/to/mypage while be transparantly translated to /htdocs/path/to/mypage because the chroot are /srv/www1/chroot.

Example 2: if you setup your apache like that.

ServerRoot /srv/www1
DocumentRoot /srv/www1/htdocs
ChrootFixRoot on
# using server root to set chrootdir.

all request to the server like http://mysserver/path/to/mypage while be transparantly translated to /htdocs/path/to/mypage because the chroot are /srv/www1.

Example 3: if you setup your apache like that.

ServerRoot /srv/www1
DocumentRoot /srv/www1/htdocs
ChrootFixRoot on
# using globale document root to set chrootdir.

all request to the server like http://mysserver/path/to/mypage while be transparantly translated to /path/to/mypage because the chroot are /srv/www1/htdocs.


Date Timezone database

Every time you use the date function in the chroot jail, you while get an error saying the timezone db is corrupt. Is why you need localtime and zoneinfo file in the chroot jail.

# mkdir -p /chroot/etc
# cp /etc/localtime /chroot/etc/localtime
# mkdir -p /chroot/usr/share
# cp /usr/share/zoneinfo /chroot/usr/share/zoneinfo

DNS lookups

Libresolv uses /etc/resolv.conf to find your DNS server. If this file doesn't exist, libresolv uses as the DNS server. You can run a small caching server listening on (which may be a good idea anyway), or use your operating system's firewall to transparently redirect queries to to your real DNS server.

Note that this is only necessary if you do DNS lookups - probably this can be avoided?

You may need to copy /etc/nsswitch.conf and /etc/resolv.conf in chroot jail :

# mkdir -p /chroot/etc
# cp /etc/nsswitch.conf /chroot/etc/nsswitch.conf
# cp /etc/resolv.conf /chroot/etc/resolv.conf

Please also read the libraries section below because libc resolver load some additional dymanics library at first DNS resolution.

If you whant to use nscd in the chroot jail you must map nscd socket in the chroot jail.

# mkdir -p /chroot/var/run
# mount -o bind /var/run/nscd /chroot/var/run/nscd


If your mySQL/PostgreSQL accepts connections on a Unix socket which is outside of your chroot jail, reconfigure it to listen on a loopback address (

With mysql if you try to connect local db you must use in place of localhost in connect string, without that mysql client try to use mysql unix sockets in place of using tcp, without that in the chroot jail you cannot connect to db because the socket are generaly outside of the chroot.

You can also map socket in the chroot jail with mount bind like that :

# mkdir -p /chroot_dir/path/to/mysql/sockets/
# mount -o bind /path/to/mysql/sockets/ /chroot_dir/path/to/mysql

PHP mail() function

Under Unix, PHP requires a sendmail binary to send mail. Putting this file inside your chroot jail may not be sufficient, you would probably need to move your mail queue as well.

To avoid that you have three options here:

  • Don't use mail()! Use a class/function that knows how to send directly via SMTP :
  • Use esmtp pecl module based on libesmtp.
  • Install a SMTP-only sendmail clone like sSMTP or mini_sendmail. You can then put a single binary inside your jail, and deliver mail via a smarthost (warning see PHP Exec function section beceause php need /bin/sh to execute sendmail clone in the chroot jail),

PHP exec Functions

PHP system(), exec(), shell_exec(), ou proc_open() execution function uses /bin/sh -c internally. Beceause of that you need to include /bin/sh in chroot jail to use that function in PHP.

Try to use busybox or toybox minimalist and staticly linked shell (no need other thing that the executable in chroot jail), or dash light posix shell (that need few dependancy in the chroot jail).

Instead of placing a static sh or copying libs there, knzl had hacked up a small wrapper that supports being called with -c command to launch sendmail. It doesn’t support escaping of parameters with quotation marks, but that’s not necessary. It’s called sh.c:

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#define MAXARG 64
int main( int argc, char* const argv[] ) {
    char* args[ MAXARG ] = {};
    if( argc < 3 || strcmp( argv[1], "-c" ) != 0 ) {
        fprintf( stderr, "Usage: %s -c <cmd>\n", argv[0] );  
        return 1;
        char* token;
        int i = 0;  
        char* argStr = strdup( argv[2] );
        while( ( token = strsep( &argStr, " " ) ) != NULL ) {
            if( token && strlen( token ) )
                args[ i++ ] = token;
            if( i >= MAXARG )
                return 2;
    return execvp( args[0], args );

Compile it by calling

# gcc sh.c -o sh -static

and place it as bin/sh in the chroot.

PHP iconv

iconv() function need /usr/lib/x86_64-linux-gnu/gconv that must be in chroot, and also /usr/lib/locale. without that you may have PHP Notice: iconv(): Wrong charset, conversion from UTF-8' to CP1252' is not allowed in log.

Shared libraries

Shared libraries are libraries which are linked to a program at run-time. Nowadays, most programs require some shared libraries to run - is most common. You can see a list of shared libraries a program requires by running ldd /path/to/program. Loading of these libraries is done automagically by at startup. mod_chroot doesn't interfere with this mechanism.

A program may also explicitly load a shared library by calling dlopen() and dlsym(). This might cause troubles in a chrooted environment - after a process is chrooted, libraries (usually stored in /lib) might be no longer accessible. This doesn't happen very often, but if it does - there is a solution: you can preload these libraries before chrooting. Apache has a handy directive for that: LoadFile.

For exemple:

  • DNS lookups - GNU libc tries to load and some time when a first DNS lookup is done. Solution:
 LoadFile /lib/
 LoadFile /lib/
  • Apache 2.0 with mpm_worker on Linux 2.6 - GNU libc tries to load when pthread_cancel is called. Solution:
 LoadFile /lib/
  • Curl need to preload nss library and trusted certificat or may occur Problem with the SSL CA cert (path? access rights?) error :
  LoadFile /usr/lib64/
  LoadFile /usr/lib64/
  LoadFile /usr/lib64/

and map trusted store in the chroot jail with mount bind :

# mkdir -p /chroot/etc/pki/tls
# mount -o bind /etc/pki/tls        /chroot/etc/pki/tls       
# mkdir -p /chroot/etc/pki/ca-trust
# mount -o bind /etc/pki/ca-trust/  /chroot/etc/pki/ca-trust

To determine what librarie they need to preload before chroot or to include in the chroot jail

Use ldd to see library linked with an executable:

# ldd /bin/sh =>  (0x00007ffe65357000) => /lib64/ (0x0000003444e00000) => /lib64/ (0x0000003435200000) => /lib64/ (0x0000003434e00000)
        /lib64/ (0x0000003434a00000)

When executing :

# strace -e trace=file /bin/sh 2>&1  | egrep "open|stat"
open("/etc/", O_RDONLY)      = 3
open("/lib64/", O_RDONLY)  = 3
open("/lib64/", O_RDONLY)     = 3
open("/lib64/", O_RDONLY)      = 3
open("/dev/tty", O_RDWR|O_NONBLOCK)     = 3
open("/usr/lib/locale/locale-archive", O_RDONLY) = 3
open("/proc/meminfo", O_RDONLY|O_CLOEXEC) = 3
stat("/mnt/www/etc", {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
stat(".", {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
open("/usr/lib64/gconv/gconv-modules.cache", O_RDONLY) = 3

Or on already running executable use strace -p <pid>[,<pid>,...] (a running process pool) :

#  strace  -p $(ps -ef | grep <process pool name> | awk '{a[i++]=$2}END{printf(a[0]);for(n=1;n<i;n++)printf(","a[n])}') 2>&1 | egrep "open|stat"
You can’t perform that action at this time.