An updated Apache DOS advisory [LWN.net]

From:		dirkx-AT-apache.org (Dirk-Willem van Gulik)
To:		announce-AT-httpd.apache.org
Subject:		Advisory: Range header DoS vulnerability Apache HTTPD 1.3/2.x (CVE-2011-3192)
Date:		Fri, 26 Aug 2011 10:35:31 +0000 (UTC)
Message-ID:		<20110826103531.998348F82@minotaur.apache.org>
Archive-link:		Article, Thread

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

          Apache HTTPD Security ADVISORY
          ==============================
                    UPDATE 2

Title:       Range header DoS vulnerability Apache HTTPD 1.3/2.x

CVE:         CVE-2011-3192
Last Change: 20110826 1030Z
Date:        20110824 1600Z
Product:     Apache HTTPD Web Server
Versions:    Apache 1.3 all versions, Apache 2 all versions

Changes since last update
=========================
In addition to the 'Range' header - the 'Range-Request' header is equally
affected. Furthermore various vendor updates, improved regexes (speed and
accommodating a different and new attack pattern).

Description:
============

A denial of service vulnerability has been found in the way the multiple 
overlapping ranges are handled by the Apache HTTPD server:

     http://seclists.org/fulldisclosure/2011/Aug/175 

An attack tool is circulating in the wild. Active use of this tool has 
been observed.

The attack can be done remotely and with a modest number of requests can 
cause very significant memory and CPU usage on the server. 

The default Apache HTTPD installation is vulnerable.

There is currently no patch/new version of Apache HTTPD which fixes this 
vulnerability. This advisory will be updated when a long term fix 
is available. 

A full fix is expected in the next 24 hours. 

Background and the 2007 report
==============================

There are two aspects to this vulnerability. One is new, is Apache specific; 
and resolved with this server side fix. The other issue is fundamentally a 
protocol design issue dating back to 2007:

      http://seclists.org/bugtraq/2007/Jan/83 

The contemporary interpretation of the HTTP protocol (currently) requires a 
server to return multiple (overlapping) ranges; in the order requested. This 
means that one can request a very large range (e.g. from byte 0- to the end) 
100's of times in a single request. 

Being able to do so is an issue for (probably all) webservers and currently 
subject of an IETF discussion to change the protocol:

      http://trac.tools.ietf.org/wg/httpbis/trac/ticket/311

This advisory details a problem with how Apache httpd and its so called 
internal 'bucket brigades' deal with serving such "valid" request. The
problem is that currently such requests internally explode into 100's of 
large fetches, all of which are kept in memory in an inefficient way. This
is being addressed in two ways. By making things more efficient. And by 
weeding out or simplifying requests deemed too unwieldy.

Mitigation:
===========

There are several immediate options to mitigate this issue until a full fix 
is available. Below examples handle both the 'Range' and the legacy
'Request-Range' with various levels of care. 

Note that 'Request-Range' is a legacy name dating back to Netscape Navigator 
2-3 and MSIE 3. Depending on your user community - it is likely that you
can use option '3' safely for this older 'Request-Range'.

1) Use SetEnvIf or mod_rewrite to detect a large number of ranges and then
   either ignore the Range: header or reject the request.

   Option 1: (Apache 2.2)

          # Drop the Range header when more than 5 ranges.
          # CVE-2011-3192
          SetEnvIf Range (?:,.*?){5,5} bad-range=1
          RequestHeader unset Range env=bad-range

          # We always drop Request-Range; as this is a legacy
          # dating back to MSIE3 and Netscape 2 and 3.
          RequestHeader unset Request-Range

          # optional logging.
          CustomLog logs/range-CVE-2011-3192.log common env=bad-range
          CustomLog logs/range-CVE-2011-3192.log common env=bad-req-range

   Above may not work for all configurations. In particular situations
   mod_cache and (language) modules may act before the 'unset'
   is executed upon during the 'fixup' phase.

   Option 2: (Pre 2.2 and 1.3)

          # Reject request when more than 5 ranges in the Range: header.
          # CVE-2011-3192
          #
          RewriteEngine on
          RewriteCond %{HTTP:range} !(bytes=[^,]+(,[^,]+){0,4}$|^$)
          # RewriteCond %{HTTP:request-range} !(bytes=[^,]+(?:,[^,]+){0,4}$|^$)
          RewriteRule .* - [F]

          # We always drop Request-Range; as this is a legacy
          # dating back to MSIE3 and Netscape 2 and 3.
          RequestHeader unset Request-Range

   The number 5 is arbitrary. Several 10's should not be an issue and may be
   required for sites which for example serve PDFs to very high end eReaders
   or use things such complex http based video streaming.

2) Limit the size of the request field to a few hundred bytes. Note that while 
   this keeps the offending Range header short - it may break other headers; 
   such as sizeable cookies or security fields. 

          LimitRequestFieldSize 200

   Note that as the attack evolves in the field you are likely to have
   to further limit this and/or impose other LimitRequestFields limits.

   See: http://httpd.apache.org/docs/2.2/mod/core.html#limitreque...

3) Use mod_headers to completely dis-allow the use of Range headers:

          RequestHeader unset Range 

   Note that this may break certain clients - such as those used for
   e-Readers and progressive/http-streaming video.

   Furthermore to ignore the Netscape Navigator 2-3 and MSIE 3 specific
   legacy header - add:

          RequestHeader unset Request-Range 

   Unlike the commonly used 'Range' header - dropping the 'Request-Range' 
   is not likely to affect many clients.

4) Deploy a Range header count module as a temporary stopgap measure:

     http://people.apache.org/~dirkx/mod_rangecnt.c

   Precompiled binaries for some platforms are available at:

     http://people.apache.org/~dirkx/BINARIES.txt

5) Apply any of the current patches under discussion - such as:


http://mail-archives.apache.org/mod_mbox/httpd-dev/201108...
   http://svn.apache.org/viewvc?view=revision&sortby=dat...

OS and Vendor specific information
==================================

Red Hat:        Option 1 cannot be used on Red Hat Enterprise Linux 4.
                https://bugzilla.redhat.com/show_bug.cgi?id=732928

NetWare:        Pre compiled binaries available.

mod_security:   Has updated their rule set; see

http://blog.spiderlabs.com/2011/08/mitigation-of-apache-r...


Actions:
========

Apache HTTPD users who are concerned about a DoS attack against their server 
should consider implementing any of the above mitigations immediately. 

When using a third party attack tool to verify vulnerability - note that most 
of the versions in the wild currently check for the presence of mod_deflate; 
and will (mis)report that your server is not vulnerable if this module is not 
present. This vulnerability is not dependent on presence or absence of 
that module.

Planning:
=========

This advisory will be updated when new information, a patch or a new release 
is available. A patch or new Apache release for Apache 2.0 and 2.2 is expected 
in the next 24 hours. Note that, while popular, Apache 1.3 is deprecated.

- -- end of advisory - update 2 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (Darwin)

iEYEARECAAYFAk5Xdu8ACgkQ/W+IxiHQpxvN8ACgwsUJ6oYMq3SyoPHCR7rqsbP6
DFkAoKhZcF87F96h40tQdM1SZsiVX9N5
=07sc
-----END PGP SIGNATURE-----

(Log in to post comments)

An updated Apache DOS advisory

Posted Aug 26, 2011 18:50 UTC (Fri) by cesarb (subscriber, #6266) [Link]

> I can't believe they would get that wrong

I can. They are under quite a lot of time pressure, and due to the importance of httpd, they have to get the definitive fix right. And for this bug, it is not a simple one-liner.

Take a look at https://mail-archives.apache.org/mod_mbox/httpd-dev/20110..., where both the fix and these advisories are being developed, and note the amount of email exchanged over this issue on the last couple of days. From a quick look, the amount of messages about this issue is around double the amount of messages for the rest of the month.

> Isn't there a env=bad-req-range missing at the end of "RequestHeader unset Request-Range" because the second logging rule needs it?

No, they meant to always remove the obsolete Request-Range header. Instead, what is missing is the SetEnvIf which would set that variable for the logging rule to use.

An updated Apache DOS advisory

Posted Aug 26, 2011 18:54 UTC (Fri) by dskoll (subscriber, #1630) [Link]

Should the regexp not be (?:,.*?){5,} instead of (?:,.*?){5,5} so that it matches 5 or more ranges and not just exactly 5?

No. The expression is not anchored, so you might as well quit after 5. {5,5} (or better {5}) is faster than {5,}

An updated Apache DOS advisory

Posted Aug 26, 2011 19:01 UTC (Fri) by cesarb (subscriber, #6266) [Link]

As to

> Should the regexp not be (?:,.*?){5,} instead of (?:,.*?){5,5} so that it matches 5 or more ranges and not just exactly 5?

It depends on whether the regex is anchored or not. If it is not anchored to the start and the end of the string, it will match on any sequence of exactly 5 ranges within the header, even if it has more.[*]

From the examples at https://httpd.apache.org/docs/2.2/mod/mod_setenvif.html#s..., it seems the regex is not anchored (some of the examples there anchor it explicitly).

[*] Actually, if I am reading it right, it should match whenever there are at least _6_ ranges, since it will ignore the first one because it is not preceded by a comma.