Re:Tool causes Apache web server to freeze
by blameless on Thursday August 25, 2011 @05:13AM
(#2924278)
"A previously unknown flaw in the code for processing byte range headers allows version 2.2.x of the Apache Web Server to be crippled from a single PC."
Re:Tool causes Apache web server to freeze
by toong on Friday August 26, 2011 @07:03AM
(#2928070)
This doesn't seem to be very effective on my test system (apache2-mpm-prefork 2.2.14-5ubuntu8.4) regarding memory usage triggering the oom-killer.
Memory seems to be (almost?) stable, but apache cpu usage is through the roof, with the client barely consuming any cpu. The moment you kill the client, the apache server is back ok.
I'll let it run for another hour or so, to see that this actually gets somewhere or not.
Are the actual memory increments really small and/or slow ?
Or isn't this supposed to work with the apache2-mpm-prefork model, but only with the worker-model ?
Re:Tool causes Apache web server to freeze
by ominous_prime on Thursday August 25, 2011 @10:02AM
(#2924914)
yes, to quote Willy:
What is needed is to remove the Range header when there are too
many occurrences of it.
Their attack puts up to 1300 Range values. Let's remove the header if
there are more than 2 :
reqidel ^Range if { hdr_cnt(Range) gt 2 }
That should reliably defeat the attack.
Re:Tool causes Apache web server to freeze
by kwantam on Thursday August 25, 2011 @10:29AM
(#2925002)
It's not just too many occurrences of the header, it's too many ranges specified in a particular header. The zero-day script submits a header that looks like this:
Range:bytes=0-,5-1,5-2,5-3,5-4,...,5-1299
I don't know if you were implying something different with your answer since I'm not familiar enough with haproxy to know the semantics of hdr_cnt; I wanted to guard against ambiguity.
A simple test of your own machine would be something like this:
Re:Tool causes Apache web server to freeze
by ominous_prime on Thursday August 25, 2011 @10:36AM
(#2925025)
Multi-valued headers and multiple headers are interchangeable as long long as the order of values is maintained. Haproxy therefor parses the two into the same internal representation, and the rule works for either.
Re:Tool causes Apache web server to freeze
by thwarted on Friday August 26, 2011 @03:27AM
(#2927717)
This is documented in haproxy's configuration.txt:
hdr <string>
hdr(header) <string>
Note: all the "hdr*" matching criteria either apply to all headers, or to a
particular header whose name is passed between parenthesis and without any
space. The header name is not case-sensitive. The header matching complies
with RFC2616, and treats as separate headers all values delimited by commas.
Use the shdr() variant for response headers sent by the server.
It was pointed out in 2007 http://seclists.org/bugtraq/2007/Jan/83
It also affects Apache 1.3.
Memory seems to be (almost?) stable, but apache cpu usage is through the roof, with the client barely consuming any cpu. The moment you kill the client, the apache server is back ok.
I'll let it run for another hour or so, to see that this actually gets somewhere or not.
Are the actual memory increments really small and/or slow ? Or isn't this supposed to work with the apache2-mpm-prefork model, but only with the worker-model ?
A simple test of your own machine would be something like this: