Subject: Re: [vserver] test patch for 2.6.30.7
From: Ed W <lists@wildgooses.com>
Date: Tue, 29 Sep 2009 13:23:33 +0100
Tue, 29 Sep 2009 13:23:33 +0100
Natanael Copa wrote:
> On Mon, 2009-09-28 at 14:27 +0200, ADNET Ghislain wrote:
>
>   
>> i think :
>>
>>              Use dietlibc: no (you have been warned)
>>
>>  is B.A.D you should use dietlibc for vserver utils, at least if not "you have been
warned" :)
>>     
>
> IIRC that was due to glibc will pull in libs dynamically, even if linked
> statically. I don't think uclibc have this problem so I don't see any
> point in building dietlibc.
>
> I actually tried to build dietlibc but it looks like I'd have to disable
> SSP then.
>
> Does anyone have an exploit/proof-of-concept testcase I could try?
>   

On gentoo I recently posted some results with a very recent 
vserver-tools and basically nearly all tools segfaulted quite quickly if 
I built dietlibc using gcc 3.4.6 hardened (obviously you are using 
something slightly different, but...)

However, there is some possibility this is actually due to some "fault" 
in the dietlibc/vserver-utils of course, which is simply not picked up 
without compiling with SSP...  I didn't try under valgrind to try and 
catch things and I guess we really need Daniel to offer better insight?

Good luck

Ed W



Natanael Copa wrote:
On Mon, 2009-09-28 at 14:27 +0200, ADNET Ghislain wrote:

  
i think :

             Use dietlibc: no (you have been warned)

 is B.A.D you should use dietlibc for vserver utils, at least if not "you have been warned" :)
    

IIRC that was due to glibc will pull in libs dynamically, even if linked
statically. I don't think uclibc have this problem so I don't see any
point in building dietlibc.

I actually tried to build dietlibc but it looks like I'd have to disable
SSP then.

Does anyone have an exploit/proof-of-concept testcase I could try?
  

On gentoo I recently posted some results with a very recent vserver-tools and basically nearly all tools segfaulted quite quickly if I built dietlibc using gcc 3.4.6 hardened (obviously you are using something slightly different, but...)

However, there is some possibility this is actually due to some "fault" in the dietlibc/vserver-utils of course, which is simply not picked up without compiling with SSP...  I didn't try under valgrind to try and catch things and I guess we really need Daniel to offer better insight?

Good luck

Ed W