On Fri, Apr 28, 2006 at 03:03:23PM +0100, Paulo Marques wrote:
> Martin Bligh wrote:
> >>[...]
> >I don't want to boot it, as that gets into security nightmares, but I
> >should be able to provide something that does static testing.
>
> Actually, booting might not be that bad using a virtual machine with qemu.
Honestly, the security nightmare begins with the compile. A patch to
the build system can result in arbitrarily insecure commands being run
during the compile - way easier than doing something that affects the
compiled kernel. A machine doing automatic compiles of untrusted
patches should be viewed as completely sacrificial from the beginning.
-VAL
Valerie Henson wrote:
> On Fri, Apr 28, 2006 at 03:03:23PM +0100, Paulo Marques wrote:
>
>>Martin Bligh wrote:
>>
>>>>[...]
>>>
>>>I don't want to boot it, as that gets into security nightmares, but I
>>>should be able to provide something that does static testing.
>>
>>Actually, booting might not be that bad using a virtual machine with qemu.
>
>
> Honestly, the security nightmare begins with the compile. A patch to
> the build system can result in arbitrarily insecure commands being run
> during the compile - way easier than doing something that affects the
> compiled kernel. A machine doing automatic compiles of untrusted
> patches should be viewed as completely sacrificial from the beginning.
True - good point ... but it's easier to chroot jail. And I'm lazy ;-)
If anyone wants to make autotest (http://test.kernel.org/autotest)
support some sort of virtual boot via creating a UML instance or
something, that'd be great. But I won't hold my breath ;-)
M.
On Mon, May 01, 2006 at 02:35:05PM -0700, Martin Bligh wrote:
> Valerie Henson wrote:
> >
> >Honestly, the security nightmare begins with the compile. A patch to
> >the build system can result in arbitrarily insecure commands being run
> >during the compile - way easier than doing something that affects the
> >compiled kernel. A machine doing automatic compiles of untrusted
> >patches should be viewed as completely sacrificial from the beginning.
>
> True - good point ... but it's easier to chroot jail. And I'm lazy ;-)
> If anyone wants to make autotest (http://test.kernel.org/autotest)
> support some sort of virtual boot via creating a UML instance or
> something, that'd be great. But I won't hold my breath ;-)
I think you should do this, security issues be darned. Just wanted to
point out where the real concern was. And thanks in advance!
-VAL