[patch] Adding stack markings to the x86 assembly for not using executable stack

[Andrew John Hughes]

2009/8/28 Kees Cook <kees at ubuntu.com>:
> Hi,
>
> On Thu, Aug 27, 2009 at 06:01:06PM +0100, Andrew John Hughes wrote:
>> 2009/8/27 Kees Cook <kees at ubuntu.com>:
>> > On Thu, Aug 27, 2009 at 12:04:07PM +0100, Andrew John Hughes wrote:
>> >> 2009/8/27 Matthias Klose <doko at ubuntu.com>:
>> >> > This was reported as https://edge.launchpad.net/bugs/409736
>> >> >
>> >> > Java is marked to have an executable stack[1]. This is potentially
>> >> > dangerous, and is simply an oversight from one of the compiled assembly
>> >> > files. Adding stack markings to the assembly solves the issue.
>> >> >
>> >> > sun/security/ssl/javax/net/ssl/NewAPIs/SessionCacheSizeTests.java passes
>> >> > both stock and and with non-exec-stack.
>> >> >
>> >> > gcc -fstack-protector is the default on Ubuntu. I'd like to see this patch
>> >> > for the IcedTea 1.6 release as well.
>> >> >
>> >> >  Matthias
>> >> >
>> >>
>> >> I've heard about this issue before from Gentoo users and the fix, if
>> >> it truly is this simple, would be good to have.
>> >
>> > The question tends to be one of portability.  In cases were non-gcc is
>> > used, ifdef's need to be built around the flag line.  I can provide some
>> > examples, if needed.
>> >
>>
>> I don't see an immediate problem, as they only affect x86/linux and
>> x86_64/linux where the compiler is gcc.
>
> Okay, sounds good.
>
>> >> Are you sending this patch upstream?  It would be good to have some
>> >> feedback from the HotSpot developers before we commit this for a
>> >> release.
>> >>
>> >> Does this affect SPARC too?
>> >
>> > I'm not familiar with SPARC hardware, but if it supports "execute" memory
>> > protections, then it is a valuable change there too.  It it doesn't, it
>> > won't hurt anything, IIUC.
>>
>> Do you have an SCA, either via Ubuntu or personally? A webrev needs to
>> be prepared against one of the HotSpot forests and posted to
>> hotspot-dev.  If this is the compiler, hotspot-comp is appropriate and
>> twisti can review it ;)
>
> I haven't signed it yet, but these two (identical) lines are unlikely
> to be attributable to me anyway, they're common knowledge for this area
> of work.
>
> I'll go figure out what I need to do for the SCA for future stuff, though.
>

I agree with you, and we'd have allowed it for GNU Classpath.
But Sun require an SCA for everything.

> Thanks!
>
> -Kees
>
> --
> Kees Cook
> Ubuntu Security Team
>



-- 
Andrew :-)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

Support Free Java!
Contribute to GNU Classpath and the OpenJDK
http://www.gnu.org/software/classpath
http://openjdk.java.net

PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8