Re: -fno-stack-protector

[Mike Frysinger <vapier at gentoo dot org>]

On Tuesday 06 May 2008, Mark Seaborn wrote:
> Mike Frysinger <vapier@gentoo.org> wrote:
> > On Monday 05 May 2008, Mark Seaborn wrote:
> > > It would be useful if glibc knew how to turn gcc's stack-protector
> > > option off.  This patch should do the trick.  If -fno-stack-protector
> > > is available it adds it to CFLAGS:
> >
> > ugh, no.  this is the opposite of what we should do.  glibc should
> > detect that it's being built with SSP and so account for it.  we
> > build glibc with SSP in Hardened Gentoo.  we havent bothered posting
> > the changes to the libc lists though as we've always been told that
> > what we want to do (build glibc with SSP) is stupid.
>
> OK, I found the changes you were referring to.  This file makes ld.so
> build with -fstack-protector:
> http://sources.gentoo.org/viewcvs.py/gentoo-x86/sys-libs/glibc/files/2.6/gl
>ibc-2.6-gentoo-stack_chk_fail.c?rev=1.4&view=markup and this adds it into
> the glibc source tree:
> http://sources.gentoo.org/viewcvs.py/gentoo-x86/sys-libs/glibc/glibc-2.7.eb
>uild?rev=1.12&view=markup
>
> Not sure I like the idea of having ld.so know about syslog.

actually no, that isnt what i was referring to.  i doubt those changes will 
ever get merged.  the only thing that does is make the handler actually 
useful by logging a message which a stack smashing attack occurs.  this is 
how the original SSP implementation did things and it is something redhat 
removed when they re-implemented SSP for gcc-4.  the current implementation 
silently crashes and no one is the wiser that a stack smashing attack 
occurred.

> I think you should post changes like this.  It helps the rest of us
> understand glibc if nothing else.

this list was *just created*.  we've been doing Hardened Gentoo for years.  i 
also pointed out that we've attempted in the past to get things merged only 
to have drepper tell us to  toss off.
-mike

Attachment: signature.asc
Description: This is a digitally signed message part.