Re: [RFC 37/37] fs/binfmt_elf: Block old shstk elf bit

[Edgecombe, Rick P]

On Mon, 2022-11-07 at 13:47 -0800, H.J. Lu wrote:
> On Mon, Nov 7, 2022 at 1:34 PM Edgecombe, Rick P
> <rick.p.edgecombe@xxxxxxxxx> wrote:
> > 
> > On Mon, 2022-11-07 at 13:21 -0800, H.J. Lu wrote:
> > > > > Some applications and libraries are compiled with -fcf-
> > > > > protection,
> > > > > but
> > > > > they manipulate the stack in such a way that they aren't
> > > > > compatible
> > > > > with the shadow stack.   However, if the build/test setup
> > > > > doesn't
> > > > > support
> > > > > shadow stack, it is impossible to validate.
> > > > > 
> > > > 
> > > > When we have everything in place, the problems would be much
> > > > more
> > > > obvious when distros started turning it on. But we can't turn
> > > > it on
> > > > as
> > > 
> > > Not necessarily.  The problem will show up only in a CET enabled
> > > environment since build/test setup may not be on a CET capable
> > > hardware.
> > 
> > Well, I'm not sure of the details of distro testing, but there are
> > plenty of TGL and later systems out there today. With kernel
> > support,
> > I'm thinking these types of problems couldn't lurk for years like
> > they
> > have.
> 
> If this is the case, we would have nothing to worry about since the
> CET
> enabled applications won't pass validation if they aren't CET
> compatible.

Hmm, I think you couldn't have already forgotten the problem binaries
are already shipped...

> 
> > > 
> > > > planned without breaking things for existing binaries. We can
> > > > have
> > > > both
> > > > by:
> > > > 1. Choosing a new bit, adding it to the tools, and never
> > > > supporting
> > > > the
> > > > old bit in glibc.
> > > > 2. Providing the option to have the kernel block the old bit,
> > > > so
> > > > upgraded users can decide what experience they would like. Then
> > > > distros
> > > > can find the problems and adjust their packages. I'm starting
> > > > to
> > > > think
> > > > a default off sysctl toggle might be better than a Kconfig.
> > > > 3. Any other ideas?
> > > 
> > > Don't enable CET in glibc until we can validate CET
> > > functionality.
> > 
> > Can you elaborate on what you mean by this? Not upstream glibc CET
> > support? Or have users not enable it? If the latter, how would they
> > know about all these problems.
> 
> The current glibc doesn't support CET.  To enable CET in an
> application,
> one should validate it together with the CET enabled glibc under the
> CET
> enabled kernel on a CET capable machine.

Agreed that this is how it should have gone.

> 
> > 
> > And what is wrong with the cleanest option, number 1? The ABI
> > document
> > can be updated.
> 
> It doesn't help resolve any issues.

Please read the coverletter if you are unsure of what issues this is
trying to address. I should have put more in the commit log.