On Thu, Jun 27, 2019 at 2:39 AM Florian Weimer [off-list ref] wrote:

* Andy Lutomirski:

quoted

Also, I don't think there's any actual requirement that the upstream
kernel recognize existing CET-enabled RHEL 8 binaries as being
CET-enabled.  I tend to think that RHEL 8 jumped the gun here.

The ABI was supposed to be finalized and everyone involved thought it
had been reviewed by the GNU gABI community and other interested
parties.  It had been included in binutils for several releases.

From my point of view, the kernel is just a consumer of the ABI.  The
kernel would not change an instruction encoding if it doesn't like it
for some reason, either.

I read the only relevant gABI thing I could find easily, and it seems
to document the "gnu property" thing.  I have no problem with that.

quoted

While the upstream kernel should make some reasonble effort to make
sure that RHEL 8 binaries will continue to run, I don't see why we
need to go out of our way to keep the full set of mitigations
available for binaries that were developed against a non-upstream
kernel.

They were developed against the ABI specification.

I do not have a strong opinion what the kernel should do going forward.
I just want to make clear what happened.

I admit that I'm not really clear on exactly what RHEL 8 shipped.
Some of this stuff is very much an ELF ABI that belongs to the
toolchain, but some if it is kernel API.  For example, the IBT legacy
bitmap API is very much in flux, and I don't think anything credible
has been submitted for upstream inclusion.  Does RHEL 8's glibc
attempt to cope with the case where some libraries are CET-compatible
and some are not?  If so, how does this work?  What, if any, services
does the RHEL 8 kernel provide in this direction?