GCC 103863, We need a warning for loss of no-exec stacks

GCC 103863, We need a warning for loss of no-exec stacks

[Jeffrey Walton]

Would anyone like to comment on this GCC feature request:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103863 .

The reason I'm bringing it here is, AP suggested a warning should be
in the linker (which combines object files), and not the compiler
(which creates object files).

Re: GCC 103863, We need a warning for loss of no-exec stacks

[Andrew Pinski]

On Wed, Dec 29, 2021 at 6:25 PM Jeffrey Walton via Binutils
<binutils@sourceware.org> wrote:
>
> Would anyone like to comment on this GCC feature request:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103863 .
>
> The reason I'm bringing it here is, AP suggested a warning should be
> in the linker (which combines object files), and not the compiler
> (which creates object files).

GCC does not create the object files from assembly code either (since
it is the case you are worried about really) :). It is the assembler
(gas) which is still part of binutils.

Thanks,
Andrew

Re: GCC 103863, We need a warning for loss of no-exec stacks

[Jeffrey Walton]

On Wed, Dec 29, 2021 at 9:27 PM Andrew Pinski <pinskia@gmail.com> wrote:
>
> On Wed, Dec 29, 2021 at 6:25 PM Jeffrey Walton via Binutils
> <binutils@sourceware.org> wrote:
> >
> > Would anyone like to comment on this GCC feature request:
> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103863 .
> >
> > The reason I'm bringing it here is, AP suggested a warning should be
> > in the linker (which combines object files), and not the compiler
> > (which creates object files).
>
> GCC does not create the object files from assembly code either (since
> it is the case you are worried about really) :). It is the assembler
> (gas) which is still part of binutils.

Thanks Andrew.

In the case of ASM files, that's true. But GCC is often the driver.
People interact with GCC, not GAS. Since GCC is the compiler driver
and supervises the build, GCC seems like a natural place to enable a
warning for users.

I think the linker would be a good place, too. I think the more
chances you give someone to catch a mistake, the better.

Jeff

Re: GCC 103863, We need a warning for loss of no-exec stacks

[Alan Modra]

I suggest that you, Jeffrey, implement this feature if you feel "we
need it".  That's the way open source works!  Good luck!

-- 
Alan Modra
Australia Development Lab, IBM

Re: GCC 103863, We need a warning for loss of no-exec stacks

[Hans-Peter Nilsson]

On Wed, 29 Dec 2021, Jeffrey Walton via Binutils wrote:

> Would anyone like to comment on this GCC feature request:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103863 .
>
> The reason I'm bringing it here is, AP suggested a warning should be
> in the linker (which combines object files), and not the compiler
> (which creates object files).

When you implement that warning, be advised that targets have
different defaults; some require a marking to enable executable
stack, some to disable it.  Grep for DEFAULT_STACK_PERMS in
glibc and observe presence of PF_X for different targets.

This is complicated by gcc target-specific
NEED_INDICATE_EXEC_STACK macro possibly being out of sync.

It was some time before I did more than (as now) grep for these
things so there may be additional factors, like a global
override.

brgds, H-P