Re: [PATCH 1/4] exec: Change uselib(2) IS_SREG() failure to EACCES

[Christian Brauner]

On Tue, May 19, 2020 at 04:32:38PM +0200, Geert Uytterhoeven wrote:
> Hi Christian,
> 
> On Tue, May 19, 2020 at 3:15 PM Christian Brauner
> <christian.brauner@xxxxxxxxxx> wrote:
> > On Tue, May 19, 2020 at 06:56:36AM -0500, Eric W. Biederman wrote:
> > > Andreas Schwab <schwab@xxxxxxxxxxxxxx> writes:
> > > > On Mai 18 2020, Eric W. Biederman wrote:
> > > >> If it was only libc4 and libc5 that used the uselib system call then it
> > > >> can probably be removed after enough time.
> > > >
> > > > Only libc4 used it, libc5 was already ELF.
> > >
> > > binfmt_elf.c supports uselib.  In a very a.out ish way.  Do you know if
> > > that support was ever used?
> > >
> > > If we are truly talking a.out only we should be able to make uselib
> > > conditional on a.out support in the kernel which is strongly mostly
> > > disabled at this point.
> >
> > The only ones that even allow setting AOUT:
> >
> > arch/alpha/Kconfig:     select HAVE_AOUT
> > arch/m68k/Kconfig:      select HAVE_AOUT if MMU
> >
> > and x86 deprecated it March 2019:
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=eac616557050737a8d6ef6fe0322d0980ff0ffde
> 
> Quoting myself (for the second time this month):
>    "I think it's safe to assume no one still runs a.out binaries on m68k."
>     http://lore.kernel.org/r/CAMuHMdW+m0Q+j3rsQdMXnrEPm+XB5Y2AQrxW5sD1mZAKgmEqoA@xxxxxxxxxxxxxx

Well, it's just a great thing to hear. :)

Fyi, I once tried to give a demo with a Slackware version in a container
which pre-dated ELF on a x86 kernel with AOUT. It didn't go well...

Christian