On Fri, 8 Dec 2017, Linus Torvalds wrote:
> On Fri, Dec 8, 2017 at 12:02 PM, Thomas Gleixner <t...@linutronix.de> wrote:
> >
> > The reason why this happens is that init_new_context_ldt() is called from
> > init_new_context() which obviously needs to be called for both fork() and
> > exec().
> >
> > It would be surprising if anything relies on that behaviour, so it seems to
> > be safe to remove that misfeature.
>
> Looks sane. That said, can't we separate this out into the copy_mm()
> phase only?
>
> We have "arch_dup_mmap()" that is called on fork() only, so that could
> do the LDT copy from the old mm, and the actual init_new_context would
> just zero it out.
>
> Then there wouldn't be any odd "check if this is an execve" because
> the copying would be done in the right place.
Yes, that should work. It just needs to change arch_dup_mmap() so it can
return an error code.
Thanks,
tglx
