Re: [PATCH] mm: larger stack guard gap, between vmas

From: Michal Hocko
Date: Tue Jul 04 2017 - 04:41:42 EST

On Mon 03-07-17 17:05:27, Linus Torvalds wrote:
> On Mon, Jul 3, 2017 at 4:55 PM, Ben Hutchings <ben@xxxxxxxxxxxxxxx> wrote:
> >
> > Firstly, some Rust programs are crashing on ppc64el with 64 KiB pages.
> > Apparently Rust maps its own guard page at the lower limit of the stack
> > (determined using pthread_getattr_np() and pthread_attr_getstack()). I
> > don't think this ever actually worked for the main thread stack, but it
> > now also blocks expansion as the default stack size of 8 MiB is smaller
> > than the stack gap of 16 MiB. Would it make sense to skip over
> > PROT_NONE mappings when checking whether it's safe to expand?

This is what my workaround for the older patch was doing, actually. We
have deployed that as a follow up fix on our older code bases. And this
has fixed verious issues with Java which was doing the similar thing.

> Hmm. Maybe.
> Also, the whole notion that the gap should be relative to the page
> size never made sense to me. So I think we could/should just make the
> default gap size be one megabyte, not that "256 pages" abortion.

The reason for having this in page units was that MAX_ARG_STRLEN is in
page units as well. And this is used as an on stack variable quite
often. 1MB wouldn't be sufficient for that to cover - we could go with a
larger gap but who knows how many other traps are there.

> > Secondly, LibreOffice is crashing on i386 when running components
> > implemented in Java. I don't have a diagnosis for this yet.
> Ugh. Nobody seeing this inside SuSe/Red Hat? I don't think I've heard
> about this..

No reports yet but we do not support 32b kernels on newer kernels which
had the upstream fix.

Michal Hocko