Re: [lkp-robot] [mm] 1be7107fbe: kernel_BUG_at_mm/mmap.c

[Linus Torvalds]

On Wed, Jun 21, 2017 at 12:33 PM, Oleg Nesterov <oleg@xxxxxxxxxx> wrote:
> -               if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {
> +if (0)         if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {

This smells bad.

That test is not about grow-down or even the guard page. That test is
that it's always wrong to grow down the stack below %esp.

Except we allow some slop, because certain instructions take the page
fault before actually updating %rsp.

So that patch is not correct. We want a page fault (and *not* expand
the stack) if somebody accesses below the stack pointer.

If we had a regression, it's due to something else.

                  Linus