Re: [PATCH] Don't mlock guardpage if the stack is growing up

[Linus Torvalds]

On Mon, May 9, 2011 at 4:08 PM, Tony Luck <tony.luck@xxxxxxxxx> wrote:
>
> Definitely not for normal processes - I'm not sure how both stacks are
> set up for threads.

We don't actually allow user space to set the growsup/growsdown bits
any more (we have PROT_GROWSUP and PROT_GROWSDOWN, but that is to
allow mprotect to not give an exact range, but say "apply this to the
end of a growsup/growsdown segment").

So the only thing that has those bits are things that the kernel sets
explicitly at exec time. So if ia64 doesn't set it, we're all good.

>> One thing I did want to verify: did the mlockall() actually change the
>> stack size without that patch? Just to double-check that the patch
>> actually did change semantics visibly.
>
> On an unpatched system I see this (lots more than one page of growth -
> pages are 64K on this config):
> 6007fffffff50000-6007fffffff70000 rw-p 00000000 00:00 0
> 6007fffffff50000-6008000000750000 rw-p 00000000 00:00 0
>
> On a patched system I see (this one has 16K pages - no growth)
> 600007ffff9d0000-600007ffff9d4000 rw-p 00000000 00:00 0
> 600007ffff9d0000-600007ffff9d4000 rw-p 00000000 00:00 0

Ok, I'll consider it tested. I'll commit it with Mikulas as author,
but note that I edited it so he won't get the blame if there's some
problem.

                             Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/