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

[Ben Hutchings]

On Sat, 2017-06-24 at 02:11 -0700, Hugh Dickins wrote:
> On Thu, 22 Jun 2017, Hugh Dickins wrote:
> > On Thu, 22 Jun 2017, Ben Hutchings wrote:
> >Â
> > > Here's my attempt at a backport to 3.2. This is only tested on
> > > x86_64 and I think I should introduce local variables for
> > > vma_start_gap() in a few places. I had to cherry-pick commit
> > > 09884964335e "mm: do not grow the stack vma just because of an overrun
> > > on preceding vma" before this one (which was a clean cherry-pick).
> >Â
> > Both your speed and your stamina are much better than mine; and your
> > patch belies your Sturgeon's law signature. I haven't got beyond the
> > architectures yet in my parallel attempt, and you do appear to be
> > doing everything right (but a local variable often welcome, yes).
> >Â
> > I'm giving up for the night, will contine tomorrow.
> > The only discrepancy I notice so far is that I have
> > arch/alpha/kernel/osf_sys.c
> > arch/ia64/mm/hugetlbpage.c
> > arch/sparc/kernel/sys_sparc_32.c
> > in my list of changed files, but they're not in yours.
> 
> And here's my attempt at a backport to 3.2.89, at last.
> I know it builds and boots and runs on x86 64 and 32,
> but that's about all that I've tried.
> 
> If you diff against yours (I preferred not to send that diff,
> because of the couple of rejects in yours against 3.2.89),
> you'll find most of the difference is just noise from where
> I used a variable, but you had not yet done so in yours.

Thanks, this is much nicer.

> But there are those three missing files, and there are a few
> places where I have a little "if (prev) {" block at the head of
> the loop after find_vma_prev(): I think those loops start off
> wrongly without that.

I also failed to update prev.

[...]
> So please at least take a look through the diff from yours, I
> think you'll find a few things to bring in, but a lot to ignore.

I think I'll take most of yours, thanks.

Ben.

-- 
Ben Hutchings
Sturgeon's Law: Ninety percent of everything is crap.

Attachment: signature.asc
Description: This is a digitally signed message part