Re: [PATCH 1/2] mm: Take all anon_vma locks in anon_vma_lock

[Andrea Arcangeli]

On Mon, May 03, 2010 at 07:11:19PM +0200, Peter Zijlstra wrote:
> On Mon, 2010-05-03 at 19:02 +0200, Andrea Arcangeli wrote:
> > On Mon, May 03, 2010 at 06:55:12PM +0200, Peter Zijlstra wrote:
> > > This does leave me worrying about concurrent faults poking at
> > > vma->vm_end without synchronization.
> > 
> > I didn't check this patch in detail yet. I agree it can be removed and
> > I think it can be safely replaced with the page_table_lock.
> 
> Sure, it could probably be replaced with the ptl, but a single
> anon_vma->lock would I think be better since there's more of them.

ptl not enough, or it'd break if stack grows fast more than the size
of one pmd, page_table_lock enough instead.

Keeping anon_vma lock is sure fine with me ;), I was informally asked
if it was a must have, and I couldn't foresee any problem in
_replacing_ it (not removing) with page_table_lock (which I hope I
mentioned in my answer ;). But I never had an interest to remove it,
just I couldn't find any good reason to keep it either other than
"paranoid just in case", which is good enough justification to me ;)
considering these archs are uncommon and by definition gets less
testing.
--
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/