On Fri, 30 Nov 2012 22:56:27 -0800
Michel Lespinasse <wal...@google.com> wrote:
> expand_stack() runs with a shared mmap_sem lock. Because of this, there
> could be multiple concurrent stack expansions in the same mm, which may
> cause problems in the vma gap update code.
>
> I propose to solve this by taking the mm->page_table_lock around such vma
> expansions, in order to avoid the concurrency issue. We only have to worry
> about concurrent expand_stack() calls here, since we hold a shared mmap_sem
> lock and all vma modificaitons other than expand_stack() are done under
> an exclusive mmap_sem lock.
>
> I previously tried to achieve the same effect by making sure all
> growable vmas in a given mm would share the same anon_vma, which we
> already lock here. However this turned out to be difficult - all of the
> schemes I tried for refcounting the growable anon_vma and clearing
> turned out ugly. So, I'm now proposing only the minimal fix.
>
I think I don't understand the problem fully. Let me demonstrate:
a) vma_lock_anon_vma() doesn't take a lock which is specific to
"this" anon_vma. It takes anon_vma->root->mutex. That mutex is
shared with vma->vm_next, yes? If so, we have no problem here?
(which makes me suspect that the races lies other than where I think
it lies).
b) I can see why a broader lock is needed in expand_upwards(): it
plays with a different vma: vma->vm_next. But expand_downwards()
doesn't do that - it only alters "this" vma. So I'd have thought
that vma_lock_anon_vma("this" vma) would be sufficient.
What are the performance costs of this change?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
