Re: [PATCH 2/6] ksm: dont allow overlap memory addresses registrations.

[Andrea Arcangeli <aarcange@redhat.com>]

On Wed, May 06, 2009 at 02:28:33PM +0100, Hugh Dickins wrote:
> There may prove to be various reasons why it wouldn't work out in practice;
> but when thinking of swapping them, it is worth considering if those KSM
> pages can just be assigned to a tmpfs file, then leave the swapping to that.

Not sure if I understand but the vma handled by KSM is anonymous, how
can you assign those pages to a tmpfs file, the anon vma won't permit
that, all regular anon methods will be called for swapin etc... What I
mean is that some change in core VM looks required and I plan those to
be external-rmap kind, KSM agnostic. But perhaps we can reuse some
shmem code yes, I didn't think about that yet. Anyway I'd rather
discuss this later, this isn't the time yet. I'm quite optimistic that
to make KSM swap it won't be a big change. For now there's a limit on
the max number of ksm pages that can be allocated at any given time so
to avoid OOM conditions, like the swap-compress logic that limits the
swapdevice size to less than ram.