On Tue, 9 Nov 2004, Hugh Dickins wrote:
> 
> Not to say that your patch or Adam's will go further (I've no objection
> to the way Adam is using tmpfs, but no opinion on the future of devfs),
> but they're two hints that I should rework that to get out of people's
> way.  I'll do a patch for that, then another something like yours on
> top, for you to go back and check.
> 
> I think the option should be "mpol=interleave" rather than just
> "interleave", who knows what baroque mpols we might want to support
> there in future?

Okay, two pattachments.

The first (against 2.6.10-rc1-mm5) being my reversion of NULL sbinfo
in shmem.c, to make it easier for others to add things into sbinfo
without having to worry about NULL cases.  So that goes back to
allocating an sbinfo even for the internal mount: I've rounded up to
L1_CACHE_BYTES to avoid false sharing, but even so, please test it out
on your 512-way to make sure I haven't screwed up the scalability we
got before - thanks.  If you find it okay, I'll send to akpm soonish.

The second (against the first) being my take on your patch, with
mpol=interleave, and minor alterations which may irritate you so much
you'll revert them immediately! (mainly, using MPOL_INTERLEAVE and
MPOL_DEFAULT within shmem.c rather than defining separate flags).
Only slightly tested at this end.

Hugh