Re: cmwq and dm-crypt devices?

[Andi Kleen <andi@firstfloor.org>]

> Please fix this regression the patch causes then or give some
> fundamental arguments why we can't satisfy the two requirements
> simultaneously: I just don't understand why it is necessary to cause a
> regression in stacking/low memory support in order to provide the
> performance boost in the way you desire.  The two things ought to be
> independent of each other.

The old stacking relies on per device threads. Per device threads
cannot be combined with per CPU threads, because that would lead
to a thread explosion for large device counts.

I thought I had handled stacking, but my implementation doesn't
seem to fully work. Most likely it needs a fallback thread
or using one of the available workqueues.
 
> (And I remain sceptical about the suitability of the patch even with a
> stack of just two or three layers as I don't see anything in the failed
> tests that shows why similar failures couldn't in principle affect

Yes most likely the problem happens with any stacking > 0.

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only.