Re: [PATCH 0/2] pdflush fix and enhancement

[Andi Kleen <andi@firstfloor.org>]

On Wed, Dec 31, 2008 at 09:08:26AM -0700, Peter W. Morreale wrote:
> On Wed, 2008-12-31 at 14:27 +0100, Andi Kleen wrote:
> > > I say most because the assumption would be that we will be successful in
> > > creating the new thread.  Not that bad an assumption I think.  Besides,
> > 
> > And that the memory read is not reordered (rmb()).
> > 
> 
> At the risk of showing my b*tt here...  I'm not very clear on memory
> barriers, is this necessary even inside a critical region?  (recall
> we're protected by the spin lock).

You're right the implied barriers in the spinlock are probably enough.
Never mind.

> If so, does the barrier go after the
> read, or before?   (Thanks for not laughing, however grins are allowed)

Before.

BTW on x86 it's a nop either way, but not on all other architectures.

> 
> 
> > 
> > Ok it probably needs some kind of feedback mechanism.
> > 
> 
> Actually, I tend to think we need an entirely different approach to
> flushing, please see my post to David Chinner which outlines some
> thoughts.  Basically a flushing heuristic that takes into account the
> characteristics of the various block devices.  

Ideally discovered at runtime (e.g. by watching queue lengths/service
times etc.) though. Otherwise the kernel would need to have knowledge
about the properties of all kinds of devices.

-Andi
-- 
ak@linux.intel.com