Re: [2.6 patch] let 4KSTACKS depend on EXPERIMENTAL (fwd)

[Dave Jones]

On Tue, Aug 03, 2004 at 03:13:39PM +0200, Adrian Bunk wrote:

 > > So put && !4KSTACKS in the XFS configuration ?
 > I originally did this additionally (including moving 4KSTACKS
 > above  XFS).
 > 
 > But independent of the XFS problem, 4kb stacks currently risk additional 
 > breakage without real benefits for most users.

Just before the Fedora kernel got 4K stacks (which was before mainline),
in stress testing, I was hitting memory allocation bugs far sooner than
I was hitting stack overflows, so I don't think this claim has any bearing on reality.
It was far more commonplace for the kernel to struggle to find a free pair
of contiguous pages under extreme load.  And as already mentioned,
those overflows _can_ be hit with an 8KB stack too, you just have to
try harder.

The 'real benefits' you aren't seeing are lots of failing order-1 allocations
under moderate to heavy load. You don't even need big iron boxes to see this,
(in fact, its easier to hit this problem on smaller underpowered boxes).

		Dave

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/