> > Who reviewed this change?
> 
> This change didn't seem like one needing review.

Given that people are objecting, clearly it should have.

> > this is what stack gap allocation is for!
> 
> Stack gap allocation is hack which should die. Primary reason is that
> it's not MT-safe.

that's true, but the appropriate replacement for stackgap allocation
is some *other* way for the kernel to allocate a chunk of user address
space for temporary use.

> > Why are we slowing down our native syscall path for the convenience of
> > emulations?
> 
> E.g. MSG_COMPAT handling costs more than this.

"A is slow, therefore it's ok to slow B down"??

					- Bill