Re: softirq bugs in pre2

[Andrew Morton]

Linus Torvalds wrote:
> 
> On Mon, 11 Jun 2001, Andrea Arcangeli wrote:
> >
> > Since I mentioned the copy-user latency fixes (even if offtopic with the
> > above) this is the URL for trivial merging:
> 
> The copy-user latency fixes only make sense for out-of-line copies. If
> we're going to have a conditional function call to "schedule()", we do not
> want to inline the dang thing any more - we've just destroyed our register
> set etc anyway.

It's overkill.  This adds many hundreds of scheduling points
to the kernel, of which we need only five.  It makes more
sense to simply open-code those five.

- generic_file_read/write
- read /dev/zero, /dev/mem
- memcpy_to_iovec()

This will by no means provide a low-latency kernel, but it will
fix the most common causes of poor interactivity in normal
use.

Just doing generic_file_read/write would suffice, actually.