On Tue, 2010-07-27 at 07:55 +0300, Michael S. Tsirkin wrote:

Peter, could you please indicate whether you think this is the way to
go, too?

I really dislike it, as you indicated, you now want priority too..

It seems the problem is that we normally don't consider work done by
kernel threads for user processes part of that process.

I'm not sure what work you're doing, but I'm pretty sure there's similar
things already in the kernel -- think about the work done by encryption
threads for encrypted sockets and stuff.

If you want proper containment of work caused by a process, I'd suggest
you start by looking at curing the general problem, instead of special
casing this one case.