[uml-devel] Re: SMP skas requirements

[Jeff D. <jd...@ad...>]

On Tue, Mar 28, 2006 at 12:32:30AM +0200, Blaisorblade wrote:
> There are tons of places where spinlocks are used improperly. We need 
> stricter locking, however that is complicated by the addition of atomic 
> sections where UML can't sleep.
> 
> For instance, sigio_lock needs to be irqsave (I'll test the patch for this 
> shortly). On the other side, uml_console_write shouldn't take any lock, it 
> seems (see Documentation/tty.txt).

Yeah, the locking needs work.  We just need to make an SMP pass over
the arch to clean this up.

				Jeff