ppc(4) locking of isa_dma*

[John Baldwin]

On Wednesday 15 July 2009 6:49:35 am Roman Divacky wrote:
> hi
> 
> the isa_dma* functions seem to require Giant. ppc(4) uses those
> but does not seem to lock Giant or anything. is this ok?
> is it a bug?

Doh, that is a bug.

> I looked at the i386/isa/isa_dma.c and it seems pretty selfcontained
> (ie. it operates on hw and it's own variables). would it be sufficient
> to just introduce a new lock and lock with it all the isa_dma* functions
> ? or is there some indirect dependency on Giant that I cant see?

That would probably work.  If it can be self-contained in the isa_dma API that 
would so that consumers do not need fixing that would be ideal.

-- 
John Baldwin