Re: ACPI from FreeBSD-STABLE

[David Rhodus <drhodus@xxxxxxxxx>]

On Sunday, August 31, 2003, at 11:29 AM, Hiten Pandya wrote:

> Matthew Dillon wrote:
> >     Consider the mess that -stable is in right now with the PAE 
> > patches
> >     still apparently crashing systems.
> >     ACPI is too critical for us to track -stable with until it has    
> >  stabilized in -stable (or -current).  I am not at all reassured
> >     by the current state of affairs in -stable.
> >     So I would say we should wait at least a few months, though I am 
> > not
> >     against trying to identify why the original poster's particular 
> > system
> >     doesn't boot and trying to fix that one thing.
>
> 	Well, as I suggested earlier on, we could create a section on
> 	the web site where we track patchsets like ACPI and PAE.
For now that might be a good path to take. The ACPI should be
brought in right away. It doesn't change any critical code paths
by pulling it in, unless you compile your kernel to use it. I wouldn't
look for a working implementation to ever find its way into the 4.x tree
at this point. Perhaps someone can pull in the framework at start fixing
some of the known problems. Something that might be interesting would
be a audit of the ACPI code because of the large entry vector it creates
into kernel space.

As for the PAE work, that's not a hard of a thing. I had to implement 
that
for a large scale web server running FreeBSD back when I was still in 
school.
Maybe someone would be interested in working with PSE-36 ?
I don't have a problem waiting, most people like myself that use 
machines with
more than 4G's of RAM should be able to handle making it work on there 
own
for now. I'd rather wait until someone that has time to commit working 
code into
the DF cvs that understands what they are really changing and not have 
the same
problems that we've just seen on -stable with the PAE commits.


-DR