Re: [XenPPC] [pushed] [ppc] serial port discovery and zilog device driver

[Maria Butrico <butrico@xxxxxxxxxxxxx>]

Hollis Blanchard wrote:
> On Fri, 2006-05-26 at 13:13 -0400, Maria Butrico wrote:
>   
> > I re-read the email and I think I adopted your suggestions.  Some are
> > in 
> > the last patch I sent and some in the one that was committed.  What I 
> > did not do is to invoke the serial port initialization code from the 
> > platform specific initialization.
> >     
> Ignore the specific "platform" vs "serial port" detection for a moment.
> It is the same scenario: you have a property, and from that value you
> need to decide what code to call. Your approach (in both cases) was to
> define a global structure and an enum. My counter-proposal is explained
> at
> http://lists.xensource.com/archives/html/xen-ppc-devel/2006-04/msg00042.html . 
> Despite that, you continued to use the global structure/enum approach in the 
> patch which was just committed.
>
>   
Your main objection is to the global data structure that is to the fact 
that we remember in some location all this stuff about the serial port.  
You advocate simply finding, using it right away, and then forgetting 
it.  For the serial port, provided that initialization can happen even 
earlier we can do it either way, but for the platform type, I think we 
need to remember in some data structure, not a global one, one 
restricted to the platform specific code 
(http://lists.xensource.com/archives/html/xen-ppc-devel/2006-05/msg00145.html).    
Also are you advocating the we recognize the serial device as Zilog vs. 
NS by the platform or by the bus location?
> >   Unfortunately the detection of the serial port and the
> > identification of its address does not change much one way or another.
> >     
> Is this a problem? If so I don't understand the issue.
>   
The bulk of the code deals with discovery of the serial device address.  
One has to do this regardless of whether the information is stored in a 
global data structure or is on  the stack and used immediately to 
initialize the device.
>   
> > Beside Jimi, Hollis and myself, who else is interested in this issue
> > and wants to participate in a meeting to discuss this? 
> >     
> I didn't realize there was still confusion. Maybe we can try to clear it
> up by email first?
>
>   
There is no confusion, but there are remaining issues.  The issue of 
whether we should key our device discovery code from the platform, or by 
device specific methods.  The issue you raised on the mpic discovery.  
Is a simple probe needed at all given that we have a canonical probe.  
The same can be said about the dart and the serial port although we 
really do not have yet a canonical probe for either of these, but we 
could write them.   We should decide on an approach.  I advocate doing 
things pretty much as they are done in the current state of things for 
mpic because the simple probe will find the device in 90% of the cases.  
Among these, there will be a small number that will not be found by the 
canonical probe (e. g., for mpic, on mambo).
_______________________________________________
Xen-ppc-devel mailing list
Xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ppc-devel