[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [XenPPC] [Patch] Detect non hypervisor hardware



On Mon, 2006-04-17 at 14:11 -0400, Maria Butrico wrote:
> I am missing all the point as well.  You don't want an initialization 
> for these variables?  You are arguing that we get it for free?

Yes.

> My preference is to initialize everything, according to the convention 
> used by the code.  What are Xen conventions for initalizing global 
> variables?   If there aren't any already, then any ideas about what we want?

xen/arch/x86/setup.c contains quite a few globals. It has a couple "= 0"
initializations, but the majority is left for .bss.

Linux omits such initializations. Since we need to work in both worlds,
I think it would be helpful to keep consistent style where possible.

I believe some versions of GCC place "= 0" globals into .data,
increasing the size of the executable. Later gcc versions notice that
it's 0 (as opposed to some other value) and place them into .bss
anyways.

> So I prefer to initialize, but I am willing to do it in any reasonable 
> way. 

I prefer not to, which I think is reasonable. ;)

> Hollis Blanchard wrote:
> > On Sat, 2006-04-15 at 08:37 -0400, Jimi Xenidis wrote:
> >   
> >> On Apr 14, 2006, at 11:36 AM, Hollis Blanchard wrote:
> >>     
> >>>>>> +/* opt_nohv: If true, hardware has no hypervisor (HV) mode or  
> >>>>>> nohv boot
> >>>>>> + * parameter was specified.
> >>>>>> + */
> >>>>>> +int opt_nohv = 0;
> >>>>>>             
> >>> Don't need to explicitly initialize this.
> >>>       
> >> Hollis, this is global processor state.  We don;t track processor  
> >> capabilities blobally yet, untill we do we need this.  I don't think  
> >> that we should expect boot_of.c to init it.
> >>
> >>     
> >>>>>> +boolean_param("nohv", opt_nohv);
> >>>>>> +
> >>>>>>  int opt_earlygdb = 0;
> >>>>>>             
> >>> (Or this; it's not your code but if you submit a patch I'd happily
> >>> apply.)
> >>>       
> >> No, I explicitly want this, I'd like to be able to force "nohv" mode  
> >> on the command line for testing and profiling purposes.
> >>     
> >
> > I think you're missing my point: I'm talking about initializing opt_nohv
> > and opt_earlygdb. These are global variables, and so they will land in
> > the .bss section.
> >
> >   
> 
> 
> 
> _______________________________________________
> Xen-ppc-devel mailing list
> Xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-ppc-devel
-- 
Hollis Blanchard
IBM Linux Technology Center


_______________________________________________
Xen-ppc-devel mailing list
Xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ppc-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.