Re: [PATCH 2.6.12 1/1] pcdp: add PCDP pci interface support

From: Mark Maule <>
Date: 2005-04-22 08:05:01
On Thu, 21 Apr 2005, Bjorn Helgaas wrote:

> On Thu, 2005-04-21 at 15:26 -0500, Mark Maule wrote:
> >  	{
> >  		extern int sn_serial_console_early_setup(void);
> >  		if (!sn_serial_console_early_setup())
> > -			return 0;
> > +			earlycons++;
> >  	}
> If SGI is going to support the PCDP, why don't you add a new
> PCDP device type for the L1 console?  It feels like a nuisance
> to have two places to look for console device info.  How do
> you say "use VGA but not the L1 console"?

Our prom today only knows how to generate PCDP for vga consoles, and even then,
older proms don't know anything about PCDP.  So for now the above would need
to stay.  I'd be willing to look into this in a future prom.

As for VGA vs. L1, my current plan is to use add_preferred_console() in 
sn_setup() to identify tty0 as the preferred default console if usable vga
exists in the system (as advertised through PCDP).

Note that there will be a follown altix-specific patch to take advantage of
this, but I figured it made more sense to iron out the generic PCDP approach

> > +unsigned long	pcdp_vga_io_base;
> > +unsigned long	pcdp_vga_mem_base;
> Apart from the extra space that snuck into the definitions, I
> don't really like these being associated with the PCDP.  Isn't
> this the sort of thing that Ben H's VGA arbitrator[1] would
> also change?
> So I vote for something with a more neutral name, possibly in
> arch/ia64/kernel/setup.c.

I don't have any problem creating non-pcdp globals.  It seemed to make more
sense to me to tie them to pcdp since that's the code that set's them up.

> And pcdp_vga_io_base isn't used anywhere, so I'd remove it.

This would be used in the above-mentioned altix patch for vga ioport access.
I added it in this patch for consistency with the mmio base.

> > +setup_vga_console(struct pcdp_device *dev)
> >  {
> > +
> >  #if defined(CONFIG_VT) && defined(CONFIG_VGA_CONSOLE)
> > -	if (efi_mem_type(0xA0000) == EFI_CONVENTIONAL_MEMORY) {
> > +	u8			*if_ptr;
> This isn't ACPI, so we don't need extra tabs in the middle
> of variable declarations (or blank lines after the beginning
> of the function).  :-)
> > +		memcpy(&if_pci, if_ptr, sizeof(struct pcdp_if_pci));
> The above is correct, but it's easier to verify this:
> 	memcpy(&if_pci, if_ptr, sizeof(if_pci));

Ok, will clean up the style issues.


To unsubscribe from this list: send the line "unsubscribe linux-ia64" in
the body of a message to
More majordomo info at
Received on Thu Apr 21 18:05:58 2005

This archive was generated by hypermail 2.1.8 : 2005-08-02 09:20:37 EST