On Fri, Feb 20, 2009 at 01:55:57PM +0100, Ingo Molnar wrote:
> The principle of probing + tearing down the mem64 resource if no
> bridge is 64-bit looks worth trying, but i'd suggest a different
> code structure to better handle true 32-bit systems built with
> 32-bit resource_size_t:
>
> - move this new code into a new file, arch/x86/pci/dac_64bit.c
>
> - build this file only if CONFIG_PHYS_ADDR_T_64BIT is set
>
> - add inline functions for pcibios_pci64_setup() and
> pcibios_pci64_verify() in the !CONFIG_PHYS_ADDR_T_64BIT case.
Agreed on all points, thanks Ingo.
> Furthermore, the filenames are very confusing in the
> arch/x86/pci/ directory and should be cleaned up ASAP. At
> minimum we should do:
>
> git mv arch/x86/pci/pcbios.c arch/x86/pci/bios_32.c
> git mv arch/x86/pci/i386.c arch/x86/pci/bios.c
>
> (and fix up the Makefile.) There's nothing 32-bit about
> "i386.c", the 64-bit x86 platform code uses it too!
Makes sense, but that's perhaps up to Jesse.
> Also, this one:
>
> > + if (insert_resource(root, r64)) {
> > + printk(KERN_WARNING "PCI: Failed to allocate PCI64 space\n");
> > + return;
> > + }
>
> Should be a more prominent WARN(), not a printk. We can only hit
> this path if the x86 platform and the PCI probing code is buggy,
> and we want to fix such bugs.
OK.
Ivan.
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
