Re: Always zero test in arch/i386/pci/mmconfig-shared.c

Re: Always zero test in arch/i386/pci/mmconfig-shared.c

[Robert Hancock]

Scott Ashcroft wrote:
> I don't think it makes any change to subsequent flow but the code in 
> pci_mmcfg_check_hostbridge is:
> 
> pci_mmcfg_config_num = 0;
> 

You missed this code:

name = pci_mmcfg_probes[i].probe();

This calls one of the probe functions above, which will set 
pci_mmcfg_config_num to something else, as with the name variable. It 
may set the name but not the config num, if the chipset is recognized 
but MMCONFIG isn't enabled, etc.

> 
> if (name) {
>     printk(KERN_INFO "PCI: Found %s %s MMCONFIG support.\n",
>            name, pci_mmcfg_config_num ? "with" : "without");
> }
> 
> So it will always prints "without".
> 
> I'd make a patch but I'm not sure what this fix is. Just drop the test, 
> the printk or make it say "which might possibly have"?
> 
> Cheers,
> Scott

-- 
Robert Hancock      Saskatoon, SK, Canada
To email, remove "nospam" from hancockr@nospamshaw.ca
Home Page: http://www.roberthancock.com/

Re: Always zero test in arch/i386/pci/mmconfig-shared.c

[Scott Ashcroft]

Robert Hancock wrote:
  > You missed this code:
> 
> name = pci_mmcfg_probes[i].probe();
> 
> This calls one of the probe functions above, which will set 
> pci_mmcfg_config_num to something else, as with the name variable. It 
> may set the name but not the config num, if the chipset is recognized 
> but MMCONFIG isn't enabled, etc.

Obvious now you've pointed it out. Bit of of a weird way to do things.
I might brew up a patch which adds the names to the 
pci_mmcfg_hostbridge_probe struct, since they appear to be always 
returned if the probe function is called, and make the probe functions 
return the config num.

Cheers,
Scott

Always zero test in arch/i386/pci/mmconfig-shared.c

[Scott Ashcroft]

I don't think it makes any change to subsequent flow but the code in 
pci_mmcfg_check_hostbridge is:

pci_mmcfg_config_num = 0;

...

if (name) {
	printk(KERN_INFO "PCI: Found %s %s MMCONFIG support.\n",
	       name, pci_mmcfg_config_num ? "with" : "without");
}

So it will always prints "without".

I'd make a patch but I'm not sure what this fix is. Just drop the test, 
the printk or make it say "which might possibly have"?

Cheers,
Scott