WARNING - OLD ARCHIVES

This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-devel

RE: [Xen-devel] Re: [PATCH][VTD] pci mmcfg patch for x86-64 -version 3

>>> "Kay, Allen M" <allen.m.kay@xxxxxxxxx> 12.12.08 03:47 >>>
>>>+/* used by mmcfg */
>>>+static inline unsigned char mmio_config_readb(void __iomem *pos)
>>>+{
>>>...
>>
>>This is preceded by a rather important comment regarding AMD Fam10
>>CPUs in Linux. Without that comment, no-one will easily understand why
>>you must use eax/ax/al here. I'm also surprised you didn't copy over
>>pci_mmcfg_amd_fam10h() from Linux...
>>
>
>OK, I will add the comments back in.  Where is pci_mmcfg_amd_family10h() 
>defined?  I'm having trouble finding it.

The name really is pci_mmcfg_amd_fam10h(), and it's in 2.6.27's
arch/x86/pci/mmconfig-shared.c, right along with the other host bridge
probe functions.

>>Oh, and one more thing I remembered just now: Linux verifies the mmcfg
>>values it gets from ACPI against the E820 map - shouldn't Xen 
>>do this too?
>
>My original intent was to keep mmconfig-shared.c and mmconfig_64.c as closely 
>to Linux as possible.  However, I soon found out
>including everything in mmconfig-shared.c involved pulling in a lot of code 
>from Linux.

Understandable. But this particular check may save us from dying on systems
where Linux works. And it shouldn't be that much code that needs adding
here if you restrict this to the E820 check; I'm not certain how significant
the ACPI check is (which would require Dom0 assistance in my opinion, as
it requires the ACPI interpreter to be available).

>As my main goal of this round is enabling ATS, I tried to limited the scope of 
>mmconfig work to be somewhat manageable for now
>and then add more stuff to it as needed.  As I'm planning to work on 
>multi-segment PCI support in Q1, this stuff will get revisited
>again.

Is that going to come through Linux again, or is this a Xen-only plan?

>Other than E820 checking, do you see anything else in mmconfig-shared.c need 
>to be included for this round?

I didn't notice anything.

Jan


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