Re: [Qemu-devel] [PATCH 1/2] pc: add pci64 memory hole

2012-05-02 Thread Gerd Hoffmann
On 05/02/12 17:31, Avi Kivity wrote:
 On 05/02/2012 05:02 PM, Gerd Hoffmann wrote:
 This patch adds a address space hole for 64bit PCI ressources.
 It starts at 0x80 (512 GB) and ends at 0x100 (1 TB),
 thus has 512 GB in size.  This matches what the seabios is doing
 (latest master branch).
 
 We should communicate this to seabios via fwcfg

A dsdt entry is needed for the pci64 window (see
http://code.coreboot.org/p/seabios/source/commit/482a020ec25f4cec655ddcb16b67c6f38b0844c0/),
which makes it non-trivial to turn this into a runtime option ...

 It should also only apply to -M old.

Is this the only reason you want this be runtime-switchable?

cheers,
  Gerd




Re: [Qemu-devel] [PATCH 1/2] pc: add pci64 memory hole

2012-05-02 Thread Avi Kivity
On 05/02/2012 06:46 PM, Gerd Hoffmann wrote:
 On 05/02/12 17:31, Avi Kivity wrote:
  On 05/02/2012 05:02 PM, Gerd Hoffmann wrote:
  This patch adds a address space hole for 64bit PCI ressources.
  It starts at 0x80 (512 GB) and ends at 0x100 (1 TB),
  thus has 512 GB in size.  This matches what the seabios is doing
  (latest master branch).
  
  We should communicate this to seabios via fwcfg

 A dsdt entry is needed for the pci64 window (see
 http://code.coreboot.org/p/seabios/source/commit/482a020ec25f4cec655ddcb16b67c6f38b0844c0/),
 which makes it non-trivial to turn this into a runtime option ...

It's a function, isn't it?  So all you need is a ASL driver for fwcfg
and an If().

  It should also only apply to -M old.

 Is this the only reason you want this be runtime-switchable?

No.  The BIOS and qemu shouldn't be making too many assumptions. 
Certainly for Q35 we'd have the bios program qemu for the 64 bit hole.

-- 
error compiling committee.c: too many arguments to function




Re: [Qemu-devel] [PATCH 1/2] pc: add pci64 memory hole

2012-05-02 Thread Avi Kivity
On 05/02/2012 05:02 PM, Gerd Hoffmann wrote:
 This patch adds a address space hole for 64bit PCI ressources.
 It starts at 0x80 (512 GB) and ends at 0x100 (1 TB),
 thus has 512 GB in size.  This matches what the seabios is doing
 (latest master branch).

We should communicate this to seabios via fwcfg (really seabios should
communicate this to the chipset, but our chipset doesn't support this at
all).

It should also only apply to -M old.

-- 
error compiling committee.c: too many arguments to function