RFC: Use CONFIG_HAS_IOPORT around vcpu->pio_data

Signed-off-by Hollis Blanchard <hollisb-r/[email protected]>

---
pio_data is obviously arch-specific, but we might be able to get by with an ifdef in kvm_vcpu_nopage() rather than needing an arch hook. Thoughts?

1 file changed, 2 insertions(+)

drivers/kvm/kvm_main.c | 2 ++

drivers/kvm/kvm_main.c +2 -1

@@ -674,8 +674,10 @@ static struct page *kvm_vcpu_nopage(stru
 	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
 	if (pgoff == 0)
 		page = virt_to_page(vcpu->run);
+#ifdef CONFIG_HAS_IOPORT
 	else if (pgoff == KVM_PIO_PAGE_OFFSET)
 		page = virt_to_page(vcpu->pio_data);
+#endif /* CONFIG_HAS_IOPORT */
 	else
 		return NOPAGE_SIGBUS;
 	get_page(page);
 
-------------------------------------------------------------------------

SF.Net email is sponsored by: The Future of Linux Business White Paper from Novell. From the desktop to the data center, Linux is going mainstream. Let it simplify your IT future.
http://altfarm.mediaplex.com/ad/ck/8857-50307-18918-4
_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

Actually, (now that I try to build with it :) this doesn't work for PowerPC, which defines (or rather, doesn't not define) CONFIG_HAS_IOPORT.

Quoted 5 lines

Signed-off-by Hollis Blanchard <hollisb-r/[email protected]> ---

pio_data is obviously arch-specific, but we might be able to get by with an ifdef in kvm_vcpu_nopage() rather than needing an arch hook. Thoughts?

drivers/kvm/kvm_main.c --- static struct page *kvm_vcpu_nopage(stru · line 674

Quoted 8 lines

1 file changed, 2 insertions(+)

drivers/kvm/kvm_main.c | 2 ++

drivers/kvm/kvm_main.c --- renamed from drivers/kvm/kvm_main.c +0 -0

a/drivers/kvm/kvm_main.c +++ b/drivers/kvm/kvm_main.c

drivers/kvm/kvm_main.c --- renamed from drivers/kvm/kvm_main.c +0 -0

@@ -674,8 +674,10 @@ static struct page *kvm_vcpu_nopage(stru
 	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +

drivers/kvm/kvm_main.c --- static struct page *kvm_vcpu_nopage(stru · line 683

Quoted 10 lines

drivers/kvm/kvm_main.c --- renamed from drivers/kvm/kvm_main.c +2 -0

 		if (pgoff == 0) page = virt_to_page(vcpu->run);
+#ifdef CONFIG_HAS_IOPORT
 	else if (pgoff == KVM_PIO_PAGE_OFFSET)
 		page = virt_to_page(vcpu->pio_data);
+#endif /* CONFIG_HAS_IOPORT */
 	else
 		return NOPAGE_SIGBUS;
 	get_page(page);
 

Quoted 8 lines

SF.Net email is sponsored by: The Future of Linux Business White Paper from Novell. From the desktop to the data center, Linux is going mainstream. Let it simplify your IT future.
http://altfarm.mediaplex.com/ad/ck/8857-50307-18918-4
_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

SF.Net email is sponsored by: The Future of Linux Business White Paper from Novell. From the desktop to the data center, Linux is going mainstream. Let it simplify your IT future.
http://altfarm.mediaplex.com/ad/ck/8857-50307-18918-4
_______________________________________________
kvm-ppc-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-ppc-devel

Quoted 3 lines · Hollis Blanchard

Actually, (now that I try to build with it :) this doesn't work for PowerPC, which defines (or rather, doesn't not define) CONFIG_HAS_IOPORT.

You mean ppc has ioport , but you don't use current kvm io infrastructure to handle it ?

Quoted 7 lines · Xiantao

Quoted 3 lines · Hollis Blanchard

Actually, (now that I try to build with it :) this doesn't work for PowerPC, which defines (or rather, doesn't not define) CONFIG_HAS_IOPORT.

You mean ppc has ioport , but you don't use current kvm io infrastructure to handle it ?

It's a matter of perspective. :)

The PowerPC architecture has no "IO" instructions; therefore all IO accesses are MMIO.
However, you can place devices with IO ports in a PowerPC system. In this case, the bridges are configured to translate MMIO accesses coming from the processor into IO port accesses on the bus. It is quite common to find standard x86 ISA devices, such as a SuperIO

In Linux, CONFIG_HAS_IOPORT doesn't really mean "does the architecture use IO ports," which is what I thought it meant. Instead, it means "should we track IO resources as well as MMIO resources," and that's true on PowerPC because we need manage the ISA bus as mentioned above. (The only current user of the symbol is lib/devres.c.)

In summary, it's an inappropriate ifdef for this situation.

Quoted 3 lines · Hollis Blanchard

Quoted 2 lines · Xiantao

Quoted 1 line · Hollis Blanchard

Actually, (now that I try to build with it :) this doesn't work

for

Quoted 6 lines

Quoted 5 lines

Quoted 2 lines

PowerPC, which defines (or rather, doesn't not define) CONFIG_HAS_IOPORT.

You mean ppc has ioport , but you don't use current kvm io infrastructure to handle it ?

Quoted 8 lines

It's a matter of perspective. :)

The PowerPC architecture has no "IO" instructions; therefore all IO accesses are MMIO.
However, you can place devices with IO ports in a PowerPC system. In this case, the bridges are configured to translate MMIO accesses coming from the processor into IO port accesses on the bus. It is quite common to find standard x86 ISA devices, such as a SuperIO

Thank you for your elaboration.
So, bridge emulation in qemu will take over the responsibility of the translation for ppc?

It's a matter of perspective. :)

The PowerPC architecture has no "IO" instructions; therefore all IO accesses are MMIO.

However, you can place devices with IO ports in a PowerPC system. In this case, the bridges are configured to translate MMIO accesses coming from the processor into IO port accesses on the bus. It is quite common to find standard x86 ISA devices, such as a SuperIO chip, in PowerPC systems.

Actually, (now that I try to build with it :) this doesn't work for PowerPC, which defines (or rather, doesn't not define) CONFIG_HAS_IOPORT.

So I guess a callout to arch code will do. I'll send a patch sometime.