Re: [kvm-ppc-devel] [PATCH] PowerPC snapshot, integrated with upstream

[Christian E. <ehr...@li...>]

Hi Zhang Wei,
this is a known issue but avtually it should be solved in the snapshot. It is already discussed on kvm-devel because it is a common issue (ppc & s390 & ?).
You might want to look at the threads "[kvm-ppc-devel] [PATCH 1 of 2] Define and use CONFIG_KVM_HAS_PIO so that we don't	need pio_data in kvm_arch_vcpu" and the old one "[kvm-ppc-devel] [PATCH] RFC: Use CONFIG_HAS_IOPORT around vcpu->pio_data". 
So it is on the way to upstream anyway and you might just want to use Hollis last patch to that out of the first mail thread I listed above until it is upstream (this helps you however Avi eventually implements it).

-- 

Grüsse / regards, 
Christian Ehrhardt
IBM Linux Technology Center, Open Virtualization

Zhang Wei wrote:
> Hi, Hollis,
> 
> A make error is here:
> arch/powerpc/kvm/../../../virt/kvm/kvm_main.c: In function
> 'kvm_vcpu_fault':
> arch/powerpc/kvm/../../../virt/kvm/kvm_main.c:681: error: 'struct
> kvm_vcpu_arch' has no member named 'pio_data' 
> 
> I have to add pio_data to kvm_vcpu_arch structure to pass the kvm_main.c
> compile.
> 
> diff --git a/include/asm-powerpc/kvm_host.h
> b/include/asm-powerpc/kvm_host.h
> index f3bb913..ef57aec 100644
> --- a/include/asm-powerpc/kvm_host.h
> +++ b/include/asm-powerpc/kvm_host.h
> @@ -125,6 +125,8 @@ struct kvm_vcpu_arch {
> 
>  	struct timer_list dec_timer;
>  	int pending_dec;
> +
> +	void *pio_data;
>  };
> 
>  struct kvm_tlb {
> 
> Did you get that error?
> 
> Cheers!
> Wei.
>> -----Original Message-----
>> From: kvm...@li... 
>> [mailto:kvm...@li...] On 
>> Behalf Of Hollis Blanchard
>> Sent: Saturday, January 19, 2008 2:08 AM
>> To: kvm-ppc-devel
>> Subject: Re: [kvm-ppc-devel] [PATCH] PowerPC 
>> snapshot,integrated with upstream
>>
>> On Thu, 2008-01-17 at 00:33 -0600, Hollis Blanchard wrote:
>>> There are still a few things we're debugging, but a few 
>> people expressed
>>> interest so I'm sending out a snapshot now.
>>>
>>> We will be posting qemu patches in the near future as well, 
>> just as soon
>>> as we've beaten it into submission. Linux+Qemu is what doesn't work
>>> today. For now this code seems to be working fine with the 
>> testcases in
>>> the kvm-userspace/user directory.
>> OK, this is the fixed patch that can run an unmodified Linux guest. In
>> addition to some qemu changes, I had to fix "mtdcr" behavior so the PC
>> would actually advance past the instruction, and also create a TLB
>> mapping to cover the UART (which Linux expects u-boot has left around,
>> as specified in the device tree).
>>
>> We will need to work on getting qemu patches into an emailable state,
>> since the kvmctl stuff in kvm-userspace, at least in its current form,
>> can't run Linux. Since we still don't do any IO (or interrupts!), UART
>> MMIO and SDRAM controller DCR emulation is pretty much the only device
>> emulation necessary at this point.
>>
>>
>>
>> PowerPC 440 KVM implementation.
>>
>> Signed-off-by: Hollis Blanchard <ho...@us...>
>>
[...]