From mboxrd@z Thu Jan  1 00:00:00 1970
From: Christian Zigotzky <chzigotzky@xenosoft.de>
Date: Fri, 30 May 2014 15:44:06 +0000
Subject: Does KVM support the P.A. Semi PA6T cpu?
Message-Id: <5388A746.6090307@xenosoft.de>
List-Id: <kvm-ppc.vger.kernel.org>
References: <53511C7B.5010208@xenosoft.de>
In-Reply-To: <53511C7B.5010208@xenosoft.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="windows-1252"
Content-Transfer-Encoding: quoted-printable
To: kvm-ppc@vger.kernel.org

On 30.05.2014 14:47, Alexander Graf wrote:
>
> On 30.05.14 14:42, Christian Zigotzky wrote:
>> Am 30.05.14 14:21, schrieb Alexander Graf:
>>>
>>> On 30.05.14 13:39, Christian Zigotzky wrote:
>>>> Hi Alex,
>>>>
>>>> Thank you for your answer. Here are the traces:
>>>>
>>>> http://www.xenosoft.de/trace_mol_mac_os_x_tiger.txt.tar.gz
>>>> http://www.xenosoft.de/trace_mol_mac_os_x_jaguar.txt.tar.gz
>>>
>>> Looking at the Jaguar log, the guest seems to be stuck trying to map=20
>>> a page:
>>>
>>> mol-4087 [001] .... 252.368002: kvm_exit: exit=DATA_STORAGE |=20
>>> pc=3D0x900074c4 | msr=3D0x4000d030 | dar=3D0xa000a434 |=20
>>> srr1=3D0x100000000000d032 | last_inst=3D0x912b0
>>> 000
>>>
>>> If you look at the trace, you will see that this PC with the exact=20
>>> same DAR happens over and over again - every time the guest thinks=20
>>> it's mapped the page.
>>>
>>> I don't know why that's happening. Try to enable the debug prints=20
>>> (or convert them to trace points ;)) in the book3s_32 guest and=20
>>> book3s_64 host emulation. Maybe that tells us something.
>>>
>>>
>>> Alex
>> Thanks Alex for your effort. How can I enable the debug prints?
>
> This should do the trick. Output will be in dmesg:
>
> diff --git a/arch/powerpc/kvm/book3s_32_mmu.c=20
> b/arch/powerpc/kvm/book3s_32_mmu.c
> index 93503bb..939b14f 100644
> --- a/arch/powerpc/kvm/book3s_32_mmu.c
> +++ b/arch/powerpc/kvm/book3s_32_mmu.c
> @@ -27,8 +27,8 @@
> #include <asm/kvm_ppc.h>
> #include <asm/kvm_book3s.h>
>
> -/* #define DEBUG_MMU */
> -/* #define DEBUG_MMU_PTE */
> +#define DEBUG_MMU
> +#define DEBUG_MMU_PTE
> /* #define DEBUG_MMU_PTE_IP 0xfff14c40 */
>
> #ifdef DEBUG_MMU
Hi Alex,

Unfortunately, the KVM modul doesn't compile with DEBUG_MMU.

Kernel 3.14.4:

make modules
CHK include/config/kernel.release
CHK include/generated/uapi/linux/version.h
CHK include/generated/utsrelease.h
CALL scripts/checksyscalls.sh
AS [M] arch/powerpc/kvm/fpu.o
CC [M] arch/powerpc/kvm/book3s_paired_singles.o
CC [M] arch/powerpc/kvm/book3s_pr.o
CC [M] arch/powerpc/kvm/book3s_pr_papr.o
CC [M] arch/powerpc/kvm/book3s_emulate.o
AS [M] arch/powerpc/kvm/book3s_interrupts.o
CC [M] arch/powerpc/kvm/book3s_mmu_hpte.o
CC [M] arch/powerpc/kvm/book3s_64_mmu_host.o
CC [M] arch/powerpc/kvm/book3s_64_mmu.o
CC [M] arch/powerpc/kvm/book3s_32_mmu.o
arch/powerpc/kvm/book3s_32_mmu.c: In function=20
=91kvmppc_mmu_book3s_32_get_pteg=92:
arch/powerpc/kvm/book3s_32_mmu.c:133:2: error: =91struct=20
kvmppc_vcpu_book3s=92 has no member named =91vcpu=92
make[1]: *** [arch/powerpc/kvm/book3s_32_mmu.o] Fehler 1
make: *** [arch/powerpc/kvm] Fehler 2

>
>
> There's also always the chance that split real mode is the real=20
> culprit for breakage here. Have you verified that the same kernel=20
> source (KVM code) works on a PPC32 host with MOL?

My PPC32 host (Sam440ep-Flex 733MHz) is too slow for MOL.

Rgds,

Christian