Re: [Qemu-ppc] system-ppcemb -M bamboo, troubles with RTC

[Alexander Graf]

On 07/02/2013 06:26 AM, nello martuscielli wrote:
> On Mon, Jul 1, 2013 at 11:22 AM, Alexander Graf<address@hidden>  wrote:
>
> _cut__
> > > [...]
> > > KVM: invalid SPR write: 1013
> > > mtspr: unknown spr 3f5
> > > Couldn't emulate instruction 0x7c75fba6 (op 31 xop 467)
> > > kvmppc_handle_exit: emulation at 700 failed (7c75fba6)
> > >
> > > in my host dmesg and a kernel panic booting my guest.
> > Yes, that is also expected. KVM is only able to do very limited 
> > cross-virtualization. You can run a G3/G4/Gekko guest on a G5/P7. But that's 
> > about as far as it goes. 440 is a completely different specification, rendering 
> > it vastly incompatible with your G4. They are only mostly compatible from a 
> > user space perspective.
> >
> > So if you want to use KVM on your G4, I'd be more than happy to see you doing 
> > so, but please use -M g3beige or -M mac99 with -cpu G3 or G4 :).
> well, it happens also with host (POWER4+, Linux-3.4.51) and (G4, Linux-3.10).
> As i know POWER4+ is a full book3s specs compliant.

Yes, for the most part a POWER4+ system behaves very similar to a G5. 
The only thing that won't work on that one (yet) is -M pseries, as 
you're most likely running on pHyp which filters hypercalls from us.

> Instead that's from host (G4,linux-3.10) and guest (mac99,
> Linux-3.9.7) when i use -enable-kvm:
>
> from qemu-1.5.1 console:
> [...]
> Please wait, loading kernel...
> Amiga partition table corrupted at block 0
> block type is not partition but 00000000
>     Elf32 kernel loaded...
> zImage starting: loaded at 0x00600000 (sp: 0x0fde7a80)
> Allocating 0x636e58 bytes for kernel ...
> OF version = 'OpenFirmware 3'
> Trying to claim from 0x600000 to 0x8a4c70 (0x2a4c70) got ffffffff
> gunzipping (0x00a00000<- 0x00607000:0x00898237)...done 0x51b260 bytes
> Linux/PowerPC load: root=/dev/sda1 ro console=ttyPZ0,57600
> Finalizing device tree... using OF tree (promptr=fff025a4)
> OF stdout device is: 
> /address@hidden/address@hidden/address@hidden/address@hidden
> Preparing to boot Linux version 3.9.7-mac99 (address@hidden) (gcc version
> 4.7.3 (CRUX PPC) ) #3 Mon Jul 1 06:32:42 CEST 2013
> Detected machine type: 00000400
> command line:
> memory layout at init:
>    memory_limit : 00000000 (16 MB aligned)
>    alloc_bottom : 0103b000
>    alloc_top    : 10000000
>    alloc_top_hi : 10000000
>    rmo_top      : 10000000
>    ram_top      : 10000000
> found display   : /address@hidden/QEMU,address@hidden, opening... done
> copying OF device tree...
> Building dt strings...
> Building dt structure...
> Device tree strings 0x0103c000 ->  0x0103c518
> Device tree struct  0x0103d000 ->  0x0103f000
> Calling quiesce...
> returning from prom_init
> console [ttyPZ0] enabled, bootconsole disabled
> pid_max: default: 32768 minimum: 301
> Mount-cache hash table entries: 512
> Oops: Exception in kernel mode, sig: 4 [#1]
> PowerMac
> Modules linked in:
> NIP: 00000700 LR: c00077a4 CTR: c0048f90
> REGS: c050fea0 TRAP: 0700   Not tainted  (3.9.7-mac99)
> MSR: 00040000<>   CR: 22000022  XER: 00000000
> TASK = c04e43c0[0] 'swapper' THREAD: c050e000
> GPR00: c03c5788 c050ff50 c04e43c0 00000000 00000000 00000001 00000000 cf840638
> GPR08: 00000000 00000000 00000000 00000000 22000024 00000000 0fde7ac8 00000002
> GPR16: 00000000 00000000 0fc5c91c c04e4518 c0520000 c04e9304 c03d0000 c03d0000
> GPR24: 00000001 00000004 c0510000 c050e000 cf840600 c04ea30c c04e8f20 cf840600
> NIP [00000700] 0x700
> LR [c00077a4] __switch_to+0xa8/0xb4
> Call Trace:
> [c050ff50] [c0520000] 0xc0520000 (unreliable)
> [c050ff60] [c03c5788] __schedule+0x2c8/0x34c
> [c050ffa0] [c03c5a8c] schedule_preempt_disabled+0x10/0x20
> [c050ffb0] [c00045d8] rest_init+0x60/0x70
> [c050ffc0] [c04b3978] start_kernel+0x280/0x288
> [c050fff0] [00003444] 0x3444
> Instruction dump:
> 00000000 XXXXXXXX XXXXXXXX XXXXXXXX 00000000 XXXXXXXX XXXXXXXX XXXXXXXX
> 00000000 XXXXXXXX XXXXXXXX XXXXXXXX 7d5043a6 XXXXXXXX XXXXXXXX XXXXXXXX
> ---[ end trace 31fd0ba7d8756001 ]---
> Kernel panic - not syncing: Attempted to kill the idle task!
>
> and from host dmesg:
> [...]
> [  470.988993] KVM: invalid SPR write: 1013
> [  470.989004] mtspr: unknown spr 0x3f5
> [  470.989009] Couldn't emulate instruction 0x7c75fba6 (op 31 xop 467)
> [  470.990466] kvmppc_handle_exit: emulation at 700 failed (7c75fba6)

Oh, the guest kernel is trying to access DABR. Great. Does the following 
host patch help?


Alex

diff --git a/arch/powerpc/kvm/book3s_emulate.c 
b/arch/powerpc/kvm/book3s_emulate.c
index 1f6344c..360ce68 100644
--- a/arch/powerpc/kvm/book3s_emulate.c
+++ b/arch/powerpc/kvm/book3s_emulate.c
@@ -458,6 +458,7 @@ int kvmppc_core_emulate_mtspr(struct kvm_vcpu *vcpu, 
int sprn, ulong spr_val)
        case SPRN_PMC4_GEKKO:
        case SPRN_WPAR_GEKKO:
        case SPRN_MSSSR0:
+       case SPRN_DABR:
                break;
 unprivileged:
        default:
@@ -555,6 +556,7 @@ int kvmppc_core_emulate_mfspr(struct kvm_vcpu *vcpu, 
int sprn, ulong *spr_val)
        case SPRN_PMC4_GEKKO:
        case SPRN_WPAR_GEKKO:
        case SPRN_MSSSR0:
+       case SPRN_DABR:
                *spr_val = 0;
                break;
        default: