Re: [Qemu-devel] Aarch64: Qemu master or 2.9.0-rc1 breaks compatibility with 4.10 kernel.

2017-03-28 Thread Auger Eric 
Hi Peter,

On 28/03/2017 13:29, Peter Maydell wrote:
> On 28 March 2017 at 12:22, Prakash B  wrote:
>> Qemu master or v2.9.0-rc1 doesn't launch guest  when host kernel
>> version 4.10 or  lower on Aarch64 (cavium ThunderX), its failing with
>> abort  "qemu-system-aarch64: KVM_GET_DEVICE_ATTR failed: No such
>> device or address"
>>
>> "GICv3 live migration support"  patch breaks the  compatibility b/w
>> kernel and qemu.
>>
>> Is this expected behaviour,  can we provide a fix without  breaking
>> compatibility b/w kernel and qemu .
> 
> This is indeed not supposed to happen. Eric, can you take a look?

Yes I start the investigations.

Thanks

Eric
> 
> thanks
> -- PMM
> 
___
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

Re: Aarch64: Qemu master or 2.9.0-rc1 breaks compatibility with 4.10 kernel.

2017-03-28 Thread Peter Maydell 
On 28 March 2017 at 12:22, Prakash B  wrote:
> Qemu master or v2.9.0-rc1 doesn't launch guest  when host kernel
> version 4.10 or  lower on Aarch64 (cavium ThunderX), its failing with
> abort  "qemu-system-aarch64: KVM_GET_DEVICE_ATTR failed: No such
> device or address"
>
> "GICv3 live migration support"  patch breaks the  compatibility b/w
> kernel and qemu.
>
> Is this expected behaviour,  can we provide a fix without  breaking
> compatibility b/w kernel and qemu .

This is indeed not supposed to happen. Eric, can you take a look?

thanks
-- PMM
___
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

Aarch64: Qemu master or 2.9.0-rc1 breaks compatibility with 4.10 kernel.

2017-03-28 Thread Prakash B 
Hi ,

Qemu master or v2.9.0-rc1 doesn't launch guest  when host kernel
version 4.10 or  lower on Aarch64 (cavium ThunderX), its failing with
abort  "qemu-system-aarch64: KVM_GET_DEVICE_ATTR failed: No such
device or address"

"GICv3 live migration support"  patch breaks the  compatibility b/w
kernel and qemu.

Is this expected behaviour,  can we provide a fix without  breaking
compatibility b/w kernel and qemu .

For  more information  below is gdb back trace .
---

Thread 1 "qemu-system-aar" received signal SIGABRT, Aborted.
0xa4bd7528 in __GI_raise (sig=sig@entry=6) at
../sysdeps/unix/sysv/linux/raise.c:54
54../sysdeps/unix/sysv/linux/raise.c: No such file or directory.
(gdb) bt
#0  0xa4bd7528 in __GI_raise (sig=sig@entry=6) at
../sysdeps/unix/sysv/linux/raise.c:54
#1  0xa4bd89e0 in __GI_abort () at abort.c:89
#2  0x00479d3c in kvm_device_access (fd=23, group=6, attr=50788,
val=0x15c80d8, write=false)
at /home/ubuntu/workarea/sources/qemu-stable/qemu-2.9/kvm-all.c:2225
#3  0x004ed948 in arm_gicv3_icc_reset (env=0x13a8ff8, ri=0x1629ed0) at
/home/ubuntu/workarea/sources/qemu-stable/qemu-2.9/hw/intc/arm_gicv3_kvm.c:618
#4  0x00610414 in cp_reg_reset (key=0x1629cf0, value=0x1629ed0,
opaque=0x13a0cf0) at
/home/ubuntu/workarea/sources/qemu-stable/qemu-2.9/target/arm/cpu.c:75
#5  0xa4fb0878 in g_hash_table_foreach () from
/lib/aarch64-linux-gnu/libglib-2.0.so.0
Backtrace stopped: previous frame identical to this frame (corrupt stack?)



Verified guest  booting on host 4.10 kernel by reverting "GICv3:
support state save/restore" patches  and  guest booted without issues.

Regards,
Prakash B
___
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm