Re: [Qemu-devel] [PATCH] target-ppc: ppc64 target's virtio can be either endian.

[Alexander Graf]

> Am 10.12.2013 um 18:37 schrieb Greg Kurz <address@hidden>:
> 
> On Mon, 9 Dec 2013 18:18:26 +0100
> Alexander Graf <address@hidden> wrote:
>> 
>>> On 09.12.2013, at 18:11, Greg Kurz <address@hidden> wrote:
>>> 
>>> On Mon, 9 Dec 2013 16:33:59 +0100
>>> Alexander Graf <address@hidden> wrote:
>>>> 
>>>>> On 25.11.2013, at 16:35, Greg Kurz <address@hidden> wrote:
>>>>> 
>>>>> We base it on the OS endian, as reflected by the endianness of the
>>>>> interrupt vectors (handled through the ILE bit in the LPCR register).
>>>>> 
>>>>> This patch does two things:
>>>>> - make LPCR a KVM register
>>>>> - implement virtio_get_byteswap() over LPCR
>>>>> 
>>>>> This patch requires to have the following defined in the linux
>>>>> headers:
>>>>> 
>>>>> $ grep LPCR linux-headers/asm-powerpc/kvm.h
>>>>> #define KVM_REG_PPC_LPCR        (KVM_REG_PPC | KVM_REG_SIZE_U32 |
>>>>> 0xb5)
>>>>> 
>>>>> Suggested-by: Benjamin Herrenschmidt <address@hidden>
>>>>> Signed-off-by: Rusty Russell <address@hidden>
>>>>> Signed-off-by: Greg Kurz <address@hidden>
>>>>> ---
>>>>> target-ppc/kvm.c         |    4 ++++
>>>>> target-ppc/misc_helper.c |   14 ++++++++++++++
>>>>> 2 files changed, 18 insertions(+)
>>>>> 
>>>>> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
>>>>> index 10d0cd9..b450a22 100644
>>>>> --- a/target-ppc/kvm.c
>>>>> +++ b/target-ppc/kvm.c
>>>>> @@ -869,6 +869,8 @@ int kvm_arch_put_registers(CPUState *cs, int
>>>>> level) DPRINTF("Warning: Unable to set VPA information to
>>>>> KVM\n"); }
>>>>>       }
>>>>> +
>>>>> +        kvm_put_one_spr(cs, KVM_REG_PPC_LPCR, SPR_LPCR);
>>>>> #endif /* TARGET_PPC64 */
>>>>>   }
>>>>> 
>>>>> @@ -1091,6 +1093,8 @@ int kvm_arch_get_registers(CPUState *cs)
>>>>>               DPRINTF("Warning: Unable to get VPA information from
>>>>> KVM\n"); }
>>>>>       }
>>>>> +
>>>>> +        kvm_get_one_spr(cs, KVM_REG_PPC_LPCR, SPR_LPCR);
>>>>> #endif
>>>>>   }
>>>>> 
>>>>> diff --git a/target-ppc/misc_helper.c b/target-ppc/misc_helper.c
>>>>> index 616aab6..0e0743a 100644
>>>>> --- a/target-ppc/misc_helper.c
>>>>> +++ b/target-ppc/misc_helper.c
>>>>> @@ -20,6 +20,8 @@
>>>>> #include "helper.h"
>>>>> 
>>>>> #include "helper_regs.h"
>>>>> +#include "hw/virtio/virtio.h"
>>>>> +#include "sysemu/kvm.h"
>>>>> 
>>>>> /*****************************************************************************/
>>>>> /* SPR accesses */
>>>>> @@ -116,3 +118,15 @@ void ppc_store_msr(CPUPPCState *env, target_ulong
>>>>> value) {
>>>>>   hreg_store_msr(env, value, 0);
>>>>> }
>>>>> +
>>>>> +bool virtio_get_byteswap(void)
>>>>> +{
>>>>> +    PowerPCCPU *cp = POWERPC_CPU(first_cpu);
>>>>> +    CPUPPCState *env = &cp->env;
>>>>> +
>>>>> +    if (kvm_enabled()) { 
>>>>> +        kvm_arch_get_registers(first_cpu);
>>>> 
>>>> This function is not defined when CONFIG_KVM is disabled.
>>>> 
>>>> Please do
>>>> 
>>>>   cpu_synchronize_state(first_cpu);
>>>> 
>>>> instead.
>>> 
>>> Oups... I'll fix that.
>>> 
>>>> Also can't virtio_get_byteswap pass in the CPU pointer of the CPU
>>>> that's calling in this moment? I'm not sure how racy it is to
>>>> synchronize the first cpu while we're not in the first cpu's execution
>>>> thread.
>>> 
>>> I kept the choices made by Rusty in the original serie. According to
>>> these, I am not sure if we can use current_cpu:
>>> 
>>> https://lists.gnu.org/archive/html/qemu-devel/2013-08/msg01156.html
>>> https://lists.gnu.org/archive/html/qemu-devel/2013-08/msg01504.html
>>> 
>>>> Either way, if we do this "right" we don't even have to jump through
>>>> these hoops, as little endian setting simply happens steered from QEMU,
>>>> so QEMU will have all knowledge about the guest's little endian mode
>>>> without the need to synchronize any state.
>>> 
>>> Sure. This is just primary work to get virtio working for cross-endian
>>> cases. :)
>> 
>> Yeah, but so far LE switching code is not upstream in KVM - and for TCG
>> we don't need the cpu_synchronize_state() call at all.
>> 
>> 
>> Alex
> 
> Sure, but for TCG, cpu_synchronize_state() does nothing since KVM is
> not enabled... what would be wrong with that ?

That it could potentially break kvm.

Alex

> 
> Cheers.
> 
> --
> Greg
>