Re: [Qemu-devel] [PATCH 2/3] hd-geometry.c/s390: Disable geometry translation

[Alexander Graf]

On 19.11.2012, at 11:59, Christian Borntraeger wrote:

> On 18/11/12 20:14, Blue Swirl wrote:
>> On Sun, Nov 18, 2012 at 4:10 PM, Paolo Bonzini <address@hidden> wrote:
>>> Il 17/11/2012 17:07, Blue Swirl ha scritto:
>>>> On Tue, Nov 13, 2012 at 8:50 AM, Christian Borntraeger
>>>> <address@hidden> wrote:
>>>>> From: Einar Lueck <address@hidden>
>>>>> 
>>>>> This patch disables the translation of geometry information read from
>>>>> disks on s390. On s390 such translations lead to wrong geometries being
>>>>> advertized to the guest because there is no entity doing these kinds
>>>>> of translation on this architecture.
>>>>> 
>>>>> Signed-off-by Einar Lueck <address@hidden>
>>>>> Signed-off-by Christian Borntraeger <address@hidden>
>>>>> ---
>>>>> hw/hd-geometry.c |    5 +++++
>>>>> 1 file changed, 5 insertions(+)
>>>>> 
>>>>> diff --git a/hw/hd-geometry.c b/hw/hd-geometry.c
>>>>> index 4cf040d..db1dc81 100644
>>>>> --- a/hw/hd-geometry.c
>>>>> +++ b/hw/hd-geometry.c
>>>>> @@ -144,6 +144,10 @@ static int guess_disk_msdosgeo(BlockDriverState *bs,
>>>>>                 continue;
>>>>>             }
>>>>> 
>>>>> +#ifdef __s390__
>>>> 
>>>> No, this would make all system emulators (e.g. x86) on s390 host to
>>>> use this translation, not just s390. I think you want to use #ifdef
>>>> CONFIG_S390X instead.
>>> 
>>> This symbol is not available because this file is compiled just once.
>>> Which non-x86 targets actually care about translation stuff?
>> 
>> I meant TARGET_S390X, sorry. I have no idea, but hosts and targets
>> should not be confused.
> 
> Would a modified patch with TARGET_S390X be acceptable?

If TARGET_S390X is defined in that file (read: it's compiled individually for 
every target), yes. Otherwise please make this a runtime check and set the 
runtime checked variable in code that is target specific.


Alex