Re: [Qemu-ppc] [PATCH] openpic: Added BRR1 register

[Scott Wood]

On 07/16/2012 03:09 PM, Alexander Graf wrote:
> 
> On 16.07.2012, at 21:32, Scott Wood wrote:
> 
>> On 07/16/2012 02:29 PM, Yoder Stuart-B08248 wrote:
>>>
>>>
>>>> -----Original Message-----
>>>> From: address@hidden [mailto:address@hidden On Behalf Of Scott Wood
>>>> Sent: Monday, July 16, 2012 12:12 PM
>>>> To: Bhushan Bharat-R65777
>>>> Cc: Wood Scott-B07421; Alexander Graf; address@hidden; address@hidden
>>>> Subject: Re: [PATCH] openpic: Added BRR1 register
>>>>
>>>> On 07/16/2012 12:09 PM, Bhushan Bharat-R65777 wrote:
>>>>>
>>>>>
>>>>>> -----Original Message-----
>>>>>> From: Wood Scott-B07421
>>>>>> Sent: Monday, July 16, 2012 10:34 PM
>>>>>> To: Bhushan Bharat-R65777
>>>>>> Cc: Alexander Graf; address@hidden; address@hidden
>>>>>> Subject: Re: [PATCH] openpic: Added BRR1 register
>>>>>>
>>>>>> On 07/16/2012 11:21 AM, Bhushan Bharat-R65777 wrote:
>>>>>>>>> +    case 0x00: /* BRR1 */
>>>>>>>>> +        retval = 0x00400200;
>>>>>>>>
>>>>>>>> Please unmagicify this one :)
>>>>>>>
>>>>>>> /* BRR1 ( Block revision register ) */
>>>>>>>
>>>>>>> #define IPID 0x00400000 /* IP-block ID */ #define IPMJ 0x00000200 /*
>>>>>>> IP major number */ #define IPMN 0x00000200 /* IP minor number */
>>>>>>
>>>>>> IPMN looks wrong.
>>>>>
>>>>> Opps : copy paste error here :)
>>>>>
>>>>> Btw, I am not aware of all MPIC IP versions, differences in them and what 
>>>>> is the latest/best version
>>>> this emulated code supports. I drive this value from mpc8544 Reference 
>>>> Manual.
>>>>>
>>>>> I requested Alex to suggest the most updated version.
>>>>
>>>> QEMU does not have a complete implementation of any version of the FSL 
>>>> MPIC.
>>>
>>> So what revision numbers should we use?
>>
>> How about 0.0 for now?
> 
> Would guests care (read: break)?

I can't answer that question for "guests" in general.  They could just
as well break due to the missing functionality.

Linux currently does not look at the version information, though there
are a couple patches posted that do depend on it to check for certain
features.  If it sees 0.0, it won't use features that aren't there.

One thing that might break if we claim 2.0 is that there was a patch
posted over the weekend that assumes an MSI erratum on version 2.0.
This would break if we're directly assigning a non-broken MSI block to
the guest, since the rest of the MPIC is still emulated.  That said, it
wouldn't be QEMU's fault.  It's a combination that doesn't exist in real
hardware and it shouldn't be surprising that the guest might need some
fixing to accommodate it.

-Scott