Re: [PATCH 2/2] xen/ppc: Drop support for pseries/OpenFirmware

[Shawn Anastasio <sanastasio@xxxxxxxxxxxxxxxxxxxxx>]

On 9/6/23 1:44 AM, Jan Beulich wrote:
> On 05.09.2023 23:46, Shawn Anastasio wrote:
>> Since QEMU's PowerNV support has matured to the point where it is
>> now suitable for development, drop support for booting on the
>> paravirtualized pseries machine type and its associated interfaces.
> 
> I think you want to mention that the pseries functionality was actually
> broken by the earlier change, both verbally and by means of a Fixes:
> tag. The breakage actually may also want mentioning in patch 1 as well
> as the cover letter.
>

Will do. I'll add mentions to the breakage in both patches and add a
Fixes: tag to this patch specifically.

>> --- a/xen/arch/ppc/setup.c
>> +++ b/xen/arch/ppc/setup.c
>> @@ -14,17 +14,12 @@ void __init noreturn start_xen(unsigned long r3, 
>> unsigned long r4,
>>  {
>>      if ( r5 )
>>      {
>> -        /* OpenFirmware boot protocol */
>> -        boot_of_init(r5);
>> +        /* Unsupported OpenFirmware boot protocol */
>> +        __builtin_trap();
>>      }
>>      else
>>      {
>> -        /*
>> -         * kexec boot protocol
>> -         *
>> -         * TODO: This currently assumes an OPAL/PowerNV system, but it's 
>> also
>> -         * possible to be kexec'd on an OF system.
>> -         */
>> +        /* kexec boot protocol */
>>          boot_opal_init((void *)r3);
> 
> At least part of the comment may want retaining, as the code only handles
> the OPAL case?
>

Since we're now explicitly only supporting Opal/PowerNV I don't think
there's any need to point out that this code path doesn't properly
handle the pseries+kexec case. boot_opal_init will gracefully
fail under that circumstance anyway when it tries to find an /ibm,opal
device tree node.

> Jan

Thanks,
Shawn