[Skiboot] [EXTERNAL] [PATCH v6 13/29] xive: make endian-clean

[Nicholas Piggin]

Cédric Le Goater's on November 7, 2019 5:38 pm:
> On 07/11/2019 07:32, Nicholas Piggin wrote:
>> Nicholas Piggin's on November 7, 2019 4:26 pm:
>>> Cédric Le Goater's on November 7, 2019 1:38 am:
>>>> On 06/11/2019 13:10, Nicholas Piggin wrote:
>>>>> Convert xive opal calls, dt construction, and in-memory hardware tables
>>>>> to use explicit endian conversions.
>>>>
>>>> Thanks for reworking the patch with helper routines. This is a huge 
>>>> work. You are setting the standards pretty high for xive2 !
>>>
>>> Well once I took your nice helpers then it's only mindless conversions.
>>> I thought I'd better do it rather than pull you away from useful work :)
>>>
>>>> Did you run a KVM guest ? 
>>>
>>> I just pulled the latest qemu and tried. Had some bugs in the
>>> conversion, so here is an updated patch. I'll not re-send the series
>>> until you get a chance to ack it, at least.
>> 
>> Oh, one other thing while I was fixing one of my bugs, what do you
>> think of this?
>> 
>>     xive: improve cache watch type checking
>>     
>>     The cache update functions take void * pointers and repeat the same
>>     parameters. Make these more specific by adding another function, and
>>     so the data pointer can be typed, and some arguments can be inferred.
> 
> This is good. I tripped over the same problem on QEMU side when modeling
> the cache updates.
> 
> But, I would prefer if it came as cleanup on the current code before adding 
> the BE helpers.

I was going to put it after, there is no conflict with the endian
fixes, but no strong preference.

The series is already too big really so I don't like to make a series
of xive patches within this series. I don't mind if you take the xive
bits into your tree, but I'd like to get LE working as soon as possible
because there's more dependencies on it I'd like to send. I'll see
what Oliver is thinking.

Thanks,
Nick