On Nov 5, 2005, at 12:25 PM, Iain Hibbert wrote:

> On Sun, 6 Nov 2005, YAMAMOTO Takashi wrote:
>
>>>> IPL_* are the appropriate values to use, I think.  IPL_xxx should
>>>> correspond directly to splxxx(), so there is no need for additional
>>>> constants that would basically be aliases of IPL_*.
>>>
>>> if that is so, then why is splraiseipl() necessary?
>>
>> because it's the cleanest way i can think of
>> to parametrize IPLs in MI subsystems.
>> please suggest an alternative, if you have any better idea.
>
> Whats wrong with splraise() for the name?  It looks to me like some  
> architectures are already using that in the same way that you are  
> proposing use of splraiseipl() - do any use it differently?

The contents of argument passed to the splraise() functions is not MI  
-- mostly, it's not always an IPL_* constant, but rather some machine- 
dependent value (like imask[IPL_BIO], or something).

>
>> From your patch, I see eg hp700 does use splraise with a mask value
> instead of the IPL value, but that could be solved easily enough by
> use of _splraise at MD level instead as seems to be done in eg evpmips
>
> ?
>
> iain

-- thorpej