>> On some architectures the spl "level" is a bit mask that you can't compare
>> this way.
>I am well aware of that. I was simply suggesting to put the MD/MI boundary
>at a different level (i.e. by implementing curspl()), so that one could
>use the normal higher level assert functions.

	curspl() model assumes that there's total order defined between assert
	levels, and it is possible to detect the current spl level from the
	hardware setting.  it seems to me a strong assumption.
	for instance, if i look at i386/include/intr.h, you see splfoo()
	implemented as splraise(imask[level]), and the "level" is not recorded
	anywhere.  i think it difficult to know the current spl level
	from the cpl value (= imask[level]).

itojun