On Jun 7, 2005, at 10:27 AM, Martin Husemann wrote:

> On Tue, Jun 07, 2005 at 01:19:37PM -0400, Thor Lancelot Simon wrote:
>
>> Is this correct?
>>
>
> 20      STD MPSAFE      { pid_t sys_getpid(void); }
>
> Can we do MP safe vnlocks already? That would be needed for some of  
> the
> others, I guess.

Nothing in the top half that takes locks can safely run outside of  
the kernel_lock perimeter until everything in the bottom half runs  
outside of the kernel_lock perimeter.

Bill Sommerfeld and I came to this conclusion a long time ago when we  
were both doing a lot of work on NetBSD's SMP support.

-- thorpej