i386 cpu_reset_real: code/comment mismatch

[John Baldwin]

On Tuesday 20 May 2008 09:37:54 am Andriy Gapon wrote:
> on 20/05/2008 00:51 John Baldwin said the following:
> > So, the comment is correct and not the code.  Curiously enough, 
OpenSolaris
> > does the same thing (it writes 0x2 followed by 0x6), but it has some sort 
of 
> > comment which implies that you have to do a write to set or clear bit 1 
> > before setting bit 2.  Linux only uses 0xcf9 on a specific x86 machine 
(View 
> > workstation or some such) in which case it just does a single write of 
0x6.  
> > I'll test locally to make sure 0x4 and 0x6 both work and if so I will 
commit 
> > the fix.
> 
> And this code is most likely never reached in majority of the cases,
> reset via keyboard controller should just work.

Except on boxes where it doesn't (and hence why I added it).

> BTW, I understand that there is a difference between hard and soft reset
> in terms of hardware signals being asserted, but I don't quite
> understand general consequences. I.e. what is a practical difference
> between hard and soft reset?

I've no idea.  It may be that we should just always do a hard reset as that is 
in effect what both Solaris and Linux do.  I'll probably just change it to do 
that.

-- 
John Baldwin