Re: [PATCH] NMI request/release

[Corey Minyard <minyard@acm.org>]

John Levon wrote:

>On Tue, Oct 22, 2002 at 08:02:11AM -0500, Corey Minyard wrote:
>
>>Ok.  I'd be inclined to leave the high-usage things where they are, 
>>although it would be nice to be able to make the NMI watchdog a module. 
>>oprofile should probably stay where it is.  Do you have an alternate 
>>implementation that would be more efficient?
>>    
>>
>I'm beginning to think you're right. You should ask Keith Owens if kdb
>etc. can use your API successfully.
>
Ok.  Good thought, that would decouple kdb a little.

>>>>dev_name could be removed, although it would be nice for reporting 
>>>>
>>>Reporting what ? from where ?
>>>
>>Registered NMI users in procfs.
>>    
>>
>Then if you add such code, you can add dev_name ... I hate code that
>does nothing ...
>
Ok, I'll add a procfs interface then :-).  IMHO, there's a different 
between stuff in an interface that is looking forward and dead code, 
though.  If I added it later, I would break all the users.  But there is 
a balance.

>>Yes.  But I don't understand why they would be used in the notifier code.
>>    
>>
>I'm trying to reduce code duplication - you do basically the same thing
>notifier register/unregister does.
>
Ah.  Yes, there is some stuff that looks the same but is subtly 
different.  I'll see what I can do.

>btw, the stuff you add to header files should all be in asm-i386/nmi.h
>IMHO.
>
Ok, I agree.

>It would make it clear that there's a fast-path "set nmi handler" and
>the slow one, and you can document the difference there, if that's what
>we're going to do.
>
>regards
>john
>
>  
>
Thanks,

-Corey