Re: [patch] Require _KERNEL/_KERNEL_STRUCTURES for all kernel headers

Chris Pressey wrote:
> On Tue, 21 Dec 2004 23:54:08 +0100
> Joerg Sonnenberger <[e-mail]> wrote:
> 
> 
>>On Tue, Dec 21, 2004 at 12:38:27PM -0800, Chris Pressey wrote:
>>
>>>I just wanted to pass it over submit@ for one last review.  If there are
>>>no objections in the next few days, I'll commit it:
>>>
>>>  http://catseye.webhop.net/DragonFlyBSD/patch/kernel_barricade.diff
>>
>>- sys/device_port.h already included
>>- in bus/cam/scsi_da.c, can you merge the _KERNEL includes and the !_KERNEL
>>  includes?
>>- in sys/lockf, the functions should be in one group and the variable in
>>  another
>>- in power.h, don't add _KERNEL, it doesn't make that much sense
>>- sys/systimer.h is _KERNEL only, if world break, we should fix it
>>- sys/taskqueue.h should stay kernel only
>>- sys/timepps.h can move the include for sys/systimer.h into the _KERNEL
>>  section?
>>- sys/tprintf.h should stay _KERNEL only 
>>- sys/xwait.h can be removed, if Matt doesn't want to resurrect it. It's
>>  only used by kern/kern_sync.c and there only by #if 0'd code.
>>
>>Joerg
>>
> 
> 
> Hope I caught all that, updated patch at
> 
>   http://catseye.webhop.net/DragonFlyBSD/patch/kernel_barricade2.diff
> 
> 'make installer_release' underway, I'll let you know if it breaks.
> 
> -Chris
> 

	Correct me if I am wrong, but if a sysutil port requires one of
	the sys headers, does that mean I need to define BOTH _KERNEL
	and _KERNEL_STRUCTURES?  I thought the point of having
	_KERNEL_STRUCTURES was to compartmentalize *kernel structures*
	and other dangerous code in header files from userland so
	that you could define _KERNEL_STRUCTURES and get what it says
	instead of the other cruft as well.

	I could be just beating a dead chicken here... :-)

		-Hiten