[PATCH] x86: simplify HIGHMEM-related Kconfig entries

[PATCH] x86: simplify HIGHMEM-related Kconfig entries

[Jan Beulich]

Additionally remove the dependency of X86_PAE on !HIGHMEM4G - at least
I can't understand why that dependency existed.

Signed-off-by: Jan Beulich <jbeulich@novell.com>

---
 arch/x86/Kconfig |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

--- linux-2.6.29-rc1/arch/x86/Kconfig	2009-01-14 11:36:05.000000000 +0100
+++ 2.6.29-rc1-x86-kconfig-highmem/arch/x86/Kconfig	2009-01-08 09:03:21.000000000 +0100
@@ -896,8 +896,8 @@ config X86_CPUID
 
 choice
 	prompt "High Memory Support"
-	default HIGHMEM4G if !X86_NUMAQ
 	default HIGHMEM64G if X86_NUMAQ
+	default HIGHMEM4G
 	depends on X86_32
 
 config NOHIGHMEM
@@ -1004,7 +1004,7 @@ config HIGHMEM
 
 config X86_PAE
 	bool "PAE (Physical Address Extension) Support"
-	depends on X86_32 && !HIGHMEM4G
+	depends on X86_32
 	help
 	  PAE is required for NX support, and furthermore enables
 	  larger swapspace support for non-overcommit purposes. It
@@ -1147,7 +1147,7 @@ source "mm/Kconfig"
 
 config HIGHPTE
 	bool "Allocate 3rd-level pagetables from highmem"
-	depends on X86_32 && (HIGHMEM4G || HIGHMEM64G)
+	depends on HIGHMEM
 	help
 	  The VM uses one page table entry for each page of physical memory.
 	  For systems with a lot of RAM, this can be wasteful of precious

Re: [PATCH] x86: simplify HIGHMEM-related Kconfig entries

[H. Peter Anvin]

Jan Beulich wrote:
> Additionally remove the dependency of X86_PAE on !HIGHMEM4G - at least
> I can't understand why that dependency existed.

At least originally, the meaning of the x86 highmem entries were:

	No highmem	- No highmem, no PAE
	HIGHMEM4G	- Highmem, no PAE
	HIGHMEM64G	- Highmem, PAE

So X86_PAE and HIGHMEM4G is a bit of a contradiction.  I haven't looked 
at the logic in detail, or remember offhand if there have been any 
weakening of the definitions above; e.g. someone could have implemented 
a way to do PAE without highmem, to get access to the NX bits.

	-hpa

Re: [PATCH] x86: simplify HIGHMEM-related Kconfig entries

[Jan Beulich]

>>> "H. Peter Anvin" <hpa@zytor.com> 14.01.09 20:39 >>>
>Jan Beulich wrote:
>> Additionally remove the dependency of X86_PAE on !HIGHMEM4G - at least
>> I can't understand why that dependency existed.
>
>At least originally, the meaning of the x86 highmem entries were:
>
>	No highmem	- No highmem, no PAE
>	HIGHMEM4G	- Highmem, no PAE
>	HIGHMEM64G	- Highmem, PAE
>
>So X86_PAE and HIGHMEM4G is a bit of a contradiction.  I haven't looked 
>at the logic in detail, or remember offhand if there have been any 
>weakening of the definitions above; e.g. someone could have implemented 
>a way to do PAE without highmem, to get access to the NX bits.

Exactly - .23 made PAE an independently selectable option (in particular,
the no-highmem+PAE combination is now valid), but I can't see why it
added the !HIGHMEM4G dependency.

Jan

Re: [PATCH] x86: simplify HIGHMEM-related Kconfig entries

[Ingo Molnar]

* Jan Beulich <jbeulich@novell.com> wrote:

> >>> "H. Peter Anvin" <hpa@zytor.com> 14.01.09 20:39 >>>
> >Jan Beulich wrote:
> >> Additionally remove the dependency of X86_PAE on !HIGHMEM4G - at least
> >> I can't understand why that dependency existed.
> >
> >At least originally, the meaning of the x86 highmem entries were:
> >
> >	No highmem	- No highmem, no PAE
> >	HIGHMEM4G	- Highmem, no PAE
> >	HIGHMEM64G	- Highmem, PAE
> >
> >So X86_PAE and HIGHMEM4G is a bit of a contradiction.  I haven't looked 
> >at the logic in detail, or remember offhand if there have been any 
> >weakening of the definitions above; e.g. someone could have implemented 
> >a way to do PAE without highmem, to get access to the NX bits.
> 
> Exactly - .23 made PAE an independently selectable option (in 
> particular, the no-highmem+PAE combination is now valid), but I can't 
> see why it added the !HIGHMEM4G dependency.

There's no deep reason, it just has never really been tested that way. I 
think we should try it.

	Ingo

Re: [PATCH] x86: simplify HIGHMEM-related Kconfig entries

[H. Peter Anvin]

Ingo Molnar wrote:
>>>
>>> 	No highmem	- No highmem, no PAE
>>> 	HIGHMEM4G	- Highmem, no PAE
>>> 	HIGHMEM64G	- Highmem, PAE
>>>
>>> So X86_PAE and HIGHMEM4G is a bit of a contradiction.  I haven't looked 
>>> at the logic in detail, or remember offhand if there have been any 
>>> weakening of the definitions above; e.g. someone could have implemented 
>>> a way to do PAE without highmem, to get access to the NX bits.
>> Exactly - .23 made PAE an independently selectable option (in 
>> particular, the no-highmem+PAE combination is now valid), but I can't 
>> see why it added the !HIGHMEM4G dependency.
> 
> There's no deep reason, it just has never really been tested that way. I 
> think we should try it.

If so, what would be the difference between HIGHMEM4G and HIGHMEM64G in
that case at all?

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

Re: [PATCH] x86: simplify HIGHMEM-related Kconfig entries

[Jan Beulich]

>>> "H. Peter Anvin" <hpa@zytor.com> 15.01.09 18:10 >>>
>Ingo Molnar wrote:
>>>>
>>>> 	No highmem	- No highmem, no PAE
>>>> 	HIGHMEM4G	- Highmem, no PAE
>>>> 	HIGHMEM64G	- Highmem, PAE
>>>>
>>>> So X86_PAE and HIGHMEM4G is a bit of a contradiction.  I haven't looked 
>>>> at the logic in detail, or remember offhand if there have been any 
>>>> weakening of the definitions above; e.g. someone could have implemented 
>>>> a way to do PAE without highmem, to get access to the NX bits.
>>> Exactly - .23 made PAE an independently selectable option (in 
>>> particular, the no-highmem+PAE combination is now valid), but I can't 
>>> see why it added the !HIGHMEM4G dependency.
>> 
>> There's no deep reason, it just has never really been tested that way. I 
>> think we should try it.
>
>If so, what would be the difference between HIGHMEM4G and HIGHMEM64G in
>that case at all?

For example being able to select the DMADEVICES config option.

Jan

Re: [PATCH] x86: simplify HIGHMEM-related Kconfig entries

[H. Peter Anvin]

Jan Beulich wrote:
>> If so, what would be the difference between HIGHMEM4G and HIGHMEM64G in
>> that case at all?
> 
> For example being able to select the DMADEVICES config option.

WTF?!  Why would DMA device support be banned on HIGHMEM64G?  That makes 
no sense whatsoever...

	-hpa

Re: [PATCH] x86: simplify HIGHMEM-related Kconfig entries

[Ingo Molnar]

* H. Peter Anvin <hpa@zytor.com> wrote:

> Jan Beulich wrote:
>>> If so, what would be the difference between HIGHMEM4G and HIGHMEM64G in
>>> that case at all?
>>
>> For example being able to select the DMADEVICES config option.
>
> WTF?!  Why would DMA device support be banned on HIGHMEM64G?  That makes 
> no sense whatsoever...

yes - but i guess the point is that there is some assymetry there which 
needs to be investigated and fixed.

	Ingo