[PATCH] RFC - move CAR config to socket config

[Warren Turkal]

This patch moves one of the CAR configs to the socket from the single
mainboard that uses it. Would this be a reasonable cleanup to perform
across the tree?

Thanks,
wt

Warren Turkal (1):
  Move CAR config from mainboard config to socket config.

src/cpu/intel/socket_441/Kconfig     |   16 ++++++++++++++++
 src/mainboard/intel/d945gclf/Kconfig |    9 ---------
 2 files changed, 16 insertions(+), 9 deletions(-)

[PATCH] Move CAR config from mainboard config to socket config.

[Warren Turkal]

Signed-off-by: Warren Turkal 
---
 src/cpu/intel/socket_441/Kconfig     |   16 ++++++++++++++++
 src/mainboard/intel/d945gclf/Kconfig |    9 ---------
 2 files changed, 16 insertions(+), 9 deletions(-)

diff --git a/src/cpu/intel/socket_441/Kconfig b/src/cpu/intel/socket_441/Kconfig
index 76cbb1f..1306656 100644
--- a/src/cpu/intel/socket_441/Kconfig
+++ b/src/cpu/intel/socket_441/Kconfig
@@ -1,5 +1,21 @@
 config CPU_INTEL_SOCKET_441
    bool
+
+if CPU_INTEL_SOCKET_441
+
+config SOCKET_SPECIFIC_OPTIONS # dummy
+	def_bool y
    select CPU_INTEL_MODEL_106CX
    select MMX
    select SSE
+	select CACHE_AS_RAM
+
+config DCACHE_RAM_BASE
+	hex
+	default 0xffdf8000
+
+config DCACHE_RAM_SIZE
+	hex
+	default 0x8000
+
+endif # CPU_INTEL_SOCKET_441
diff --git a/src/mainboard/intel/d945gclf/Kconfig b/src/mainboard/intel/d945gclf/Kconfig
index f1ec575..2faca40 100644
--- a/src/mainboard/intel/d945gclf/Kconfig
+++ b/src/mainboard/intel/d945gclf/Kconfig
@@ -40,7 +40,6 @@ config BOARD_SPECIFIC_OPTIONS # dummy
    select HAVE_ACPI_TABLES
    select HAVE_SMI_HANDLER
    select BOARD_ROMSIZE_KB_512
-	select CACHE_AS_RAM
    select GFXUMA
    select TINY_BOOTBLOCK

@@ -48,14 +47,6 @@ config MAINBOARD_DIR
    string
    default intel/d945gclf

-config DCACHE_RAM_BASE
-	hex
-	default 0xffdf8000
-
-config DCACHE_RAM_SIZE
-	hex
-	default 0x8000
-
 config MAINBOARD_PART_NUMBER
    string
    default "D945GCLF"

[Patrick Georgi]

Am 26.09.2010 08:11, schrieb Warren Turkal:
> This patch moves one of the CAR configs to the socket from the single
> mainboard that uses it. Would this be a reasonable cleanup to perform
> across the tree?

I like the change a lot for DCACHE_RAM_*, but I'm undecided on
CACHE_AS_RAM - it's a simple way to determine if a board uses romcc or not.
Also, we used to have chipsets where some boards were brought over to
CAR and some weren't. Adding the DCACHE_RAM_* values won't hurt there,
changing the build mode will - but given that this is a manual
per-chipset change to the build system, I guess that could be
accomodated for (by not migrating those chipsets, or even better, by
moving the remaining boards to CAR)

Thanks,
Patrick

[Warren Turkal]

I would say that the CACHE_AS_RAM should be moved in cases where all
boards support it simply because it is not really a property of the
board at that point.

Given that, I should probably also move things like ARCH_X86 into the
socket config as well.

What do you all think?

Thanks,
wt

On Sat, Sep 25, 2010 at 11:58 PM, Patrick Georgi  wrote:
> Am 26.09.2010 08:11, schrieb Warren Turkal:
> > This patch moves one of the CAR configs to the socket from the single
> > mainboard that uses it. Would this be a reasonable cleanup to perform
> > across the tree?
>
> I like the change a lot for DCACHE_RAM_*, but I'm undecided on
> CACHE_AS_RAM - it's a simple way to determine if a board uses romcc or not.
> Also, we used to have chipsets where some boards were brought over to
> CAR and some weren't. Adding the DCACHE_RAM_* values won't hurt there,
> changing the build mode will - but given that this is a manual
> per-chipset change to the build system, I guess that could be
> accomodated for (by not migrating those chipsets, or even better, by
> moving the remaining boards to CAR)
>
> Thanks,
> Patrick

[Stefan Reinauer]

On 9/26/10 9:16 AM, Warren Turkal wrote:
> I would say that the CACHE_AS_RAM should be moved in cases where all
> boards support it simply because it is not really a property of the
> board at that point.

I agree. And it should not be, unless it's a workaround.
> Given that, I should probably also move things like ARCH_X86 into the
> socket config as well.

Sounds reasonable.

[Warren Turkal]

Can I get the current patch committed if we agree that it is correct?
I'd rather do the bulk moves of ARCH_X86 and other options in
different patches.

Thanks,
wt

On Sun, Sep 26, 2010 at 8:27 AM, Stefan Reinauer
 wrote:
> On 9/26/10 9:16 AM, Warren Turkal wrote:
> > I would say that the CACHE_AS_RAM should be moved in cases where all
> > boards support it simply because it is not really a property of the
> > board at that point.
>
> I agree. And it should not be, unless it's a workaround.
> > Given that, I should probably also move things like ARCH_X86 into the
> > socket config as well.
>
> Sounds reasonable.