Trying to add Kconfig support for M57SLI - build error

[Harald Gutmann]

Hello,

recently I'm trying to add Kconfig support to the M57SLI board from gigabyte, 
but right now i stumbled into an error where I can't figure out what is wrong.

This problem is not directly related to Kconfig, but it seems that I 
misunderstand something in the compiler errors, or that I'm not able to fix 
this one.

The log of make:

hargut@benchvice:~/cb-v2$ make
    GEN        build/build.h
    CC         build/northbridge/amd/amdk8/misc_control.o
In file included from src/northbridge/amd/amdk8/amdk8.h:6,
                 from src/northbridge/amd/amdk8/misc_control.c:22:
src/northbridge/amd/amdk8/amdk8_f.h:524:21: error: macro "hard_reset" passed 1 
arguments, but takes just 0
src/northbridge/amd/amdk8/misc_control.c: In function ‘mcf3_read_resources’:
src/northbridge/amd/amdk8/misc_control.c:51: warning: passing argument 2 of 
‘get_option’ discards qualifiers from pointer target type
src/northbridge/amd/amdk8/misc_control.c: In function ‘misc_control_init’:
src/northbridge/amd/amdk8/misc_control.c:113: warning: unused variable 
‘f2_dev’
make: *** [/home/hargut/cb-v2/build/northbridge/amd/amdk8/misc_control.o] 
Error 1
hargut@benchvice:~/cb-v2$ grep hard_reset 
src/northbridge/amd/amdk8/misc_control.c
#include <part/hard_reset.h>
                hard_reset();
hargut@benchvice:~/cb-v2$ grep hard_reset src/northbridge/amd/amdk8/amdk8_f.h
void hard_reset(void);
                hard_reset();
hargut@benchvice:~/cb-v2$

EOL (End of Log)

The error itself is:
src/northbridge/amd/amdk8/amdk8_f.h:524:21: error: macro "hard_reset" passed 1 
arguments, but takes just 0

but where gets the hard_reset passed one argument?
And or why is hard_reset used as in line 23 of cache_as_ram_auto.c there is 
the "#define __ROMCC__" which should cause that soft_reset() in amdk8_f.h is 
used.

"#ifdef __ROMCC__
static void soft_reset(void);
#else
void hard_reset(void);
#endif"

What of that stuff do I misunderstand that I'm not able to fix that compilation 
error? Or am I unable to see the wood because of so much trees?

Many thanks for any tips on solving that error/understanding problem.

Kind regards,
Harald

Trying to add Kconfig support for M57SLI - build error

[ron minnich]

let's start over. can you tell me what changes you have made to the
tree at this time?

ron

Trying to add Kconfig support for M57SLI - build error

[Harald Gutmann]

On Friday 21 August 2009 00:06:44 ron minnich wrote:
> let's start over. can you tell me what changes you have made to the
> tree at this time?

The tree itself has no changes in source.
I just added a few Kconfig and Makefile.inc files.

I think it's not necessary to post the patch right now, the .config should be 
enough to track that down, or is a full diff against svn needed?

The .config which is generated by Kconfig is attached.
> ron

Thanks, 
regards Harald

[patch] Trying to add Kconfig support for M57SLI - build error

[Cristi Magherusan]

On Fri, 2009-08-21 at 13:27 +0200, Harald Gutmann wrote:
> On Friday 21 August 2009 00:06:44 ron minnich wrote:
> > let's start over. can you tell me what changes you have made to the
> > tree at this time?
>
> The tree itself has no changes in source.
> I just added a few Kconfig and Makefile.inc files.
>
> I think it's not necessary to post the patch right now, the .config should be 
> enough to track that down, or is a full diff against svn needed?
>
> The .config which is generated by Kconfig is attached.

I think the problem is caused by your socket AM2 CPU, which is missing
from Kbuild.

The attached patch may help.
This is
Signed-off by: Cristi Magherusan

[patch] Trying to add Kconfig support for M57SLI - build error

[Harald Gutmann]

On Friday 21 August 2009 14:20:56 Cristi Magherusan wrote:
> On Fri, 2009-08-21 at 13:27 +0200, Harald Gutmann wrote:
> > On Friday 21 August 2009 00:06:44 ron minnich wrote:
> > > let's start over. can you tell me what changes you have made to the
> > > tree at this time?
> >
> > The tree itself has no changes in source.
> > I just added a few Kconfig and Makefile.inc files.
> >
> > I think it's not necessary to post the patch right now, the .config
> > should be enough to track that down, or is a full diff against svn
> > needed?
> >
> > The .config which is generated by Kconfig is attached.
>
> I think the problem is caused by your socket AM2 CPU, which is missing
> from Kbuild.

Thanks for that hint, there have been some errors in that files, but your patch 
doesn't solve the problem.
> The attached patch may help.

Good point, but doesn't solve it.
> This is
> Signed-off by: Cristi Magherusan

Thanks!
Regards Harald

[patch] Trying to add Kconfig support for M57SLI - build error

[Cristi Magherusan]

On Fri, 2009-08-21 at 14:40 +0200, Harald Gutmann wrote:
> On Friday 21 August 2009 14:20:56 Cristi Magherusan wrote:
> > On Fri, 2009-08-21 at 13:27 +0200, Harald Gutmann wrote:
> > > On Friday 21 August 2009 00:06:44 ron minnich wrote:
> > > > let's start over. can you tell me what changes you have made to the
> > > > tree at this time?
> > >
> > > The tree itself has no changes in source.
> > > I just added a few Kconfig and Makefile.inc files.
> > >
> > > I think it's not necessary to post the patch right now, the .config
> > > should be enough to track that down, or is a full diff against svn
> > > needed?
> > >
> > > The .config which is generated by Kconfig is attached.
> >
> > I think the problem is caused by your socket AM2 CPU, which is missing
> > from Kbuild.
>
> Thanks for that hint, there have been some errors in that files, but your patch 
> doesn't solve the problem.
> > The attached patch may help.
>
> Good point, but doesn't solve it.

I've seen you had already implemented what I sent, and my patch was
wrong (just copy/pasted from socket F), sorry for this.
Your big patch is ok to help us replicate your work, but can't be
committed as a whole, which is what we should have as a purpose.

So let's stop reinventing the wheel for each new board, and make smaller
patches that add a single piece of functionality which can be commented
on, fixed and committed to the tree.

For example the attached patch adds support for AM2 CPUs (I fixed the
0x11 issue).

There's still a debatable issue remaining, which is the inclusion of the
src/cpu/amd/model_fxx/Kconfig file.

Is it fine if it's done like in here? Adding it to all the CPU sockets
that may need it causes errors due to double inclusions.

Thanks,
Cristi

The patch is 
Signed-off by: Cristi Magherusan

[patch] Trying to add Kconfig support for M57SLI - build error

[ron minnich]

On Fri, Aug 21, 2009 at 9:17 AM, Cristi
Magherusan wrote:
> There's still a debatable issue remaining, which is the inclusion of the
> src/cpu/amd/model_fxx/Kconfig file.

it's ok, Kconfig rules are not makefile rules. Let's just assume we source
all subdirectory Kconfigs from a higher level directory.

ron

[patch] Trying to add Kconfig support for M57SLI - build error

[Myles Watson]

> The patch is
> Signed-off by: Cristi Magherusan

Acked-by: Myles Watson

Rev. 4565.

Thanks,
Myles

Trying to add Kconfig support for M57SLI - build error

[Harald Gutmann]

On Friday 21 August 2009 00:06:44 ron minnich wrote:
> let's start over. can you tell me what changes you have made to the
> tree at this time?

Okay, and here is the full svn diff (hopefully I didn't forget to svn add files) 
of what I've already done in order to get Kconfig working on that mainboard.
> ron

Regards,
Harald

Trying to add Kconfig support for M57SLI - build error

[Harald Gutmann]

On Friday 21 August 2009 14:09:35 Harald Gutmann wrote:
> On Friday 21 August 2009 00:06:44 ron minnich wrote:
> > let's start over. can you tell me what changes you have made to the
> > tree at this time?
>
> Okay, and here is the full svn diff (hopefully I didn't forget to svn add
> files) of what I've already done in order to get Kconfig working on that
> mainboard.

New version with some corrected errors.
Thanks Christi for the hint!
> > ron
>
> Regards,
> Harald

Trying to add Kconfig support for M57SLI - build error

[ron minnich]

This is *almost* perfect. Questions and changes.

Index: src/cpu/amd/socket_AM2/Kconfig
===================================================================
--- src/cpu/amd/socket_AM2/Kconfig	(revision 0)
+++ src/cpu/amd/socket_AM2/Kconfig	(revision 0)
@@ -0,0 +1,30 @@
+config CPU_AMD_SOCKET_AM2
+	bool
+	default false
+
+config CPU_SOCKET_TYPE
+	hex
+	default 0x10
+	depends on CPU_AMD_SOCKET_AM2

It's a type 0x10?

+
+config K8_REV_F_SUPPORT
+	int	
+	default 1
+	depends on CPU_AMD_SOCKET_AM2

Socket AM2 supports REV F? I don't know, just asking.

+config CPU_SOCKET_TYPE
+	hex
+	default 0x11
+	depends on CPU_AMD_SOCKET_AM2

it's in here twice.

Index: src/cpu/amd/socket_AM2/Makefile.inc
===================================================================
--- src/cpu/amd/socket_AM2/Makefile.inc	(revision 0)
+++ src/cpu/amd/socket_AM2/Makefile.inc	(revision 0)
@@ -0,0 +1,14 @@
+obj-$(CONFIG_CPU_AMD_SOCKET_AM2) += socket_AM.o
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../model_fxx
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../dualcore
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../mtrr
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/tsc
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/mtrr
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/fpu
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/mmx
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/sse
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/lapic
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/cache
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/mtrr
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/pae
+subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/smm

Just make these -y. This Makefile.inc is *conditionally* included from
the src/cpu/amd/Makefile.inc. There's no reason to make
them conditional in turn.

Let's go around one more time taking into account my comments. We are
almost there.

ron

Trying to add Kconfig support for M57SLI - build error

[Harald Gutmann]

New patch version attached, but the compile error stays at the same state than 
before.

On Friday 21 August 2009 17:31:17 ron minnich wrote:
> This is *almost* perfect. Questions and changes.

I think it would be possible to sparse some values in the mainboard directory 
out, but that's a todo after fixing compilation.
> Index: src/cpu/amd/socket_AM2/Kconfig
>
> --- src/cpu/amd/socket_AM2/Kconfig	(revision 0)
> +++ src/cpu/amd/socket_AM2/Kconfig	(revision 0)
> @@ -0,0 +1,30 @@
> +config CPU_AMD_SOCKET_AM2
> bool
> default false
>
> +config CPU_SOCKET_TYPE
> hex
> default 0x10
> depends on CPU_AMD_SOCKET_AM2
>
> It's a type 0x10?
>
> +config K8_REV_F_SUPPORT
> int
> default 1
> depends on CPU_AMD_SOCKET_AM2
>
> Socket AM2 supports REV F? I don't know, just asking.
>
> +config CPU_SOCKET_TYPE
> hex
> default 0x11
> depends on CPU_AMD_SOCKET_AM2
>
> it's in here twice.
> Index: src/cpu/amd/socket_AM2/Makefile.inc
>
> --- src/cpu/amd/socket_AM2/Makefile.inc	(revision 0)
> +++ src/cpu/amd/socket_AM2/Makefile.inc	(revision 0)
> @@ -0,0 +1,14 @@
> +obj-$(CONFIG_CPU_AMD_SOCKET_AM2) += socket_AM.o
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../model_fxx
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../dualcore
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../mtrr
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/tsc
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/mtrr
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/fpu
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/mmx
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/sse
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/lapic
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/cache
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/mtrr
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/pae
> +subdirs-$(CONFIG_CPU_AMD_SOCKET_AM2) += ../../x86/smm
>
> Just make these -y. This Makefile.inc is *conditionally* included from
> the src/cpu/amd/Makefile.inc. There's no reason to make
> them conditional in turn.
>
> Let's go around one more time taking into account my comments. We are
> almost there.
>
> ron

Trying to add Kconfig support for M57SLI - build error

[ron minnich]

I'm missing
src/southbridge/nvidia/ck804/Kconfig
ron

Trying to add Kconfig support for M57SLI - build error

[Harald Gutmann]

On Friday 21 August 2009 18:10:37 ron minnich wrote:
> I'm missing
> src/southbridge/nvidia/ck804/Kconfig

Yes, because that's not needed.
The m57sli board has a mcp55 southbridge and therefore ck804 is not needed.
I removed this because I wanted to clean out and sort up the patch as far as 
possible.

Index: src/southbridge/nvidia/Kconfig
===================================================================
--- src/southbridge/nvidia/Kconfig      (revision 4560)
+++ src/southbridge/nvidia/Kconfig      (working copy)
@@ -1,2 +1 @@
-source src/southbridge/nvidia/ck804/Kconfig
 source src/southbridge/nvidia/mcp55/Kconfig

This sould fix also that little line, but I don't think that this affects my 
problem, as Kconfig would complain if it needs that file.
> ron

Regards,
Harald

Trying to add Kconfig support for M57SLI - build error

[Myles Watson]

> This problem is not directly related to Kconfig, but it seems that I
> misunderstand something in the compiler errors, or that I'm not able to fix
> this one.
>
> The log of make:
>
> hargut@benchvice:~/cb-v2$ make
>    GEN        build/build.h
>    CC         build/northbridge/amd/amdk8/misc_control.o
> In file included from src/northbridge/amd/amdk8/amdk8.h:6,
>                 from src/northbridge/amd/amdk8/misc_control.c:22:
> src/northbridge/amd/amdk8/amdk8_f.h:524:21: error: macro "hard_reset" passed 1
> arguments, but takes just 0

I'm having similar problems (at least it looks similar to me).  My
ck804_reset.c file never gets compiled.

Here is my src/southbridge/nvidia/ck804/Makefile.inc:

driver-y += ck804.o
driver-y += ck804_usb.o
driver-y += ck804_lpc.o
driver-y += ck804_smbus.o
driver-y += ck804_ide.o
driver-y += ck804_sata.o
driver-y += ck804_usb2.o
driver-y += ck804_ac97.o
driver-y += ck804_nic.o
driver-y += ck804_pci.o
driver-y += ck804_pcie.o
driver-y += ck804_ht.o
object-y += ck804_reset.o

object-$(CONFIG_HAVE_ACPI_TABLES) +=ck804_fadt.o

All of the drivers get compiled, but the two objects are never compiled.

Any suggestions?

Thanks,
Myles

Trying to add Kconfig support for M57SLI - build error

[Myles Watson]

> ck804_reset.c file never gets compiled.
> object-y += ck804_reset.o
>
> object-$(CONFIG_HAVE_ACPI_TABLES) +=ck804_fadt.o

obj-y is the correct syntax.  Sorry for the noise.

Myles

Trying to add Kconfig support for M57SLI - build error

[ron minnich]

On Fri, Aug 21, 2009 at 9:44 AM, Myles Watson wrote:
> > ck804_reset.c file never gets compiled.
> > object-y += ck804_reset.o
> >
> > object-$(CONFIG_HAVE_ACPI_TABLES) +=ck804_fadt.o
>
> obj-y is the correct syntax.  Sorry for the noise.

one last correction:
obj-$(CONFIG_HAVE_ACPI_TABLES) +=ck804_fadt.o

since it is conditional on that variable.

ron

Trying to add Kconfig support for M57SLI - build error

[Myles Watson]

> > hargut@benchvice:~/cb-v2$ make
> >    GEN        build/build.h
> >    CC         build/northbridge/amd/amdk8/misc_control.o
> > In file included from src/northbridge/amd/amdk8/amdk8.h:6,
> >                 from src/northbridge/amd/amdk8/misc_control.c:22:
> > src/northbridge/amd/amdk8/amdk8_f.h:524:21: error: macro "hard_reset" passed 1
> > arguments, but takes just 0
>
> I'm having similar problems (at least it looks similar to me).  My
> ck804_reset.c file never gets compiled.

I applied your patch, and it doesn't look like you uncomment the line
in src/southbridge/nvidia/Makefile.inc so that mcp55 gets built.

Myles

Trying to add Kconfig support for M57SLI - build error

[ron minnich]

let's slow down on these patches.

We're moving way too fast and it's getting confusing, and errors and
mistakes are creeping in. Cristi is right.

Let's first do AM2. Then nvidia south. And so on.

OK, I am going to apply Cristi's patch, and, if it looks good, commit.
Then we can do nvidia south.

ron

Trying to add Kconfig support for M57SLI - build error

[Myles Watson]

> Let's first do AM2. Then nvidia south. And so on.
>
> OK, I am going to apply Cristi's patch, and, if it looks good, commit.
> Then we can do nvidia south.
> From Cristi's patch:

--- /dev/null
+++ b/src/cpu/amd/microcode/Makefile.inc
@@ -0,0 +1 @@
+obj-y += microcode.o

Let's not create a whole file for one obj.  Can we just put that in
socket_AM2/Makefile.inc like the intel sockets do?

obj-y += ../microcode/microcode.o

Thanks,
Myles

Trying to add Kconfig support for M57SLI - build error

[ron minnich]

On Fri, Aug 21, 2009 at 11:09 AM, Myles Watson wrote:
> > Let's first do AM2. Then nvidia south. And so on.
> >
> > OK, I am going to apply Cristi's patch, and, if it looks good, commit.
> > Then we can do nvidia south.
>
> From Cristi's patch:
> --- /dev/null
> +++ b/src/cpu/amd/microcode/Makefile.inc
> @@ -0,0 +1 @@
> +obj-y += microcode.o
>
> Let's not create a whole file for one obj.  Can we just put that in
> socket_AM2/Makefile.inc like the intel sockets do?
>
> obj-y += ../microcode/microcode.o

The reason to create a whole file for one object is that, over time,
kif we get more than one object, nobody has to change their makefiles
-- we add to microcode/Makefile.inc.

I think the subdir usage and extra Makefile.inc are fine, if we're
going to have the directory.

At same time, I'm not overly concerned either way. I have to go out
rest of the day, the AM2 patch is almost there, Myles, can you and
Cristi wrap it up?

ron

Trying to add Kconfig support for M57SLI - build error

[Myles Watson]

> > Let's not create a whole file for one obj.  Can we just put that in
> > socket_AM2/Makefile.inc like the intel sockets do?
> >
> > obj-y += ../microcode/microcode.o
>
> The reason to create a whole file for one object is that, over time,
> kif we get more than one object, nobody has to change their makefiles
> -- we add to microcode/Makefile.inc.
>
> I think the subdir usage and extra Makefile.inc are fine, if we're
> going to have the directory.

OK.
> At same time, I'm not overly concerned either way. I have to go out
> rest of the day, the AM2 patch is almost there, Myles, can you and
> Cristi wrap it up?

I've made so many Kconfig mistakes I'm not sure I'm the right person to Ack
& commit.  But I don't have any objections to the patch.

Myles

Trying to add Kconfig support for M57SLI - build error

[Harald Gutmann]

On Friday 21 August 2009 20:00:23 ron minnich wrote:
> let's slow down on these patches.
>
> We're moving way too fast and it's getting confusing, and errors and
> mistakes are creeping in. Cristi is right.
>
> Let's first do AM2. Then nvidia south. And so on.
>
> OK, I am going to apply Cristi's patch, and, if it looks good, commit.
> Then we can do nvidia south.

Okay, sounds good.

I've reviewed the mcp55 Kconfig once more and attached it.
It seems for me that it is fine, and should work normally.

This patch is:
Signed-off-by: Harald Gutmann
> ron

Harald

Trying to add Kconfig support for M57SLI - builderror

[Myles Watson]

> > OK, I am going to apply Cristi's patch, and, if it looks good, commit.
> > Then we can do nvidia south.
>
> Okay, sounds good.

Lets not create a new CONFIG variable SOUTHBRIDGE_NVIDIA_MCP55.  I think
that whole file can go.  And the files that source it.
Index: nvidia/mcp55/Kconfig
Index: nvidia/Kconfig

I'd rather you didn't remove the ck804 references, since my patch will have
to put them back.

+driver-$(CONFIG_HAVE_ACPI_TABLES) += mcp55_fadt.o
I think you mean obj here.

Thanks,
Myles

Trying to add Kconfig support for M57SLI - builderror

[Harald Gutmann]

On Friday 21 August 2009 20:21:27 Myles Watson wrote:
> > > OK, I am going to apply Cristi's patch, and, if it looks good, commit.
> > > Then we can do nvidia south.
> >
> > Okay, sounds good.
>
> Lets not create a new CONFIG variable SOUTHBRIDGE_NVIDIA_MCP55.  I think
> that whole file can go.  And the files that source it.
> Index: nvidia/mcp55/Kconfig
> Index: nvidia/Kconfig

Are you sure that this will work as in the mainboard/{vendor}/{model}/Kconfig 
it is used to select which sb is used. (Maybe I'm wrong with this, but I used 
the amd_serengeti target as starting point, and there it's done the same way.)

For example here on my board it looks like:
config BOARD_GIGABYTE_M57SLI
        bool "M57SLI"
        select ARCH_X86
        select CPU_AMD_K8
        select CPU_AMD_SOCKET_AM2
        select NORTHBRIDGE_AMD_AMDK8
        select NORTHBRIDGE_AMD_AMDK8_ROOT_COMPLEX
        select SOUTHBRIDGE_NVIDIA_MCP55
        select SUPERIO_WINBOND_IT8716F
        select PIRQ_TABLE
        select USE_PRINTK_IN_CAR
        help
        Gigabyte M57SLI mainboard
endchoice
> I'd rather you didn't remove the ck804 references, since my patch will have
> to put them back.

Done.
> +driver-$(CONFIG_HAVE_ACPI_TABLES) += mcp55_fadt.o
> I think you mean obj here.

Yes.

v2 of the patch attached, with some more (little) modifications.
> Thanks,
> Myles

Thanks,
Harald

Trying to add Kconfig support for M57SLI - builderror

[Myles Watson]

On Fri, Aug 21, 2009 at 12:33 PM, Harald Gutmann wrote:
> On Friday 21 August 2009 20:21:27 Myles Watson wrote:
> > > > OK, I am going to apply Cristi's patch, and, if it looks good, commit.
> > > > Then we can do nvidia south.
> > >
> > > Okay, sounds good.
> >
> > Lets not create a new CONFIG variable SOUTHBRIDGE_NVIDIA_MCP55.  I think
> > that whole file can go.  And the files that source it.
> > Index: nvidia/mcp55/Kconfig
> > Index: nvidia/Kconfig
>
> Are you sure that this will work as in the mainboard/{vendor}/{model}/Kconfig
> it is used to select which sb is used. (Maybe I'm wrong with this, but I used
> the amd_serengeti target as starting point, and there it's done the same way.)

You're right.  The reason I was questioning it was that I grepped the
tree and didn't find any CONFIG_SOUTHBRIDGE_NVIDIA*

I think the correct fix for that problem is to change
src/southbridge/nvidia/Makefile.inc
subdirs-y should be subdirs-$(CONFIG_SOUTHBRIDGE_NVIDIA_MCP55).

Does that sound better?
> > I'd rather you didn't remove the ck804 references, since my patch will have
> > to put them back.
>
> Done.
> > +driver-$(CONFIG_HAVE_ACPI_TABLES) += mcp55_fadt.o
> > I think you mean obj here.
>
> Yes.
>
> v2 of the patch attached, with some more (little) modifications.

Thanks,
Myles

Trying to add Kconfig support for M57SLI - builderror

[ron minnich]

On Fri, Aug 21, 2009 at 12:14 PM, Myles Watson wrote:
> I think the correct fix for that problem is to change
> src/southbridge/nvidia/Makefile.inc
> subdirs-y should be subdirs-$(CONFIG_SOUTHBRIDGE_NVIDIA_MCP55).

yes!

This is the general rule. It's even in my small Kconfig.tex doco :-)

ron

Trying to add Kconfig support for M57SLI - builderror

[Myles Watson]

On Fri, Aug 21, 2009 at 2:41 PM, ron minnich wrote:
> On Fri, Aug 21, 2009 at 12:14 PM, Myles Watson wrote:
> > I think the correct fix for that problem is to change
> > src/southbridge/nvidia/Makefile.inc
> > subdirs-y should be subdirs-$(CONFIG_SOUTHBRIDGE_NVIDIA_MCP55).
>
> yes!
>
> This is the general rule. It's even in my small Kconfig.tex doco :-)

Unfortunately it fits into the category of "the tree speaks louder
than the docs."  We shouldn't have missed it, but we did.

Myles