[PATCH] X86_NUMAQ build fix

[PATCH] X86_NUMAQ build fix

[OGAWA Hirofumi]

Hi,

This patch fixes the build breakage of X86_NUMAQ. And this declares
xquad_portio on only X86_NUMAQ.

Please apply.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
---

 arch/i386/boot/compressed/misc.c |    4 +++-
 arch/i386/kernel/smpboot.c       |    2 +-
 arch/i386/pci/Makefile           |    3 ++-
 3 files changed, 6 insertions(+), 3 deletions(-)

diff -puN arch/i386/pci/Makefile~numaq-build-fix arch/i386/pci/Makefile
--- linux-2.6/arch/i386/pci/Makefile~numaq-build-fix	2006-04-22 00:54:29.000000000 +0900
+++ linux-2.6-hirofumi/arch/i386/pci/Makefile	2006-04-22 00:55:27.000000000 +0900
@@ -5,10 +5,11 @@ obj-$(CONFIG_PCI_MMCONFIG)	+= mmconfig.o
 obj-$(CONFIG_PCI_DIRECT)	+= direct.o
 
 pci-y				:= fixup.o
-pci-$(CONFIG_ACPI)		+= acpi.o
 pci-y				+= legacy.o irq.o
 
 pci-$(CONFIG_X86_VISWS)		:= visws.o fixup.o
 pci-$(CONFIG_X86_NUMAQ)		:= numa.o irq.o
 
+pci-$(CONFIG_ACPI)		+= acpi.o
+
 obj-y				+= $(pci-y) common.o
diff -puN arch/i386/boot/compressed/misc.c~numaq-build-fix arch/i386/boot/compressed/misc.c
--- linux-2.6/arch/i386/boot/compressed/misc.c~numaq-build-fix	2006-04-22 00:54:29.000000000 +0900
+++ linux-2.6-hirofumi/arch/i386/boot/compressed/misc.c	2006-04-22 00:54:29.000000000 +0900
@@ -122,7 +122,9 @@ static int vidport;
 static int lines, cols;
 
 #ifdef CONFIG_X86_NUMAQ
-static void * xquad_portio = NULL;
+/* hack to avoid using xquad_portio=NULL */
+#undef outb_p
+#define outb_p		outb_local_p
 #endif
 
 #include "../../../../lib/inflate.c"
diff -puN arch/i386/kernel/smpboot.c~numaq-build-fix arch/i386/kernel/smpboot.c
--- linux-2.6/arch/i386/kernel/smpboot.c~numaq-build-fix	2006-04-22 00:54:29.000000000 +0900
+++ linux-2.6-hirofumi/arch/i386/kernel/smpboot.c	2006-04-22 00:54:29.000000000 +0900
@@ -1116,9 +1116,9 @@ static void smp_tune_scheduling (void)
  */
 
 static int boot_cpu_logical_apicid;
+#ifdef CONFIG_X86_NUMAQ
 /* Where the IO area was mapped on multiquad, always 0 otherwise */
 void *xquad_portio;
-#ifdef CONFIG_X86_NUMAQ
 EXPORT_SYMBOL(xquad_portio);
 #endif
 
_

Re: [PATCH] X86_NUMAQ build fix

[Dave Hansen]

On Sat, 2006-04-22 at 01:06 +0900, OGAWA Hirofumi wrote:
> This patch fixes the build breakage of X86_NUMAQ. And this declares
> xquad_portio on only X86_NUMAQ.

What bug does this patch fix?  What is your .config?  I'm not having any
compile problems on my NUMAQ lately.

>  obj-$(CONFIG_PCI_DIRECT)       += direct.o
>  
>  pci-y                          := fixup.o
> -pci-$(CONFIG_ACPI)             += acpi.o
>  pci-y                          += legacy.o irq.o
>  
>  pci-$(CONFIG_X86_VISWS)                := visws.o fixup.o
>  pci-$(CONFIG_X86_NUMAQ)                := numa.o irq.o
>  
> +pci-$(CONFIG_ACPI)             += acpi.o

Am I reading this wrong, or does this just move the option down a bit?
Did you need to change the link order?  Why?

> +++ linux-2.6-hirofumi/arch/i386/boot/compressed/misc.c 2006-04-22 00:54:29.000000000 +0900
> @@ -122,7 +122,9 @@ static int vidport;
>  static int lines, cols;
>  
>  #ifdef CONFIG_X86_NUMAQ
> -static void * xquad_portio = NULL;
> +/* hack to avoid using xquad_portio=NULL */
> +#undef outb_p
> +#define outb_p         outb_local_p
>  #endif

It's really weird, but I'd hope that there was a reason for having two
xquad_portio.  Are you sure that this has no other consequences?

That said, it does boot on my 16-way NUMAQ, but I still have no real
idea what problem this is solving or what it is actually doing ;)

-- Dave

Re: [PATCH] X86_NUMAQ build fix

[OGAWA Hirofumi]

Dave Hansen <haveblue@us.ibm.com> writes:

> On Sat, 2006-04-22 at 01:06 +0900, OGAWA Hirofumi wrote:
>> This patch fixes the build breakage of X86_NUMAQ. And this declares
>> xquad_portio on only X86_NUMAQ.
>
> What bug does this patch fix?  What is your .config?  I'm not having any
> compile problems on my NUMAQ lately.

Ah, sorry for stuipd description...

>>  obj-$(CONFIG_PCI_DIRECT)       += direct.o
>>  
>>  pci-y                          := fixup.o
>> -pci-$(CONFIG_ACPI)             += acpi.o
>>  pci-y                          += legacy.o irq.o
>>  
>>  pci-$(CONFIG_X86_VISWS)                := visws.o fixup.o
>>  pci-$(CONFIG_X86_NUMAQ)                := numa.o irq.o
>>  
>> +pci-$(CONFIG_ACPI)             += acpi.o
>
> Am I reading this wrong, or does this just move the option down a bit?
> Did you need to change the link order?  Why?

No, this is not link order. Note that CONFIG_X86_VISWS/CONFIG_X86_NUMAQ
uses ":=", not "+=".  In case of ACPI=y", it breaks build.

Maybe NUMAQ shouldn't allow ACPI=y in Kconfig (very old Makefile
doesn't allow to build acpi.o), but I think Makefile is ok with this,
and it also fixes CONFIG_X86_VISWS case (very old Makefile also allow to
build acpi.o).

>> +++ linux-2.6-hirofumi/arch/i386/boot/compressed/misc.c 2006-04-22 00:54:29.000000000 +0900
>> @@ -122,7 +122,9 @@ static int vidport;
>>  static int lines, cols;
>>  
>>  #ifdef CONFIG_X86_NUMAQ
>> -static void * xquad_portio = NULL;
>> +/* hack to avoid using xquad_portio=NULL */
>> +#undef outb_p
>> +#define outb_p         outb_local_p
>>  #endif
>
> It's really weird, but I'd hope that there was a reason for having two
> xquad_portio.  Are you sure that this has no other consequences?

Note that this is boot/compressed/misc.c. We can't share vmlinux's
data in here. However we can share macros, so probably, it's using
"static void * xquad_portio = NULL" hack.

But, in include/asm-i386/io.h, it is declared as "extern". At least,
gcc-4.0 doesn't allow "static" and "extern" mismatch.

> That said, it does boot on my 16-way NUMAQ, but I still have no real
> idea what problem this is solving or what it is actually doing ;)

Oh, good. Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH] X86_NUMAQ build fix

[Dave Hansen]

On Sat, 2006-04-22 at 03:50 +0900, OGAWA Hirofumi wrote:
> Dave Hansen <haveblue@us.ibm.com> writes:
> >>  pci-y                          := fixup.o
> >> -pci-$(CONFIG_ACPI)             += acpi.o
> >>  pci-y                          += legacy.o irq.o
> >>  
> >>  pci-$(CONFIG_X86_VISWS)                := visws.o fixup.o
> >>  pci-$(CONFIG_X86_NUMAQ)                := numa.o irq.o
> >>  
> >> +pci-$(CONFIG_ACPI)             += acpi.o
> >
> > Am I reading this wrong, or does this just move the option down a bit?
> > Did you need to change the link order?  Why?
> 
> No, this is not link order. Note that CONFIG_X86_VISWS/CONFIG_X86_NUMAQ
> uses ":=", not "+=".  In case of ACPI=y", it breaks build.

Ahh, so NUMAQ and VISWS don't seem to allow or want ACPI code.
Especially for the NUMAQ, I don't think we should even mess with this
too much.  Just fix it in the build system and directly disallow NUMAQ
&& ACPI. 

We already have this:

menu "ACPI (Advanced Configuration and Power Interface) Support"
        depends on !X86_VISWS

Just add the NUMAQ in there, too.

-- Dave

Re: [PATCH] X86_NUMAQ build fix

[OGAWA Hirofumi]

Dave Hansen <haveblue@us.ibm.com> writes:

> On Sat, 2006-04-22 at 03:50 +0900, OGAWA Hirofumi wrote:
>> Dave Hansen <haveblue@us.ibm.com> writes:
>
> Ahh, so NUMAQ and VISWS don't seem to allow or want ACPI code.
> Especially for the NUMAQ, I don't think we should even mess with this
> too much.  Just fix it in the build system and directly disallow NUMAQ
> && ACPI. 
>
> We already have this:
>
> menu "ACPI (Advanced Configuration and Power Interface) Support"
>         depends on !X86_VISWS
>
> Just add the NUMAQ in there, too.

Oh, thanks a lot for noticeing this. Fixed patch is here.

[compile test only. I don't have NUMAQ, and I just wanted to see
"objdump -S" of NUMA's code.]
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>



Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
---

 arch/i386/boot/compressed/misc.c |    4 +++-
 arch/i386/kernel/smpboot.c       |    2 +-
 drivers/acpi/Kconfig             |    1 +
 3 files changed, 5 insertions(+), 2 deletions(-)

diff -puN arch/i386/pci/Makefile~numaq-build-fix arch/i386/pci/Makefile
diff -puN arch/i386/boot/compressed/misc.c~numaq-build-fix arch/i386/boot/compressed/misc.c
--- linux-2.6/arch/i386/boot/compressed/misc.c~numaq-build-fix	2006-04-22 04:15:04.000000000 +0900
+++ linux-2.6-hirofumi/arch/i386/boot/compressed/misc.c	2006-04-22 04:15:04.000000000 +0900
@@ -122,7 +122,9 @@ static int vidport;
 static int lines, cols;
 
 #ifdef CONFIG_X86_NUMAQ
-static void * xquad_portio = NULL;
+/* hack to avoid using xquad_portio=NULL */
+#undef outb_p
+#define outb_p		outb_local_p
 #endif
 
 #include "../../../../lib/inflate.c"
diff -puN arch/i386/kernel/smpboot.c~numaq-build-fix arch/i386/kernel/smpboot.c
--- linux-2.6/arch/i386/kernel/smpboot.c~numaq-build-fix	2006-04-22 04:15:04.000000000 +0900
+++ linux-2.6-hirofumi/arch/i386/kernel/smpboot.c	2006-04-22 04:15:04.000000000 +0900
@@ -1116,9 +1116,9 @@ static void smp_tune_scheduling (void)
  */
 
 static int boot_cpu_logical_apicid;
+#ifdef CONFIG_X86_NUMAQ
 /* Where the IO area was mapped on multiquad, always 0 otherwise */
 void *xquad_portio;
-#ifdef CONFIG_X86_NUMAQ
 EXPORT_SYMBOL(xquad_portio);
 #endif
 
diff -puN drivers/acpi/Kconfig~numaq-build-fix drivers/acpi/Kconfig
--- linux-2.6/drivers/acpi/Kconfig~numaq-build-fix	2006-04-22 04:16:35.000000000 +0900
+++ linux-2.6-hirofumi/drivers/acpi/Kconfig	2006-04-22 04:16:41.000000000 +0900
@@ -4,6 +4,7 @@
 
 menu "ACPI (Advanced Configuration and Power Interface) Support"
 	depends on !X86_VISWS
+	depends on !X86_NUMAQ
 	depends on !IA64_HP_SIM
 	depends on IA64 || X86
 
_

Re: [PATCH] X86_NUMAQ build fix

[Dave Hansen]

Do you really need anything other than the Kconfig fix?  I thought that
it alone would be a complete fix.

-- Dave

Re: [PATCH] X86_NUMAQ build fix

[OGAWA Hirofumi]

Dave Hansen <haveblue@us.ibm.com> writes:

> Do you really need anything other than the Kconfig fix?  I thought that
> it alone would be a complete fix.

Yes. I thought I explained those...

>  #ifdef CONFIG_X86_NUMAQ
> -static void * xquad_portio = NULL;
> +/* hack to avoid using xquad_portio=NULL */
> +#undef outb_p
> +#define outb_p		outb_local_p
>  #endif

This fixes the following error.

make -C /devel/linux/works/linux-2.6 O=/devel/linux/works/linux-2.6-devron
  GEN    /devel/linux/works/linux-2.6-devron/Makefile
  CHK     include/linux/version.h
  Using /devel/linux/works/linux-2.6 as source for kernel
  CHK     include/linux/compile.h
  LD      vmlinux
  SYSMAP  System.map
  SYSMAP  .tmp_System.map
  Building modules, stage 2.
  LD      arch/i386/boot/setup
  CC      arch/i386/boot/compressed/misc.o
  OBJCOPY arch/i386/boot/compressed/vmlinux.bin
  GZIP    arch/i386/boot/compressed/vmlinux.bin.gz
/devel/linux/works/linux-2.6/arch/i386/boot/compressed/misc.c:125: error: static declaration of 'xquad_portio' follows non-static declaration
include2/asm/io.h:303: error: previous declaration of 'xquad_portio' was here
make[4]: *** [arch/i386/boot/compressed/misc.o] Error 1
make[4]: *** Waiting for unfinished jobs....
  MODPOST
make[3]: *** [arch/i386/boot/compressed/vmlinux] Error 2
make[2]: *** [bzImage] Error 2
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [_all] Error 2
make: *** [all] Error 2

>  static int boot_cpu_logical_apicid;
> +#ifdef CONFIG_X86_NUMAQ
>  /* Where the IO area was mapped on multiquad, always 0 otherwise */
>  void *xquad_portio;
> -#ifdef CONFIG_X86_NUMAQ
>  EXPORT_SYMBOL(xquad_portio);
>  #endif

xquad_portio is needed to only X86_NUMAQ.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>