From: Judith Lebzelter <[email protected]>
These warnings occur during modpost for ia64/allmodconfig in 2.6.19-rc1
WARNING: "empty_zero_page" [arch/ia64/hp/sim/simscsi.ko] undefined!
WARNING: "ia64_ssc" [arch/ia64/hp/sim/simscsi.ko] undefined!
This patch exports those variables and stops the warnings.
Signed-off-by: Judith Lebzelter <[email protected]>
---
Index: linux/arch/ia64/kernel/ia64_ksyms.c
===================================================================
--- linux.orig/arch/ia64/kernel/ia64_ksyms.c 2006-10-09 10:11:47.000000000 -0700
+++ linux/arch/ia64/kernel/ia64_ksyms.c 2006-10-09 10:15:18.000000000 -0700
@@ -111,3 +111,9 @@
#endif
extern char ia64_ivt[];
EXPORT_SYMBOL(ia64_ivt);
+
+#include "../hp/sim/hpsim_ssc.h"
+EXPORT_SYMBOL(ia64_ssc);
+#include <asm/pgtable.h>
+EXPORT_SYMBOL(empty_zero_page);
+
Judith,
+++ linux/arch/ia64/kernel/ia64_ksyms.c 2006-10-09 10:15:18.000000000 -0700
These exports are only needed for the HP simulator ... it seems
probable that the more likely fix is change Kconfig to prevent simscsi
from being built as a module. I assume that any remaining SKI users
build this into the kernel ... arch/ia64/configs/sim_defconfig sets
CONFIG_HP_SIMSCSI=y for example.
-Tony
On Wed, 11 Oct 2006 15:52:28 -0700
"Luck, Tony" <[email protected]> wrote:
> Judith,
>
> +++ linux/arch/ia64/kernel/ia64_ksyms.c 2006-10-09 10:15:18.000000000 -0700
>
> These exports are only needed for the HP simulator ... it seems
> probable that the more likely fix is change Kconfig to prevent simscsi
> from being built as a module. I assume that any remaining SKI users
> build this into the kernel ... arch/ia64/configs/sim_defconfig sets
> CONFIG_HP_SIMSCSI=y for example.
>
The problem is that ia64 allmodconfig now bombs out, since depmod treats
this as a hard error.
IOW, please make allmodconfig work ;)
On Thu, Oct 12, 2006 at 12:11:39AM -0700, Andrew Morton wrote:
> The problem is that ia64 allmodconfig now bombs out, since depmod treats
> this as a hard error.
>
> IOW, please make allmodconfig work ;)
I tried simply swapping "tristate" for "bool":
--- a/arch/ia64/hp/sim/Kconfig 2006-10-12 10:45:18.000000000 -0700
+++ b/arch/ia64/hp/sim/Kconfig 2006-10-12 09:43:30.000000000 -0700
@@ -13,7 +13,7 @@
depends on HP_SIMSERIAL
config HP_SIMSCSI
- tristate "Simulated SCSI disk"
+ bool "Simulated SCSI disk"
depends on SCSI
endmenu
... and now it fails in a new an diferent way:
LD .tmp_vmlinux1
arch/ia64/hp/sim/built-in.o(.init.text+0x9d2): In function `simscsi_init':
arch/ia64/hp/sim/simscsi.c:407: undefined reference to `scsi_host_alloc'
arch/ia64/hp/sim/built-in.o(.init.text+0xa02):arch/ia64/hp/sim/simscsi.c:411: undefined reference to `scsi_add_host'
arch/ia64/hp/sim/built-in.o(.init.text+0xa22):arch/ia64/hp/sim/simscsi.c:413: undefined reference to `scsi_scan_host'
presumably because we have CONFIG_HP_SIMSCSI=y but CONFIG_SCSI=m
[Which I don't understand ... HP_SIM_SCSI "depends on SCSI", so
how did make allmodconfig come up with this combination?].
-Tony
P.S. Next layer of the onion is CONFIG_BLK_DEV_AMD74XX ... perhaps
that needs to be "depends on X86"?
On Thu, Oct 12, 2006 at 10:55:36AM -0700, Luck, Tony wrote:
> P.S. Next layer of the onion is CONFIG_BLK_DEV_AMD74XX ... perhaps
> that needs to be "depends on X86"?
I sent patches to do that a few months ago.
They were shot down due to the alleged value of compile testing on
architectures which will never run those drivers.
Dave
--
http://www.codemonkey.org.uk
On Thu, 12 Oct 2006 10:55:36 -0700
"Luck, Tony" <[email protected]> wrote:
> On Thu, Oct 12, 2006 at 12:11:39AM -0700, Andrew Morton wrote:
> > The problem is that ia64 allmodconfig now bombs out, since depmod treats
> > this as a hard error.
> >
> > IOW, please make allmodconfig work ;)
>
> I tried simply swapping "tristate" for "bool":
> --- a/arch/ia64/hp/sim/Kconfig 2006-10-12 10:45:18.000000000 -0700
> +++ b/arch/ia64/hp/sim/Kconfig 2006-10-12 09:43:30.000000000 -0700
> @@ -13,7 +13,7 @@
> depends on HP_SIMSERIAL
>
> config HP_SIMSCSI
> - tristate "Simulated SCSI disk"
> + bool "Simulated SCSI disk"
> depends on SCSI
>
> endmenu
>
> ... and now it fails in a new an diferent way:
>
> LD .tmp_vmlinux1
> arch/ia64/hp/sim/built-in.o(.init.text+0x9d2): In function `simscsi_init':
> arch/ia64/hp/sim/simscsi.c:407: undefined reference to `scsi_host_alloc'
> arch/ia64/hp/sim/built-in.o(.init.text+0xa02):arch/ia64/hp/sim/simscsi.c:411: undefined reference to `scsi_add_host'
> arch/ia64/hp/sim/built-in.o(.init.text+0xa22):arch/ia64/hp/sim/simscsi.c:413: undefined reference to `scsi_scan_host'
>
> presumably because we have CONFIG_HP_SIMSCSI=y but CONFIG_SCSI=m
> [Which I don't understand ... HP_SIM_SCSI "depends on SCSI", so
> how did make allmodconfig come up with this combination?].
This happens a lot and I always forget what the fix is.
Something like
depends on SCSI=y || (m && SCSI)
but probably not exactly that.
[IA64] Fix allmodconfig build
The HP_SIMSCSI driver can't be built as a module (unhealthy
dependencies on things that shouldn't really be exported).
AMD and nVidia IDE support doesn't sound too useful for ia64
either :-)
Signed-off-by: Tony Luck <[email protected]>
---
With these two patches allmodconfig builds (but takes 11m24s, ouch!)
diff --git a/arch/ia64/hp/sim/Kconfig b/arch/ia64/hp/sim/Kconfig
index 18ccb12..f92306b 100644
--- a/arch/ia64/hp/sim/Kconfig
+++ b/arch/ia64/hp/sim/Kconfig
@@ -13,8 +13,8 @@ config HP_SIMSERIAL_CONSOLE
depends on HP_SIMSERIAL
config HP_SIMSCSI
- tristate "Simulated SCSI disk"
- depends on SCSI
+ bool "Simulated SCSI disk"
+ depends on SCSI=y
endmenu
diff --git a/drivers/ide/Kconfig b/drivers/ide/Kconfig
index 0c68d0f..12ab7c6 100644
--- a/drivers/ide/Kconfig
+++ b/drivers/ide/Kconfig
@@ -486,6 +486,7 @@ config WDC_ALI15X3
config BLK_DEV_AMD74XX
tristate "AMD and nVidia IDE support"
+ depends on X86
help
This driver adds explicit support for AMD-7xx and AMD-8111 chips
and also for the nVidia nForce chip. This allows the kernel to
On Thu, Oct 12, 2006 at 02:00:33PM -0700, Luck, Tony wrote:
> [IA64] Fix allmodconfig build
>
> The HP_SIMSCSI driver can't be built as a module (unhealthy
> dependencies on things that shouldn't really be exported).
>
> AMD and nVidia IDE support doesn't sound too useful for ia64
> either :-)
Thanks for doing this. I run cross compiles and do not always
know what is the best fix for the architecture. I have been trying
to address some of these errors so that the compiles for 'allmodconfig'
will be less 'noisy' and easier to get value out of.
Judith