[PATCH linux-next] x86_32: fix VisualWS and Voyager kexec build failures

[PATCH linux-next] x86_32: fix VisualWS and Voyager kexec build failures

[Randy Dunlap]

From: Randy Dunlap <randy.dunlap@oracle.com>

cc: Eric Biederman <ebiederm@xmission.com>

Both Visual WS and Voyager builds fail in almost the same way (in
linux-next) without this patch:

VOYAGER:
kernel/built-in.o: In function `crash_kexec':
(.text+0x28588): undefined reference to `machine_crash_shutdown'

VISWS:
kernel/built-in.o: In function `crash_kexec':
/next-20080401/kernel/kexec.c:1074: undefined reference to `machine_crash_shutdown'
make[1]: *** [.tmp_vmlinux1] Error 1

because arch/x86/kernel/reboot.c isn't built since CONFIG_X86_BIOS_REBOOT=n,
so machine_crash_shutdown() isn't available.

This patch does seem a small bit odd since the KEXEC help text says that
kexec is independent of the system firmware.

Eric, is there some other way that this should be handled?

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
---
 arch/x86/Kconfig |    1 +
 1 file changed, 1 insertion(+)

--- next-20080401.orig/arch/x86/Kconfig
+++ next-20080401/arch/x86/Kconfig
@@ -1094,6 +1094,7 @@ source kernel/Kconfig.hz
 
 config KEXEC
 	bool "kexec system call"
+	depends on X86_64 || X86_BIOS_REBOOT
 	help
 	  kexec is a system call that implements the ability to shutdown your
 	  current kernel, and to start another kernel.  It is like a reboot

Re: [PATCH linux-next] x86_32: fix VisualWS and Voyager kexec build failures

[Eric W. Biederman]

Randy Dunlap <randy.dunlap@oracle.com> writes:

> From: Randy Dunlap <randy.dunlap@oracle.com>
>
> cc: Eric Biederman <ebiederm@xmission.com>
>
> Both Visual WS and Voyager builds fail in almost the same way (in
> linux-next) without this patch:
>
> VOYAGER:
> kernel/built-in.o: In function `crash_kexec':
> (.text+0x28588): undefined reference to `machine_crash_shutdown'
>
> VISWS:
> kernel/built-in.o: In function `crash_kexec':
> /next-20080401/kernel/kexec.c:1074: undefined reference to
> `machine_crash_shutdown'
> make[1]: *** [.tmp_vmlinux1] Error 1
>
> because arch/x86/kernel/reboot.c isn't built since CONFIG_X86_BIOS_REBOOT=n,
> so machine_crash_shutdown() isn't available.

Weird.  I haven't had a chance to update to the devel kernels lately.
And in the older kernel I have machine_crash_shutdown is in crash.c
and is indeed not dependent xyz.

I get the feeling someone refactored something and ran afoul of the
x86_32 weird subarchitecture stuff in their testing.

> This patch does seem a small bit odd since the KEXEC help text says that
> kexec is independent of the system firmware.

Yes.

> Eric, is there some other way that this should be handled?

Yes.  Move machine_crash_shutdown back into crash.c
Or find some other way to accomplish whatever cleanup was done,
so that we still compile.

Eric

Re: [PATCH linux-next] x86_32: fix VisualWS and Voyager kexec build failures

[Randy Dunlap]

On Wed, 02 Apr 2008 00:49:15 -0600 Eric W. Biederman wrote:

> Randy Dunlap <randy.dunlap@oracle.com> writes:
> 
> > From: Randy Dunlap <randy.dunlap@oracle.com>
> >
> > cc: Eric Biederman <ebiederm@xmission.com>
> >
> > Both Visual WS and Voyager builds fail in almost the same way (in
> > linux-next) without this patch:
> >
> > VOYAGER:
> > kernel/built-in.o: In function `crash_kexec':
> > (.text+0x28588): undefined reference to `machine_crash_shutdown'
> >
> > VISWS:
> > kernel/built-in.o: In function `crash_kexec':
> > /next-20080401/kernel/kexec.c:1074: undefined reference to
> > `machine_crash_shutdown'
> > make[1]: *** [.tmp_vmlinux1] Error 1
> >
> > because arch/x86/kernel/reboot.c isn't built since CONFIG_X86_BIOS_REBOOT=n,
> > so machine_crash_shutdown() isn't available.
> 
> Weird.  I haven't had a chance to update to the devel kernels lately.
> And in the older kernel I have machine_crash_shutdown is in crash.c
> and is indeed not dependent xyz.
> 
> I get the feeling someone refactored something and ran afoul of the
> x86_32 weird subarchitecture stuff in their testing.
> 
> > This patch does seem a small bit odd since the KEXEC help text says that
> > kexec is independent of the system firmware.
> 
> Yes.
> 
> > Eric, is there some other way that this should be handled?
> 
> Yes.  Move machine_crash_shutdown back into crash.c
> Or find some other way to accomplish whatever cleanup was done,
> so that we still compile.

Well.  This is a problem in linux-next and in -mm, not yet in
mainline.  Hopefully this patch (wherever it is) won't be merged
into mainline in its present form.

---
~Randy