Re: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

Re: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Natalie Protasevich]

>  > Can you please try the attached patch? I have compiled it with
> CONFIG_ES7000=y
>  > and CONFIG_X86_GENERICARCH=n. But don't have any ES7000 machine to
> test it.
>
> I actually don't have anything remotely like ES7000 iether.  I just
> found the problem while tracking down another randconfig failure.
>
>  > Subarch build procedure is not straightforward and this patch adds to
> the
>  > misery. Please suggest if there is a better way of handling things.
>
>  > +/*
>  > + * This file also gets compiled if CONFIG_X86_GENERICARCH is set.
> Generic
>  > + * arch already has got following function definitions
> (asm-generic/es7000.c)
>  > + * hence no need to define these for that case.
>  > + */
>  > +#ifndef CONFIG_X86_GENERICARCH
>
> Seems it would be cleaner to figure out some way to build es7000.c for
> if CONFIG_X86_ES7000 is set?  Or just move them here all the time?
>
>  > ---
> linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_mpparse.h~i386-e
> s7000-build-breakage-fix        2007-06-13 22:52:14.000000000 +0530
>  > +++
> linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h
> 2007-06-13 22:52:14.000000000 +0530
>  > @@ -18,6 +18,12 @@ extern int parse_unisys_oem (char *oempt
>  >  extern int find_unisys_acpi_oem_table(unsigned long *oem_addr);
>  >  extern void setup_unisys(void);
>  >
>  > +#ifndef CONFIG_X86_GENERICARCH
>  > +extern int acpi_madt_oem_check(char *oem_id, char *oem_table_id);
>  > +extern int mps_oem_check(struct mp_config_table *mpc, char *oem,
>  > +                            char *productid);
>  > +#endif
>
> It seems that this #ifndef is not needed -- even if there is another
> declaration of these functions that is visible, the signatures should
> match so it should be OK.

I just finished my test builds with and without GENERIC, so I can
suggest just move all functions to arch/i386/mach-es7000/es7000plat.c
- es7000_sw_apic() ,es7000_enable_apic_mode(),acpi_madt_oem_check(),
mps_oem_check() and make them non-static - from
arch/i386/mach-generic/es7000.c and I think I externed couple things
in asm/mach-es7000/mach_apic.h. So the file
arch/i386/mach-generic/es7000.c looks very simple:
/*  * APIC driver for the Unisys ES7000 chipset.
 */
#define APIC_DEFINITION 1
#include <linux/threads.h>
#include <linux/cpumask.h>
#include <asm/smp.h>
#include <asm/mpspec.h>
#include <asm/genapic.h>
#include <asm/fixmap.h>
#include <asm/apicdef.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/smp.h>
#include <linux/init.h>
#include <asm/mach-es7000/mach_apicdef.h>
#include <asm/mach-es7000/mach_apic.h>
#include <asm/mach-es7000/mach_ipi.h>
#include <asm/mach-es7000/mach_mpparse.h>
#include <asm/mach-es7000/mach_wakecpu.h>

static int probe_es7000(void)
{
        /* probed later in mptable/ACPI hooks */
        return 0;
}

struct genapic apic_es7000 = APIC_INIT("es7000", probe_es7000);


and no ifdefs. Sorry for this skimpy description without actual patch...
I can send a patch later on, or you guys can try it yourselves :)
sorry have such a tight time this afternoon, only can back to my
computer after 7-8 pm...
Regards,
--Natalie

Re: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Vivek Goyal]

On Wed, Jun 13, 2007 at 03:22:55PM -0700, Natalie Protasevich wrote:
[..]
> >Seems it would be cleaner to figure out some way to build es7000.c for
> >if CONFIG_X86_ES7000 is set?  Or just move them here all the time?
> >
> > > ---
> >linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_mpparse.h~i386-e
> >s7000-build-breakage-fix        2007-06-13 22:52:14.000000000 +0530
> > > +++
> >linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h
> >2007-06-13 22:52:14.000000000 +0530
> > > @@ -18,6 +18,12 @@ extern int parse_unisys_oem (char *oempt
> > >  extern int find_unisys_acpi_oem_table(unsigned long *oem_addr);
> > >  extern void setup_unisys(void);
> > >
> > > +#ifndef CONFIG_X86_GENERICARCH
> > > +extern int acpi_madt_oem_check(char *oem_id, char *oem_table_id);
> > > +extern int mps_oem_check(struct mp_config_table *mpc, char *oem,
> > > +                            char *productid);
> > > +#endif
> >
> >It seems that this #ifndef is not needed -- even if there is another
> >declaration of these functions that is visible, the signatures should
> >match so it should be OK.
> 
> I just finished my test builds with and without GENERIC, so I can
> suggest just move all functions to arch/i386/mach-es7000/es7000plat.c
> - es7000_sw_apic() ,es7000_enable_apic_mode(),acpi_madt_oem_check(),
> mps_oem_check() and make them non-static - from
> arch/i386/mach-generic/es7000.c

I think this will also break with CONFIG_X86_GENERIC=y. The moment you
make es7000 specific acpi_madt_oem_check() non static, It will conflict
with generic definition present in mach-generic/probe.c. That function is
a generic function designed to probe for underlying subarch.

Thanks
Vivek

Re: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Natalie Protasevich]

On 6/15/07, Vivek Goyal <vgoyal@in.ibm.com> wrote:
> On Wed, Jun 13, 2007 at 03:22:55PM -0700, Natalie Protasevich wrote:
> [..]
> > >Seems it would be cleaner to figure out some way to build es7000.c for
> > >if CONFIG_X86_ES7000 is set?  Or just move them here all the time?
> > >
> > > > ---
> > >linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_mpparse.h~i386-e
> > >s7000-build-breakage-fix        2007-06-13 22:52:14.000000000 +0530
> > > > +++
> > >linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h
> > >2007-06-13 22:52:14.000000000 +0530
> > > > @@ -18,6 +18,12 @@ extern int parse_unisys_oem (char *oempt
> > > >  extern int find_unisys_acpi_oem_table(unsigned long *oem_addr);
> > > >  extern void setup_unisys(void);
> > > >
> > > > +#ifndef CONFIG_X86_GENERICARCH
> > > > +extern int acpi_madt_oem_check(char *oem_id, char *oem_table_id);
> > > > +extern int mps_oem_check(struct mp_config_table *mpc, char *oem,
> > > > +                            char *productid);
> > > > +#endif
> > >
> > >It seems that this #ifndef is not needed -- even if there is another
> > >declaration of these functions that is visible, the signatures should
> > >match so it should be OK.
> >
> > I just finished my test builds with and without GENERIC, so I can
> > suggest just move all functions to arch/i386/mach-es7000/es7000plat.c
> > - es7000_sw_apic() ,es7000_enable_apic_mode(),acpi_madt_oem_check(),
> > mps_oem_check() and make them non-static - from
> > arch/i386/mach-generic/es7000.c
>
> I think this will also break with CONFIG_X86_GENERIC=y. The moment you
> make es7000 specific acpi_madt_oem_check() non static, It will conflict
> with generic definition present in mach-generic/probe.c. That function is
> a generic function designed to probe for underlying subarch.

I am not sure why it needs to be static, since the genapic structure
in include/asm-i386/ just being a switch table for the corresponding
functions of the right type. It compiles for me, and i just need to
extern those in include/asm/mach-es7000/mach_apic.h:

...
extern void es7000_sw_apic(void);
extern void __init enable_apic_mode(void);

extern __init int mps_oem_check(struct mp_config_table *mpc, char
*oem, char *productid);
extern int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id);
...

or do this in mach_mpparse.h - wherever I removed them from. I
apologize again, will make a descent patch shortly just so we can look
closer and evaluate possible approach.

Regards,
--Natalie

RE: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Natalie Protasevich]

> On Wed, Jun 13, 2007 at 01:39:29PM -0700, Roland Dreier wrote:
> >  > Can you please try the attached patch? I have compiled it with
> CONFIG_ES7000=y
> >  > and CONFIG_X86_GENERICARCH=n. But don't have any ES7000 machine to
> test it.
> >
> > I actually don't have anything remotely like ES7000 iether.  I just
> > found the problem while tracking down another randconfig failure.
> >
> >  > Subarch build procedure is not straightforward and this patch adds
> to the
> >  > misery. Please suggest if there is a better way of handling things.
> >
> >  > +/*
> >  > + * This file also gets compiled if CONFIG_X86_GENERICARCH is set.
> Generic
> >  > + * arch already has got following function definitions
> (asm-generic/es7000.c)
> >  > + * hence no need to define these for that case.
> >  > + */
> >  > +#ifndef CONFIG_X86_GENERICARCH
> >
> > Seems it would be cleaner to figure out some way to build es7000.c for
> > if CONFIG_X86_ES7000 is set?  Or just move them here all the time?
> >
> >  > ---
> linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_mpparse.h~i386-e
> s7000-build-breakage-fix        2007-06-13 22:52:14.000000000 +0530
> >  > +++
> linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h
> 2007-06-13 22:52:14.000000000 +0530
> >  > @@ -18,6 +18,12 @@ extern int parse_unisys_oem (char *oempt
> >  >  extern int find_unisys_acpi_oem_table(unsigned long *oem_addr);
> >  >  extern void setup_unisys(void);
> >  >
> >  > +#ifndef CONFIG_X86_GENERICARCH
> >  > +extern int acpi_madt_oem_check(char *oem_id, char *oem_table_id);
> >  > +extern int mps_oem_check(struct mp_config_table *mpc, char *oem,
> >  > +                          char *productid);
> >  > +#endif
> >
> > It seems that this #ifndef is not needed -- even if there is another
> > declaration of these functions that is visible, the signatures should
> > match so it should be OK.
> >
>
> Compiler does throw warning as one set of functions is static and
> other is not.
>
> arch/i386/mach-generic/es7000.c:37: warning: static declaration of
> 'mps_oem_check' follows non-static declaration
> include/asm/mach-es7000/mach_mpparse.h:23: warning: previous declaration
> of 'mps_oem_check' was here
> arch/i386/mach-generic/es7000.c:50: warning: static declaration of
> 'acpi_madt_oem_check' follows non-static declaration
> include/asm/mach-es7000/mach_mpparse.h:21: warning: previous declaration
> of 'acpi_madt_oem_check' was here

Here is the patch that does work - but... I do have warnings that I
haven't noticed initially when compile for ES7000 case:

...
  CC      arch/i386/kernel/acpi/boot.o
include/asm-i386/mach-es7000/mach_mpparse.h:56: warning:
'mps_oem_check' defined but not used
  CC      arch/i386/kernel/mpparse.o
...
include/asm-i386/mach-es7000/mach_mpparse.h:35: warning:
'acpi_madt_oem_check' defined but not used
  CC      arch/i386/kernel/apic.o
...
  CC      arch/i386/kernel/nmi.o

I'm afraid I will have to do something ugly or ifdef to get rid of
this warning, so I guess I'd rather take yours, Vivec :) since it
compiles and runs cleanly.
Thanks!
--Natalie

---------------------------------------------------------------------------------
diff -Naur linux-2.6/arch/i386/mach-es7000/es7000plat.c
linux-2.6-mod/arch/i386/mach-es7000/es7000plat.c---
linux-2.6/arch/i386/mach-es7000/es7000plat.c        2007-06-12
21:16:15.000000000 -0700
+++ linux-2.6-mod/arch/i386/mach-es7000/es7000plat.c    2007-06-17
06:56:49.000000000 -0700
@@ -277,3 +277,11 @@
                return;
        }
 }
+
+void __init
+enable_apic_mode(void)
+{
+       es7000_sw_apic();
+       return;
+}
+
diff -Naur linux-2.6/arch/i386/mach-generic/es7000.c
linux-2.6-mod/arch/i386/mach-generic/es7000.c
--- linux-2.6/arch/i386/mach-generic/es7000.c   2007-06-12
21:16:15.000000000 -0700
+++ linux-2.6-mod/arch/i386/mach-generic/es7000.c       2007-06-17
06:57:20.000000000 -0700
@@ -26,44 +26,6 @@
 }

 extern void es7000_sw_apic(void);
-static void __init enable_apic_mode(void)
-{
-       es7000_sw_apic();
-       return;
-}
-
-static __init int mps_oem_check(struct mp_config_table *mpc, char *oem,
-               char *productid)
-{
-       if (mpc->mpc_oemptr) {
-               struct mp_config_oemtable *oem_table =
-                       (struct mp_config_oemtable *)mpc->mpc_oemptr;
-               if (!strncmp(oem, "UNISYS", 6))
-                       return parse_unisys_oem((char *)oem_table);
-       }
-       return 0;
-}
-
-#ifdef CONFIG_ACPI
-/* Hook from generic ACPI tables.c */
-static int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
-{
-       unsigned long oem_addr;
-       if (!find_unisys_acpi_oem_table(&oem_addr)) {
-               if (es7000_check_dsdt())
-                       return parse_unisys_oem((char *)oem_addr);
-               else {
-                       setup_unisys();
-                       return 1;
-               }
-       }
-       return 0;
-}
-#else
-static int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
-{
-       return 0;
-}
-#endif
+extern void enable_apic_mode(void);

 struct genapic apic_es7000 = APIC_INIT("es7000", probe_es7000); diff
-Naur linux-2.6/include/asm-i386/mach-es7000/mach_apic.h
linux-2.6-mod/include/asm-i386/mach-es7000/mach_apic.h
--- linux-2.6/include/asm-i386/mach-es7000/mach_apic.h  2007-06-12
21:16:18.000000000 -0700
+++ linux-2.6-mod/include/asm-i386/mach-es7000/mach_apic.h
2007-06-17 06:51:48.000000000 -0700
@@ -199,4 +199,6 @@
        return cpuid_apic >> index_msb;
 }

+extern void enable_apic_mode(void);
+
 #endif /* __ASM_MACH_APIC_H */ diff -Naur
linux-2.6/include/asm-i386/mach-es7000/mach_mpparse.h
linux-2.6-mod/include/asm-i386/mach-es7000/mach_mpparse.h
--- linux-2.6/include/asm-i386/mach-es7000/mach_mpparse.h
2007-06-12 21:16:18.000000000 -0700
+++ linux-2.6-mod/include/asm-i386/mach-es7000/mach_mpparse.h
2007-06-17 06:54:54.000000000 -0700
@@ -29,6 +29,38 @@
                return 1;
        return 0;
 }
+
+/* Hook from generic ACPI tables.c */
+static int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
+{
+       unsigned long oem_addr;
+       if (!find_unisys_acpi_oem_table(&oem_addr)) {
+               if (es7000_check_dsdt())
+                       return parse_unisys_oem((char *)oem_addr);
+               else {
+                       setup_unisys();
+                       return 1;
+               }
+       }
 #endif

+static __init int mps_oem_check(struct mp_config_table *mpc, char *oem,
+               char *productid)
+{
+       if (mpc->mpc_oemptr) {
+               struct mp_config_oemtable *oem_table =
+                       (struct mp_config_oemtable *)mpc->mpc_oemptr;
+               if (!strncmp(oem, "UNISYS", 6))
+                       return parse_unisys_oem((char *)oem_table);
+       }
+       return 0;
+}
+
 #endif /* __ASM_MACH_MPPARSE_H */

RE: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Natalie Protasevich]

> From: Vivek Goyal [mailto:vgoyal@in.ibm.com]
> Sent: Wednesday, June 13, 2007 11:38 AM
> To: Roland Dreier
> Cc: linux-kernel@vger.kernel.org; Protasevich, Natalie;
> akpm@linux-foundation.org
> Subject: Re: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n,
> CONFIG_ACPI=y build broken
>
> On Tue, Jun 12, 2007 at 12:13:50PM -0700, Roland Dreier wrote:
> > (sending to Vivek since he seems to be one of the last people to touch
> > this code)
> >
> > For some strange reason I found myself compiling a kernel with
> > CONFIG_X86_ES7000 and CONFIG_ACPI set and CONFIG_X86_GENERICARCH not
> > set.  It died with
> >
> >     arch/i386/kernel/built-in.o: In function `acpi_parse_madt':
> >     boot.c:(.init.text+0x68b7): undefined reference to
> `acpi_madt_oem_check'
> >     arch/i386/kernel/built-in.o: In function `get_smp_config':
> >     (.init.text+0x9f98): undefined reference to `mps_oem_check'
> >     arch/i386/kernel/built-in.o: In function `connect_bsp_APIC':
> >     (.init.text+0xa340): undefined reference to `enable_apic_mode'
> >     arch/i386/kernel/built-in.o: In function `connect_bsp_APIC':
> >     (.init.text+0xa370): undefined reference to `enable_apic_mode'
> >
> > this seems to be because eg acpi_madt_oem_check() and mps_oem_check()
> > are defined in arch/i386/mach-generic/es7000.c, but arch/i386/Makefile
> > only goes into mach-generic if CONFIG_X86_GENERICARCH is set.
> > I didn't track down why enable_apic_mode is broken yet.
> >
>

Vivek,
I also have a patch that seems to work... I did remove the functions
from mach-generic/es7000.c and added them same place you did, so the
code is not duplicated and I didn't have to make any ifdefs.
As soon as your patch works though I think it's fine.
Thanks,
--Natalie

> Hi Roland,
>
> Can you please try the attached patch? I have compiled it with
> CONFIG_ES7000=y
> and CONFIG_X86_GENERICARCH=n. But don't have any ES7000 machine to test
> it.
>
> I have compiled and booted the kernel on my machine with
> CONFIG_X86_GENERICARCH=y.
>
> Subarch build procedure is not straightforward and this patch adds to
> the
> misery. Please suggest if there is a better way of handling things.
>
>
>
> o Commit 1833d6bc72893265f22addd79cf52e6987496e0f broke the build if
>   compiled with CONFIG_ES7000=y and CONFIG_X86_GENERICARCH=n
>
> o Fix the build issue. Provided the definitions of missing functions.
>
>
> Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
> ---
>
>  linux-2.6.22-rc4-git4-vivek/arch/i386/mach-es7000/es7000plat.c
> |   48 ++++++++++
>  linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_apic.h
> |    4
>  linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h
> |    6 +
>  3 files changed, 58 insertions(+)
>
> diff -puN
> arch/i386/mach-es7000/es7000plat.c~i386-es7000-build-breakage-fix
> arch/i386/mach-es7000/es7000plat.c
> ---
> linux-2.6.22-rc4-git4/arch/i386/mach-es7000/es7000plat.c~i386-es7000-bui
> ld-breakage-fix 2007-06-13 22:52:14.000000000 +0530
> +++ linux-2.6.22-rc4-git4-vivek/arch/i386/mach-es7000/es7000plat.c
> 2007-06-13 22:53:03.000000000 +0530
> @@ -40,6 +40,7 @@
>  #include <asm/smp.h>
>  #include <asm/apicdef.h>
>  #include "es7000.h"
> +#include <mach_mpparse.h>
>
>  /*
>   * ES7000 Globals
> @@ -174,6 +175,53 @@ find_unisys_acpi_oem_table(unsigned long
>  }
>  #endif
>
> +/*
> + * This file also gets compiled if CONFIG_X86_GENERICARCH is set.
> Generic
> + * arch already has got following function definitions
> (asm-generic/es7000.c)
> + * hence no need to define these for that case.
> + */
> +#ifndef CONFIG_X86_GENERICARCH
> +void es7000_sw_apic(void);
> +void __init enable_apic_mode(void)
> +{
> +       es7000_sw_apic();
> +       return;
> +}
> +
> +__init int mps_oem_check(struct mp_config_table *mpc, char *oem,
> +               char *productid)
> +{
> +       if (mpc->mpc_oemptr) {
> +               struct mp_config_oemtable *oem_table =
> +                       (struct mp_config_oemtable *)mpc->mpc_oemptr;
> +               if (!strncmp(oem, "UNISYS", 6))
> +                       return parse_unisys_oem((char *)oem_table);
> +       }
> +       return 0;
> +}
> +#ifdef CONFIG_ACPI
> +/* Hook from generic ACPI tables.c */
> +int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
> +{
> +       unsigned long oem_addr;
> +       if (!find_unisys_acpi_oem_table(&oem_addr)) {
> +               if (es7000_check_dsdt())
> +                       return parse_unisys_oem((char *)oem_addr);
> +               else {
> +                       setup_unisys();
> +                       return 1;
> +               }
> +       }
> +       return 0;
> +}
> +#else
> +int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
> +{
> +       return 0;
> +}
> +#endif
> +#endif /* COFIG_X86_GENERICARCH */
> +
>  static void
>  es7000_spin(int n)
>  {
> diff -puN
> include/asm-i386/mach-es7000/mach_apic.h~i386-es7000-build-breakage-fix
> include/asm-i386/mach-es7000/mach_apic.h
> ---
> linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_apic.h~i386-es70
> 00-build-breakage-fix   2007-06-13 22:52:14.000000000 +0530
> +++ linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_apic.h
> 2007-06-13 22:52:14.000000000 +0530
> @@ -73,6 +73,10 @@ static inline void init_apic_ldr(void)
>         apic_write_around(APIC_LDR, val);
>  }
>
> +#ifndef CONFIG_X86_GENERICARCH
> +extern void enable_apic_mode(void);
> +#endif
> +
>  extern int apic_version [MAX_APICS];
>  static inline void setup_apic_routing(void)
>  {
> diff -puN
> include/asm-i386/mach-es7000/mach_mpparse.h~i386-es7000-build-breakage-f
> ix include/asm-i386/mach-es7000/mach_mpparse.h
> ---
> linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_mpparse.h~i386-e
> s7000-build-breakage-fix        2007-06-13 22:52:14.000000000 +0530
> +++
> linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h
> 2007-06-13 22:52:14.000000000 +0530
> @@ -18,6 +18,12 @@ extern int parse_unisys_oem (char *oempt
>  extern int find_unisys_acpi_oem_table(unsigned long *oem_addr);
>  extern void setup_unisys(void);
>
> +#ifndef CONFIG_X86_GENERICARCH
> +extern int acpi_madt_oem_check(char *oem_id, char *oem_table_id);
> +extern int mps_oem_check(struct mp_config_table *mpc, char *oem,
> +                               char *productid);
> +#endif
> +
>  #ifdef CONFIG_ACPI
>
>  static inline int es7000_check_dsdt(void)
> _
>

CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Roland Dreier]

(sending to Vivek since he seems to be one of the last people to touch
this code)

For some strange reason I found myself compiling a kernel with
CONFIG_X86_ES7000 and CONFIG_ACPI set and CONFIG_X86_GENERICARCH not
set.  It died with

    arch/i386/kernel/built-in.o: In function `acpi_parse_madt':
    boot.c:(.init.text+0x68b7): undefined reference to `acpi_madt_oem_check'
    arch/i386/kernel/built-in.o: In function `get_smp_config':
    (.init.text+0x9f98): undefined reference to `mps_oem_check'
    arch/i386/kernel/built-in.o: In function `connect_bsp_APIC':
    (.init.text+0xa340): undefined reference to `enable_apic_mode'
    arch/i386/kernel/built-in.o: In function `connect_bsp_APIC':
    (.init.text+0xa370): undefined reference to `enable_apic_mode'

this seems to be because eg acpi_madt_oem_check() and mps_oem_check()
are defined in arch/i386/mach-generic/es7000.c, but arch/i386/Makefile
only goes into mach-generic if CONFIG_X86_GENERICARCH is set.
I didn't track down why enable_apic_mode is broken yet.

No patch because I don't understand how this subarch stuff is supposed
to build :(

 - R.

Re: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Vivek Goyal]

On Tue, Jun 12, 2007 at 12:13:50PM -0700, Roland Dreier wrote:
> (sending to Vivek since he seems to be one of the last people to touch
> this code)
> 
> For some strange reason I found myself compiling a kernel with
> CONFIG_X86_ES7000 and CONFIG_ACPI set and CONFIG_X86_GENERICARCH not
> set.  It died with
> 
>     arch/i386/kernel/built-in.o: In function `acpi_parse_madt':
>     boot.c:(.init.text+0x68b7): undefined reference to `acpi_madt_oem_check'
>     arch/i386/kernel/built-in.o: In function `get_smp_config':
>     (.init.text+0x9f98): undefined reference to `mps_oem_check'
>     arch/i386/kernel/built-in.o: In function `connect_bsp_APIC':
>     (.init.text+0xa340): undefined reference to `enable_apic_mode'
>     arch/i386/kernel/built-in.o: In function `connect_bsp_APIC':
>     (.init.text+0xa370): undefined reference to `enable_apic_mode'
> 
> this seems to be because eg acpi_madt_oem_check() and mps_oem_check()
> are defined in arch/i386/mach-generic/es7000.c, but arch/i386/Makefile
> only goes into mach-generic if CONFIG_X86_GENERICARCH is set.
> I didn't track down why enable_apic_mode is broken yet.
>

Hi Roland,

Can you please try the attached patch? I have compiled it with CONFIG_ES7000=y
and CONFIG_X86_GENERICARCH=n. But don't have any ES7000 machine to test it.

I have compiled and booted the kernel on my machine with CONFIG_X86_GENERICARCH=y.

Subarch build procedure is not straightforward and this patch adds to the
misery. Please suggest if there is a better way of handling things.

 

o Commit 1833d6bc72893265f22addd79cf52e6987496e0f broke the build if
  compiled with CONFIG_ES7000=y and CONFIG_X86_GENERICARCH=n

o Fix the build issue. Provided the definitions of missing functions.


Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
---

 linux-2.6.22-rc4-git4-vivek/arch/i386/mach-es7000/es7000plat.c          |   48 ++++++++++
 linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_apic.h    |    4 
 linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h |    6 +
 3 files changed, 58 insertions(+)

diff -puN arch/i386/mach-es7000/es7000plat.c~i386-es7000-build-breakage-fix arch/i386/mach-es7000/es7000plat.c
--- linux-2.6.22-rc4-git4/arch/i386/mach-es7000/es7000plat.c~i386-es7000-build-breakage-fix	2007-06-13 22:52:14.000000000 +0530
+++ linux-2.6.22-rc4-git4-vivek/arch/i386/mach-es7000/es7000plat.c	2007-06-13 22:53:03.000000000 +0530
@@ -40,6 +40,7 @@
 #include <asm/smp.h>
 #include <asm/apicdef.h>
 #include "es7000.h"
+#include <mach_mpparse.h>
 
 /*
  * ES7000 Globals
@@ -174,6 +175,53 @@ find_unisys_acpi_oem_table(unsigned long
 }
 #endif
 
+/*
+ * This file also gets compiled if CONFIG_X86_GENERICARCH is set. Generic
+ * arch already has got following function definitions (asm-generic/es7000.c)
+ * hence no need to define these for that case.
+ */
+#ifndef CONFIG_X86_GENERICARCH
+void es7000_sw_apic(void);
+void __init enable_apic_mode(void)
+{
+	es7000_sw_apic();
+	return;
+}
+
+__init int mps_oem_check(struct mp_config_table *mpc, char *oem,
+		char *productid)
+{
+	if (mpc->mpc_oemptr) {
+		struct mp_config_oemtable *oem_table =
+			(struct mp_config_oemtable *)mpc->mpc_oemptr;
+		if (!strncmp(oem, "UNISYS", 6))
+			return parse_unisys_oem((char *)oem_table);
+	}
+	return 0;
+}
+#ifdef CONFIG_ACPI
+/* Hook from generic ACPI tables.c */
+int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
+{
+	unsigned long oem_addr;
+	if (!find_unisys_acpi_oem_table(&oem_addr)) {
+		if (es7000_check_dsdt())
+			return parse_unisys_oem((char *)oem_addr);
+		else {
+			setup_unisys();
+			return 1;
+		}
+	}
+	return 0;
+}
+#else
+int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
+{
+	return 0;
+}
+#endif
+#endif /* COFIG_X86_GENERICARCH */
+
 static void
 es7000_spin(int n)
 {
diff -puN include/asm-i386/mach-es7000/mach_apic.h~i386-es7000-build-breakage-fix include/asm-i386/mach-es7000/mach_apic.h
--- linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_apic.h~i386-es7000-build-breakage-fix	2007-06-13 22:52:14.000000000 +0530
+++ linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_apic.h	2007-06-13 22:52:14.000000000 +0530
@@ -73,6 +73,10 @@ static inline void init_apic_ldr(void)
 	apic_write_around(APIC_LDR, val);
 }
 
+#ifndef CONFIG_X86_GENERICARCH
+extern void enable_apic_mode(void);
+#endif
+
 extern int apic_version [MAX_APICS];
 static inline void setup_apic_routing(void)
 {
diff -puN include/asm-i386/mach-es7000/mach_mpparse.h~i386-es7000-build-breakage-fix include/asm-i386/mach-es7000/mach_mpparse.h
--- linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_mpparse.h~i386-es7000-build-breakage-fix	2007-06-13 22:52:14.000000000 +0530
+++ linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h	2007-06-13 22:52:14.000000000 +0530
@@ -18,6 +18,12 @@ extern int parse_unisys_oem (char *oempt
 extern int find_unisys_acpi_oem_table(unsigned long *oem_addr);
 extern void setup_unisys(void);
 
+#ifndef CONFIG_X86_GENERICARCH
+extern int acpi_madt_oem_check(char *oem_id, char *oem_table_id);
+extern int mps_oem_check(struct mp_config_table *mpc, char *oem,
+				char *productid);
+#endif
+
 #ifdef CONFIG_ACPI
 
 static inline int es7000_check_dsdt(void)
_

Re: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Roland Dreier]

 > Can you please try the attached patch? I have compiled it with CONFIG_ES7000=y
 > and CONFIG_X86_GENERICARCH=n. But don't have any ES7000 machine to test it.

I actually don't have anything remotely like ES7000 iether.  I just
found the problem while tracking down another randconfig failure.

 > Subarch build procedure is not straightforward and this patch adds to the
 > misery. Please suggest if there is a better way of handling things.

 > +/*
 > + * This file also gets compiled if CONFIG_X86_GENERICARCH is set. Generic
 > + * arch already has got following function definitions (asm-generic/es7000.c)
 > + * hence no need to define these for that case.
 > + */
 > +#ifndef CONFIG_X86_GENERICARCH

Seems it would be cleaner to figure out some way to build es7000.c for
if CONFIG_X86_ES7000 is set?  Or just move them here all the time?

 > --- linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_mpparse.h~i386-es7000-build-breakage-fix	2007-06-13 22:52:14.000000000 +0530
 > +++ linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h	2007-06-13 22:52:14.000000000 +0530
 > @@ -18,6 +18,12 @@ extern int parse_unisys_oem (char *oempt
 >  extern int find_unisys_acpi_oem_table(unsigned long *oem_addr);
 >  extern void setup_unisys(void);
 >  
 > +#ifndef CONFIG_X86_GENERICARCH
 > +extern int acpi_madt_oem_check(char *oem_id, char *oem_table_id);
 > +extern int mps_oem_check(struct mp_config_table *mpc, char *oem,
 > +				char *productid);
 > +#endif

It seems that this #ifndef is not needed -- even if there is another
declaration of these functions that is visible, the signatures should
match so it should be OK.

 - R.

Re: CONFIG_X86_ES7000=y, CONFIG_X86_GENERICARCH=n, CONFIG_ACPI=y build broken

[Vivek Goyal]

On Wed, Jun 13, 2007 at 01:39:29PM -0700, Roland Dreier wrote:
>  > Can you please try the attached patch? I have compiled it with CONFIG_ES7000=y
>  > and CONFIG_X86_GENERICARCH=n. But don't have any ES7000 machine to test it.
> 
> I actually don't have anything remotely like ES7000 iether.  I just
> found the problem while tracking down another randconfig failure.
> 
>  > Subarch build procedure is not straightforward and this patch adds to the
>  > misery. Please suggest if there is a better way of handling things.
> 
>  > +/*
>  > + * This file also gets compiled if CONFIG_X86_GENERICARCH is set. Generic
>  > + * arch already has got following function definitions (asm-generic/es7000.c)
>  > + * hence no need to define these for that case.
>  > + */
>  > +#ifndef CONFIG_X86_GENERICARCH
> 
> Seems it would be cleaner to figure out some way to build es7000.c for
> if CONFIG_X86_ES7000 is set?  Or just move them here all the time?
> 
>  > --- linux-2.6.22-rc4-git4/include/asm-i386/mach-es7000/mach_mpparse.h~i386-es7000-build-breakage-fix	2007-06-13 22:52:14.000000000 +0530
>  > +++ linux-2.6.22-rc4-git4-vivek/include/asm-i386/mach-es7000/mach_mpparse.h	2007-06-13 22:52:14.000000000 +0530
>  > @@ -18,6 +18,12 @@ extern int parse_unisys_oem (char *oempt
>  >  extern int find_unisys_acpi_oem_table(unsigned long *oem_addr);
>  >  extern void setup_unisys(void);
>  >  
>  > +#ifndef CONFIG_X86_GENERICARCH
>  > +extern int acpi_madt_oem_check(char *oem_id, char *oem_table_id);
>  > +extern int mps_oem_check(struct mp_config_table *mpc, char *oem,
>  > +				char *productid);
>  > +#endif
> 
> It seems that this #ifndef is not needed -- even if there is another
> declaration of these functions that is visible, the signatures should
> match so it should be OK.
> 

Compiler does throw warning as one set of functions is static and
other is not.

arch/i386/mach-generic/es7000.c:37: warning: static declaration of 'mps_oem_check' follows non-static declaration
include/asm/mach-es7000/mach_mpparse.h:23: warning: previous declaration of 'mps_oem_check' was here
arch/i386/mach-generic/es7000.c:50: warning: static declaration of 'acpi_madt_oem_check' follows non-static declaration
include/asm/mach-es7000/mach_mpparse.h:21: warning: previous declaration of 'acpi_madt_oem_check' was here

Thanks
Vivek