[Xenomai-help] Xenomai for SH4

[Xenomai-help] Xenomai for SH4

[Pavel Bartusek]

Hello,

I'd like to use Xenomai on SH4 architecture and it seems it is not in
the list of supported architectures now. There were some attempt in the
past to port Xenomai to SH4 architecture
(http://article.gmane.org/gmane.linux.real-time.xenomai.users/858/match=xenomai+sh4)
Unfortunately, it seems it the patch hasn't been integrated to the
mainstream Xenomai. Does somebody have the patch or is somebody working
on the port  for SH4 now ?

Thanks for your answers
Pavel Bartůšek

[Xenomai-help] compile error xenomai 2.4.5

[George Pitich]

[-- Attachment #1: Type: text/plain, Size: 559 bytes --]

Tried to compile linux 2.6.26.5 kernel with xenomai 2.4.5 (also ran ./configure --disable-x86-tsc --enable-x86-sep)

./prepare-kernel.sh  --arch=i386 --adeos=/usr/src/xenomai-2.4.5/ksrc/arch/x86/patches/adeos-ipipe-2.6.26-x86-2.0-09.patch --linux=/usr/src/linux-2.6.26.5

make gave the error:

arch/x86/xenomai/hal_32.c: In function ârthal_latency_above_maxâ:
arch/x86/xenomai/hal_32.c:139: error: implicit declaration of function âdie_nmiâ
make[1]: *** [arch/x86/xenomai/hal_32.o] Error 1
make: *** [arch/x86/xenomai] Error 2

Thanks



      

[-- Attachment #2: Type: text/html, Size: 703 bytes --]

Re: [Xenomai-help] compile error xenomai 2.4.5

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1554 bytes --]

[ Don't start new topics by replying to unrelated ones! ]

George Pitich wrote:
> Tried to compile linux 2.6.26.5 kernel with xenomai 2.4.5 (also ran ./configure --disable-x86-tsc --enable-x86-sep)
> 
> ./prepare-kernel.sh  --arch=i386 --adeos=/usr/src/xenomai-2.4.5/ksrc/arch/x86/patches/adeos-ipipe-2.6.26-x86-2.0-09.patch --linux=/usr/src/linux-2.6.26.5
> 
> make gave the error:
> 
> arch/x86/xenomai/hal_32.c: In function ârthal_latency_above_maxâ:
> arch/x86/xenomai/hal_32.c:139: error: implicit declaration of function âdie_nmiâ
> make[1]: *** [arch/x86/xenomai/hal_32.o] Error 1
> make: *** [arch/x86/xenomai] Error 2

That's an issue of the 2.6.26 ipipe patch. Philippe, you dropped the
nmi_32.h hunk:

diff --git a/include/asm-x86/nmi_32.h b/include/asm-x86/nmi_32.h
index 70a958a..da25e1c 100644
--- a/include/asm-x86/nmi_32.h
+++ b/include/asm-x86/nmi_32.h
@@ -29,7 +29,8 @@ extern void setup_apic_nmi_watchdog (void *);
 extern void stop_apic_nmi_watchdog (void *);
 extern void disable_timer_nmi_watchdog(void);
 extern void enable_timer_nmi_watchdog(void);
-extern int nmi_watchdog_tick (struct pt_regs * regs, unsigned reason);
+extern int (*nmi_watchdog_tick)(struct pt_regs * regs, unsigned reason);
+extern void die_nmi(struct pt_regs *, const char *msg);

 extern atomic_t nmi_active;
 extern unsigned int nmi_watchdog;

Should be merged into the unified nmi.h now.

George, in the meantime you can simply disable NMI watchdog support in
Xenomai, and then the build should succeed again.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]

Re: [Xenomai-help] compile error xenomai 2.4.5

[Philippe Gerum]

Jan Kiszka wrote:
> [ Don't start new topics by replying to unrelated ones! ]
> 
> George Pitich wrote:
>> Tried to compile linux 2.6.26.5 kernel with xenomai 2.4.5 (also ran ./configure --disable-x86-tsc --enable-x86-sep)
>>
>> ./prepare-kernel.sh  --arch=i386 --adeos=/usr/src/xenomai-2.4.5/ksrc/arch/x86/patches/adeos-ipipe-2.6.26-x86-2.0-09.patch --linux=/usr/src/linux-2.6.26.5
>>
>> make gave the error:
>>
>> arch/x86/xenomai/hal_32.c: In function ârthal_latency_above_maxâ:
>> arch/x86/xenomai/hal_32.c:139: error: implicit declaration of function âdie_nmiâ
>> make[1]: *** [arch/x86/xenomai/hal_32.o] Error 1
>> make: *** [arch/x86/xenomai] Error 2
> 
> That's an issue of the 2.6.26 ipipe patch. Philippe, you dropped the
> nmi_32.h hunk:
>

No I didn't, but the die_nmi() prototype is now different between the 32 and 64bit
implementations; the i386 decl is missing. Here is the proper fix:

diff --git a/include/asm-x86/nmi.h b/include/asm-x86/nmi.h
index 765a6d6..f0fc7c5 100644
--- a/include/asm-x86/nmi.h
+++ b/include/asm-x86/nmi.h
@@ -42,6 +42,7 @@ extern void die_nmi(char *str, struct pt_regs *regs, int do_panic);
 extern void nmi_watchdog_default(void);
 #else
 #define nmi_watchdog_default() do {} while (0)
+extern void die_nmi(struct pt_regs *regs, const char *msg);
 #endif

 extern int check_nmi_watchdog(void);


> diff --git a/include/asm-x86/nmi_32.h b/include/asm-x86/nmi_32.h
> index 70a958a..da25e1c 100644
> --- a/include/asm-x86/nmi_32.h
> +++ b/include/asm-x86/nmi_32.h
> @@ -29,7 +29,8 @@ extern void setup_apic_nmi_watchdog (void *);
>  extern void stop_apic_nmi_watchdog (void *);
>  extern void disable_timer_nmi_watchdog(void);
>  extern void enable_timer_nmi_watchdog(void);
> -extern int nmi_watchdog_tick (struct pt_regs * regs, unsigned reason);
> +extern int (*nmi_watchdog_tick)(struct pt_regs * regs, unsigned reason);
> +extern void die_nmi(struct pt_regs *, const char *msg);
> 
>  extern atomic_t nmi_active;
>  extern unsigned int nmi_watchdog;
> 
> Should be merged into the unified nmi.h now.
> 
> George, in the meantime you can simply disable NMI watchdog support in
> Xenomai, and then the build should succeed again.
> 
> Jan
> 


-- 
Philippe.

Re: [Xenomai-help] Xenomai for SH4

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 948 bytes --]

Pavel Bartusek wrote:
> Hello,
> 
> I'd like to use Xenomai on SH4 architecture and it seems it is not in
> the list of supported architectures now. There were some attempt in the
> past to port Xenomai to SH4 architecture
> (http://article.gmane.org/gmane.linux.real-time.xenomai.users/858/match=xenomai+sh4)
> Unfortunately, it seems it the patch hasn't been integrated to the
> mainstream Xenomai. Does somebody have the patch or is somebody working
> on the port  for SH4 now ?

As far as I know, that patch never saw any public light, so we were
unable to even review it. Dunno what happened to this effort, if it was
stopped or if there is now some Xenomai-based SH4 product being shipped
without providing the GPL sources... Maybe you want to contact that guy
directly, asking for the state.

Otherwise, a new effort would be welcome, and I can only encourage doing
it differently (release often, release early).

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]