Anyone success with 440GR Rev.B (1.1)?

Anyone success with 440GR Rev.B (1.1)?

[Andriy Korud]

Hi,
I have complete system (U-Boot, Linux, userspace) running on 440GR =
Rev.A.=20
However after mounting newer chip revision (Rev.B, 1.1) U-Boot and Linux =
boots fine, however when trying to execute first userland application =
(init), I get:

Freeing unused kernel memory: 112k init
init has generated signal 4 but has no handler for it
Kernel panic - not syncing: Attempted to kill init!

I've tried everything - different compilers (3.x, 4.1.x), deifferent =
kernels (2.6.18, 19) - problem is 100% reproducible. Even when replaced =
busybox with printf("hello world") - the same.=20
According to AMCC changelog and erratas there are no changes in CPU core =
between revisions.=20

Anybody has an idea what should I check/try?

Thanks in advance,
--
Andriy Korud

Re: Anyone success with 440GR Rev.B (1.1)?

[Roland Dreier]

 > I have complete system (U-Boot, Linux, userspace) running on 440GR Rev.A. 
 > However after mounting newer chip revision (Rev.B, 1.1) U-Boot and Linux boots fine, however when trying to execute first userland application (init), I get:
 > 
 > Freeing unused kernel memory: 112k init
 > init has generated signal 4 but has no handler for it
 > Kernel panic - not syncing: Attempted to kill init!

signal 4 is SIGILL.  Just a guess but do you have a cputable entry for
440GR rev B in your arch/powerpc/kernel/cputable.c?  If you don't then
I think the default kernel settings will result in the wrong cacheline
size being used, which leads to userspace often dying with illegal
instruction traps.

(I don't see any 440GR entries in cputable.c, neither rev A nor rev B,
in the upstream kernel so I'm assuming you're patching your kernel
somehow.  So checking that your patch covers rev B would be the first
thing to do)

 - R.

Re: Anyone success with 440GR Rev.B (1.1)?

[Stefan Roese]

On Friday 02 February 2007 16:45, Roland Dreier wrote:
>  > I have complete system (U-Boot, Linux, userspace) running on 440GR
>  > Rev.A. However after mounting newer chip revision (Rev.B, 1.1) U-Boot
>  > and Linux boots fine, however when trying to execute first userland
>  > application (init), I get:
>  >
>  > Freeing unused kernel memory: 112k init
>  > init has generated signal 4 but has no handler for it
>  > Kernel panic - not syncing: Attempted to kill init!
>
> signal 4 is SIGILL.  Just a guess but do you have a cputable entry for
> 440GR rev B in your arch/powerpc/kernel/cputable.c?

Good idea. That's probably it.

> If you don't then 
> I think the default kernel settings will result in the wrong cacheline
> size being used, which leads to userspace often dying with illegal
> instruction traps.
>
> (I don't see any 440GR entries in cputable.c, neither rev A nor rev B,
> in the upstream kernel so I'm assuming you're patching your kernel
> somehow.  So checking that your patch covers rev B would be the first
> thing to do)

440GR has the same PVR as 440EP:

440EP Rev B == 440GR Rev A
440EP Rev C == 440GR Rev B

So your kernel should support 440EP Rev. C. The denx kernel has support for 
those CPU's included. I'll try to send some patches in the next few days to 
support them in the kernel.org version.

Best regards,
Stefan

RE: Anyone success with 440GR Rev.B (1.1)?

[Andriy Korud]

Thanks,
I've discovered this few hours ago. After adding entry to cputable =
everything works fine.

Anyway thanks and sorry for bothering :)
regards,
--
Andriy Korud

-----Original Message-----
From: Stefan Roese [mailto:sr@denx.de]
Sent: =F0=D4 2007-02-02 21:26
To: Andriy Korud
Cc: linuxppc-embedded@ozlabs.org; Roland Dreier
Subject: Re: Anyone success with 440GR Rev.B (1.1)?
=20
On Friday 02 February 2007 16:45, Roland Dreier wrote:
>  > I have complete system (U-Boot, Linux, userspace) running on 440GR
>  > Rev.A. However after mounting newer chip revision (Rev.B, 1.1) =
U-Boot
>  > and Linux boots fine, however when trying to execute first userland
>  > application (init), I get:
>  >
>  > Freeing unused kernel memory: 112k init
>  > init has generated signal 4 but has no handler for it
>  > Kernel panic - not syncing: Attempted to kill init!
>
> signal 4 is SIGILL.  Just a guess but do you have a cputable entry for
> 440GR rev B in your arch/powerpc/kernel/cputable.c?

Good idea. That's probably it.

> If you don't then=20
> I think the default kernel settings will result in the wrong cacheline
> size being used, which leads to userspace often dying with illegal
> instruction traps.
>
> (I don't see any 440GR entries in cputable.c, neither rev A nor rev B,
> in the upstream kernel so I'm assuming you're patching your kernel
> somehow.  So checking that your patch covers rev B would be the first
> thing to do)

440GR has the same PVR as 440EP:

440EP Rev B =3D=3D 440GR Rev A
440EP Rev C =3D=3D 440GR Rev B

So your kernel should support 440EP Rev. C. The denx kernel has support =
for=20
those CPU's included. I'll try to send some patches in the next few days =
to=20
support them in the kernel.org version.

Best regards,
Stefan