initramfs /init freeze

initramfs /init freeze

[Roni Feldman]

Hello everyone,

I'm experiencing odd behavior on an SH4 setup: I have an embedded
initramfs in the kernel image.
The kernel boots fine but the sequence freezes after
run_init_process("/init") is executed.

I say "freeze" because I tried code that only does "return 0;" and I
even wrote an init.s with
a _start function that jmp's to address 0 (and was supposed to cause
an exception, but nothing
appeared in ttySC1, not even a panic for the exit of the init process).

All the printk()'s are visible, and the console handover from sercon0
to ttySC1 seems to work
fine (I have no trouble doing printk's before the call to kernel_execve).

I also tried both static and dynamic linking when I wrote the empty main().

Busybox as a shell and an /init shell script seems to suffer from the
same symptoms.

My setup:
cpu: SH7751R LE configuration
kernel: 2.6.25-rc8
gcc: Gentoo 4.1.2 p1.0.2 with glibc 2.7

I started to suspect my too-new glibc, so now I'm struggling with
crossdev to compile a uclibc
toolchain.

Do you guys have any tips for me?

Best regards,
Roni

Re: initramfs /init freeze

[Kristoffer Ericson]

On Mon, 28 Apr 2008 21:04:49 +0200
"Roni Feldman" <roni.feldman+lkml@gmail.com> wrote:

> Hello everyone,
> 
> I'm experiencing odd behavior on an SH4 setup: I have an embedded
> initramfs in the kernel image.
> The kernel boots fine but the sequence freezes after
> run_init_process("/init") is executed.
> 
> I say "freeze" because I tried code that only does "return 0;" and I
> even wrote an init.s with
> a _start function that jmp's to address 0 (and was supposed to cause
> an exception, but nothing
> appeared in ttySC1, not even a panic for the exit of the init process).
> 
> All the printk()'s are visible, and the console handover from sercon0
> to ttySC1 seems to work
> fine (I have no trouble doing printk's before the call to kernel_execve).
> 
> I also tried both static and dynamic linking when I wrote the empty main().

I believe the dynamic linking issue was in pre glibc 2.3.4 (atleast on SH3) so
shouldn't be an issue atm.

> 
> Busybox as a shell and an /init shell script seems to suffer from the
> same symptoms.
> 
> My setup:
> cpu: SH7751R LE configuration
> kernel: 2.6.25-rc8
> gcc: Gentoo 4.1.2 p1.0.2 with glibc 2.7
> 
> I started to suspect my too-new glibc, so now I'm struggling with
> crossdev to compile a uclibc
> toolchain.

I know glibc 2.3.x / 2.4.x / 2.5.x works well, but never tried 2.7. Usually though its gcc
thats causing the issues. But you should step back and build it with older gcc/glibc just to make sure.

For reference I usually use gcc 3.4.5 & glibc 2.3.6 for kernel builds.

> 
> Do you guys have any tips for me?
> 
> Best regards,
> Roni
> --
> To unsubscribe from this list: send the line "unsubscribe linux-sh" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html


-- 
Kristoffer Ericson <Kristoffer.Ericson@Gmail.com>

Re: initramfs /init freeze

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 1875 bytes --]

On Monday 28 April 2008, Kristoffer Ericson wrote:
> On Mon, 28 Apr 2008 21:04:49 "Roni Feldman" wrote:
> > I'm experiencing odd behavior on an SH4 setup: I have an embedded
> > initramfs in the kernel image.
> > The kernel boots fine but the sequence freezes after
> > run_init_process("/init") is executed.
> >
> > I say "freeze" because I tried code that only does "return 0;" and I
> > even wrote an init.s with
> > a _start function that jmp's to address 0 (and was supposed to cause
> > an exception, but nothing
> > appeared in ttySC1, not even a panic for the exit of the init process).
> >
> > All the printk()'s are visible, and the console handover from sercon0
> > to ttySC1 seems to work
> > fine (I have no trouble doing printk's before the call to kernel_execve).
> >
> > I also tried both static and dynamic linking when I wrote the empty
> > main().
>
> I believe the dynamic linking issue was in pre glibc 2.3.4 (atleast on SH3)
> so shouldn't be an issue atm.

and static seems to fail the same way ...

you could try enabling magic sysrq on your system.  otherwise, jtag/kgdb would 
be very useful just to see where the processor is stuck.

> > Busybox as a shell and an /init shell script seems to suffer from the
> > same symptoms.
> >
> > My setup:
> > cpu: SH7751R LE configuration
> > kernel: 2.6.25-rc8
> > gcc: Gentoo 4.1.2 p1.0.2 with glibc 2.7
> >
> > I started to suspect my too-new glibc, so now I'm struggling with
> > crossdev to compile a uclibc
> > toolchain.
>
> I know glibc 2.3.x / 2.4.x / 2.5.x works well, but never tried 2.7. Usually
> though its gcc thats causing the issues. But you should step back and build
> it with older gcc/glibc just to make sure.
>
> For reference I usually use gcc 3.4.5 & glibc 2.3.6 for kernel builds.

glibc-2.7 seems to work fine on my systems ...
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 827 bytes --]

Re: initramfs /init freeze

[Roni Feldman]

Hi,

My fault indeed! I use the sh4-unknown-linux-gnu toolchain, so the gcc
crt*.o files were with FPU support (with FPU instructions). Only I
disabled FPU support in the kernel since I'm not going to use it. So
an FPU opcode probably cause the CPU freeze. Everything works
now (with FPU support enabled), but I really don't want the FPU. I
tried generating an sh4-softfloat-linux-gnu toolchain (gcc 4.1.2 libc
2.7) with crossdev but I failed.
I'm probably missing something obvious, but I didn't work on it for too long...

Thanks a lot!
Roni Feldman

On Fri, May 2, 2008 at 4:22 AM, Mike Frysinger <vapier@gentoo.org> wrote:
> On Monday 28 April 2008, Kristoffer Ericson wrote:
>
> > On Mon, 28 Apr 2008 21:04:49 "Roni Feldman" wrote:
>  > > I'm experiencing odd behavior on an SH4 setup: I have an embedded
>  > > initramfs in the kernel image.
>  > > The kernel boots fine but the sequence freezes after
>  > > run_init_process("/init") is executed.
>  > >
>  > > I say "freeze" because I tried code that only does "return 0;" and I
>  > > even wrote an init.s with
>  > > a _start function that jmp's to address 0 (and was supposed to cause
>  > > an exception, but nothing
>  > > appeared in ttySC1, not even a panic for the exit of the init process).
>  > >
>  > > All the printk()'s are visible, and the console handover from sercon0
>  > > to ttySC1 seems to work
>  > > fine (I have no trouble doing printk's before the call to kernel_execve).
>  > >
>  > > I also tried both static and dynamic linking when I wrote the empty
>  > > main().
>  >
>  > I believe the dynamic linking issue was in pre glibc 2.3.4 (atleast on SH3)
>  > so shouldn't be an issue atm.
>
>  and static seems to fail the same way ...
>
>  you could try enabling magic sysrq on your system.  otherwise, jtag/kgdb would
>  be very useful just to see where the processor is stuck.
>
>
>  > > Busybox as a shell and an /init shell script seems to suffer from the
>  > > same symptoms.
>  > >
>  > > My setup:
>  > > cpu: SH7751R LE configuration
>  > > kernel: 2.6.25-rc8
>  > > gcc: Gentoo 4.1.2 p1.0.2 with glibc 2.7
>  > >
>  > > I started to suspect my too-new glibc, so now I'm struggling with
>  > > crossdev to compile a uclibc
>  > > toolchain.
>  >
>  > I know glibc 2.3.x / 2.4.x / 2.5.x works well, but never tried 2.7. Usually
>  > though its gcc thats causing the issues. But you should step back and build
>  > it with older gcc/glibc just to make sure.
>  >
>  > For reference I usually use gcc 3.4.5 & glibc 2.3.6 for kernel builds.
>
>  glibc-2.7 seems to work fine on my systems ...
>  -mike
>

Re: initramfs /init freeze

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 470 bytes --]

On Friday 02 May 2008, Roni Feldman wrote:
> I tried generating an sh4-softfloat-linux-gnu toolchain (gcc 4.1.2 libc
> 2.7) with crossdev but I failed.
> I'm probably missing something obvious, but I didn't work on it for too
> long...

ive only ever attempted softfloat on arm targets and with gcc-3.4.x.  i'd 
suggest you do a quicktest and see if gcc 3.4.x works with 
sh4-softfloat-linux-gnu ... if it doesnt, then something needs a little 
TLC ...
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 827 bytes --]