Kernel hang solid

[Vincent Legoll]

Could someone explain the differences in linux kernel boot
logs for coreboot vs BIOS that are attached. The coreboot
log is here too, just in case.

The total absence of reserved memory ranges seems
unnatural. The DMA zone range starting at 0x00000001
looks fishy too.

I'm completely lost, the hang happen at init/main.c in
console_init()

[Corey Osgood]

On Thu, Sep 4, 2008 at 8:00 PM, Vincent Legoll  wrote:
> Could someone explain the differences in linux kernel boot
> logs for coreboot vs BIOS that are attached. The coreboot
> log is here too, just in case.
>
> The total absence of reserved memory ranges seems
> unnatural. The DMA zone range starting at 0x00000001
> looks fishy too.
>
> I'm completely lost, the hang happen at init/main.c in
> console_init()

It looks like you're IRQ routing is broken, at least on the surface.
Is this a currently supported board, or a new port you're working on?

-Corey

[Vincent Legoll]

On Fri, Sep 5, 2008 at 4:50 AM, Corey Osgood  wrote:
> It looks like you're IRQ routing is broken, at least on the surface.
> Is this a currently supported board, or a new port you're working on?

New port, IRQ routing just has landed. So it may still be broken.

But, kernel "apic=debug" output is now the same for coreboot-based
or BIOS-based boots.

Where could I have more infos on other IRQs not listed there (for
example the IRQs 3 & 4 that are being used for serial port when
booted with legacy bios). Kernel is silent about them at that stage
of the boot process.

[yhlu]

On Thu, Sep 4, 2008 at 5:00 PM, Vincent Legoll wrote:
> Could someone explain the differences in linux kernel boot
> logs for coreboot vs BIOS that are attached. The coreboot
> log is here too, just in case.
>
> The total absence of reserved memory ranges seems
> unnatural. The DMA zone range starting at 0x00000001
> looks fishy too.
>
> I'm completely lost, the hang happen at init/main.c in
> console_init()

does previous kernel work? like 2.6.24

YH

[Vincent Legoll]

On Fri, Sep 5, 2008 at 5:03 AM, yhlu  wrote:
> does previous kernel work? like 2.6.24

I've not tried a kernel that old, but I will.

I've tried vanilla -git for last few days only.

[Vincent Legoll]

On Fri, Sep 5, 2008 at 11:38 AM, Vincent Legoll
 wrote:
> On Fri, Sep 5, 2008 at 5:03 AM, yhlu  wrote:
> > does previous kernel work? like 2.6.24
>
> I've not tried a kernel that old, but I will.
>
> I've tried vanilla -git for last few days only.

Tested 2.6.{23,24,25} they all hang the same way at the same place:
the first console_initcall() (which I don't know if it is the serial one or
the vga one)

Anyone can explain the strange things (at least strange to me) in
linux_coreboot.txt ? (attached to 1st post in thread)

[Vincent Legoll]

On Fri, Sep 5, 2008 at 10:21 PM, Vincent Legoll
 wrote:
> Tested 2.6.{23,24,25} they all hang the same way at the same place:
> the first console_initcall() (which I don't know if it is the serial one or
> the vga one)

I've narrowed it to the
down(&console_sem) in acquire_console_sem() in kernel/printk.c
called from con_init() in drivers/char/vt.c

I suspect the spin_lock_irqsave() in there hanging...

I have a lot of CONNFIG_DEBUG_ options activated, especially
CONFIG_DEBUG_SPINLOCK, but in vain...

So is there something I can do to narrow the problem.
How can I validate my irq setup ?

[ron minnich]

On Fri, Sep 5, 2008 at 4:54 PM, Vincent Legoll  wrote:
> On Fri, Sep 5, 2008 at 10:21 PM, Vincent Legoll
>  wrote:
> > Tested 2.6.{23,24,25} they all hang the same way at the same place:
> > the first console_initcall() (which I don't know if it is the serial one or
> > the vga one)
>
> I've narrowed it to the
> down(&console_sem) in acquire_console_sem() in kernel/printk.c
> called from con_init() in drivers/char/vt.c
>
> I suspect the spin_lock_irqsave() in there hanging...
>
> I have a lot of CONNFIG_DEBUG_ options activated, especially
> CONFIG_DEBUG_SPINLOCK, but in vain...
>
> So is there something I can do to narrow the problem.
> How can I validate my irq setup ?

I don't know but when it gets this weird it can actually be memory ...

ron