2.5.44-ac2: stack overflow in acpi_initialize_objects

2.5.44-ac2: stack overflow in acpi_initialize_objects

[Eric Buddington]

2.5.44-ac2 compiled for Athlon with gcc-3.2 fails to boot with a
really exciting stack overflow that dumps hordes of stack trace on the
screen. I'm too lazy to write it all down, but the last line before
'init' refers to acpi_initialize_objects.

I can write down more of it if needed.

-Eric

Re: 2.5.44-ac2: stack overflow in acpi_initialize_objects

[David C. Hansen]

On Thu, 2002-10-24 at 07:20, Eric Buddington wrote:
> 2.5.44-ac2 compiled for Athlon with gcc-3.2 fails to boot with a
> really exciting stack overflow that dumps hordes of stack trace on the
> screen. I'm too lazy to write it all down, but the last line before
> 'init' refers to acpi_initialize_objects.
> 
> I can write down more of it if needed.

Does it panic, or just print out a lot of the traces?  

If it panics, then you were able to chew through 7.5K of stack while it
warned during the last 3.5k.  Could you give us an idea of what was
being called?  Something like "a bunch of ACPI routines" or "the same
function a bunch of times" would be very helpful.  For right now, just
turn the CONFIG_X86_STACK_CHECK config option off.

-- 
Dave Hansen
haveblue@us.ibm.com

Re: 2.5.44-ac2: stack overflow in acpi_initialize_objects

[Eric Buddington]

On Thu, Oct 24, 2002 at 09:56:59AM -0700, David C. Hansen wrote:
> On Thu, 2002-10-24 at 07:20, Eric Buddington wrote:
> > 2.5.44-ac2 compiled for Athlon with gcc-3.2 fails to boot with a
> > really exciting stack overflow that dumps hordes of stack trace on the
> > screen. I'm too lazy to write it all down, but the last line before
> > 'init' refers to acpi_initialize_objects.
> 
> Does it panic, or just print out a lot of the traces?  

It panics, and yes, it's a very long list (So I can't tell what's at
the top of it). I'm trying to get a proper record via serial-console,
but can't find my adapters. I'll post it as soon as I can get it...

-Eric

RE: 2.5.44-ac2: stack overflow in acpi_initialize_objects

[Grover, Andrew]

> From: David C. Hansen [mailto:haveblue@us.ibm.com] 
> On Thu, 2002-10-24 at 07:20, Eric Buddington wrote:
> > 2.5.44-ac2 compiled for Athlon with gcc-3.2 fails to boot with a
> > really exciting stack overflow that dumps hordes of stack 
> trace on the
> > screen. I'm too lazy to write it all down, but the last line before
> > 'init' refers to acpi_initialize_objects.
> > 
> > I can write down more of it if needed.
> 
> Does it panic, or just print out a lot of the traces?  

The ACPI-related call path to acpi_initialize_objects is only 3 deep:

acpi_init
acpi_bus_init
acpi_initialize_objects

...none of which have big variables on the stack.

Need more info...

-- Andy