ACPI+Inspiron 1545 = crash at xrstor_state()

ACPI+Inspiron 1545 = crash at xrstor_state()

[Jody Bruchon]

I previously reported a boot hang on an Inspiron 1545 laptop which is 
cured by disabling CONFIG_ACPI.  I didn't know about the "earlyprintk" 
boot option, and upon using it, I was able to see the actual crash.  
Hopefully the following information will help.  Note that I have turned 
off SMP and APIC/LAPIC support, but the bug persists even with these 
disabled (I noticed a lot of APIC/IO-APIC code that is eliminated in 
early boot if they're disabled.)

The crash was in function xsave_init with EIP at xstate_enable_boot_cpu 
(the last normal message output before this was "Initializing CPU#0").  
To narrow this down further, I edited xsave.c to be full of printk() 
messages for each called function.  The last printk message indicates 
that the function in xsave.c which crashes out is:

xrstor_state(init_xstate_buf, -1);

The result is a general protection fault.  xrstor_state is just a tiny 
piece of asm code, and at this point I've reached the extent of my 
current capabilities, so I don't know where to go from here.  I can 
manually type out chunks of the crash info on request.  I'm clueless as 
to how the inclusion of ACPI causes this.

Please CC my email if you respond since I don't subscribe to LKML.  
Thanks in advance.

Jody Bruchon

Re: ACPI+Inspiron 1545 = crash at xrstor_state()

[Borislav Petkov]

On Fri, Nov 12, 2010 at 07:06:37PM -0500, Jody Bruchon wrote:
> I previously reported a boot hang on an Inspiron 1545 laptop which
> is cured by disabling CONFIG_ACPI.  I didn't know about the
> "earlyprintk" boot option, and upon using it, I was able to see the
> actual crash.  Hopefully the following information will help.  Note
> that I have turned off SMP and APIC/LAPIC support, but the bug
> persists even with these disabled (I noticed a lot of APIC/IO-APIC
> code that is eliminated in early boot if they're disabled.)
> 
> The crash was in function xsave_init with EIP at
> xstate_enable_boot_cpu (the last normal message output before this
> was "Initializing CPU#0").  To narrow this down further, I edited
> xsave.c to be full of printk() messages for each called function.
> The last printk message indicates that the function in xsave.c which
> crashes out is:
> 
> xrstor_state(init_xstate_buf, -1);
> 
> The result is a general protection fault.  xrstor_state is just a
> tiny piece of asm code, and at this point I've reached the extent of
> my current capabilities, so I don't know where to go from here.  I
> can manually type out chunks of the crash info on request.  I'm
> clueless as to how the inclusion of ACPI causes this.
> 
> Please CC my email if you respond since I don't subscribe to LKML.
> Thanks in advance.

Ok, first of all, which kernel are you using? Second, send us your
.config. Then, please try booting the latest kernel (2.6.36) without
your printk's in it and with earlyprintk enabled on the kernel cmd line
and when the oops appears on the screen try to photograph it with a
digicam or similar and upload the pic somewhere. Before that, make sure
the pic is readable.

Thanks.

-- 
Regards/Gruss,
    Boris.