Odd APM oops

[Mike]

Power off on shutdown is consistently making my box oops as it shuts down
in 2.3.1[34].  This worked fine before 2.3.13.

This is an SMP box.

Since the oops happens at power off, it doesn't get logged, but the
important bits are:

general protection fault: 0000
CPU: 0
EIP: 0010:[<c0107a26>]  (this is apm_bios_call+50)
EFLAGS: 00010046
eax: 00005037 ebx: 00000001 ecx: 00000003 adx: 00000000
esi: c4a06000 edi: fee1dead epb: bffffd74 esp: c4a07eb0
ds:  0000 es: 0000 ss: 00`8

The call trace goes:
set_power_state, apm_set_power_sate, apm_power_off, machine_power_off,
sys_reboot, ....

which is hardly suprising.

Dump of assembler code for function apm_bios_call_simple:
0xc01079f4 <apm_bios_call_simple>:      subl   $0xc,%esp
0xc01079f7 <apm_bios_call_simple+3>:    pushl  %edi
0xc01079f8 <apm_bios_call_simple+4>:    pushl  %esi
0xc01079f9 <apm_bios_call_simple+5>:    pushl  %ebx
0xc01079fa <apm_bios_call_simple+6>:    movl   0x20(%esp,1),%ebx
0xc01079fe <apm_bios_call_simple+10>:   movl   0x24(%esp,1),%ecx
0xc0107a02 <apm_bios_call_simple+14>:   pushf
0xc0107a03 <apm_bios_call_simple+15>:   popl   0xc(%esp,1)
0xc0107a07 <apm_bios_call_simple+19>:   cli
0xc0107a08 <apm_bios_call_simple+20>:   movw   %fs,0x14(%esp,1)
0xc0107a0c <apm_bios_call_simple+24>:   movw   %gs,0x10(%esp,1)
0xc0107a10 <apm_bios_call_simple+28>:   movl   0x1c(%esp,1),%eax
0xc0107a14 <apm_bios_call_simple+32>:   pushl  %ds
0xc0107a15 <apm_bios_call_simple+33>:   pushl  %es
0xc0107a16 <apm_bios_call_simple+34>:   xorl   %edx,%edx
0xc0107a18 <apm_bios_call_simple+36>:   movw   %dx,%ds
0xc0107a1b <apm_bios_call_simple+39>:   movw   %dx,%es
0xc0107a1e <apm_bios_call_simple+42>:   movw   %dx,%fs
0xc0107a21 <apm_bios_call_simple+45>:   movw   %dx,%gs
0xc0107a24 <apm_bios_call_simple+48>:   pushl  %edi
0xc0107a25 <apm_bios_call_simple+49>:   pushl  %ebp
0xc0107a26 <apm_bios_call_simple+50>:   lcall  *%cs:0xc021dc4c
(0xc021dc4c corresponds to apm_bios_entry, as expected).
0xc0107a2d <apm_bios_call_simple+57>:   setb   %bl
0xc0107a30 <apm_bios_call_simple+60>:   popl   %ebp
0xc0107a31 <apm_bios_call_simple+61>:   popl   %edi
0xc0107a32 <apm_bios_call_simple+62>:   popl   %es
0xc0107a33 <apm_bios_call_simple+63>:   popl   %ds
0xc0107a34 <apm_bios_call_simple+64>:   movl   0x28(%esp,1),%edi
0xc0107a38 <apm_bios_call_simple+68>:   movl   %eax,(%edi)
0xc0107a3a <apm_bios_call_simple+70>:   movw   0x14(%esp,1),%fs
0xc0107a3e <apm_bios_call_simple+74>:   movw   0x10(%esp,1),%gs
0xc0107a42 <apm_bios_call_simple+78>:   pushl  0xc(%esp,1)
0xc0107a46 <apm_bios_call_simple+82>:   popf
0xc0107a47 <apm_bios_call_simple+83>:   movzbl %bl,%edx
0xc0107a4a <apm_bios_call_simple+86>:   movl   %edx,%eax
0xc0107a4c <apm_bios_call_simple+88>:   popl   %ebx
0xc0107a4d <apm_bios_call_simple+89>:   popl   %esi
0xc0107a4e <apm_bios_call_simple+90>:   popl   %edi
0xc0107a4f <apm_bios_call_simple+91>:   addl   $0xc,%esp
0xc0107a52 <apm_bios_call_simple+94>:   ret
0xc0107a53 <apm_bios_call_simple+95>:   nop

-- 
Mike <[e-mail]>

If it is a Miracle, any sort of evidence will answer, but if it is a Fact,
proof is necessary.
		-- Samuel Clemens