2005-09-08 07:41:14

by Guennadi Liakhovetski

[permalink] [raw]
Subject: [OOPS] vanilla 2.6.13 + "rmmod processor"

Hi

Just booted a 2.6.13 compiled with UP, ACPI, APIC, LAPIC, sensor modules
with "nolapic noapic acpi=off". The processor module was still loaded by
the hotplug. On rmmod it Oopsed:

Linux version 2.6.13 (gl@X) (gcc version 3.3.3 (SuSE Linux)) #10 Mon Sep 5 13:25:58 CEST 2005

...

Processor #0 6:13 APIC version 17
I/O APIC #2 Version 17 at 0xFEC00000.
Enabling APIC mode: Flat. Using 1 I/O APICs
Processors: 1
Allocating PCI resources starting at 0e000000 (gap: 0e000000:f0c00000)
Built 1 zonelists
Kernel command line: root=/dev/hda2 ro vga=785 acpi=off noapic nolapic
mapped APIC to ffffd000 (fee00000)
mapped IOAPIC to ffffc000 (fec00000)
Initializing CPU#0

...

CPU: Intel(R) Celeron(R) M processor 1.00GHz stepping 08

...

ACPI: Subsystem revision 20050408
ACPI: Interpreter disabled.

...

shpchp: acpi_shpchprm:get_device PCI ROOT HID fail=0x1001
Unable to handle kernel NULL pointer dereference at virtual address 00000004
printing eip:
c020f944
*pde = 00000000
Oops: 0002 [#1]
Modules linked in: evdev snd_intel8x0 snd_ac97_codec snd_pcm snd_timer snd soundcore snd_page_alloc pci_hotplug ehci_hcd autofs4 af_packet vfat fat e100 mii i2c_i801 i2c_core parport_pc parport ohci1394 ieee1394 usbhid uhci_hcd usbcore processor
CPU: 0
EIP: 0060:[<c020f944>] Not tainted VLI
EFLAGS: 00010246 (2.6.13)
EIP is at acpi_bus_unregister_driver+0x21/0x38
eax: 00000000 ebx: ce93d3c0 ecx: ccb09f68 edx: 00000000
esi: 00000000 edi: 00000880 ebp: ccb09f58 esp: ccb09f50
ds: 007b es: 007b ss: 0068
Process rmmod (pid: 2991, threadinfo=ccb08000 task=cd640530)
Stack: ce93d780 00000000 ccb09f60 ce93b058 ccb09fb4 c0133368 00000000 636f7270
6f737365 b7ef0072 b7efc000 ccb09fa0 c014cbdc b7efc000 b7efd000 cd3e1280
c139b074 cd3e1280 cd3e12b0 b7efc000 ccb09fb4 0014cc6d b7efc000 bfc13a98
Call Trace:
[<c0103db6>] show_stack+0xa6/0xb0
[<c0103f29>] show_registers+0x149/0x1c0
[<c010410b>] die+0xbb/0x140
[<c0117732>] do_page_fault+0x512/0x693
[<c0103a03>] error_code+0x4f/0x54
[<ce93b058>] acpi_processor_exit+0x1c/0x2e [processor]
[<c0133368>] sys_delete_module+0x138/0x150
[<c0102f59>] syscall_call+0x7/0xb
Code: 04 89 d0 e8 fd fe ff ff 5d c3 55 89 e5 56 31 f6 85 c0 53 89 c3 74 20 e8 56 ff ff ff 8b 83 a8 00 00 00 85 c0 75 16 8b 53 04 8b 03 <89> 50 04 89 02 89 5b 04 89 1b eb 05 be ea ff ff ff 5b 89 f0 5e

Thanks
Guennadi
---------------------------------
Guennadi Liakhovetski, Ph.D.
DSA Daten- und Systemtechnik GmbH
Pascalstr. 28
D-52076 Aachen
Germany


2005-09-08 08:23:07

by Brown, Len

[permalink] [raw]
Subject: RE: [OOPS] vanilla 2.6.13 + "rmmod processor"


>Just booted a 2.6.13 compiled with UP, ACPI, APIC, LAPIC,
>sensor modules
>with "nolapic noapic acpi=off".

Huh, I don't see I don't see the processor module checking
for acpi_disabled anyplace...

I assume the oops goes away when you
do not boot with "acpi=off"?

> The processor module was still loaded by
>the hotplug. On rmmod it Oopsed:

Note other processor rmmod fix here, maybe unrelated:
http://bugzilla.kernel.org/show_bug.cgi?id=5021


2005-09-08 10:47:17

by Guennadi Liakhovetski

[permalink] [raw]
Subject: RE: [OOPS] vanilla 2.6.13 + "rmmod processor"

On Thu, 8 Sep 2005, Brown, Len wrote:

>> Just booted a 2.6.13 compiled with UP, ACPI, APIC, LAPIC,
>> sensor modules
>> with "nolapic noapic acpi=off".
>
> Huh, I don't see I don't see the processor module checking
> for acpi_disabled anyplace...
>
> I assume the oops goes away when you
> do not boot with "acpi=off"?

Yep. With apic lapic without acpi= thermal (and other ACPI modules) get
also loaded successfully by hotplug, so, one has to rmmod thermal
processor, then no Oops. Actually, should it (processor) at all load
successfully with acpi=off? It is the only module that loads.

>> The processor module was still loaded by
>> the hotplug. On rmmod it Oopsed:
>
> Note other processor rmmod fix here, maybe unrelated:
> http://bugzilla.kernel.org/show_bug.cgi?id=5021

Are you sure it is this bug? Cannot see how those patches can be related
to rmmod path, it said something about races, which doesn't seem to be the
case here. I might be wrong, of course.

Thanks
Guennadi
---------------------------------
Guennadi Liakhovetski, Ph.D.
DSA Daten- und Systemtechnik GmbH
Pascalstr. 28
D-52076 Aachen
Germany