2.6.16-rc1 kernel init oops

2.6.16-rc1 kernel init oops

[Randy.Dunlap]

Hi,

I'm trying to boot 2.6.16-rc1 on a T42 Thinkpad notebook.
No serial port for serial console.  I don't think that networking
is alive yet (for network console ?).

Anyone recognize this?  got patch?

This is just typed in, so could contain a few errors.

Unable to handle kernel NULL pointer dereference at virtual address 00000001
printing eip:
00000001
*pde = 00000000
Oops: 0000 [#1]
SMP DEBUG_PAGEALLOC
Modules linked in:
CPU:	0
EIP:	0060:[<00000001>]   Not tainted VLI
EFLAGS: 00010202   (2.6.16-rc1)
EIP is at 0x1
<skip reg. dump>
<skip stack dump>
Call trace:
	show_stack_log_lvl+0xa5/0xad
	show_registers+0xf9/0x162
	die+0xfe/0x179
	do_page_fault+0x399/0x4d8
	error_code+0x4f/0x54
	device_register+0x13/0x18
	platform_bus_init+0xd/0x19
	driver_init+0x1c/0x2d
	do_basic_setup+0x12/0x1e
	init+0x95/0x195
	kernel_thread_helper+0x5/0xb
Code:  Bad EIP value.

Thanks,
---
~Randy

Re: 2.6.16-rc1 kernel init oops

[Randy.Dunlap]

On Sat, 28 Jan 2006 17:18:41 -0800 Randy.Dunlap wrote:

> Hi,
> 
> I'm trying to boot 2.6.16-rc1 on a T42 Thinkpad notebook.
> No serial port for serial console.  I don't think that networking
> is alive yet (for network console ?).
> 
> Anyone recognize this?  got patch?
> 
> This is just typed in, so could contain a few errors.
> 
> Unable to handle kernel NULL pointer dereference at virtual address 00000001
> printing eip:
> 00000001
> *pde = 00000000
> Oops: 0000 [#1]
> SMP DEBUG_PAGEALLOC
> Modules linked in:
> CPU:	0
> EIP:	0060:[<00000001>]   Not tainted VLI
> EFLAGS: 00010202   (2.6.16-rc1)
> EIP is at 0x1
> <skip reg. dump>
> <skip stack dump>
> Call trace:
> 	show_stack_log_lvl+0xa5/0xad
> 	show_registers+0xf9/0x162
> 	die+0xfe/0x179
> 	do_page_fault+0x399/0x4d8
> 	error_code+0x4f/0x54
> 	device_register+0x13/0x18
> 	platform_bus_init+0xd/0x19
> 	driver_init+0x1c/0x2d
> 	do_basic_setup+0x12/0x1e
> 	init+0x95/0x195
> 	kernel_thread_helper+0x5/0xb
> Code:  Bad EIP value.

Both 2.6.15 and 2.6.15.1 boot OK for me.
.config for 2.6.16-rc1 is at
  http://www.xenotime.net/linux/doc/config-2616rc1

---
~Randy

Re: 2.6.16-rc1 kernel init oops

[Greg KH]

On Sat, Jan 28, 2006 at 05:55:11PM -0800, Randy.Dunlap wrote:
> On Sat, 28 Jan 2006 17:18:41 -0800 Randy.Dunlap wrote:
> 
> > Hi,
> > 
> > I'm trying to boot 2.6.16-rc1 on a T42 Thinkpad notebook.
> > No serial port for serial console.  I don't think that networking
> > is alive yet (for network console ?).
> > 
> > Anyone recognize this?  got patch?
> > 
> > This is just typed in, so could contain a few errors.
> > 
> > Unable to handle kernel NULL pointer dereference at virtual address 00000001
> > printing eip:
> > 00000001
> > *pde = 00000000
> > Oops: 0000 [#1]
> > SMP DEBUG_PAGEALLOC
> > Modules linked in:
> > CPU:	0
> > EIP:	0060:[<00000001>]   Not tainted VLI
> > EFLAGS: 00010202   (2.6.16-rc1)
> > EIP is at 0x1
> > <skip reg. dump>
> > <skip stack dump>
> > Call trace:
> > 	show_stack_log_lvl+0xa5/0xad
> > 	show_registers+0xf9/0x162
> > 	die+0xfe/0x179
> > 	do_page_fault+0x399/0x4d8
> > 	error_code+0x4f/0x54
> > 	device_register+0x13/0x18
> > 	platform_bus_init+0xd/0x19
> > 	driver_init+0x1c/0x2d
> > 	do_basic_setup+0x12/0x1e
> > 	init+0x95/0x195
> > 	kernel_thread_helper+0x5/0xb
> > Code:  Bad EIP value.
> 
> Both 2.6.15 and 2.6.15.1 boot OK for me.
> .config for 2.6.16-rc1 is at
>   http://www.xenotime.net/linux/doc/config-2616rc1

If you disable CONFIG_PNP and CONFIG_ISAPNP options does that help?

thanks,

greg k-h

Re: 2.6.16-rc1 kernel init oops

[Randy.Dunlap]

On Sun, 29 Jan 2006 11:00:29 -0800 Greg KH wrote:

> On Sat, Jan 28, 2006 at 05:55:11PM -0800, Randy.Dunlap wrote:
> > On Sat, 28 Jan 2006 17:18:41 -0800 Randy.Dunlap wrote:
> > 
> > > Hi,
> > > 
> > > I'm trying to boot 2.6.16-rc1 on a T42 Thinkpad notebook.
> > > No serial port for serial console.  I don't think that networking
> > > is alive yet (for network console ?).
> > > 
> > > Anyone recognize this?  got patch?
> > > 
> > > This is just typed in, so could contain a few errors.
> > > 
> > > Unable to handle kernel NULL pointer dereference at virtual address 00000001
> > > printing eip:
> > > 00000001
> > > *pde = 00000000
> > > Oops: 0000 [#1]
> > > SMP DEBUG_PAGEALLOC
> > > Modules linked in:
> > > CPU:	0
> > > EIP:	0060:[<00000001>]   Not tainted VLI
> > > EFLAGS: 00010202   (2.6.16-rc1)
> > > EIP is at 0x1
> > > <skip reg. dump>
> > > <skip stack dump>
> > > Call trace:
> > > 	show_stack_log_lvl+0xa5/0xad
> > > 	show_registers+0xf9/0x162
> > > 	die+0xfe/0x179
> > > 	do_page_fault+0x399/0x4d8
> > > 	error_code+0x4f/0x54
> > > 	device_register+0x13/0x18
> > > 	platform_bus_init+0xd/0x19
> > > 	driver_init+0x1c/0x2d
> > > 	do_basic_setup+0x12/0x1e
> > > 	init+0x95/0x195
> > > 	kernel_thread_helper+0x5/0xb
> > > Code:  Bad EIP value.
> > 
> > Both 2.6.15 and 2.6.15.1 boot OK for me.
> > .config for 2.6.16-rc1 is at
> >   http://www.xenotime.net/linux/doc/config-2616rc1
> 
> If you disable CONFIG_PNP and CONFIG_ISAPNP options does that help?

Nope, no change.  Any other suggestions?

I just booted with a KOBJECT_DEBUG
built kernel and it's failing after:

kobject platform: registering, parent: <NULL>, set: devices

thanks,
---
~Randy

Re: 2.6.16-rc1 kernel init oops

[Greg KH]

On Sun, Jan 29, 2006 at 11:19:34AM -0800, Randy.Dunlap wrote:
> On Sun, 29 Jan 2006 11:00:29 -0800 Greg KH wrote:
> 
> > On Sat, Jan 28, 2006 at 05:55:11PM -0800, Randy.Dunlap wrote:
> > > On Sat, 28 Jan 2006 17:18:41 -0800 Randy.Dunlap wrote:
> > > 
> > > > Hi,
> > > > 
> > > > I'm trying to boot 2.6.16-rc1 on a T42 Thinkpad notebook.
> > > > No serial port for serial console.  I don't think that networking
> > > > is alive yet (for network console ?).
> > > > 
> > > > Anyone recognize this?  got patch?
> > > > 
> > > > This is just typed in, so could contain a few errors.
> > > > 
> > > > Unable to handle kernel NULL pointer dereference at virtual address 00000001
> > > > printing eip:
> > > > 00000001
> > > > *pde = 00000000
> > > > Oops: 0000 [#1]
> > > > SMP DEBUG_PAGEALLOC
> > > > Modules linked in:
> > > > CPU:	0
> > > > EIP:	0060:[<00000001>]   Not tainted VLI
> > > > EFLAGS: 00010202   (2.6.16-rc1)
> > > > EIP is at 0x1
> > > > <skip reg. dump>
> > > > <skip stack dump>
> > > > Call trace:
> > > > 	show_stack_log_lvl+0xa5/0xad
> > > > 	show_registers+0xf9/0x162
> > > > 	die+0xfe/0x179
> > > > 	do_page_fault+0x399/0x4d8
> > > > 	error_code+0x4f/0x54
> > > > 	device_register+0x13/0x18
> > > > 	platform_bus_init+0xd/0x19
> > > > 	driver_init+0x1c/0x2d
> > > > 	do_basic_setup+0x12/0x1e
> > > > 	init+0x95/0x195
> > > > 	kernel_thread_helper+0x5/0xb
> > > > Code:  Bad EIP value.
> > > 
> > > Both 2.6.15 and 2.6.15.1 boot OK for me.
> > > .config for 2.6.16-rc1 is at
> > >   http://www.xenotime.net/linux/doc/config-2616rc1
> > 
> > If you disable CONFIG_PNP and CONFIG_ISAPNP options does that help?
> 
> Nope, no change.  Any other suggestions?
> 
> I just booted with a KOBJECT_DEBUG
> built kernel and it's failing after:
> 
> kobject platform: registering, parent: <NULL>, set: devices

Can you enable CONFIG_DEBUG_DRIVER and see if that helps?

thanks,

greg k-h

Re: 2.6.16-rc1 kernel init oops

[Randy.Dunlap]

On Sun, 29 Jan 2006 12:19:23 -0800 Greg KH wrote:

> On Sun, Jan 29, 2006 at 11:19:34AM -0800, Randy.Dunlap wrote:
> > On Sun, 29 Jan 2006 11:00:29 -0800 Greg KH wrote:
> > 
> > > On Sat, Jan 28, 2006 at 05:55:11PM -0800, Randy.Dunlap wrote:
> > > > On Sat, 28 Jan 2006 17:18:41 -0800 Randy.Dunlap wrote:
> > > > 
> > > > > Hi,
> > > > > 
> > > > > I'm trying to boot 2.6.16-rc1 on a T42 Thinkpad notebook.
> > > > > No serial port for serial console.  I don't think that networking
> > > > > is alive yet (for network console ?).
> > > > > 
> > > > > Anyone recognize this?  got patch?
> > > > > 
> > > > > This is just typed in, so could contain a few errors.
> > > > > 
> > > > > Unable to handle kernel NULL pointer dereference at virtual address 00000001
> > > > > printing eip:
> > > > > 00000001
> > > > > *pde = 00000000
> > > > > Oops: 0000 [#1]
> > > > > SMP DEBUG_PAGEALLOC
> > > > > Modules linked in:
> > > > > CPU:	0
> > > > > EIP:	0060:[<00000001>]   Not tainted VLI
> > > > > EFLAGS: 00010202   (2.6.16-rc1)
> > > > > EIP is at 0x1
> > > > > <skip reg. dump>
> > > > > <skip stack dump>
> > > > > Call trace:
> > > > > 	show_stack_log_lvl+0xa5/0xad
> > > > > 	show_registers+0xf9/0x162
> > > > > 	die+0xfe/0x179
> > > > > 	do_page_fault+0x399/0x4d8
> > > > > 	error_code+0x4f/0x54
> > > > > 	device_register+0x13/0x18
> > > > > 	platform_bus_init+0xd/0x19
> > > > > 	driver_init+0x1c/0x2d
> > > > > 	do_basic_setup+0x12/0x1e
> > > > > 	init+0x95/0x195
> > > > > 	kernel_thread_helper+0x5/0xb
> > > > > Code:  Bad EIP value.
> > > > 
> > > > Both 2.6.15 and 2.6.15.1 boot OK for me.
> > > > .config for 2.6.16-rc1 is at
> > > >   http://www.xenotime.net/linux/doc/config-2616rc1
> > > 
> > > If you disable CONFIG_PNP and CONFIG_ISAPNP options does that help?
> > 
> > Nope, no change.  Any other suggestions?
> > 
> > I just booted with a KOBJECT_DEBUG
> > built kernel and it's failing after:
> > 
> > kobject platform: registering, parent: <NULL>, set: devices
> 
> Can you enable CONFIG_DEBUG_DRIVER and see if that helps?

Yes.  Transcribing that:

kobject platform: registering, parent: <NULL>, set: devices
DEV: registering device: ID = 'platform'
kobject_uevent
PM: Adding info for No Bus:platform
<Oops happens>


---
~Randy

Re: 2.6.16-rc1 kernel init oops

[Randy.Dunlap]

On Sun, 29 Jan 2006 13:08:12 -0800 Randy.Dunlap wrote:

> On Sun, 29 Jan 2006 12:19:23 -0800 Greg KH wrote:
> 
> > On Sun, Jan 29, 2006 at 11:19:34AM -0800, Randy.Dunlap wrote:
> > > On Sun, 29 Jan 2006 11:00:29 -0800 Greg KH wrote:
> > > 
> > > > On Sat, Jan 28, 2006 at 05:55:11PM -0800, Randy.Dunlap wrote:
> > > > > On Sat, 28 Jan 2006 17:18:41 -0800 Randy.Dunlap wrote:
> > > > > 
> > > > > > Hi,
> > > > > > 
> > > > > > I'm trying to boot 2.6.16-rc1 on a T42 Thinkpad notebook.
> > > > > > No serial port for serial console.  I don't think that networking
> > > > > > is alive yet (for network console ?).
> > > > > > 
> > > > > > Anyone recognize this?  got patch?
> > > > > > 
> > > > > > This is just typed in, so could contain a few errors.
> > > > > > 
> > > > > > Unable to handle kernel NULL pointer dereference at virtual address 00000001
> > > > > > printing eip:
> > > > > > 00000001
> > > > > > *pde = 00000000
> > > > > > Oops: 0000 [#1]
> > > > > > SMP DEBUG_PAGEALLOC
> > > > > > Modules linked in:
> > > > > > CPU:	0
> > > > > > EIP:	0060:[<00000001>]   Not tainted VLI
> > > > > > EFLAGS: 00010202   (2.6.16-rc1)
> > > > > > EIP is at 0x1
> > > > > > <skip reg. dump>
> > > > > > <skip stack dump>
> > > > > > Call trace:
> > > > > > 	show_stack_log_lvl+0xa5/0xad
> > > > > > 	show_registers+0xf9/0x162
> > > > > > 	die+0xfe/0x179
> > > > > > 	do_page_fault+0x399/0x4d8
> > > > > > 	error_code+0x4f/0x54
> > > > > > 	device_register+0x13/0x18
> > > > > > 	platform_bus_init+0xd/0x19
> > > > > > 	driver_init+0x1c/0x2d
> > > > > > 	do_basic_setup+0x12/0x1e
> > > > > > 	init+0x95/0x195
> > > > > > 	kernel_thread_helper+0x5/0xb
> > > > > > Code:  Bad EIP value.
> > > > > 
> > > > > Both 2.6.15 and 2.6.15.1 boot OK for me.
> > > > > .config for 2.6.16-rc1 is at
> > > > >   http://www.xenotime.net/linux/doc/config-2616rc1
> > > > 
> > > > If you disable CONFIG_PNP and CONFIG_ISAPNP options does that help?
> > > 
> > > Nope, no change.  Any other suggestions?
> > > 
> > > I just booted with a KOBJECT_DEBUG
> > > built kernel and it's failing after:
> > > 
> > > kobject platform: registering, parent: <NULL>, set: devices
> > 
> > Can you enable CONFIG_DEBUG_DRIVER and see if that helps?
> 
> Yes.  Transcribing that:
> 
> kobject platform: registering, parent: <NULL>, set: devices
> DEV: registering device: ID = 'platform'
> kobject_uevent
> PM: Adding info for No Bus:platform
> <Oops happens>

odd.  in device_add(), <platform_notify> (a function ptr)
has a value of 1.  garbahze.

---
~Randy

Re: 2.6.16-rc1 kernel init oops

[Randy.Dunlap]

On Sun, 29 Jan 2006 15:07:37 -0800 Randy.Dunlap wrote:

> On Sun, 29 Jan 2006 13:08:12 -0800 Randy.Dunlap wrote:
> 
> > On Sun, 29 Jan 2006 12:19:23 -0800 Greg KH wrote:
> > 
> > > On Sun, Jan 29, 2006 at 11:19:34AM -0800, Randy.Dunlap wrote:
> > > > On Sun, 29 Jan 2006 11:00:29 -0800 Greg KH wrote:
> > > > 
> > > > > On Sat, Jan 28, 2006 at 05:55:11PM -0800, Randy.Dunlap wrote:
> > > > > > On Sat, 28 Jan 2006 17:18:41 -0800 Randy.Dunlap wrote:
> > > > > > 
> > > > > > > Hi,
> > > > > > > 
> > > > > > > I'm trying to boot 2.6.16-rc1 on a T42 Thinkpad notebook.
> > > > > > > No serial port for serial console.  I don't think that networking
> > > > > > > is alive yet (for network console ?).
> > > > > > > 
> > > > > > > Anyone recognize this?  got patch?
> > > > > > > 
> > > > > > > This is just typed in, so could contain a few errors.
> > > > > > > 
> > > > > > > Unable to handle kernel NULL pointer dereference at virtual address 00000001
> > > > > > > printing eip:
> > > > > > > 00000001
> > > > > > > *pde = 00000000
> > > > > > > Oops: 0000 [#1]
> > > > > > > SMP DEBUG_PAGEALLOC
> > > > > > > Modules linked in:
> > > > > > > CPU:	0
> > > > > > > EIP:	0060:[<00000001>]   Not tainted VLI
> > > > > > > EFLAGS: 00010202   (2.6.16-rc1)
> > > > > > > EIP is at 0x1
> > > > > > > <skip reg. dump>
> > > > > > > <skip stack dump>
> > > > > > > Call trace:
> > > > > > > 	show_stack_log_lvl+0xa5/0xad
> > > > > > > 	show_registers+0xf9/0x162
> > > > > > > 	die+0xfe/0x179
> > > > > > > 	do_page_fault+0x399/0x4d8
> > > > > > > 	error_code+0x4f/0x54
> > > > > > > 	device_register+0x13/0x18
> > > > > > > 	platform_bus_init+0xd/0x19
> > > > > > > 	driver_init+0x1c/0x2d
> > > > > > > 	do_basic_setup+0x12/0x1e
> > > > > > > 	init+0x95/0x195
> > > > > > > 	kernel_thread_helper+0x5/0xb
> > > > > > > Code:  Bad EIP value.
> > > > > > 
> > > > > > Both 2.6.15 and 2.6.15.1 boot OK for me.
> > > > > > .config for 2.6.16-rc1 is at
> > > > > >   http://www.xenotime.net/linux/doc/config-2616rc1
> > > > > 
> > > > > If you disable CONFIG_PNP and CONFIG_ISAPNP options does that help?
> > > > 
> > > > Nope, no change.  Any other suggestions?
> > > > 
> > > > I just booted with a KOBJECT_DEBUG
> > > > built kernel and it's failing after:
> > > > 
> > > > kobject platform: registering, parent: <NULL>, set: devices
> > > 
> > > Can you enable CONFIG_DEBUG_DRIVER and see if that helps?
> > 
> > Yes.  Transcribing that:
> > 
> > kobject platform: registering, parent: <NULL>, set: devices
> > DEV: registering device: ID = 'platform'
> > kobject_uevent
> > PM: Adding info for No Bus:platform
> > <Oops happens>
> 
> odd.  in device_add(), <platform_notify> (a function ptr)
> has a value of 1.  garbahze.

furthermore, this happens on 2.6.16-rc1 and 2.6.16-rc1-git4,
but <platform_notify> is not at the same memory address in
my builds, so it smells more like a bad ptr reference
somewhere than like bad memory IMO.

---
~Randy

Re: 2.6.16-rc1 kernel init oops

[Greg KH]

On Sun, Jan 29, 2006 at 04:57:32PM -0800, Randy.Dunlap wrote:
> furthermore, this happens on 2.6.16-rc1 and 2.6.16-rc1-git4,
> but <platform_notify> is not at the same memory address in
> my builds, so it smells more like a bad ptr reference
> somewhere than like bad memory IMO.

Hm, can you do a git bisect from 2.6.15 to 2.6.16-rc1 to see if you can
find this?

thanks,

greg k-h

Re: 2.6.16-rc1 kernel init oops

[Randy.Dunlap]

On Sun, 29 Jan 2006 20:44:36 -0800 Greg KH wrote:

> On Sun, Jan 29, 2006 at 04:57:32PM -0800, Randy.Dunlap wrote:
> > furthermore, this happens on 2.6.16-rc1 and 2.6.16-rc1-git4,
> > but <platform_notify> is not at the same memory address in
> > my builds, so it smells more like a bad ptr reference
> > somewhere than like bad memory IMO.
> 
> Hm, can you do a git bisect from 2.6.15 to 2.6.16-rc1 to see if you can
> find this?

follow-up:  for now I guess I'll blame the oopsen that I was seeing
on a bad compiler or just karma.  I don't have proof except that
I installed a new distro + gcc and no longer can cause/create the
problems.  (was using gcc 3.4.3 from Mandriva 2005)

thanks,
---
~Randy