Re: 2.6.16-rc1 kernel init oops

["Randy.Dunlap" <rdunlap@xenotime.net>]

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