Oops in vanilla 2.4.22 serial-usb driver

Oops in vanilla 2.4.22 serial-usb driver

[Philippe Troin]

This happened at the end of a Palm sync.
The machine went on, but USB is not irresponsive to USB attach/detach
since khubd is dead. The USB low-level driver is UHCI_ALT, compiled
in-kernel.

BTW, is there any way to restart khubd without rebooting?

Phil.

Kernel info:

vanilla 2.4.22 + vfs-lock + lvm-1.0.7 + kmsgdump 0.4.4
Linux ceramic 2.4.22 #1 SMP Tue Sep 16 18:14:44 PDT 2003 i686 unknown

Oops:

Unable to handle kernel NULL pointer dereference at virtual address 000009a4
ed341d05
*pde = 00000000
Oops: 0002
CPU:    1
EIP:    0010:[<ed341d05>]    Tainted: P 
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010246
eax: 00000000   ebx: dae4bc90   ecx: 00000046   edx: 00000001
esi: 00000002   edi: dae4bc00   ebp: 00000000   esp: e7d1df48
ds: 0018   es: 0018   ss: 0018
Process khubd (pid: 10, stackpage=e7d1d000)
Stack: ed342e20 d16591e0 ed342e40 c01f2fd0 ce5eac00 dae4bc00 00000100 00000002 
       e7d2b000 00000001 00000000 ce5eac00 c01f539e e7d2b114 00000001 00000002 
       e7d2b000 e7d1f8a0 00000001 00000001 0000000a c01f565d e7d1f8a0 00000001 
Call Trace:    [<ed342e20>] [<ed342e40>] [<c01f2fd0>] [<c01f539e>] [<c01f565d>]
  [<c01f584b>] [<c0105a54>]
Code: c7 80 a4 09 00 00 00 00 00 00 8d 4b 5c f0 ff 43 5c 0f 8e ae 


>>EIP; ed341d05 <[usbserial]usb_serial_disconnect+6d/1ec>   <=====

>>ebx; dae4bc90 <_end+1aae7f44/284a02b4>
>>edi; dae4bc00 <_end+1aae7eb4/284a02b4>
>>esp; e7d1df48 <_end+279ba1fc/284a02b4>

Trace; ed342e20 <[usbserial]usb_serial_driver+0/0>
Trace; ed342e40 <[usbserial].data.start+20/3c>
Trace; c01f2fd0 <usb_disconnect+88/128>
Trace; c01f539e <usb_hub_port_connect_change+4a/20c>
Trace; c01f565d <usb_hub_events+fd/29c>
Trace; c01f584b <usb_hub_thread+4f/f4>
Trace; c0105a54 <arch_kernel_thread+28/38>

Code;  ed341d05 <[usbserial]usb_serial_disconnect+6d/1ec>
00000000 <_EIP>:
Code;  ed341d05 <[usbserial]usb_serial_disconnect+6d/1ec>   <=====
   0:   c7 80 a4 09 00 00 00      movl   $0x0,0x9a4(%eax)   <=====
Code;  ed341d0c <[usbserial]usb_serial_disconnect+74/1ec>
   7:   00 00 00 
Code;  ed341d0f <[usbserial]usb_serial_disconnect+77/1ec>
   a:   8d 4b 5c                  lea    0x5c(%ebx),%ecx
Code;  ed341d12 <[usbserial]usb_serial_disconnect+7a/1ec>
   d:   f0 ff 43 5c               lock incl 0x5c(%ebx)
Code;  ed341d16 <[usbserial]usb_serial_disconnect+7e/1ec>
  11:   0f 8e ae 00 00 00         jle    c5 <_EIP+0xc5> ed341dca <[usbserial]usb_serial_disconnect+132/1ec>

Module list:

nfsd                   67264  12
visor                  10624   0
usbserial              16224   0 [visor]
microcode               3744   0 (autoclean)
mousedev                3904   1
lvm-mod                59680   9
quota_v2                6376   2
nfs                    69372   5
lockd                  47648   1 [nfsd nfs]
sunrpc                 65396   1 [nfsd nfs lockd]
vfat                    9308   0 (unused)
msdos                   4828   0
fat                    29944   0 [vfat msdos]
isofs                  17408   0 (unused)
udf                    79392   0 (unused)
eeprom                  3520   0 (unused)
mtp008                  8064   0 (unused)
i2c-proc                5952   0 [eeprom mtp008]
i2c-viapro              3848   0 (unused)
i2c-core               12192   0 [eeprom mtp008 i2c-proc i2c-viapro]
syncfb                 12098   0 (unused)
matroxfb_base          19040  63
matroxfb_DAC1064        5312   0 [matroxfb_base]
matroxfb_accel          7264   0 [matroxfb_base matroxfb_DAC1064]
matroxfb_misc          14528   0 [matroxfb_base matroxfb_DAC1064 matroxfb_accel]fbcon-cfb24             4192   0 [matroxfb_accel]
fbcon-cfb8              3264   0 [matroxfb_accel]
fbcon-cfb32             3616   0 [matroxfb_accel]
fbcon-cfb16             3904   0 [matroxfb_accel]
mga_vid                 8256   0
mga                    98576   1
agpgart                17696   3
emu10k1                56512   2
sound                  52940   0 [emu10k1]
ac97_codec             11648   0 [emu10k1]
soundcore               3460   7 [emu10k1 sound]
apm                     9504   0
softdog                 1764   1
loop                    8624   0
floppy                 47072   0
sg                     25092   0 (unused)
sr_mod                 11896   0
cdrom                  26976   0 [sr_mod]
st                     27344   0
3c59x                  25864   1
af_packet              13096   0 (unused)

Re: Oops in vanilla 2.4.22 serial-usb driver

[Greg KH]

On Wed, Sep 24, 2003 at 09:17:00PM -0700, Philippe Troin wrote:
> This happened at the end of a Palm sync.
> The machine went on, but USB is not irresponsive to USB attach/detach
> since khubd is dead. The USB low-level driver is UHCI_ALT, compiled
> in-kernel.

This should be fixed in the 2.4.23-pre tree now.  If you want, you can
try applying the patch at:
	http://www.kernel.org/pub/linux/kernel/people/gregkh/usb/2.4/usb-serial-02-2.4.23-pre3.patch
if you don't want the whole -pre tree.

> BTW, is there any way to restart khubd without rebooting?

Nope, sorry.

Let me know if this doesn't solve the problem for you.

thanks,

greg k-h

Re: Oops in vanilla 2.4.22 serial-usb driver

[Philippe Troin]

Greg KH <greg@kroah.com> writes:

> On Wed, Sep 24, 2003 at 09:17:00PM -0700, Philippe Troin wrote:
> > This happened at the end of a Palm sync.
> > The machine went on, but USB is not irresponsive to USB attach/detach
> > since khubd is dead. The USB low-level driver is UHCI_ALT, compiled
> > in-kernel.
> 
> This should be fixed in the 2.4.23-pre tree now.  If you want, you can
> try applying the patch at:
> 	http://www.kernel.org/pub/linux/kernel/people/gregkh/usb/2.4/usb-serial-02-2.4.23-pre3.patch
> if you don't want the whole -pre tree.

That's what I had figured out from looking at the 23rcX changelogs and
diffs.

> > BTW, is there any way to restart khubd without rebooting?
> 
> Nope, sorry.

Are there any technical reasons behind that, or that just that it is
not implemented?
 
> Let me know if this doesn't solve the problem for you.

I won't be able to test until later today, but I will let you know.

Thanks,
Phil.

Re: Oops in vanilla 2.4.22 serial-usb driver

[Greg KH]

On Thu, Sep 25, 2003 at 01:27:57PM -0700, Philippe Troin wrote:
> > > BTW, is there any way to restart khubd without rebooting?
> > 
> > Nope, sorry.
> 
> Are there any technical reasons behind that, or that just that it is
> not implemented?

It's a bit hard to restart a kernel thread that is oopsed :)

greg k-h

Re: Oops in vanilla 2.4.22 serial-usb driver

[Philippe Troin]

Greg KH <greg@kroah.com> writes:

> On Thu, Sep 25, 2003 at 01:27:57PM -0700, Philippe Troin wrote:
> > > > BTW, is there any way to restart khubd without rebooting?
> > > 
> > > Nope, sorry.
> > 
> > Are there any technical reasons behind that, or that just that it is
> > not implemented?
> 
> It's a bit hard to restart a kernel thread that is oopsed :)

Yes, but still, with a completely modular USB subsystem, removing all
the modules and reinsrting them (when possible) restarts khubd... So
if it is possible with modules, it ought to be possible with
monolithic USB...

Phil.

Re: Oops in vanilla 2.4.22 serial-usb driver

[Greg KH]

On Thu, Sep 25, 2003 at 03:31:25PM -0700, Philippe Troin wrote:
> Greg KH <greg@kroah.com> writes:
> 
> > On Thu, Sep 25, 2003 at 01:27:57PM -0700, Philippe Troin wrote:
> > > > > BTW, is there any way to restart khubd without rebooting?
> > > > 
> > > > Nope, sorry.
> > > 
> > > Are there any technical reasons behind that, or that just that it is
> > > not implemented?
> > 
> > It's a bit hard to restart a kernel thread that is oopsed :)
> 
> Yes, but still, with a completely modular USB subsystem, removing all
> the modules and reinsrting them (when possible) restarts khubd... So
> if it is possible with modules, it ought to be possible with
> monolithic USB...

Usually when khubd oopses like this, a usb module is stuck with a
incremented reference count which prevents the usbcore from being able
to be unloaded.

But hey, patches are always accepted :)

thanks,

greg k-h

Re: Oops in vanilla 2.4.22 serial-usb driver

[Philippe Troin]

Greg KH <greg@kroah.com> writes:

> On Wed, Sep 24, 2003 at 09:17:00PM -0700, Philippe Troin wrote:
> > This happened at the end of a Palm sync.
> > The machine went on, but USB is not irresponsive to USB attach/detach
> > since khubd is dead. The USB low-level driver is UHCI_ALT, compiled
> > in-kernel.
> 
> This should be fixed in the 2.4.23-pre tree now.  If you want, you can
> try applying the patch at:
> 	http://www.kernel.org/pub/linux/kernel/people/gregkh/usb/2.4/usb-serial-02-2.4.23-pre3.patch
> if you don't want the whole -pre tree.

I confirm that this patch fixes the problem.
Thanks Greg.

Phil.