Re: joydev: kernel panic accessing /dev/js0 with 2.6.2-rc1-bk1

Re: joydev: kernel panic accessing /dev/js0 with 2.6.2-rc1-bk1

[Vojtech Pavlik]

On Sun, Jan 25, 2004 at 07:07:31AM +0100, rvalles wrote:
> > If you also could get the function where EIP is, that'd be great.
> Oops! Did I really post all the stack trace and forgot about the current
> function? Well, I wonder what was I thinking. Here it is:
> 
> EIP is at hidinput_hid_event+0x8f/0x2a0
> 
> Also, had to boot from a Knoppix 3.3 for some stuff, and verified, with jstest,
> that the pad seems to be working well with its kernel (2.4.22-xfs).
> 
> I also forgot to tell you that I've got CONFIG_HID_FF and CONFIG_LOGITECH_FF
> enabled on my 2.6.2-rc1-bk1.
> 
> cc me on reply. I'm not suscribed into linux-kernel.

Ok, now I see how and where it happens. But I need the device HID
desriptor dump. Please enable DEBUG and DEBUG_DATA in hid-core.c,
and send me the output it produces into 'dmesg'. If you want to avoid
the crash, simply disable hid-input in kernel config for the experiment.

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR

Re: joydev: kernel panic accessing /dev/js0 with 2.6.2-rc1-bk1

[Vojtech Pavlik]

On Sat, Jan 24, 2004 at 10:03:54AM +0100, rvalles wrote:

> I got a new USB gamepad called "Logitech WingMan RumblePad".
> 
> I had USB with UHCI HCD working on 2.6.1, for my mouse (USB full HID) on
> kernel (not as a module). But I didn't have joystick support, so I got myself
> 2.6.2-rc1-bk1 and built it, with Joystick Interface as a module.
> 
> Then, I do:
> # modprobe joydev
> # cat /dev/js0
> 
> and I get:
> DIVIDE ERROR:0000[#1]
> 
> Calltrace:
> hid_process_event
> hid_input_field
> add_timer_randomness
> hid_input_report
> hid_irq_in
> usb_hcd_irq
> handle_IRQ_event
> do_IRQ
> rest_init
> common_interrupt
> rest_init
> <...>
> 
> Kernel Panic: Fatal exception in interrupt
> 
> I've shortened a lot of detail, through, because I had to copy it by hand. I
> hope it's enough to debug it. Tell me otherwise, and I'll copy the needed
> stuff.

If you also could get the function where EIP is, that'd be great.

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR