Re: pci_dio_32 bug?

On Tuesday 06 August 2002 19:32, you wrote:
> On Tue, Aug 06, 2002 at 12:02:06PM +0000, Dominik Kuhlen wrote:
> > Is it possible to crash the kernel with invalid commandarguments?
>
> Only if there's a bug in the driver.
This is getting strange, because this problem
occurs only if I remodprobe only then ni_pcidio driver.
And then only if I use my selfwritten testprogram. All the
comedilib/demo examples work.

If I remodprobe all drivers (comedi, mite, 8255, ni_pcidio)
everything works fine.


Here's a short extraction of the /var/log/messages.
Working (the last one):  
 ni_pcidio: command started
 ni_pcidio_interrupt: status=0x01,flags=0x16,m_status=0x00200000
 group_1_flags: 4 2 CountExpired
 group_status: DataLeft1
 CountExpired
 ni_pcidio_interrupt: status=0x00,flags=0x14,m_status=0x82000000
 group_1_flags: 4 2
 group_status:

Not working 
 ni_pcidio: command started
 ni_pcidio_interrupt: status=0x01,flags=0x35,m_status=0x00000080
 group_1_flags: Waited 4 2 TransferReady
 group_status: DataLeft1
 Waited
 ni_pcidio_interrupt: status=0x01,flags=0x35,m_status=0x82000080
 group_1_flags: Waited 4 2 TransferReady
 group_status: DataLeft1
 Waited


>
> > if i pass :end count 4 and chanlist_len 4 this looks different:
> > if I repeat reading (3-4? times) the read() call blocks
> > and then the kernel is in some critical condition.
>
> There probably was an Oops in the driver code.  Please read the
> man page for ksymoops and send me the output.
Yes, many oopses :-(
I'll try to convert them to useful information. 
>
>
>
Dominik

Received on 2002-08-07Z09:14:12