[sane-devel] sane SCSI lockup with anything other than binary mode

[Henning Meier-Geinitz]

Hi,

On Sun, Jan 12, 2003 at 09:49:14AM -0500, Carter Brey wrote:
> Running RH 8 distro with upgraded 2.4.19 kernel on a Dell Inspiron 7500 
> (PIII).
> 
> I have an Epson GT-8500 scsi scanner attached to the laptop with an 
> Adaptec 1460A SlimScsi PC card adpater. I use the aha152x kernel module 
> as a driver.

[...]

> Jan  7 15:17:59 localhost cardmgr[455]: socket 0: Adaptec APA-1460 SlimSCSI
> Jan  7 15:17:59 localhost cardmgr[455]: executing: 'modprobe aha152x_cs'

aha152x_cs? Isn't the name of the module "aha152x"?

> Jan  7 15:21:48 localhost kernel: Unable to handle kernel NULL pointer dereference at virtual address 0000001b
> Jan  7 15:21:48 localhost kernel:  printing eip:
> Jan  7 15:21:48 localhost kernel: cd0a4a21
> Jan  7 15:21:48 localhost kernel: *pde = 00000000
> Jan  7 15:21:48 localhost kernel: Oops: 0000
> Jan  7 15:21:48 localhost kernel: CPU:    0
> Jan  7 15:21:48 localhost kernel: EIP:    0010:[<cd0a4a21>]    Not tainted

Kernel oops. Run the output through ksymoops and contact the linux
kernel mailing list. But check the archives before, maybe it's already
solved in a current kernel.

It may well be triggered by the driver not expecting "big" buffers.
However, i thought this kind of bug has been solved long time ago.
Anyway, you could try seeting the environment variable
SANE_SG_BUFFERSIZE to 32768 before running a SANE frontend to decrease
the buffer size.

Bye,
  Henning