On Wed, May 03, 2000 at 11:02:21AM +0200, abel deuring wrote:
> > The panic string was "Foooooooood
> > fight!" in both cases.
>
> This is a problem inside the Linux kernel. The panic occurs in the
> aha1542 driver, if the program using the driver has declared a maximum
> buffer length of more than 32768 bytes. The problem is fixed with the
> most recent SG driver (http://www.torque.net/sg). With Sane 1.0.2, you
> can also avoid the kernel panic by either setting the environment
> variable SANE_SG_BUFFERSIZE to 32768 or by running configure with the
> parameter --enable-scsibuffersize=32768.
Thanks very much, that did it! Upgrading to the latest sg
(2.1.37) alone didn't help but limiting the buffer size did.
When fiddling with scanimage to see whether I can still get it
to panic it suddenly started scanning and filled my screen with
the most beautiful garbage I've ever seen ;-)
I'm not a photo editor, but in the few test scans I did (flatbed
only, I haven't got the transparency patch yet) the colour
balance appeared to be quite alright.
I've observed a couple of strange things, though:
Whenever I start a scan the kernel produces the message "Wrong
buffer length supplied for request sense (20)". It doesn't seem
to break anything, but anyway.
Secondly, when using xscanimage and the preview window, after
about 25 or 30% of the preview has arrived, strange things are
happening to some system timer:
- DPMS and screen blanking immediately kick in, the only way of
getting rid of the blank screen is to do an xset -dpms and
xset s off (this is XFree86 v.4.0 on a G400 MAX).
- My gkrellm and the asclock applet in the gnome panel stop
updating. The clock applet jumped ahead by more than an hour
once.
- It doesn't affect all times and timers, date still keeps
showing the correct time, and window manager animations appear
to be timed as usual. Top updates at the same old interval and
doesn't show anything unusual (IIRC).
It's gotta be an interrupt thing, hasn't it? The aha1542 adapter
uses IRQ 10, DMA 0 and IO 330. I reserved IRQ 10 and DMA 0 in
the BIOS for ISA/EISA devices, so there shouldn't be any
conflicts. The only other ISA card I'm using is a soundblaster
compatible which uses different settings (and has them reserved
in the BIOS, too).
Simply loading the scsi modules doesn't cause the problem, only
using xscanimage, so I'm not even sure about blaming
interrupts. Could it be a GNOME/gtk+ bug in the end?
The only way to get out of the "time warp" seems to be reboot,
i.e. it affects all subsequently started X servers, too.
Cheers
Steffen.
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
This archive was generated by hypermail 2b29 : Wed May 03 2000 - 18:30:16 PDT