[sane-devel] Xsane locks system w/UMAX Astra 1200S

[David Anderson]

No luck with the smaller buffer size, unfortunately.

I did find this past thread on the Sane mailing list archives:
http://lists.alioth.debian.org/pipermail/sane-devel/2002-September/004071.html

Someone had the same problem as me.  One person was able to resolve the
problem by disabling SCSI bus resets on their Adaptec card.  I checked
my SCSI card BIOS (Adaptec 2930CU) and didn't see anything that was
related to SCSI bus resets.  I tried disabling several things, powering
off, booting up, and trying again - still hung every time.

What irks me the most is that it used to work!  And I bet if I reloaded
Gentoo (a rather large task), it would work again.  Something changed
somewhere along the line that broke it.  I bet it was when I did an
'emerge system' on my box, which will automatically update Gentoo.

I work at a computer shop.  I'm going to see if I can borrow another
SCSI card and give that a try.

If anyone has any suggestions, I'm all ears.  

Thanks!

On Sun, 2004-02-22 at 20:51, abel deuring wrote:
> Dave,
> 
> > A few updates.
> > 
> > 1)  I have confirmed that the CD Writer still functions fine.
> > 
> > 2)  I forced the SCSI card to use IRQ 4 (serial ports are disabled).  No
> > change.  I pulled some extra, unused cards out of the system and also
> > moved the SCSI card to a different slot.  No change.
> > 
> > I'm going to try XSane .92 now.
> 
> As Oliver already wrote, updating XSane will probably not change very 
> much regarding the crash. You could try to use another version of 
> sane-backends, but I have doubts that this will help.
> 
> Sane frontends like XSane or scanimage are just "ordinary" user space 
> programs, and the scanner specific backends are ordinary libraries 
> linked to the frontends. While a frontend or a backend may contain bugs 
> cuasing segfault or garbled images, it is highly unlikely that a user 
> program itself can freeze a Linux box as you are experiencing. This is 
> the reason that I suspect either a hardware failure -- or you hit 
> perhaps a kernel bug.
> 
> One noticeable difference between most programs accessing CD 
> drives/writers and Sane backend for SCSI scanners is that Sane uses 
> comparatively large data sizes in its SCSI commands (typically 128 kB 
> for READ commands), while CD writing software uses probably data block 
> sizes < 32 kB.
> 
> It could be that these larger data block sizes are somehow related with 
> your bug. (though they are not the cause, I think - Sane uses these 
> block sizes since several years without any problem)
> 
> You could try to reduce the block size by setting "option 
> scsi-buffer-size-min" and "option scsi-buffer-size-max" to values like 
> 16384 or 32768.
> 
> Abel
>