Hi all,
problem found (and a tale of woe)!
For several years I've shared the SCSI scanner between two computers,
my wife's and mine. My wife has an Adaptec 152x card set to SCSI ID 7
and my computer has an Adaptec 2940 (1.23) set to ID 6. The scanner
is set to ID 2 (there is also a tape drive on ID 5).
This worked fine, running Win95 on my wife's system and Linux 2.0.36
on mine (and a really early version of SANE, 0.71 I think).
Then, sometime last year, I upgraded my system to the 2.2.14 kernel
plus all the user mode stuff, but didn't get around to installing
a new version of SANE. When I finally did install SANE, I started to
have the corrupted image problem that I had documented in an earlier
email, thinking the problem had to be SANE or the SCSI subsystem in
the 2.2.14 kernel. Based on this assumption I build kernels with
all sorts of versions of the sg driver (including the new V3) and
lots of versions of the AIC driver. Nothing worked. I never suspected
that the problem might be caused by my wife's computer (which had
been upgraded to Win98 somewhere along the way).
Finally, in desperation, I shut off my wife's computer, and voila,
the problem disappeared. Armed with that information, I quickly
tracked down the culprit, STIMON.EXE. STIMON.EXE (the WDM StillImage-
Monitor) queries the scanner every so often (about every 100 ms as
far as I am able to determine). The scanner firmware however
seems to be a strickly sequence/task oriented piece of software
that could not deal with the fact that a query would be arriving
from a different SCSI ID while a scan was initated and in progress
from another SCSI ID. Disabling STIMON.EXE solved the problem on
the Linux side, but disabled the scanning capability on the Windows
machine. Turns out though, the Control Panel applet for Cameras and
Scanners on Windows allows you to disable "Device Event Messages".
With this disabled, I could run STIMON.EXE again and scan from both
Windows and Linux without problems.
Interesting side note: The firmware of the scanner crashes completely
if two hosts attempt to scan simultaniously. It seems to need a
power cycle to resurrect it.
Thank you all very much for your suggestions!
Cheers,
Jens
>
> Jens Petersohn wrote:
> Hi Jens,
>
> I dont know how to solve the problem. But I think it must have something
> to do with the SCSI interface. The image looks like some bytes have been
> lost. Two of the strips in the image have exactly the same size of
> 393216 Bytes == 384KB. One of strips is one byte more than 416KB. Both
> values are multiple of 32KB. If you have the sources of the hp-backend,
> you can try to change the value of HP_PIPEBUF in file hp-scl.c from
> 32768 to 16384 or 4096. When the compilation was successful, you should
> see the new value in the log file with the messages "entering request to
> read num bytes" and "trying to read num bytes".
> I am not sure if this solves the problem and if so why.
>
> --Peter
>
-- 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 Aug 23 2000 - 11:58:03 PDT