[sane-devel] Debian Etch + ppscsi/microtek2 problem

Eugene Stemple gene_s at hughes.net
Wed Jun 6 19:23:44 UTC 2007

After Linux upgrade to Etch release (4.0) I have a problem with
a Microtek parallel port scanner.

Relevant version numbers are:
    scanimage 1.0.18   from sane-backends_1.0.18-6
    microtek2 backend, from sane-backends_1.0.18-6
    onscsi  module  ver. 0.91  from ppscsi-source_0.3
    ppscsi  module  ver. 0.92  from ppscsi-source_0.3
    sg driver module ver. 3.5.34  def_reserved_size=130560
    kernel  2.6.18

The above combination will scan at low resolutions o.k., and
xscanimage will acquire scan previews o.k.
However, at resolutions greater than 75 px/inch the scan
stalls immediately after writing header info to the out file.

For comparison, scanimage WORKS properly with all the above
program versions EXCEPT:
    sg driver module ver. 3.5.51  def_reserved_size=130560
    kernel  2.6.8

It seems that the backend (microtek2) is happy as long as the
"transferlength" (as reported from scsi_read_image) is less
than ~65535 as it is in previews and low resolution scans.

I have captured debug logs from both the good (kernel 2.6.8)
and the bad (kernel 2.6.18) configurations.  In the failing
configuration all appears normal up through the "readimageinfo"
and subsequent "readimagestatus" stages; then a "readimagecmd"
is attempted but the onscsi debug does NOT show it having been
sent!  The next device read returns 0 bytes (=EOF) and scanimage

Are there subtle changes in the kernel/sg upgrade to Etch that
might need changes in any of the sane components that are
using the (scsi) parallel port interface?

I didn't attach the debug logs  --  they're big  --  but if
extracts would be helpful just tell me what is needed.  And
I have the sane-backends-1.0.18.orig.tar and -6.diff files
in place ready to make any suggested changes, or to insert
debug code.

More information about the sane-devel mailing list