[sane-devel] Acard AEC6712S (atp870u.ko) and HP 6100C C2520A (sg.ko) problems

[abel deuring]

Justin,

> On Fri, Aug 18, 2006 at 11:54:59AM +0200, abel deuring wrote:
>>If possible, try different cables and another SCSI adapter, or
>>connect another SCSI device to the Acard adapter to get a better
>>idea, where the error might come from.
> 
> I took the scanner, SCSI-2 cable, and acard SCSI host adapter and
> plugged them into my roomate's windows computer, installed the
> appropriate drivers, rebooted a few times, suffered through windows'
> occult complaints but eventually got the device functioning.
> 
> Next I tried it on his FC4 system (same hardware, different OS) which,
> upon loading the appropriate kernel modules segfaulted the kernel and
> froze the machine.
> 
> Sometimes I am able to get the kernel modules loaded on my machine
> before the atp870u driver complains.  At this point sane-find-scanner
> detects the scanner
> 
> found SCSI processor "HP C2520A 3644" at /dev/sg1
> 
> but `scanimage -L` does nothing.  Perhaps it precipitates an error of
> some kind because after cancelling the command sane-find-scanner again
> detects nothing and subsequent driver reloadings print out the same set
> of SCSI errors.
> 
> I know at lease someone somewhere has gotten this scanner working before
> because http://www.sane-project.org/sane-mfgs.html says that support for
> 'ScanJet 6100C' is 'Complete'.
> 
> What can I try next?  I'm willing to start hacking hardware/drivers if
> necessary but I have absolutely no idea where to begin, nor even where
> to ask.

Sounds like a bug in the lower level part of the SCSI drivers, or
like a hardware error. I am not sure what I can honestly recommend.
I have never used an Acard SCSI adapter, and don't want to spread
FUD, but anyway: Some time ago, other users reported problems with
Acard SCSI adapters. The atp870u driver contained at least one
strange bug: the data returned for the INQUIRY command was limited
to the standard length, while several scanners describe hardware
details in an -- absolutely "legal" -- vendor-specific "extension"
of the inquiry data, thus confusing one or the other Sane backend. I
noticed the author of the backend about this bug, never got reply,
but somehow and at some time this bug was fixed in the kernel sources.

Generally, I have the impression that there are SCSI adapters which
are better supported by the Linux kernel, so, before you continue to
hunt a bug, it would be worth to try for example an adapter
supported for example by the sym53c8xx or the aic7xxx drivers. In my
experience, these drivers are very stable, well supported, and you
can buy for example a used Adaptec 2940 for a few dollar at Ebay.

Abel