Re: [PARPORT] ppscsi/epst mode question (slow scanning w/ hp5200C)
Tue, 22 Jun 1999 07:10:46 -0400 (EDT)

> If I try to change the mode by echoing mode= into /proc/scsi/epst/0, the
> driver reports the new mode, but stops working. I have to reboot to fix
> this since 'rmmod epst' doesn't work (it reports the device as in use).

When you load the driver, it probes all the possible modes, and determines
which ones actually work. Forcing the mode using /proc is OK if you force
it to a mode that passed the test, but forcing to a known bad mode is not
really a good idea - as you've discovered.

> The scanner works fine with the driver in mode 2 (for which I am very
> grateful) but is rather slow; around 4:30 minutes to scan a 300dpi color
> page. Under Windows 95 with the HP driver/backend it takes about 35
> seconds. I don't know how much, if any, of this performance difference
> is due to the ppscsi/epst drivers and how much is due to SANE.

I haven't timed it exactly, but my HP5200C scans in less than a
minute. Your problem is certainly related to the parallel port mode.

> Anyway. Any ideas as to why I can't get EPP mode to work? Have I done
> something wrong? or will it just not work with my hardware? (And, would
> it even make a difference?) I must plead laziness here since I haven't
> looked at how epst determines the port mode, I thought I'd just ask
> first.

The driver ignores any "official" methods for determining the port settings
and runs a series of tests to see if it can reliably communicate with the
adapter in each of the supported modes.

If you load the driver with the "verbose=1" option, it will log a report
about each of the port/mode combination it has tested. Included in that
report are some numbers which might help to answer your question. Please
send us this output.

You also haven't mentioned anything about the machine you are using. There
are some Thinkpad models, for instance, that appear to support EPP mode,
but don't actually work (and only the actual data transfer test can
detect this problem).

Grant R. Guenther

-- To unsubscribe, send mail to: --
-- with the single word "unsubscribe" in the body of the message. --

This archive was generated by hypermail 2.0b3 on Tue 22 Jun 1999 - 07:13:03 EDT