Re: [PARPORT] Re: Bug in ppa?


David Campbell (campbell@torque.net)
Tue, 19 Jan 1999 11:13:49 +0800


From: alan@lxorguk.ukuu.org.uk (Alan Cox)
Subject: Re: [PARPORT] Re: Bug in ppa?
To: campbell@torque.net (David Campbell)
Date sent: Tue, 19 Jan 1999 03:43:03 +0000 (GMT)
Copies to: huba@unixoft.hu, alan@redhat.com, linux-parport@torque.net

> > of the SCSI drivers under linux would suffer a similar fate if you were to
> > remove the controller card from the machine before unloading the driver
> > (theoretically it is possible to whip a PCI card out without shorting
> > anything...)
>
> Probably
>
> > To actually fix this problem would result in turning the ppa driver into a TSR
> > program that would periodically (say every 5 seconds) probe the parallel port
> > for ppa devices and automatically register/deregister the SCSI hosts. This
> > "would be nice" but the problems that this could cause would be enormous, hence
>
> You can do register/unregister of devices in user space. Am I right in thinking
> that you can detect it not responding - if so perhaps you need to return
> a scsi "media not found" type error if the port has gone walkies ? That
> seems a reasonably truthful error. SCSI alas lacks a "user tripped over
> wire" error.
 
Currently:
  a) The driver can detect the parallel to SCSI interface is missing.
  b) It waits patiently for it to return (probably not a good move)

I have to be careful what I report to the mid level driver, often the first
response the mid-level driver is to perform a request-sense command (a form of
SCSI diagnostic to determine what is really going on...).

Looks like I need to contact someone in the linux-scsi area and ask about the
"user tripped over wire" senario and how to deal with it. Preferably so the mid-
level driver does not try to either have a second attempt or reset the non-
existent device.

Finally I am starting to have real bad visions of what would happen if someone
pulled the cable out, changed the disk and stuck the cable back in again. The
result would be mega filesystem corruption, to the point where you would want
parallel port devices to be sync. (eg: no buffering, read or write).

David Campbell
=======================================================
campbell@torque.net
Check http://www.torque.net/parport for all Linux parallel port solutions.

Current project list:
a) Get ZIP Plus drive back from waranty claim
b) Maintain Linux ZIP drivers (need to update docs)

-- To unsubscribe, send mail to: linux-parport-request@torque.net --
-- with the single word "unsubscribe" in the body of the message. --



This archive was generated by hypermail 2.0b3 on Mon 18 Jan 1999 - 22:21:04 EST