Need help with diagnosing SCSI related (probably hardware?) problem (DC395 driver)

Need help with diagnosing SCSI related (probably hardware?) problem (DC395 driver)

[Oliver Feiler]

[-- Attachment #1: signed data --]
[-- Type: text/plain, Size: 1835 bytes --]

Hi,

since I couple of days I have the following problem. Though I have no idea how 
to diagnose it and what is causing it. Maybe someone has a clue if I post it 
here.

My SCSI chain:
Tekram DC395, using the driver from Kurt Garloff 
(http://www.garloff.de/kurt/linux/dc395/)
Attached is a TEAC CD-R56S4 CDR drive and Panasonic LF-D101 DVD-RAM drive. I 
use the DVD-RAM to backup stuff, usually 2 GB gets written to a medium as one 
file.

Since a couple of days however the system just freezes during the backup. As 
the drive LED of the CD burner blinks every minute or so I suspect Linux is 
trying to issue a SCSI reset, because something (presumable the DVD-RAM 
drive) got stuck. This happens endlessly until I press the reset button.

If the system is frozen, no keyboard input works, the screen is not updated 
anymore. You can ping the system, but not ssh into. There is no hard disk 
activity and everything that tries to access the hard disk seems to get stuck 
also. ssh'ing into the machine before starting the backup and using "tail -f 
/var/log/messages" doesn't give any output.

After a reset when the SCSI controller BIOS inits it gets stuck when scanning 
ID2, which is the DVD-RAM drive. Only power cycling gets it to boot again. So 
I guess this device might be the problem. Though it's kinda annoying that the 
kernel gets stuck in this endless reset loop (or whatever happens at that 
time) needing to reset the box. Which also makes it practically impossible to 
debug this problem any further and definitely find out what the problem is.

If anyone has further ideas I could try I'd be happy to hear about them. I 
only want to dump the drive if I can really be sure that it's broken.

Thanks! :)

Oliver

-- 
Oliver Feiler  <kiza@(kcore.de|lionking.org|gmx[pro].net)>

[-- Attachment #2: signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: Need help with diagnosing SCSI related (probably hardware?) problem (DC395 driver)

[Kurt Garloff]

[-- Attachment #1: Type: text/plain, Size: 1293 bytes --]

Oliver,

On Mon, Dec 01, 2003 at 12:08:39PM +0100, Oliver Feiler wrote:
Content-Description: signed data
> CC'ing to Kurt Garloff since he wrote the driver initially and might be the 
> best person to ask what the syslog output means. Are you still maintaining 
> the DC395 driver, Kurt, or has someone else taken maintainership in the 
> meantime?

For 2.6, Oliver Neukum and others have picked up the driver.
I maintain the 2.4 version still.

> Small followup to the described problem. I managed to get a "screenshot" from 
> the kernel's syslog output. Indeed the driver is queueing endless SCSI resets 
> with little success. These two pages (some parts may be missing. Taken with a 
> digital camera) scroll over the screen endlessly. Can someone tell me from 
> this what's going on and what it is causing the problem?
> 
> http://home.kcore.de/~kiza/stuff/dc395.jpg
> http://home.kcore.de/~kiza/stuff/dc395-2.jpg
> (both 1024x768 ~100KB)

Hmm, it looks quite normal; the driver is trying to recover from errors.
Apparently the write10 (2a) timed out. Can you try to catch the first
error message? 

Regards,
-- 
Kurt Garloff  <garloff@suse.de>                            Cologne, DE 
SUSE LINUX AG, Nuernberg, DE                          SUSE Labs (Head)

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]