glm and data overruns with Fujitsu M-O drive

glm and data overruns with Fujitsu M-O drive

Post by Jim Seymo » Mon, 14 Jul 2003 23:31:22



Still trying to get a Fujitsu MCM3130SS, 1.3GB magneto-optical drive
working on my Sun Sparc AXi running Sun Solaris 7.

I've installed Joerg Schily's scg driver.  I also found that by
setting the drive to device type mode "X'00' (Direct access device)"
and turning off "automatic spindle motor stop" I can get devfsadm
and boot -r to recognize the drive.

Now the problem is this: When running Schily's sformat, I get the
following after I answer yes to "Ignore old disk label?" and "Modify
Label Geometry Parameters" (sorry for the long lines):

------------------------- begin included text --------------------------

sformat: Warning: controller does not support format parameter page.
sformat: Warning: controller does not support geometry page.
Save Mode Parameters 00 03 00 08 00 00 00 00 00 00 08 00 01 0A 80 1F 28 00 00 00 01 00 00 00
sformat: I/O error. mode select: scsi sendcmd: no error
CDB:  15 01 00 00 18 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 05 00 00 00 00 18 00 00 00 00 26 00 00 00 22 51
Sense Key: 0x5 Illegal Request, Segment 0
Sense Code: 0x26 Qual 0x00 (invalid field in parameter list) Fru 0x0
Sense flags: Blk 0 (not valid)
cmd finished after 0.005s timeout 20s
Set  Mode Parameters 00 03 00 08 00 00 00 00 00 00 08 00 01 0A 80 1F 28 00 00 00 01 00 00 00
Save Mode Parameters 00 03 00 08 00 00 00 00 00 00 08 00 02 0E 80 00 00 00 00 00 00 00 00 00 00 00 00 00
sformat: I/O error. mode select: scsi sendcmd: no error
CDB:  15 01 00 00 1C 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 05 00 00 00 00 18 00 00 00 00 26 00 00 00 22 51
Sense Key: 0x5 Illegal Request, Segment 0
Sense Code: 0x26 Qual 0x00 (invalid field in parameter list) Fru 0x0
Sense flags: Blk 0 (not valid)
cmd finished after 0.005s timeout 20s
Set  Mode Parameters 00 03 00 08 00 00 00 00 00 00 08 00 02 0E 80 00 00 00 00 00 00 00 00 00 00 00 00 00

-------------------------- end included text ---------------------------

Then, after formatting, and again after verifying, I get the following:

------------------------- begin included text --------------------------

sformat: Cannot set geometry to driver (cannot map disk name).
sformat: Error 0. read defect list: scsi sendcmd: no error
CDB:  37 00 1C 00 00 00 00 F0 00 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 01 00 00 00 00 18 00 00 00 00 1C 00 00 00 26 70
Sense Key: 0x1 Recovered Error, Segment 0
Sense Code: 0x1C Qual 0x00 (defect list not found) Fru 0x0
Sense flags: Blk 0 (not valid)
resid: 126976
cmd finished after 0.005s timeout 20s
sformat: I/O error. write_g0: scsi sendcmd: no error
CDB:  0A 09 3E 96 02 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 05 00 00 00 00 18 00 00 00 00 21 00 00 00 24 00
Sense Key: 0x5 Illegal Request, Segment 0
Sense Code: 0x21 Qual 0x00 (logical block address out of range) Fru 0x0
Sense flags: Blk 0 (not valid)
resid: 4096
cmd finished after 0.003s timeout 20s
Could not write defect to disk
Defect list written on disk.

-------------------------- end included text ---------------------------

In at least the case following verifying, I get this in the "console"
window:

------------------------- begin included text --------------------------

Jul 13 00:38:33 jimsun  Resetting scsi bus, data overrun: got too much data from target from (4,0)

Jul 13 00:38:33 jimsun  SCSI transport failed: reason 'reset': retrying command

Jul 13 00:38:37 jimsun  Resetting scsi bus, data overrun: got too much data from target from (4,0)

Jul 13 00:38:37 jimsun  SCSI transport failed: reason 'reset': retrying command

-------------------------- end included text ---------------------------

I get the same messages if I try to mount a partition from the M-O
drive.

The SCSI cabling is all believed good.  SCSI chain is properly
terminated.  I've installed the latest glm patches for Sparc Solaris
7.

So close, and yet so far :(.  Any suggestions?

(Yes, Barbie, I know: They work great under Solaris 9 ;).)

TIA,
Jim
--
Jim Seymour                  | PGP Public Key available at:

http://jimsun.LinxNet.com    |

 
 
 

glm and data overruns with Fujitsu M-O drive

Post by Bruce Adle » Tue, 15 Jul 2003 07:28:18



> ... I also found that by
> setting the drive to device type mode "X'00' (Direct access device)"
> and turning off "automatic spindle motor stop" I can get devfsadm
> and boot -r to recognize the drive.

The drive probably uses a 2K blocksize. The Solaris disk driver requires
that Type 0 Direct Access devices support a 512 byte blocksize. Some
older M-O devices (but not all of them) have a an 512-2K option switch.

The various errors you're seeing are due to the fact that the driver and
the format utility both expect that a Type 0 device support many of the
commands and Mode Sense page parameters that the SCSI-2 spec says are
required for Type 0 devices. I've only checked a couple devices but I
doubt you'll find a M-O device which has an option to fully implement all
the Type 0 commands and parameters. Therefore a couple of (minor) driver
modications from Sun are pretty much required for Solaris to cleanly
support M-O devices. AFAIK, your only other option is to write your own
driver.

 
 
 

glm and data overruns with Fujitsu M-O drive

Post by Jim Seymo » Tue, 15 Jul 2003 10:48:25





>> ... I also found that by
>> setting the drive to device type mode "X'00' (Direct access device)"
>> and turning off "automatic spindle motor stop" I can get devfsadm
>> and boot -r to recognize the drive.

> The drive probably uses a 2K blocksize.

It does indeed.  Apologies for not mentioning that.  I'd asked here
in the past if that would be an issue and somehow gotten the
impression that Sun had gotten over their obsession with 512-byte
blocks.  That's why I didn't mention it.

Quote:>                                         The Solaris disk driver requires
> that Type 0 Direct Access devices support a 512 byte blocksize.

*sigh* I was afraid of that :(.  I presume Solaris 8 suffers from
the same defect?

Quote:>                                                                 Some
> older M-O devices (but not all of them) have a an 512-2K option switch.

This is one of Fujitsu's newer models and does not have that option.
I *can* switch it to a type 7 "Optical memory device" but, of course,
then Solaris doesn't acknowledge its existence at all.  (Except for
probe-scsi-all.)

[technical explanation: snip]

Quote:>                                     Therefore a couple of (minor) driver
> modications from Sun are pretty much required for Solaris to cleanly
> support M-O devices.

LOL!  That'd be the day!

Quote:>                      AFAIK, your only other option is to write your own
> driver.

I have written Unix device drivers and I did write a "floppy" driver
for a real-time kernel once.  So I suppose I could do it if I had
to.  Maybe if there was source out there for an existing working
driver I could hack on...  But to start from scratch?  Alas, too many
other irons in the fire right now.  No time for learning Solaris
device driver stuff.  Not enough interest, in any event.

I did see a post by Schily about a "hacked sd" driver.  But that post
was back in '98, I don't know if it ran under 2.7 and the URL is no
longer valid.

So it would seem my options are: Give up on the M-O drive or switch
to Sparc Linux.  Hmmm...

Thanks for the follow-up, Bruce.  I don't like the answer, but at
least now I do *have* an answer :).

--
Jim Seymour                    | PGP Public Key available at:

                               |
                               | http://jimsun.LinxNet.com

 
 
 

glm and data overruns with Fujitsu M-O drive

Post by Andrew Gabri » Tue, 15 Jul 2003 21:00:11




Quote:

>>                                                                 Some
>> older M-O devices (but not all of them) have a an 512-2K option switch.

> This is one of Fujitsu's newer models and does not have that option.
> I *can* switch it to a type 7 "Optical memory device" but, of course,
> then Solaris doesn't acknowledge its existence at all.  (Except for
> probe-scsi-all.)

I'm not familiar with your drive, but the earlier Fujitsu MO drives
(M2511A, M2512A, M2513A) could take 512byte/sector media. The media
is hard sectored, so you had to buy the right sector size in the
first place.

--
Andrew Gabriel

 
 
 

glm and data overruns with Fujitsu M-O drive

Post by Jim Seymo » Tue, 15 Jul 2003 22:05:13






>>>                                                                 Some
>>> older M-O devices (but not all of them) have a an 512-2K option switch.

>> This is one of Fujitsu's newer models and does not have that option.
>> I *can* switch it to a type 7 "Optical memory device" but, of course,
>> then Solaris doesn't acknowledge its existence at all.  (Except for
>> probe-scsi-all.)

> I'm not familiar with your drive, but the earlier Fujitsu MO drives
> (M2511A, M2512A, M2513A) could take 512byte/sector media. The media
> is hard sectored, so you had to buy the right sector size in the
> first place.

Yes, I'm aware of that.  Unfortunately, Fujitsu didn't see fit to
continue that option into the 1.3GB and 2.3GB models.

Thanks for the follow-up, though.

--
Jim Seymour                    | PGP Public Key available at:
WARNING: The "From:" address   | http://www.uk.pgp.net/pgpnet/pks-commands.html
is a spam trap.  DON'T USE IT! |

 
 
 

glm and data overruns with Fujitsu M-O drive

Post by Philip Bro » Thu, 17 Jul 2003 03:55:13



>I have written Unix device drivers and I did write a "floppy" driver
>for a real-time kernel once.  So I suppose I could do it if I had
>to.  Maybe if there was source out there for an existing working
>driver I could hack on...  But to start from scratch?  Alas, too many
>other irons in the fire right now.

sun has sample drivers, at

http://developers.sun.com/prodtech/solaris/driverdev/reference/codesa...

Presumably, you'd want to tweak "bst - Block SCSI target driver"

--
  http://www.blastwave.org/ for solaris pre-packaged binaries with pkg-get
    Organized by the author of pkg-get
[Trim the no-bots from my address to reply to me by email!]

                            http://www.spamlaws.com/state/ca1.html

 
 
 

1. vold and M-O drives

        Is it possible to use Volume Management on a removable M-O
        device?  (Maxoptix TMT3-1300).  I can mount the drive and see
        the file systems, but I'd like to enable users to change the
        media without root intervention.  The man page for vold.conf
        only mentions CDs and floppys.

--
David B. Chorlian
Neurodynamics Lab  SUNY/HSCB


2. Alpha Numeric Paging Software

3. vendor independent file system format for M-O drives

4. Commercial Linux Support Available

5. SCSI M-O drive on -current

6. Announce: FlightGear 0.7.7

7. Panasonic "PD" m-o/cd drives, Linux and OS/2

8. Which libraries to link with when using FLTK

9. Data overruns?

10. Help: ttyS3 not receiving data. (input overrun)

11. Data overruns?

12. SLIP data overruns

13. Data overrun problem