[NTLUG:Discuss] udev - how do get it to create a tape /dev?

[Richard Geoffrion]

The SCSI Generic driver is loaded, the system sees the tape drive AND 
the robotic arm (lun 1), and the tape drive is right there under 
/dev/sg10 with the changer being /dev/sg11.  The problem is that those 
device numbers won't always be the same depending on how many of the hot 
pull SCSI drives are in the bays when the system powers up.  With UDEV 
creating these devices on the fly, I need a way to have a stable 
/dev/$name created so that I don't have to hit a moving target.

I've searched in vain for examples and I'm coming up empty trying to 
find the documentation on how to make udev auto create a stable 
/dev/$name.  

While putting together documentation and proof of how all of my searches 
for help failed, I found
http://www.reactivated.net/writing_udev_rules.html  .   I can't say 
enough good things about this documentation!  The guy uses EXAMPLES!!! I 
am ALMOST  in heaven --- almost because I still seem to be stuck.

To identify my devices, I started in /var/log/messages with this...

---
Jan 19 12:27:36 gwibkp kernel: [  330.856476] scsi4 : Adaptec AIC7XXX 
EISA/VLB/PCI SCSI HBA DRIVER, Rev 7.0
Jan 19 12:27:51 gwibkp udev[4135]: removing device node '/dev/target4:0:0'
Jan 19 12:27:52 gwibkp kernel: [  347.526389]   Vendor: ARCHIVE   Model: 
Python 06241-XXX  Rev: 9100
Jan 19 12:27:52 gwibkp kernel: [  347.526401]   Type:   
Sequential-Access                  ANSI SCSI revision:
 03
Jan 19 12:27:52 gwibkp kernel: [  347.526427]  target4:0:6: Beginning 
Domain Validation
Jan 19 12:27:52 gwibkp kernel: [  347.532654]  target4:0:6: wide 
asynchronous.
Jan 19 12:27:52 gwibkp kernel: [  347.554691]  target4:0:6: Domain 
Validation skipping write tests
Jan 19 12:27:52 gwibkp kernel: [  347.555884]  target4:0:6: FAST-20 WIDE 
SCSI 40.0 MB/s ST (50 ns, offset 32)
Jan 19 12:27:52 gwibkp scsi.agent[4220]: how to add device type= at 
/devices/pci0000:00/0000:00:0e.0/host4/tar
get4:0:6/4:0:6:0 ??
Jan 19 12:27:52 gwibkp kernel: [  347.560632]  target4:0:6: Ending 
Domain Validation
Jan 19 12:27:52 gwibkp kernel: [  347.595306] Attached scsi generic sg10 
at scsi4, channel 0, id 6, lun 0,  ty
pe 1
Jan 19 12:27:52 gwibkp udev[4240]: creating device node '/dev/sg10'
Jan 19 12:27:52 gwibkp scsi.agent[4250]: how to add device type= at 
/devices/pci0000:00/0000:00:0e.0/host4/tar
get4:0:6/4:0:6:1 ??
Jan 19 12:27:52 gwibkp kernel: [  347.614455]   Vendor: ARCHIVE   Model: 
Python 06241-XXX  Rev: 9100
Jan 19 12:27:52 gwibkp kernel: [  347.614466]   Type:   Medium 
Changer                     ANSI SCSI revision:
 03
Jan 19 12:27:52 gwibkp kernel: [  347.614479]  target4:0:6: FAST-20 WIDE 
SCSI 40.0 MB/s ST (50 ns, offset 32)
Jan 19 12:27:52 gwibkp kernel: [  347.649329] Attached scsi generic sg11 
at scsi4, channel 0, id 6, lun 1,  ty
pe 8
Jan 19 12:27:53 gwibkp udev[4270]: creating device node '/dev/sg11'
----

using  devices/pci0000:00/0000:00:0e.0/host4/target4:0:6/4:0:6:0   and
devices/pci0000:00/0000:00:0e.0/host4/target4:0:6/4:0:6:1 as a starting 
point, I run udevinfo and get...

[udevinfo results]
 # udevinfo -a -p 
/sys/devices/pci0000:00/0000:00:0e.0/host4/target4:0:6/4:0:6:0

  looking at class device 
'/sys/devices/pci0000:00/0000:00:0e.0/host4/target4:0:6/4:0:6:0':
    SUBSYSTEM=="unknown"
    SYSFS{device_blocked}=="0"
    SYSFS{iocounterbits}=="32"
    SYSFS{iodone_cnt}=="0x11"
    SYSFS{ioerr_cnt}=="0x5"
    SYSFS{iorequest_cnt}=="0x11"
    SYSFS{model}=="Python 06241-XXX"
    SYSFS{queue_depth}=="2"
    SYSFS{queue_type}=="none"
    SYSFS{rev}=="9100"
    SYSFS{scsi_level}=="4"
    SYSFS{state}=="running"
    SYSFS{timeout}=="0"
    SYSFS{type}=="1"
    SYSFS{vendor}=="ARCHIVE "
[/ results]

I think I was able toproperly identify both devices this way and I wound 
up creating two rules to handle the drive.  I changed the "type" where 
appropriate.   They didn't seem to work though.

[udev.rules insert]
BUS="scsi", KERNEL="sg[0-9]*",SYSFS{model}="Python 06241-XXX", 
SYSFS{type}="1", NAME="%k", SYMLINK="tapedrive"

BUS="scsi", KERNEL="sg[0-9]*",SYSFS{model}="Python 06241-XXX", 
SYSFS{type}="8", NAME="%k", SYMLINK="tapearm"
[/insert]

When I try to diagnose my udev rule, I try to query the device by the 
name assigned to it, /dev/sg10 but I get....

[query device]
# udevinfo -q all -n /dev/sg10
device not found in database
#
[/query]

So..now I don't know where I need to go next.  I've hit the perverbial 
brick wall due to my inability to interpret/understand what udev is 
doing and why it's missing my scsi device.  Any insights??

-- 
Richard (the noobie)