Red Hat Bugzilla – Bug 203734
Firewire external HD confuses the kernel on Dell Precision M90
Last modified: 2008-04-27 01:05:02 EDT
Description of problem:
Firewire external HD confuses the kernel on Dell Precision M90. (It worked fine
once if powered on before plugging it in, but has been giving that error
The external drive is an iomega 120BG HDD, model DHD120-C, and is detected as
the following by the kernel the one time it worked:
"scsi1 : SBP-2 IEEE-1394
ieee1394: sbp2: Logged into SBP-2 device
Vendor: HDS72251 Model: 2VLAT20 Rev:
Type: Direct-Access-RBC ANSI SCSI revision"
It also mentions something about forcing the driver to serialize I/O
(serialize_io=1), and that serialize_io=0 might give better performance.
When it doesn't work, you get the following in /var/log/messages:
"kernel: ieee1394: Error parsing configrom for node 0-00:1023"
The firewire port is detected by lspci as:
03:01.0 Firewire (IEEE 1394): Ricoh Co Ltd Unknown device 0832
Version-Release number of selected component (if applicable):
Nearly always. I'm not completely sure why it worked the once (other than that I
powered it on before plugging it into the firewire port, but that didn't make it
happy with subsequent tries).
Things you could try:
- Unload all 1394 drivers, plug disk in, load ohci1394.
- After a failure, use the tool gscanbus (GUI) or 1394commander (nice CLI) to
reset the FireWire bus.
- Check if there is a firmware update for the bridge chip used by Iomega
More diagnostics would require the IEEE 1394 drivers to be recompiled with
Reassigning to correct owner, kernel-maint.
Suzanne, could you try the new stack we're shipping in rawhide and see if you
get better results with that? Thanks.
One much simpler thing I forgot to mention in comment #1:
(plug disk in and switch on)
# modprobe -r ohci1394
(wait a few seconds)
# modprobe ohci1394
If that helps it would of course be interesting to make it work without such a
procedure, but it can be tedious without direct access to the hardware. --- But
as Kristian said, maybe it works now with the rawhide kernel.
Has this bug been fixed by now?
I'm pretty sure this is fixed but will close as insufficient_info.