#77 occasional sbp2 lockup when powering down device

[Anonymous]

Kernel: 2.6.6, using in-kernel 1394 drivers
Disk: Acom data external 160GB

Sometimes when I'm done with the device and umount it,
then power it off, the sbp2 subsystem will freeze, not
allowing me to mount the device again until I reboot.

Whenever the freeze ocurrs, this is at the end of my dmesg:

ieee1394: Node changed: 0-01:1023 -> 0-00:1023
ieee1394: Node suspended: ID:BUS[0-00:1023]
GUID[00063a27d40510d2]
Synchronizing SCSI cache for disk sde:

When I umount and shut down device successfully, I see
the two messages from ieee1394, but NOT the
"Synchronizing SCSI cache for disk sde:" message.

I checked in the kernel code and found that there are
two situations when sd_shutdown won't try to
synchronize the cache:

1) Get drive data returns null, or

2) The SCSI device is no longer online.

Could there be a race condition in the code where the
sbp2 instance is destroyed AFTER these checks, causing
the cache sync attempt to lock up?

I'm no kernel hacker, this just my uneducated guess.

-Zorin <lynx at zorin.org>

[Nobody/Anonymous]

Logged In: NO

I would like to add that it seems pointless to try to
synchronize the SCSI cache for a disk that has been
disconnected. The disk is GONE, trying to flush its cache is
impossible.

Perhaps this function should simply return if the shutdown
device is removeable/hotplug?

-Zorin <lynx at zorin.org>

[Nobody/Anonymous]

Logged In: NO

I would like to add that it seems pointless to try to
synchronize the SCSI cache for a disk that has been
disconnected. The disk is GONE, trying to flush its cache is
impossible.

Perhaps this function should simply return if the shutdown
device is removeable/hotplug?

-Zorin <lynx at zorin.org>