Comment 9 for bug 899757

Sebastien kindly did some more debugging, keeping notes here. This seems to be related to media polling (we verified both in-kernel and udisks based polling trigger this, and cause the same uevents).

Correlating udev events and the corresponding udev debug output, we get:

UDEV [26112.066235] remove /devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host29/target29:0:0/29:
0:0:0/block/sdb/sdb1 (block)
--
**** REMOVING /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host29/target29:0:0/29:0:0:0/block/
sdb/sdb1
**** EMITTING REMOVED for /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host29/target29:0:0/29:
0:0:0/block/sdb/sdb1

UDEV [26112.067123] remove /devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host29/target29:0:0/29:0:0:0/block/sdb (block)
--
**** REMOVING /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host29/target29:0:0/29:0:0:0/block/
sdb
**** EMITTING REMOVED for /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host29/target29:0:0/29:
0:0:0/block/sdb
**** scsi_host IGNORING REMOVE /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host29/scsi_host/h
ost29

Now we get a rather weird change uevent with a hilariously broken device path and device name, at a time when sdb is already gone!

UDEV [26112.087644] change /sdb (block)
ACTION=change
DEVNAME=/dev/sdb
DEVPATH=/sdb
DEVTYPE=disk
DISK_MEDIA_CHANGE=1
SUBSYSTEM=block
--
**** TREATING CHANGE AS ADD /sys/sdb
**** ADDING /sys/sdb
**** UPDATING /sys/sdb
**** ADDED /sys/sdb
**** EMITTING ADDED for /sys/sdb

Now we get a similar change event again, but this time WITHOUT DISK_MEDIA_CHANGE

UDEV [26112.102230] change /sdb (block)
ACTION=change
DEVNAME=/dev/sdb
DEVPATH=/sdb
DEVTYPE=disk
SUBSYSTEM=block
--
**** CHANGING /sys/sdb
**** UPDATING /sys/sdb
**** EMITTING CHANGED for /sys/sdb
**** CHANGED /sys/sdb

Now we get the device back:

UDEV [26123.419938] add /devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host32/target32:0:0/32:0:0:0/block/sdb (block)
--
**** ADDING /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host32/target32:0:0/32:0:0:0/block/sdb
**** UPDATING /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.0/host32/target32:0:0/32:0:0:0/block/sdb

(udisks-daemon:24900): udisks-daemon-ERROR **: **** HACK: Wanting to register object at path `/org/freedesktop/UDisks/devices/sdb' but there is already an object there. This is an internal error in the daemon. Aborting.

So I think the problem is that the media polling for the /sdb device creates a device in udisks.