I wrote on 2005-12-09:
I just saw the following under Linux 2.6.14.3. A web search turns up
more reports of this kind, usually with USB devices.
These are mostly older reports though.
[sbp2 throws transport errors during inquiry]
Dec 9 22:09:56 shuttle kernel: scsi: Device offlined - not ready after
error recovery: host 0 channel 0 id 0 lun 0
Dec 9 22:09:56 shuttle kernel: ieee1394: sbp2: scsi_add_device failed
[I unloaded and reloaded sbp2]
Dec 9 22:10:31 shuttle kernel: SCSI subsystem initialized
Dec 9 22:10:31 shuttle kernel: ieee1394: sbp2: Driver forced to
serialize I/O (serialize_io=1)
Dec 9 22:10:31 shuttle kernel: ieee1394: sbp2: Try serialize_io=0 for
better performance
Dec 9 22:10:31 shuttle kernel: kmem_cache_create: duplicate cache
scsi_cmd_cache
Dec 9 22:10:31 shuttle kernel: ------------[ cut here ]------------
Dec 9 22:10:31 shuttle kernel: kernel BUG at mm/slab.c:1807!
Dec 9 22:10:31 shuttle kernel: invalid operand: 0000 [#1]
...
I had a look at how pool->users is refcounted but did not find
any obvious mistake.
I found and fixed a bug in sbp2 in the meantime. If the inquiry did not
succeed, sbp2 did not release its own data structures and did not remove
the Scsi_Host. I don't know though if this was actually the cause for
the subsequent failure in kmem_cache_create when the SBP-2 unit was
added as a new host.
--
Stefan Richter
-=====-=-=-= ==-- -==--
http://arcgraph.de/sr/
-
: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html