Alan Stern wrote:
Second, concerning soft host removal... I have in mind a situation
where a user wants to unbind usb-storage from a device in a clean, safe
manner, without unplugging anything. Existing commands should be
allowed to complete, caches should be flushed, and so on. Is there any
way to do this?
I haven't tried it with usb-storage, or any other SCSI low-level driver,
but here is something which works with sbp2 or firewire-sbp2:
# readlink /sys/bus/scsi/devices/2\:0\:0\:0
../../../devices/pci0000:00/0000:00:01.0/0000:01:00.0/0000:02:02.0/fw2/fw2.0/host2/target2:0:0/2:0:0:0
# readlink \
/sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0/0000:02:02.0/fw2/fw2.0/driver
../../../../../../../bus/firewire/drivers/sbp2
# echo -n fw2.0 > /sys/bus/firewire/drivers/sbp2/unbind
This results in:
May 2 17:43:54 stein sd 2:0:0:0: [sdd] Synchronizing SCSI cache
May 2 17:43:54 stein firewire_sbp2: released fw2.0, target 2:0:0
The code which does this is is of course the struct driver.remove
callback of fw-sbp2.c. This or another place in fw-sbp2.c drops the
last reference on the SBP-2 target representation, and
sbp2_release_target() is called. This looks up all struct scsi_device
instances which belong to the target and performs
scsi_remove_device(sdev);
scsi_device_put(sdev);
on them. This causes, among else, calls to the .shutdown() method of
the SCSI highlevel driver which is bound to the device, e.g.
sd_shutdown() which among else sends a SYNCHRONIZE CACHE command if
applicable.
After that, the Scsi_Host is being removed because we (currently)
instantiate one Scsi_Host per target, not one per initiator port.
Almost the same happens in drivers/ieee1394/sbp2.c, except that this one
has simply one Scsi_Host per logical unit (i.e. per scsi_device) rather
than per target.
It used to be that scsi_remove_host() would cancel all outstanding
commands. Isn't that still true? If so, is there any way to
accomplish what I want?
Previous iterations of sbp2 and firewire-sbp2 had, independently of each
other, also simply calls to scsi_remove_host() without previous
scsi_remove_device(). I believe the semantics of scsi_remove_host()
changed though over the time... and I gather this is why you are asking
in the first place.
I don't remember how scsi_remove_host() behaved when. I only see in the
git history that I added the scsi_remove_device() in sbp2.c between
2.6.14-rc2 and -rc3. I did this with the explicit goal to support "soft
removal" like you are asking for for usb-storage.
--
Stefan Richter
-=====-==--- -=-= ---=-
http://arcgraph.de/sr/
--
To unsubscribe from this list: 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
