Re: 1394 SBP-2 Drives

Re: 1394 SBP-2 Drives

[James Goodwin]

Hi,

> > > As there are a lot of device on a firewire bus, it 
> probably can't use the
> > > pseudo host controller hack. And shouldn't - it's a hack, which is
> > > unfortunately necessary for usb.
> >
> > I don't know where that came from, but notice I never said
> > anything about such stuff.  In fact, neither does the firewire
> > code I glanced at -- Initiator/Target is the terminology.
> 
> That's from usb-storage. It's the only currently working way
> to get a device that uses the scsi command set to support
> hotplugging.

The USB storage and IEEE-1394 SBP-2 drivers (as well as ide-scsi for that
matter) serve as both scsi low-level drivers and high-level protocol drivers
for their respective busses... this is done in order to leverage the scsi
command set and higher-level block drivers and filesystems.

The primary difference between the two drivers (scsi-wise) is that the USB
storage driver registers a new virtual SCSI host for each attached USB
storage device, while the IEEE-1394 SBP-2 driver registers one SCSI host per
IEEE-1394 bus, but maps each attached SBP-2 device as a virtual SCSI ID.

> I was referring to the SPB-2 subclass of ieee1394 devices which like 
> usb-storage use the scsi command set. The driver for them indeed
> is not part of the standard kernel. These disks (and scanners) are
> the hardest case. If they work everything works in terms of 
> hotplugging.

The main issues currently for the SBP-2 driver (and SCSI) are how to trigger
and deal with scsi device addition or removal, and how to get drives mounted
and unmounted as appropriate. The work being done on /sbin/hotplug can
certainly be leveraged, and the scsi folk are talking about hooks to allow
for event driven addition/removal of devices (in the scsi mid-level code).

One additional issue to think about is "surprise" removal of storage devices
(e.g. recovery and error handling after mounted drives with open files are
hot-unplugged).   ;-)

Cheers,
--James

_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: 1394 SBP-2 Drives

[Oliver Neukum]

> The main issues currently for the SBP-2 driver (and SCSI) are how to
> trigger and deal with scsi device addition or removal, and how to get
> drives mounted and unmounted as appropriate. The work being done on
> /sbin/hotplug can certainly be leveraged, and the scsi folk are talking
> about hooks to allow for event driven addition/removal of devices (in the
> scsi mid-level code).

Mounting drives ? Are you refering to filesystem (on a medium) on the drives ?

If so, I am quite sure that this belongs to user space.
Registering devices to the scsi layer should probably be done in kernel space.

Mounting in user space has its own set of problems which we fortunately
need not discuss here. Another possible issue would be scripts changing
/etc/fstab on hotplugging events. I am not sure I like that idea. It seems 
too risky to keep changing that file.

	Regards
		Oliver

_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: 1394 SBP-2 Drives

[James H. Cloos Jr.]

>>>>> "Oliver" = Oliver Neukum <Oliver.Neukum@lrz.uni-muenchen.de> writes:

Oliver> Another possible issue would be scripts changing /etc/fstab on
Oliver> hotplugging events. I am not sure I like that idea. It seems
Oliver> too risky to keep changing that file.

Agreed.  A better solution to that particular issue may be to
interface with autofs.

-JimC
-- 
James H. Cloos, Jr.  <http://jhcloos.com/public_key>     1024D/ED7DAEA6 
<cloos@jhcloos.com>  E9E9 F828 61A4 6EA9 0F2B  63E7 997A 9F17 ED7D AEA6


_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: 1394 SBP-2 Drives

[David Hinds]

On Sun, Jan 21, 2001 at 01:14:27PM +0100, Oliver Neukum wrote:

> Mounting in user space has its own set of problems which we fortunately
> need not discuss here. Another possible issue would be scripts changing
> /etc/fstab on hotplugging events. I am not sure I like that idea. It seems 
> too risky to keep changing that file.

Well... I won't argue whether it is good or bad... but the PCMCIA
subsystem already will automatically update /etc/fstab when new block
devices are hot plugged, if desired.

-- Dave

_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel