Re: Fwd: Composite USB devices

On 10/1/07, Matthew Dharm <[email protected]> wrote:
> For the sake of the list archives, could you provide the device/model
> number (i.e. how is it marked on the box) for this thing?
>
> I've not seen very many RAID-enabled dual-HD enclosures for USB.
>
> Matt

Done, see below.


On 10/1/07, Alan Stern <[email protected]> wrote:
> Could you summarize your off-list discussion for the benefit of the
> rest of us?
>
> Alan Stern

Hopefully Matt will not object to having his emails pasted...

This is a simplified log of the conversation between Matt Dharm and I
about the Western Digital "My Book Dual-drive Storage System with RAID
and Backup Software", order WDG2T10000N/model WD10000C033-001. I had
been digging through the kernel code, looking into TYPE_ENCLOSURE
more, and had found essentially no real code dealing with it. It
*looks* like a stub framework for later implementation, so I ask...

me:

Is this USB device type (TYPE_ENCLOSURE) supported at all, or is it
just a skeletal framework so far?

him:

TYPE_ENCLOSURE is a SCSI device type reserved for (usually large) HD
enclosures.  Things like 19" rack-mount multi-drive bays.

While the transport to and from these devices is supported (as with all
SCSI devices), there is no linux support for these devices that I am aware
of.

I've been following your thread on the mailing list.  I'm pretty sure your
device is just reporting the wrong type code.  Nobody sane makes a USB
enclosure that returns TYPE_ENCLOSURE.

me:

No offense, but I'm less concerned with it being sanely designed
and/or correct, and far more concerned with it working.

Correct me if I'm wrong, but wouldn't something so simple as it just
returning the wrong type, just be a matter of the patch which was
submitted, no matter how incorrectly, making TYPE_ENCLOSURE
essentially an alias for TYPE_DISK? Please note that that did not
work.

Also noteworthy is that there are two 500GB drives in the "enclosure"
with RAID striping handled internally. So technically, it is a large
multi-drive enclosure.

him:

Now that's interesting information.

me:

Isn't it, though? I have a knack at finding odd glitches in things,
with or without odd hardware.

I don't suppose you've got any idea where I would drop in a small
patch to simply make the USB subsystem essentially ignore the first
"I" line, and go for the second one instead? because really, that's
the main problem, as far as i can tell. It's going for the HID and
ignoring the mass storage, and I need it to do vice versa. Protocol
and standards be damned.

him:

First of all, nothing is getting ignored.  Your data shows that usb-storage
is binding to every interface which is available to it.  USB supports
binding multiple drivers to the same device via multiple "Interfaces" --
that's what the "I" stands for.

My best guess at this point is that the enclosure is actually a multi-LUN
device and you're kernel is only probing the first LUN.  The second LUN is
probably the RAID controller.

Do you have CONFIG_SCSI_MULTI_LUN turned on?

me:

I did test that, and got no results, but now that I think about it, i
was having other issues at the same time, so lemme give it another
shot and see what happens.

---

...and thus the problem was solved.

If anyone's wondering, I'm pretty sure the HID device is used to
reconfigure the device from a 1TB stripe to a 500GB mirrored set "for
data protection".

Also noteworthy, the issues I stated I was having before arose from
GREP_OPTIONS having --color=always in it. Modules won't install with
that, it just says "MODPOST 0 Modules" and runs depmod like a friggin
genius. "unset GREP_OPTIONS; export GREP_OPTIONS; make modules
modules_install" ftw.

So end result/summary: none of this would have happened at all, if I
had rememebered to try the solution I'd tried before I noticed I had a
completely unrelated problem that was preventing module installation
in the first place.

Thanks again to Matt

--
pegasus

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
[email protected]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-users