On Fri, Apr 23, 2010 at 5:36 PM, Alan Stern <stern@rowland.harvard.edu> wrote:
quoted text
>
> On Fri, 23 Apr 2010, Chouteau Fabien wrote:
>
> > > > + * When a LUN receive an "eject" SCSI request (Start/Stop Unit),
> > > > + * if the LUN is removable, the backing file is released to simulate
> > > > + * ejection.
> > > > + * The "eject" state of a LUN is available in the "ejected" file of the
> > > > + * LUN's sysfs directory (see above). The "eject" state is only updated
> > > > + * by SCSI request, not by user ejection.
> > >
> > > What's the reason for that?  With a real removable device, like a CD
> > > player, it doesn't make any difference whether the medium was ejected
> > > because of a SCSI command or because I pressed the "eject" button.
> > >
> > > I just don't see any point in keeping track of the two actions
> > > separately, since they end up having the same final result.
> > >
> >
> > By user ejection, I mean send an empty line in the "file" sysfs entry.
> > The Start/Stop request is an action from the USB host side, user
> > ejection is from the USB device side, for me it's two different
> > events.
> > Maybe my comment is not clear about this point.
>
> No; it's clear enough and I understood what you meant.  It's true that
> they are two different events, but they have the same end result.
>
> > I use a FAT disk image as LUN file, users can put some files in the
> > "fake" disk and then eject it. When I get the ejected signal, I mount
> > the disk image on loop device and perform operations on the user's
> > files.
> > So I want to know when users eject the disk and only when users do.
> >
> > I still can use the LUN ejection from device side to disable the mass
> > storage device, and in this case I don't want to mount the disk and
> > search for user's files.
>
> Why not?  Isn't it possible that the user put some files there before
> the device-side eject happened?
>
> What happens if the user and the device both try to eject the medium at
> approximately the same time?  Which event occurs first will be purely
> random chance.  That means there's a 50% probability you will end up
> doing the wrong thing.
>
> No, I think you need to do the same thing whenever an eject occurs, or
> else find a better criterion for deciding what to do.

You're right, I can do the same thing for both ejection.
I'm going to resend the patch without the "eject" sysfs entry.


--
Fabien Chouteau
EPITA GISTR 2010
--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/