Jason Thorpe wrote:

>    > For Mass Storage Devices an alternative would be not to detach the
>    > device at once, but hope that it will reappear shortly.  The serial number
>    > of the device can be used to detect if the same device reappears.
>
>   Two problems:
>
>           What happens if the device never comes back?

Well, I wasn't suggesting that you'd wait forever, but rather that there would
be a delay before you give up on the disk.


>           What about other SCSI disks (like ones plugged into PCMCIA cards)
>           which don't have this serial number facility?

Well, this feature would only work for busses where devices can be reattached.
So perhaps the whole thing should be hidden in the umass driver.  I'm not sure
how to make such a thing work in the current device framework, but then I've
not been thinking much about it.


>   I think the best option is to have an unmount happen, and to use amd to
>   mount the disk in the first place (so you only have to cd back into the
>   directory to get it back, once the device exists again).

But what about all those unflushed buffers?  It's kind of irritating to know
that you've lost a file just because the cat stepped on a cable when you know
all the data is in the memory somewhere and all that is needed is to write
it to disk once it's reattached again.

But I think we can leave this as future enhancements.  Thanks for fixing
the SCSI detach!

    -- Lennart