On Fri, Jun 07, 2002 at 01:02:30PM -0700, Scott Bronson wrote:
> >From the recent discussions it seems that there's some disagreement over
> exactly what device state should be preserved across suspend and
> resume. Is there a policy written down anywhere that tells what to do
> in situations like these?
>
>
> For simple devices like mice and webcams, it's fine to just disconenct
> when suspending and re-enumerate on resume. Printers and scanners can
> be considered simple devices--I don't think it's reasonable to expect
> your printer to resume printing if you suspend your computer in the
> middle of a job.
>
> HOWEVER, all attached devices do need to keep their device nodes.
Why? This keeps people on their toes :)
(just kidding...)
> They cannot re-order themselves arbitrarily while the machine is
> asleep. From my limited tests, it appears that this is currently the
> case. Is there code in place that guarantees this?
Nope.
> What if I move my printer from one hub to another one while the machine
> is asleep? If I re-order devices, but don't plug any new ones in or
> out, should that cause device nodes to switch? I don't have any USB
> devices that share device nodes so I can't test this now.
It could cause that to happen, yes.
> Network devices are a bit more complex because they need to keep
> connections across suspend and resume.
This can be handled by /sbin/hotplug by binding a network device number
to a specific mac address.
> PCMCIA does this currently, but it only does it if devices are NOT
> reordered while the machine is asleep. If I have multiple devices and I
> reorder them while the machine is asleep, they will often get bound to
> different interfaces (eth1 and eth0 will be swapped). Open connections
> are still associated with the old interface, so now they're associated
> with a different device. If I quickly swap them back, the connections
> still work fine. This might be an obscure security risk...?
>
> Is this behavior acceptable for USB? If devices are re-ordered while
> the machine is asleep, then can it just enumerate the moved devices from
> scratch? That seems acceptable to me provided, of course, that all open
> connections are closed before associating an existing interface with a
> different device.
>
> Is there any way to bind a specific network connection to a specific
> device, so that it doesn't matter what order I plug the device in or
> out, or move it, all connections follow it and wait for it to be plugged
> in again? This would be very useful for both PCMCIA and USB.
As I just mentioned, yes, network devices can do this. See:
http://www.kroah.com/linux/hotplug/ols_2001_hotplug_talk/html/mgp00014.html
for one way to do it.
Yes, for other types of USB devices, this would be a good thing to do.
This has been talked over in the past, and hopefully we can have the
time to implement something like this into 2.5. But we need the
driverfs/device model patches in there first :)
thanks,
greg k-h
_______________________________________________________________
Don't miss the 2002 Sprint PCS Application Developer's Conference
August 25-28 in Las Vegas -
http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
