On Mon, Dec 17, 2001 at 11:19:31AM +0100, Martin Diehl wrote:

> > > Well, as explained in the other posting, suspend/resume is pretty much
> > > different from unplug/replug because a suspended device is still drawing
> > > some power from the suspended USB. This way even a buspowered device can
> > > keep its state and the host can be sure, it's still the same device which
> > > he will resume later on.
> > 
> > How? The host will not be notified when the device is unplugged and
> > replugged while suspended.
> 
> The HC has to monitor the USB state anyway - at least if it supports
> remote wakeup. A device which signals remote wakeup executes exactly a
> simulated disconnect/reconnect cycle on the USB which triggers the host
> to resume, if enabled. If the host is unable to monitor the bus state
> (I don't know whether this applies to _any_ HC D* state), you are right,
> it can't tell. But in this case the HC is required to signal a connect
> when resuming operation anyway - it's all in chapter 9 of the USB specs.

Again, same problem as mentioned earlier. Think hubs.

> > The host should pass all the relevant information to an userspace agent,
> > which can then decide (even can let the user decide if really needed)
> > whether it is the same device and should be used as such or not.
> 
> Do you really believe in average users' ability to decide whether a device
> which was connected to a different box meanwhile is still the
> "same" device?

No. I believe in most cases this can be handled by software and always
on the safe side. And if this is not convenient, it could be configured
to ask.

> > You usually will want a replacement device just work without having to
> > restart all the programs using it.
> 
> So you wan't the application which has open files on a usb-disk to
> continue writing to these files if there is a replacement device there
> after suspend/resume (with unplug/replug)?
> I doubt that's what you wanted to say?

Yes, I didn't want to say that.

Anyway, this would be the same as with NFS soft/hard mounts. It *makes*
sense.

I must admit I was thinking about mice mostly.

> IMHO the HC+HCD can tell whether this can be treated as a resume or not.
> If not, the application shall be forced to close all open files...

Remember the 'already running port %i disabled by hub (EMI?),
re-enabling...' message? Well, that happens quite often on some machines
and/or hubs. Do you want all applications that have open files on that
device to close them in case of EMI, which can be as frequent as every
couple minutes?

-- 
Vojtech Pavlik
SuSE Labs

_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to