Re: Helping out with autosuspend

On Fri, Sep 14, 2007 at 01:34:27AM -0700, Greg KH wrote:
> On Thu, Sep 13, 2007 at 04:16:01PM -0700, Sarah Sharp wrote:
> > I'd like to get involved with USB autosuspend, and I'm looking for ideas
> > for where to start.  I've been following the recent discussions about
> > switching to a whitelist, but I don't think that changes the need for
> > autosuspend support in kernel drivers.
> 
> Exactly, we still need that.
> 
> > Currently, Windows XP supports autosuspend [1] for:
> >  - hubs
> 
> So do we :)
> 
> >  - hid devices (with a whitelist)
> >  - bluetooth (self powered devices only)
> >  - USB modems
> > 
> > Linux is lacking in bluetooth autosuspend support.  Would that be a good
> > place for me to start?  Or is there a better autosuspend project for me
> > to work on?
> 
> I think the USB bluetooth drivers have a major rewrite that has not hit
> the tree.  It will be much smaller, and it might have suspend added to
> it, you should as the maintainer of it about this.

Ok, I'll send mail to them. 

> The "whitelist" effort is all done as far as the kernel is concerned,
> but userspace needs some work that is already being planned by the
> udev/HAL developers.
> 
> There are patches outstanding for a number of usb networking drivers,
> the HID driver, and the usb-storage driver.
> 
> Any other driver is probably up for grabs if you want to try to work
> with some devices that you already have.
> 
> good luck,
> 
> greg k-h
> 
> p.s. what happend to the usbfs2 work?

The short answer is I need to jump through some hoops to get Intel to
pay me to work on it.  Otherwise it will languish in my "things to do in
my copious spare time" queue.

I talked with Zack Brown yesterday about AIO and syslets.  We worked out
that usbfs2 could work with syslets, with a slight modification to the
syslet API.  The user could set a flag to tell the syslet interface not
to clone if the operation blocks; usbfs2 could make a function call
after it submits the URB to let the syslet code know that it's ok to
clone after that point.  This would solve the data reordering problem,
which was the only reason I didn't like syslets.

It's not perfect, since it requires the user to know they need this
special flag.  It would be great to put the flag in the file struct, but
syslets have no knowledge of files; syslets only know about operations.

I'm not sure what to do with usbfs2 until syslets are in-tree.  I could
implement it with in-kernel AIO, but without the troublesome cancel
functionality.  Would you accept that, or should I work under the
assumption that usbfs2 will be released when syslets are?

Sarah

-------------------------------------------------------------------------
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-devel