On Thu, May 13, 2004 at 02:42:29PM -0400, Alan Stern wrote:
> On Thu, 13 May 2004, Greg KH wrote:
>
> > I don't object to that, I object to a new function call. If we add
> > this, then what's the use of usb_unlink_urb() at all? Why would a
> > driver ever want to call it?
>
> You mean, what's the use of synchronous usb_unlink_urb() -- asynchronous
> is a separate thing and still necessary; it's used within usb_kill_urb().
>
> > So, why not just fix usb_unlink_urb() to do the same thing that this
> > usb_kill_urb() call does? That seems much simpler to do, and no drivers
> > will have to be changed.
>
> I thought about doing that. It certainly would simplify hcd_unlink_urb()
> considerably.
>
> Here's two reasons to keep usb_kill_urb apart from usb_unlink_urb:
>
> 1. Overloading usb_unlink_urb() to handle both synchronous and
> asynchronous unlinking was a mistake from the beginning.
I agree.
> Now we can finally have two different routines for the two
> different purposes: usb_unlink_urb() for asynch. and
> usb_kill_urb() for synchronous unlinks.
Ok, fair enough.
> 2. It's just barely possible, albeit highly unlikely, that some
> driver somewhere relies on the peculiar semantics of synchronous
> usb_unlink_urb(). I can't imagine how or why, and if a driver
> does so it's probably by mistake. But on the off chance, I
> didn't want to completely replace synch. unlink_urb.
Yeah, possibly, but I really doubt it.
> On the other hand, if you feel that usb_kill_urb() can take over the job
> of asynchronous unlinking with no problem, then we can simply add this
> code to the start of usb_unlink_urb():
>
> /* Replace synchronous unlinks with usb_kill_urb() */
> if (!(usb->transfer_flags & URB_ASYNC_UNLINK)) {
> usb_kill_urb(urb);
> return 0;
> }
Ok, I like this. This gives us a transition for drivers to make over
time, and it should not break any current users, right?
Then, in 2.7 we can just get rid of the URB_ASYNC_UNLINK flag all
together, and then backport the changes to 2.6 :)
> (Notice that the return code is meaningless. That's one of the
> differences between synchronous unlink and kill_urb.)
>
> This also would allow hcd_unlink_urb() to be greatly simplified.
> Later on, drivers can be changed to call usb_kill_urb() directly.
Sounds good. Care to respin the patch with this minor change?
Anyone else have any objections?
thanks,
greg k-h
-------------------------------------------------------
This SF.Net email is sponsored by: SourceForge.net Broadband
Sign-up now for SourceForge Broadband and get the fastest
6.0/768 connection for only $19.95/mo for the first 3 months!
http://ads.osdn.com/?ad_id=2562&alloc_id=6184&op=click
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
