On Sat, 30 Aug 2003, David Brownell wrote:

> Alan Stern wrote:
> > 
> > This looks pretty good to me.  So you're still relying on the driver's 
> > start() and resume() methods to set hcd->state to RUNNING, rather than 
> > relying on checking their return codes in hcd-pci.c and doing it for them?
> 
> For now, yes.  If you're implying that those states
> should only get set in the hcd glue ... I'm not sure
> that'd work.  There are points in those routines that
> enable HC IRQs.  Since they may arrive immediately,
> there'd be a race with between IRQ arrival and
> returning to hcd-pci.c to update the state.

Okay.  So not only should start() and resume() set hcd->state to RUNNING,
they should do so _before_ enabling any interrupt requests.  And of
course, if they fail they should reset the HC and set hcd->state to HALT.

Alan Stern



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to