On Friday 06 October 2006 9:57 pm, Christopher "Monty" Montgomery wrote:
> On 10/7/06, David Brownell <[EMAIL PROTECTED]> wrote:
>
> > I'd made suggestions there ... probably the simplest way for things like
> > the audio driver is to return a positive status code from urb submit,
> > indicating a schedule lossage fault but not a submit error. Simple enough
> > to make sure usbcore can pass such codes through, and all existing driver
> > code should only be testing "retval < 0" for faults.
>
>
> In the current code, the URB isn't actually submitted in that case.
> We shouldn't return nonerror if the URB didn;t get submitted.
If it didn't get submitted, that would be an error. :)
Terminology: "fault" == "some non-mainstream case, including but not
limited to errors". "error" == "something that should not happen".
So for example a page fault is not usually an error, unless it's
because of a reference to an unmapped address.
> If we do submit the URB as well as returning a positive error code, we
> take away the possibility of the higher level driver being able to act
> on the XRUN before the next URB is processed. I can't actually think
> of a specific reason why that would be a problem, but it occurs to me
> that it might.
That was the other option, less straightforward for audio driver etc:
return an error (like -EL2NSYNC) unless some magic flag was set.
The difference is between audio/video/... drivers going
code = usb_submit_urb(...);
switch (code) {
case -EL2NSYNC:
/* report underrun */
/* add magic URB flag */
code = usb_submit_urb(...);
/* whoa, messy code goes here to handle errors */
break;
case 0:
/* success, all is fine */
break;
case -ESHUTDOWN:
/* handle device going away */
break;
case -Error1:
/* handle it */
break;
case -Error2:
/* handle it */
break;
...
default:
/* log unexpected error, probably give up */
break;
}
And the same code with simpler logic for the underrun case ... assuming here
it returns a POSITIVE code instead of negative:
... as above ...
case EL2NSYNC:
/* report underrun */
/* FALLTHROUGH
case 0:
/* success, all is fine */
break;
... as above ...
Of course one could maybe combine the two, and require the magic flag to
get that "positive" fault code behavior.
- Dave
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
linux-usb-devel@lists.sourceforge.net
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
