On Nedjelja 09 Prosinac 2001 00:51, Martin Diehl wrote:
> > With usb-uhci i receive -ECONNABORTED while submitting urbs. (currently
> > the
>
> During submitting? Strange. For me, the usb_submit_usb succeeds (like it
> always did without the patch). And the completion handler gets called with
> urb->status=0. However, with further testing I'm able now to reproduce
> the status gets changed to ECONNABORTED _after_ the completion handler
> returned (but already before doing anything else like an additional
> unlink). Even width additional 10ms sleep the status never changes from
> ECONNABORTED - and I tend to believe the URB never gets completely removed
> from the schedule.
Yes I believe that could be the problem why all aditional urbs fail (only the
first goes through).
Maybe I should call usb_unlink_urb in the completion handler?? Maybe it would
cause it to be ok? I use a one-shot interrupt urb, but maybe the error of the
usb-uhci is that he doesn't correctly handle the one-shot urbs? So I must do
it manually?
> PS:
> > if ( usb_submit_urb( outurb ) )
> > {
> > err( DRIVER_NAME ": Failed(%d) submit urb!\n", outurb->status );
> > spin_unlock ( &global_usb_nepp->io_lock );
>
> so you have this spinlock still acquired in the submit-succeed case
>
> > return -EIO;
> > }
> > interruptible_sleep_on_timeout ( &global_usb_nepp->write_wait, 1500 *
> > HZ/1000 );
>
> ... but go sleeping here?
> And sleep_on and friends are really deprecated due to races.
;-) I noticed the problems on a SMP, I have rewritten the driver three times
in the meantime, now it's ok :)
(it's very simple only some 500 lines of code)
With regards,
Mark
--
Mark Burazin
[EMAIL PROTECTED]
---<>---<>---<>---<>---<>---<>---<>---<>---<>
Lemna d.o.o.
http://www.lemna.biz - [EMAIL PROTECTED]
<>---<>---<>---<>---<>---<>---<>---<>---<>---
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
