Re: [Libusb-devel] usbfs: usb_submit_urb returned -121

[Ludovic R. <lud...@gm...>]

2010/6/26 Peter Stuge <pe...@st...>:
> Alan Stern wrote:
>> > > libusb returns -99 (LIBUSB_ERROR_OTHER)
>> > ..
>> > > libusb:warning [do_sync_bulk_transfer] unrecognised status code 1
>> >
>> > Alan's description is helpful. libusb does handle when transfer >=2
>> > in a multi-URB transfer submitted to Linux fails. libusb discards the
>> > previous transfers
>>
>> Do you mean it discards the as-yet-unsubmitted URBs and cancels the
>> outstanding URBs?
>
> Yep, that's correct.
>
>
>> >  and sets reap_action=SUBMIT_FAILED. As libusb is
>> > reaping the URBs, it notices that one could not be submitted, and
>> > returns an error.
>>
>> I gather that the difficulty is how to tell whether or not a submission
>> failure represents an error.  The criterion is simple: If the errno
>> value from the failed submission ioctl is EREMOTEIO then it's not an
>> error (it's merely a rejection because the transer has already
>> completed), otherwise it is.
>
> Thanks for clarifying that! I'm attaching a revised patch to be
> applied instead of the previous one.
>
> The former patch changed the reap code path when URB submission had
> failed. This patch instead changes the code path directly after
> submission, and if errno == EREMOTEIO then the reap code is
> instructed that all is well via reap_action = COMPLETED_EARLY.
>
> (libusb already handles this situation correctly when all URBs get
> submitted and one completes early.)
>
> While my guess is that the former patch would work for Ludovic this
> one is much more correct. Please test it!

Your patch solves my problem.
Thanks

I hope a new libusb-1.0.9 with my both problems fixed will be released
soon. The patches are available.

Regards,

-- 
 Dr. Ludovic Rousseau