Hi,
On 08/03/2012 02:46 PM, Pete Batard wrote:
Attached is my final proposal then.
Looks good, ack.
Note I did find one more issue wrt to timer handling, but that is unrelated
(and in another part of the code), so lets just push this patch and then fix
that issue with a separate patch. See my
Hi,
On 08/04/2012 10:59 AM, Hans de Goede wrote:
Hi,
On 08/03/2012 02:46 PM, Pete Batard wrote:
Attached is my final proposal then.
Looks good, ack.
Correction, you forgot to add the arm_timerfd_for_next_timeout() call we
agreed would be added to the transfer submission failure path in
Hi,
On 08/04/2012 10:59 AM, Hans de Goede wrote:
Note I did find one more issue wrt to timer handling, but that is unrelated
(and in another part of the code), so lets just push this patch and then fix
that issue with a separate patch. See my next mail on that.
Scrap that, upon typing a
On 2012.08.04 10:19, Hans de Goede wrote:
Correction, you forgot to add the arm_timerfd_for_next_timeout() call we
agreed would be added to the transfer submission failure path
Forgot about that one - Thanks for spotting it. I have now modified the
patch to include this change and pushed it to
Hi,
On 08/02/2012 01:28 AM, Pete Batard wrote:
On 2012.08.01 12:52, Hans de Goede wrote:
Hmm, I hadn't noticed the explicit disarm there, since the
handle_timeouts_locked code just cancels
transfers and does not do anything io-intensive, the run-time for
handle_timeouts_locked will be
On 2012.08.01 12:52, Hans de Goede wrote:
Hmm, I hadn't noticed the explicit disarm there, since the
handle_timeouts_locked code just cancels
transfers and does not do anything io-intensive, the run-time for
handle_timeouts_locked will be
short, so I see no value in the disarm there, and vote
On 2012.07.12 22:00, Hans de Goede wrote:
I believe that all arms / disarm should be done under the
flying transfer lock.
Same here. And my concern is that the disarm we do in
handle_timerfd_trigger() isn't.
Rather then reasoning ourselves a headache about how
it is absolutely safe in all
Hey,
Hans de Goede wrote:
doing the disarm without the flying-transfers lock held is racy, ie:
Bringing back the old behavior was a start. Yes, race also. The
attached patch on top of the previous fix avoids the race, and is
available also via
git fetch git://git.libusb.org/libusb-stuge.git