On Wednesday 15 June 2011 10:23:01 Tejun Heo wrote:
On Wed, Jun 15, 2011 at 10:18:58AM +0200, Tejun Heo wrote:
No human being can feel 120usec difference and I can't see how using HIGHPRI is justified here (which is what the code is doing _accidentally_ by using singlethread_workqueue).
Ooh, one more thing, and even if you insist on using HIGHPRI (please don't), you don't need to create workqueue for each device. You can just create one for the whole driver in init and destroy it from exit. What matters is the HIGHPRI attribute of the workqueue. The number of workqueues is completely irrelevant.
Fair enough. I'll move to create_workqueue. If we later find issues with this (in a 'live' system), we can figure out a way to fix it.
Thank you for your time on this. I'll make the changes accordingly.
Regards, Péter