Hi On Wed, May 6, 2009 at 11:58 AM, jasjit singh <[email protected]> wrote: > Hi > I am using work queue to do bottom half processing in my driver. The > function queue_work() doesn't enqueue the work if the work_struct is already > in the queue. I want to know how this phenomenon is synchronized. Suppose I > have enqueued a work which calls the function do_this_work(). Now the > do_this_work() has been invoked and not finished yet. So if there is an > interrupt while processing of this function, will the new work be enqueued > or the queue_work() still give an error ? > What I could think is when the kernel starts processing do_this_work(), it > resets some flag so that more work can be enqueued to the work queue. Does > it behave like this or not ? Please elaborate the scenario if u can.
Looking at the source, specifically kernel/workqueue.c on function run_workqueue()... I saw locking mechanism. I believe it's using spin_lock_irq(), thus for current CPU, the interrupt is temporarily disabled. Furthermore, I conclude that at this point, interrupt won't trigger the execution of work queue at this point. CMIIW people. regards, Mulyadi. -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to [email protected] Please read the FAQ at http://kernelnewbies.org/FAQ
