Hi, Rafael > From: Rafael J. Wysocki [mailto:rjw@rjwysocki.net] > Sent: Thursday, November 13, 2014 10:59 AM > > On Thursday, November 13, 2014 02:31:08 AM Zheng, Lv wrote: > > Hi, Rafael > > > > > From: Rafael J. Wysocki [mailto:rjw@rjwysocki.net] > > > Sent: Wednesday, November 12, 2014 9:17 AM > > [cut] > > > > > + > > > > +static int ec_create_event_poller(struct acpi_ec *ec) > > > > +{ > > > > + struct task_struct *t; > > > > + > > > > + t = kthread_run(acpi_ec_event_poller, ec, "ec/gpe-%lu", ec->gpe); > > > > > > Does it have to be a kernel thread? > > > > > > What about using a workqueue instead? > > > > Actually I just want to use threaded IRQ here in response to Andi Kleen's comment. > > If acpi_irq is registered as threaded IRQ, then acpi_ec_event_poller() will be the > > callback from it. > > How so? > > > Since ACPICA is not ready for threaded IRQ currently, we cannot proceed at this point. > > So I copied the threaded IRQ code from kernel/irq/manage.c here to prepare threaded IRQ logics. > > Oh dear, no. > > This isn't the way forward here. > > > Using a separate work queue, we didn't decrease the kernel thread count. > > Why does that matter at all? > > > And the code written for the work item cannot be derived when things are > > switched to the threaded IRQ. > > So I used kthread here. > > Please use a workqueue instead. If/when we need to switch over to threaded > IRQs, we'll do the work then. For now, let's not complicate things more > than necessary. It seems we need the thread because we will move polling code from ec_poll() to acpi_ec_event_poller(). This will happen right after these cleanups. That's the threaded IRQ logic - IRQ is polled in the thread. We cannot achieve this using work queue. Best regards -Lv {.n++%ݶw{.n+{G{ayʇڙ,jfhz_(階ݢj"mG?&~iOzv^m ?I