On 2016-11-07 15:48:50 [+0200], Andy Shevchenko wrote: > On Thu, 2016-11-03 at 17:29 +0100, Sebastian Andrzej Siewior wrote: > > On 2016-11-03 18:14:23 [+0200], Andy Shevchenko wrote: > > > We can't and basically don't need to call runtime PM in IRQ handler. > > > If IRQ is > > > ours, device must be powered on. Otherwise check if the device is > > > powered off > > > and return immediately. > > > > The old beagle board does PM put once it does not use the UART anymore > > / > > the UART is idle. > > And how can it generate interrupts? It has this priv->wakeirq = irq_of_parse_and_map(pdev->dev.of_node, 1); line in the driver. I *think* it is a different interrupt (same source but different PM setup and thus different number) and the driver needs to be invoked in order to ACK it. > > You need to PM get the device even if the device is not yet suspended > > because it might get suspended in the middle of your operation (maybe > > not while you serve the IRQ but we have threaded interrupts). > > I grepped code and found no driver under 8250 which is using threaded > IRQ handler. What did I miss? Documentation/kernel-parameters.txt: threadirqs [KNL] > > If the HW is off then it will be woken by an RX interrupt and I think > > it > > is a different one (I remember Tony wanted to rework that part so that > > the device/UART does not need to care about that). > > Rx interrupt from other device? > > Or OMAP UART has inbound wake source? See above. It is the "priv->wakeirq" thing. It is a wake source which is the RX pin setup differently (my memory here). Tony wanted to change PM code in regard to this as he wasn't too happy about it, too. > > This should work in general but I think that is racy with threaded > > interrupts: What stops the PM core to shutdown the HW before > > serial_port_in() ? > > Basically what I understand from this the question is: "Is > pm_runtime_suspended() synchronous with regard to > pm_runtime_get()/pm_runtime_suspend()?". I guess so. Sebastian -- To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
