[lkml]   [2017]   [Aug]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [RFC PATCH v2 1/3] PCI: rockchip: Add support for pcie wake irq
Hi Brian,

On 08/19/2017 01:01 AM, Brian Norris wrote:
> Did you test that this works out correctly as a level-triggered
> interrupt? IIUC, the dummy handler won't mask the interrupt, so it might
> keep firing. See:
> static irqreturn_t handle_threaded_wake_irq(int irq, void *_wirq)
> {
> struct wake_irq *wirq = _wirq;
> int res;
> /* Maybe abort suspend? */
> if (irqd_is_wakeup_set(irq_get_irq_data(irq))) {
> pm_wakeup_event(wirq->dev, 0);
> return IRQ_HANDLED; <--- We can return here, with the trigger still asserted
> }
> ...
> This could cause some kind of an IRQ storm, including a lockup or
> significant slowdown, I think.

hmmm, right, but as i replied at cros partner issue, this irq handle
might not be called actually...

in my test on cros 4.4 kernel, it would break by irq_may_run(returning
static bool irq_may_run(struct irq_desc *desc)
* If the interrupt is an armed wakeup source, mark it pending
* and suspended, disable it and notify the pm core about the
* event.
if (irq_pm_check_wakeup(desc))
return false;

bool irq_pm_check_wakeup(struct irq_desc *desc)
if (irqd_is_wakeup_armed(&desc->irq_data)) {
irqd_clear(&desc->irq_data, IRQD_WAKEUP_ARMED);
irq_disable(desc); <--- disabled here
return true;

and for irqd_is_wakeup_armed:

static bool suspend_device_irq(struct irq_desc *desc)
if (irqd_is_wakeup_set(&desc->irq_data)) {
irqd_set(&desc->irq_data, IRQD_WAKEUP_ARMED); <-- set
irqd_is_wakeup_armed here

void dpm_noirq_begin(void)

so unless we get an irq between device_wakeup_arm_wake_irqs and
suspend_device_irq, the irq_pm_check_wakeup would not let us get to

> BTW, in another context, Tony suggested we might need to fix up the IRQ flags
> like this:
> int dev_pm_set_dedicated_wake_irq(struct device *dev, int irq)
> {
> ...
> err = request_threaded_irq(irq, NULL, handle_threaded_wake_irq,
> - IRQF_ONESHOT, dev_name(dev), wirq);
> + IRQF_ONESHOT | irq_get_trigger_type(irq), dev_name(dev), wirq);
> But IIUC, that's not actually necessary, because __setup_irq()
> automatically configures the trigger type if the driver didn't request
> one explicitly.

actually this would not work...irq_get_trigger_type would return zero
due to a bug which we have a patch for it already:

9908207 New [tip:irq/urgent] genirq: Restore trigger settings
in irq_modify_status()

BTW, using dev_name for the name of this wake irq seems not very
convenient...maybe add a ":wake" suffix?
> Brian

 \ /
  Last update: 2017-08-18 19:48    [W:0.120 / U:46.904 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site