| Message ID | 1624543720-4170-1-git-send-email-loic.poulain@linaro.org |
|---|---|
| State | New |
| Headers | show |
| Series | PM: wakeup: Wakeup accounting for interrupts | expand |
diff --git a/drivers/base/power/wakeup.c b/drivers/base/power/wakeup.c index f0b37c1..922f701 100644 --- a/drivers/base/power/wakeup.c +++ b/drivers/base/power/wakeup.c @@ -952,8 +952,19 @@ void pm_wakeup_clear(bool reset) void pm_system_irq_wakeup(unsigned int irq_number) { if (pm_wakeup_irq == 0) { + struct wakeup_source *ws; + int srcuidx; + pm_wakeup_irq = irq_number; pm_system_wakeup(); + + /* wakeup accounting */ + srcuidx = srcu_read_lock(&wakeup_srcu); + list_for_each_entry_rcu(ws, &wakeup_sources, entry) { + if (ws->wakeirq && ws->wakeirq->irq == irq_number) + pm_wakeup_ws_event(ws, 0, false); + } + srcu_read_unlock(&wakeup_srcu, srcuidx); } }
Most of the time, system wakeup is caused by a wakeup-enabled interrupt, but the accounting is not done for the related wakeup source, causing wrong values reported by device's wakeup attributes and debugfs stats (debug/wakeup_sources). This change reports a wakeup event for any wakeup-sources the irq is attached with. Note: This only works for drivers explicitly attaching the irq to a given device (e.g. with dev_pm_set_wake_irq). Note2: Some drivers call pm_wakeup_event() in their irq handler, but not all, moreover, an interrupt can be disabled while being flagged as wakeup source, and so accounting must be performed. This solution ensures that accounting will always be done for the interrupt waking up the host. Signed-off-by: Loic Poulain <loic.poulain@linaro.org> --- drivers/base/power/wakeup.c | 11 +++++++++++ 1 file changed, 11 insertions(+) -- 2.7.4