Hi Barnabás,

Thank you for reviewing theis patch! I've Cced a new version to you.

On Thu, Oct 15, 2020 at 10:33:50AM +0000, Barnabás Pőcze wrote:

Hi,

I believe this patch causes I2C HID devices not to work with IRQs after resuming
from suspend.

quoted

[...]
 #ifdef CONFIG_PM_SLEEP

@@ -1183,7 +1292,8 @@ static int i2c_hid_suspend(struct device *dev)
 	/* Save some power */
 	i2c_hid_set_power(client, I2C_HID_PWR_SLEEP);

-	disable_irq(client->irq);
+	if (polling_mode == I2C_POLLING_DISABLED)
+		disable_irq(client->irq);

The IRQ is disabled when suspending if polling is *off*.

quoted

 	if (device_may_wakeup(&client->dev)) {
 		wake_status = enable_irq_wake(client->irq);

@@ -1216,7 +1326,7 @@ static int i2c_hid_resume(struct device *dev)

 		if (ihid->pdata.post_power_delay_ms)
 			msleep(ihid->pdata.post_power_delay_ms);
-	} else if (ihid->irq_wake_enabled) {
+	} else if (ihid->irq_wake_enabled && polling_mode != I2C_POLLING_DISABLED) {

As a side note, I'm not sure if the 'polling_mode != I2C_POLLING_DISABLED' part
is necessary (or that it's necessary *here*). It causes 'i2c_hid_resume' and
'i2c_hid_suspend' to be "asymmetric" which - I believe - may cause problems.

quoted

 		wake_status = disable_irq_wake(client->irq);
 		if (!wake_status)
 			ihid->irq_wake_enabled = false;

@@ -1225,7 +1335,8 @@ static int i2c_hid_resume(struct device *dev)
 				wake_status);
 	}

-	enable_irq(client->irq);
+	if (polling_mode != I2C_POLLING_DISABLED)
+		enable_irq(client->irq);

The IRQ is enabled when resuming if polling is *on*. It should be enabled if polling is *off*
in my opinion.

quoted

[...]

Regards,
Barnabás Pőcze

--
Best regards,
Coiby