AM62P: Suspend failure with 12.01.00.05 Linux Kernel

Hi fd, 

Are you testing on a TI EVM or on custom hardware?

Does this error occur consistently? Can you provide the DM Firmware version?

Best Regards,

Anshu

Anshu Madwesh said:

Are you testing on a TI EVM or on custom hardware?

custom HW. The issue is happening with both Verdin AM62 and with Verdin AM62P. You have the DT for both in your linux kernel tree.

it's not obvious for me to test on the TI SK boards.

Anshu Madwesh said:

Does this error occur consistently?

Yes, the issue is systematic

Anshu Madwesh said:

Can you provide the DM Firmware version?

SYSFW ABI: 4.0 (firmware rev 0x000c '12.0.2--v12.00.02 (Clever Cat)')

I have attached the full logs in the first message, you should be able to find more data there, if you need.

Hi fd, 

Let me review the test logs with the software development team and see which patches got merged between the releases. I will get back to you by end of day Thursday.

Best Regards,

Anshu

Hi fd,

Can you apply this patch and share the logs again?

diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
index 5bbfcc36eb8ef..9c4b43e9bed82 100644
--- a/drivers/base/power/main.c
+++ b/drivers/base/power/main.c
@@ -1593,9 +1593,12 @@ static void dpm_propagate_wakeup_to_parent(struct device *dev)
 
 	spin_lock_irq(&parent->power.lock);
 
-	if (device_wakeup_path(dev) && !parent->power.ignore_children)
+	if (device_wakeup_path(dev) && !parent->power.ignore_children) {
 		parent->power.wakeup_path = true;
 
+		dev_info(parent, "DEBUG: dpm_suspend: device is parent of a wakeup source\n");
+	}
+
 	spin_unlock_irq(&parent->power.lock);
 }
 
@@ -1951,9 +1954,12 @@ static void device_suspend(struct device *dev, pm_message_t state, bool async)
  End:
 	if (!error) {
 		dev->power.is_suspended = true;
-		if (device_may_wakeup(dev))
+		if (device_may_wakeup(dev)) {
 			dev->power.wakeup_path = true;
 
+			dev_info(dev, "DEBUG: dpm_suspend: device may wakeup due to being in the wakeup_path\n");
+		}
+
 		dpm_propagate_wakeup_to_parent(dev);
 		dpm_clear_superiors_direct_complete(dev);
 	}
diff --git a/drivers/pmdomain/ti/ti_sci_pm_domains.c b/drivers/pmdomain/ti/ti_sci_pm_domains.c
index 0c21a2fdec8d9..8b118b5c349bd 100644
--- a/drivers/pmdomain/ti/ti_sci_pm_domains.c
+++ b/drivers/pmdomain/ti/ti_sci_pm_domains.c
@@ -91,6 +91,8 @@ static inline void ti_sci_pd_set_wkup_constraint(struct device *dev)
 		 * If device can wakeup using IO daisy chain wakeups,
 		 * we do not want to set a constraint.
 		 */
+		dev_info(dev, "DEBUG: device may wakeup or in wakeup path\n");
+
 		if (device_out_band_wakeup(dev)) {
 			dev_dbg(dev, "%s: has out of band wakeup, not setting constraints\n", \
 					__func__);

Best Regards,

Anshu

Thanks for the suggestion, I'll have a look in the next few days (this week I am not sure I have the time).

Were you able to test on some TI EVK / SK platform? The issue is not present there?

Hi fd,

We didn't test on a EVM now, but the development team noted they observed a similar error during their work sometime ago. They started the debug process with the same patch I shared.

Best Regards,

Anshu