Samsung A5 - Touchscreen only active when screen is off with some replacement screens

[JD9999]

The device is an A500F. The device works normally as soon as it is restarted. However, as soon as the screen is turned off (with the power button), the touchscreen will only accept user input when the display is off.

I confirmed the issue on the Phosh GUI by SSHing into the device and using evtest. No SYN_REPORTs were recording when the screen was on, except on restart. I also had the same issue with Plasma Mobile, but I had issues with SSH and so I haven't ran evtest on it yet.

I am able to provide much more information if needed :)

[JD9999]

Some more information:

• I have confirmed with SSH that this issue also exists with the Plasma GUI
• On Plasma, I managed to get the screen to turn off without pressing the power button, and then I immediately pressed the power button so the screen was on, on home page not lock page, same issue exists. So it is not an issue with the lock screen interface, or the power button itself.

[wonderfulShrineMaidenOfParadise]

    reg_vdd_tsp: regulator-vdd-tsp {
        compatible = "regulator-fixed";
        regulator-name = "vdd_tsp";
        regulator-min-microvolt = <3300000>;
        regulator-max-microvolt = <3300000>;

-       gpio = <&msmgpio 73 GPIO_ACTIVE_HIGH>;
+       gpio = <&msmgpio 73 GPIO_ACTIVE_LOW>;
        enable-active-high;

        pinctrl-names = "default";
        pinctrl-0 = <&tsp_en_default>;
    };

I am not sure but changing this line could be interesting. https://github.com/msm8916-mainline/linux/blob/master/arch/arm64/boot/dts/qcom/msm8916-samsung-a2015-common.dtsi#L83

[minlexx]

On my a5lte touchscreen works fine both when screen is off or on. In Phosh :)

[JD9999]

Sorry I realised I haven't replied!

I tested your code a while ago and no difference, I'll try updating to the latest service pack and see if that fixes anything. If it doesn't I'll try installing Android (or maybe even Lineage OS) on it, see if it is maybe a hardware issue, not a software issue.

Thanks for your help :)

[JD9999]

The screen works fine on Android (Android 6.0.1, based on linux 3.10.49) so it's not a hardware issue. The 2015 A5 is not compatible with any other OS, so I can't test any other OS. While it was installing Android, it fixed a memory issue, so I'm going to try reinstalling pmOS again, and see if I have the same issue again.

[JD9999]

Still have issues. I have tried image files and pmbootstrap with Phosh on v21.12, no success. Also tried Xfce4 and weston, the screen didn't even work on the first time! In v21.12 and in edge in Plasma, worked first time but once screen turned off screen is active when display is off, and deactivates when display is on. I have tried the fix above, as well as getting rid of enable-active-high (which stops the touchscreen from even appearing in evtest).

If there are any more suggestions I will gladly try them!

[TravMurav]

Hi, is there any chance that you have an aftermarket replacement display on that device instead of original?

[JD9999]

Yes, it is definitely aftermarket. It is not installed very well either (big gap between the display and the phone). I bought it off someone else on Gumtree, the aftermarket display was already installed

[JD9999]

lk2nd says it is a ss_dsi_panel_EA8061V_AMS497EE01_HD

[stephan-gh]

Is it an AMOLED at least or does it have horrible viewing angle and backlight (lk2nd screen is more gray than black)?

[JD9999]

It definitely has horrible viewing angles and backlight

[stephan-gh]

Alright, I have some of these trashy displays and will try to reproduce the problem myself. :)
Not sure how much we can do though, trashy displays are trash. :(

The Samsung Android might accidentally work around it with some weird changes to e.g. the touchscreen code. A similar problem exists e.g. for harpia (https://github.com/msm8916-mainline/linux/issues/190) with "fake" replacement screens that we were never really able to solve. :/

[JD9999]

@stephan-gh Thanks heaps!

I'll try the same thing iyzsong tried, see if I have any improvement.

[stephan-gh]

I'll try the same thing iyzsong tried, see if I have any improvement.

The A5 uses a different touchscreen driver than motorola-harpia, it won't help. :)
I just listed this as an example that "aftermarket displays" tend to cause problems with touchscreens.

[JD9999]

You're right, it didn't help. That makes sense :)

It's a strange problem, because the issue only happens when the Plasma/Phosh screen comes up. If there's a delay between the backlight coming on, and the Plasma screen coming on, and I touch the screen then, then I can use it until I turn off the display again (and hope for another delay).

So it's definitely software related, but I am absolutely clueless what change needs to be made.

Is there a way to automate a "touch" on the screen while backlight is on but display is off? Or am I just going too far now?

[stephan-gh]

Have you tried disabling suspend like described in https://wiki.postmarketos.org/wiki/Qualcomm_Snapdragon_410/412_(MSM8916)#Suspend? It's not very well tested at the moment and provides little benefits so far.

[minlexx]

I can add unrelated info, that I have weird replacement screen (not amoled, gray bg and horrible view angles) on my second samsung-klte (Galaxy S5) - and it does not work in mainline either, same as on harpia I guess (synaptics rmi4-i2c: device was reset during probe)

Aand of course it works fine in Android.

I can also add that LCD panel itself reports slightly different panel ID than expected (using samsung's custom "read panel ID" DSI commands (DA/DB/DC))

[JD9999]

@stephan-gh I just tried disabling suspend, still had problems

@minlexx Since this is happening on lots of different chipsets and devices, maybe I should just give up on my screen working :(

Dang!

[stephan-gh]

I can't seem to reproduce this with my trash screen. I installed the latest prebuilt image: https://images.postmarketos.org/bpo/edge/samsung-a5/phosh/20220323-2153/20220323-2153-postmarketOS-edge-phosh-18-samsung-a5.img.xz

Then I pressed the power button a couple of times, waited a bit while the display was off and pressed the power button again. The lock screen showed up again and touch input works (the lock screen is a bit slow to react sometimes though).

Could you try exactly the same setup and see if it happens? If it does, I guess your fake screen is either different from mine or it's because I have SM-A500FU and not SM-A500F.

[JD9999]

@stephan-gh Tried what you did, screen does not respond after first screen lock

[stephan-gh]

I'm pretty clueless here at this point. It might just be some dumb property of the replacement screen. If you want you can try if keeping the power supplies on helps somehow:

diff --git a/arch/arm64/boot/dts/qcom/msm8916-samsung-a2015-common.dtsi b/arch/arm64/boot/dts/qcom/msm8916-samsung-a2015-common.dtsi
index 5f5880874cf3..4afb55f88112 100644
--- a/arch/arm64/boot/dts/qcom/msm8916-samsung-a2015-common.dtsi
+++ b/arch/arm64/boot/dts/qcom/msm8916-samsung-a2015-common.dtsi
@@ -85,6 +85,8 @@ reg_vdd_tsp: regulator-vdd-tsp {

        pinctrl-names = "default";
        pinctrl-0 = <&tsp_en_default>;
+
+       regulator-always-on;
    };

    i2c-muic {
@@ -353,6 +355,7 @@ l5 {
    l6 {
        regulator-min-microvolt = <1800000>;
        regulator-max-microvolt = <1800000>;
+       regulator-always-on;
    };

    l7 {
@@ -410,6 +413,7 @@ l16 {
    l17 {
        regulator-min-microvolt = <2850000>;
        regulator-max-microvolt = <2850000>;
+       regulator-always-on;
    };

    l18 {
diff --git a/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts b/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
index f29937f8dbf7..70af82a782d5 100644
--- a/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
+++ b/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
@@ -21,6 +21,8 @@ reg_vlcd_vdd3: regulator-vlcd-vdd3 {

        pinctrl-names = "default";
        pinctrl-0 = <&lcd_on_default>;
+
+       regulator-always-on;
    };

    reg_vlcd_vci: regulator-vlcd-vci {
@@ -31,6 +33,8 @@ reg_vlcd_vci: regulator-vlcd-vci {

        gpio = <&msmgpio 87 GPIO_ACTIVE_HIGH>;
        enable-active-high;
+
+       regulator-always-on;
    };

    reg_touch_key: regulator-touch-key {
@@ -44,6 +48,8 @@ reg_touch_key: regulator-touch-key {

        pinctrl-names = "default";
        pinctrl-0 = <&tkey_en_default>;
+
+       regulator-always-on;
    };
 };

[JD9999]

@stephan-gh my device works now! :)

Now time to find out which regulators needs to stay on, and which ones I can turn off :)

Thanks heaps!

[JD9999]

So adding the regulator-always-on to the regulator-vlcd-vci code fixes this issue. None of the other changes are required. Should I submit a merge request to add this in or just make the change on my device each time I update it?

[stephan-gh]

Can you try the following instead of the regulator-always-on:

diff --git a/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts b/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
index f29937f8dbf7..d9b2852de358 100644
--- a/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
+++ b/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
@@ -102,6 +102,10 @@ iris {
    };
 };

+&reg_vdd_tsp {
+   vin-supply = <&reg_vlcd_vci>;
+};
+
 &touchkey {
    vcc-supply = <&reg_touch_key>;
    vdd-supply = <&reg_touch_key>;

It's a bit dumb (does not represent the actual hardware state), but should ensure vdd3/vci are turned on together with the touchscreen supply.

[JD9999]

Can you try the following instead of the regulator-always-on:

diff --git a/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts b/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
index f29937f8dbf7..d9b2852de358 100644
--- a/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
+++ b/arch/arm64/boot/dts/qcom/msm8916-samsung-a5u-eur.dts
@@ -102,6 +102,10 @@ iris {
  };
 };

+&reg_vdd_tsp {
+ vin-supply = <&reg_vlcd_vci>;
+};
+
 &touchkey {
  vcc-supply = <&reg_touch_key>;
  vdd-supply = <&reg_touch_key>;

It's a bit dumb (does not represent the actual hardware state), but should ensure vdd3/vci are turned on together with the touchscreen supply.

Yes this also works - out of interest, do you prefer this one to the other one, and if so, why so? (I have no understanding of any of this)

[stephan-gh]

Yes this also works - out of interest, do you prefer this one to the other one, and if so, why so? (I have no understanding of any of this)

The regulators control the power supply to the components connected to them. There should be two for the display and one for the touchscreen. Normally they are independent, so if you want the touchscreen you just need the touchscreen regulator. This does not work in your case for some reason, it seems like the display regulator also supplies part of the touchscreen somehow.

This might cause the following problem: If the touchscreen is started before the display gets enabled (this depends on the timing), then one of its power supplies is still missing and it will fail to start.

With this change, the touchscreen regulator is marked as being supplied by the display regulator. This means that Linux will always turn on the display regulator before the touchscreen regulator is enabled. The advantage is that the display regulator does not stay always on, when the display and touchscreen are both disabled the display regulator still gets turned off to save some minor power.

[JD9999]

Sorry Stephen, would you like me to submit a pull request to fix this issue, or do something downstream, or not at all? I have never submitted a pull request before and don't know whether to do it or not since this seems very device specific

[stephan-gh]

I'm still undecided if we should make this change since it only seems to affect your device, and it does potentially increase power consumption a bit in some edge cases.

[JD9999]

It sounds like you are hesitant - I won't submit a request then, but if we hear of another device with this issue, then I can learn how to do a pull request.

In that case, should I close the issue?

[stephan-gh]

Let's keep it open for now I guess, it doesn't hurt and might make it easier for others to find the workaround we came up with.

[stephan-gh]

(If someone else is affected by this please comment here so we could decide to make the workaround the default for everyone)