linux-postmarketos-omap: duplicate drivers

By Andreas Kemnade on 2024-05-19T15:52:30

thanks for the notification. @MightyM17 has a 6030/32 device and will handle this

By Sicelo on 2024-05-19T16:13:18

e.g.

        twl: twl@48 {
                reg = <0x48>;
                /* IRQ# = 7 */
                interrupts = ; /* IRQ_SYS_1N casc
aded to gic */
                system-power-controller;
        };

in omap4-panda-common.dtsi

By Andreas Kemnade on 2024-05-20T18:03:06

Yep, exactly how the previous patch worked as well. I'll compare the two (mainline and older) patches tomorrow and also give it a test with just mainline to see if its working, one a quick skimming I do see the values for regs to be different, plus they've used a downstream kernel as a base for the values so will need to check before dropping our patch

By Mighty on 2024-05-20T18:03:06

besides dropping the patch you need to replace the twl6030-power subnode by the system-power-controller property in the espresso devicetrees.

By Andreas Kemnade on 2024-05-20T20:05:47

#define TWL6030_PHOENIX_DEV_ON                  0x06

is relative to

#define TWL6030_BASEADD_PM_MASTER       0x001F

so that is equivant to using

#define TWL6030_PHOENIX_DEV_ON         0x25

on top of TWL6030_MODULE_ID0 as done in the patch.

By Andreas Kemnade on 2024-05-20T20:15:55

Ah alright, I'll test it without the patch, and should I just add the prop inside twl6032.dtsi since its the common dtsi imported?

By Mighty on 2024-05-21T13:47:28

hmm, it might not necessarily apply to all devices using it. At least with the power off property for the 4030 this is not done that way. It is in the individual dts files

By Andreas Kemnade on 2024-05-21T14:26:30

As far as i know the espresso is only device having twl6032 so having the prop either in twl6032.dtsi or espresso-common.dtsi shouldnt make much of a difference, I'll follow how you've done it and add it to espresso

By Mighty on 2024-05-22T07:05:19

Seems to break everything for me, lost my wlan and usb support, neither does ui want to start now, I'll have to create a new img with fbkeyboard for further testing to get logs.

Apparently pmOS gets detected as 1-2 18d1:d001 Android ADB Recovery Interface now lol.

Here's my changes

--- a/device/community/linux-postmarketos-omap/0002-arm-dts-Add-common-dtsi-for-espresso.patch
+++ b/device/community/linux-postmarketos-omap/0002-arm-dts-Add-common-dtsi-for-espresso.patch
@@ -512,7 +512,7 @@ index 000000000..b44af61d4
 +
 +      twl: twl@48 {
 +              reg = <0x48>;
-+
++              system-power-controller;
 +              pinctrl-names = "default";
 +              pinctrl-0 = <
 +                      &twl6030_pins
@@ -523,10 +523,10 @@ index 000000000..b44af61d4
 +              interrupts = <GIC_SPI 7 IRQ_TYPE_LEVEL_HIGH>; /* IRQ_SYS_1N cascaded to gic */
 +              interrupt-parent = <&gic>;
 +
-+              twl_power: power {
-+                      compatible = "ti,twl6030-power";
-+                      ti,system-power-controller;
-+              };

and removed 0005-Add-TWL6030-power-driver-with-minimal-support-for-power-off.patch

By Mighty on 2024-05-22T07:36:14

hmm, that is strange, it just makes no sense that enabling power off has such much consequences. Are you sure you are booting what you expect to boot (i.e. modules are in place)?

There is some issue with probe order here regarding USB on 603[02]. I worked around by compiling the musb driver as builtin (CONFIG_USB_MUSB_OMAP2PLUS=y) and the phy as module (CONFIG_TWL6030_USB=m). Maybe you unveiled some issues like that.

BTW: embt2ws also has the TWL6032

By Andreas Kemnade on 2024-05-22T10:19:10

Oh didnt know about the embt2ws, I'll upstream the 6032 dtsi then :D

Shouldnt be a modules issue since i installed the linux-pmos-omap package but again without logs its hard to pinpoint the issue. I'll rebuild with those changes in defconfig that you mentioned. Again it shouldnt be an issue with just changing the power off driver, could be something else is at play (some race condition or kernel panic).

I'll try ttyescape and get back (no usb/uart/wlan :( ).

By Mighty on 2024-05-22T10:26:39

Sorry for the lack of updates, wont be able to test it till next Monday unfortunately

By Mighty on 2024-05-27T11:21:14

@akemnade tried out a build now, the fix for USB doesnt seem to work fine, i always get device id not found on my pc, and with your patches for power controller, poweroff no longer works, device was at 75% before power off and after 15-20 minutes (after running sudo poweroff assuming that the device has shut down) it was hot to the touch and dropped 11% battery doing noting :P (went to 64%). I'd still want the previous patch then to atleast make the device shut down properly

By Mighty on 2024-06-07T10:48:08

that is all strange, best would be to find out why the in-tree solution does not work

By Andreas Kemnade on 2024-06-07T12:26:49

Does it work for you on the embt2ws? Coz the older patch for sure worked for me (and other twl6030 devices. https://lore.kernel.org/lkml/20220820071659.1215-1-bavishimithil@gmail.com/T/#r4307cce5c81ec271153ef7537774da35e633fcfb

By Mighty on 2024-06-07T12:35:11

before submitting that patch I have tested it with twl6032 and twl6030. embt2ws was not getting warm, no loss of charge noticed. Do we have other stuff registering as power off handler?

By Andreas Kemnade on 2024-06-08T11:51:12

Seems like the position matters, putting system-power-controller; below the irqs made it work!

By Mighty on 2024-06-08T12:54:57

mentioned in merge request !5127 (merged)

By Andreas Kemnade on 2024-05-19T19:24:41

assigned to @MightyM17

By Sicelo on 2024-05-22T06:47:07

added device-category::community device-samsung-espresso7 labels

mentioned in issue #2874 (closed)

By Mighty on 2024-06-07T10:59:50

I am playing around with a 6.10-rc1 kernel to upstream things for embt, there I have reliable power off via echo -n o >/proc/sysrq-trigger, but not via halt command. even though I see reboot: System halted

By Andreas Kemnade on 2024-06-07T13:37:01

retested with debug-shell from original linux-postmarketos-omap initfs

 echo -n o >/proc/sysrq-trigger

does work there too.

By Andreas Kemnade on 2024-06-07T15:34:10

ok, silly, need poweroff not halt. Then it works for embt2ws.

By Andreas Kemnade on 2024-06-08T09:59:37

Oh, I'll try with latest upstream and build again, using same config from linux-postmarketos-omap and patches 0001, 0002, 0003, 0004 thats it. Then just modifying my twl node as to use system-power-controller; like shown above.

By Mighty on 2024-06-08T09:59:37

@akemnade the clock patch also seems to be broken, wifi does not work on cold boot for me

By Mighty on 2024-06-08T11:40:38

hmm, does /sys/bus/platform/drivers/twl-clk exist? And it includes a symlink named twl6032-clk?

Does /sys/kernel/debug/clk/clk_summary say something about that clock? I have here:

                                 enable  prepare  protect                                duty  hardware                            connection
   clock                          count    count    count        rate   accuracy phase  cycle    enable   consumer                         id
---------------------------------------------------------------------------------------------------------------------------------------------
 clk32kaudio                         2       2        0        32768       0          0     50000      Y   serial0-0                       ext_clock                
                                                                                                           wl12xx-pwrseq                   ext_clock

By Andreas Kemnade on 2024-06-08T12:36:20

Nope, no such clk is present in the clk_summary.

/sys/bus/platform/drivers/twl-clk just contains bind unbind and uevent

Any logs I can check in dmesg?

Here's my changes

&mmc5 { 
    status = "okay";
    #address-cells = <1>;
    #size-cells = <0>;

    non-removable;
    bus-width = <4>;
    vmmc-supply = <&reg_espresso_wlan>;
    mmc-pwrseq = <&wlan_pwrseq>;

    pinctrl-names = "default";
    pinctrl-0 = <&mmc5_pins>;

    brcmf: wifi@1 {
            compatible = "brcm,bcm4330-fmac", "brcm,bcm4329-fmac";
            reg = <1>;

            interrupt-parent = <&gpio3>;
            interrupts = <17 IRQ_TYPE_LEVEL_HIGH>;
            interrupt-names = "host-wake";
            pinctrl-names = "default";
            pinctrl-0 = <&wlan_host_wake>;
    };
};

    twl: twl@48 {
            reg = <0x48>;
            #clock-cells = <1>;

    wlan_pwrseq: wlan-pwrseq {
            compatible = "mmc-pwrseq-simple";
            clocks = <&twl 0>; // 0 is ckl32kg, 1 is clk32kaudio
            clock-names = "ext_clock";
    };

By Mighty on 2024-06-08T13:00:27

Checking the logs I do get,

`platfrom wlan-pwrseq: deferred probe pending: pwrseq_simple: external clock not ready`

Seems to be a race condition between pwrseq and clk-twl, I see you've built clk-twl as a module for your device, any such reason?

By Mighty on 2024-06-08T13:03:21

Fixes with building CLK_TWL as module but still no clk32kg in summary neither does wlan work when cold booted (mmc4 doesnt even get detected)

By Mighty on 2024-06-08T13:20:21

so the driver is there, but no device. You have #clock-cells = <1>;?

Wifi works for me with the embt2ws just out of the box, no fiddling with anything.

By Andreas Kemnade on 2024-06-08T13:33:42

ok, you have that clock-cells, is there /sys/bus/platform/devices/twl6032-clk ?

By Andreas Kemnade on 2024-06-08T13:59:03

Nope /sys/bus/platform/devices/twl6032-clk doesnt exist

By Mighty on 2024-06-08T14:14:22

Neither does twl-clk show up in lsmod

By Mighty on 2024-06-08T14:17:50

Your driver is based of the clk-twl from my repo (https://github.com/MightyM17/linux_pvr/blob/testing/drivers/clk/ti/clk-twl.c) right? The difference being that its no more required to have a subnode like mine did, so something there might be causing the issue?

By Mighty on 2024-06-08T14:26:30

Well, I think you took an older version of mine: https://lore.kernel.org/linux-clk/20230819134147.456060-3-andreas@kemnade.info/ But whatever... Add some debug output in twl-core.c probe, esp. around

        if (id->driver_data == (TWL6030_CLASS | TWL6032_SUBCLASS)) {
                status = devm_mfd_add_devices(&client->dev,
                                              PLATFORM_DEVID_NONE,
                                              twl6032_cells,
                                              ARRAY_SIZE(twl6032_cells),
                                              NULL, 0, NULL);
                if (status < 0)
                        goto free;
        }

to see what is happening. The clk device should be created there.

By Andreas Kemnade on 2024-06-08T14:55:19

Nah I checked the latest (that being v4 of the patchset), what exactly are we trying to get from the twl-core.c debug?

By Mighty on 2024-06-08T15:52:35

To see if that device registration is reached and successful, that code snipped is responsible for creating /sys/bus/platform/device/twl6032-clk

By Andreas Kemnade on 2024-06-08T15:58:25

No logs

if (id->driver_data == (TWL6030_CLASS | TWL6032_SUBCLASS)) {
    dev_info(&client->dev, "Driver data matches TWL6030_CLASS | TWL6032_SUBCLASS\n"); 
    status = devm_mfd_add_devices(&client->dev, PLATFORM_DEVID_NONE, twl6032_cells,        ARRAY_SIZE(twl6032_cells), NULL, 0, NULL);
    printk("FINDME: %d status", status); 
    if (status < 0) 
        goto free; 
}

By Mighty on 2024-06-08T16:35:11

just check for the obvious: that should only work on the espresso10 with the twl6032, not the 7 with the 6030 for now.

By Andreas Kemnade on 2024-06-08T18:03:53

Uhh? Both have twl6032 and use the omap4-samsung-espresso-common.dtsi which has 6032 node

By Mighty on 2024-06-08T19:23:15

Going back to baby steps, I just need to have CLK_TWL built (built-in/module) right? And the #clock-cells = <1> in my twl node, and finally clocks = <&twl 0>; ?

It doesnt work with sound node (clk32kaudio - wm8994 driver throws 517 when finding for this clk) and wlan_pwrseq (clk32kg - no logs)

By Mighty on 2024-06-08T19:27:38

hmm,no, looking at fdtdump omap4-samsung-espresso7.dtb I found:

   twl@48 {
                            reg = <0x00000048>;
                            pinctrl-names = "default";
                            pinctrl-0 = <0x00000079 0x0000007a>;
                            interrupts = <0x00000000 0x00000007 0x00000004>;
                            interrupt-parent = <0x00000004>;
                            compatible = "ti,twl6030";

So the compatible="ti,twl6032" might get overwritten?

That would be an explanation for a lot of trouble.

By Andreas Kemnade on 2024-06-08T19:38:53

Oh, there's a hanging include "twl6030.dtsi" thats causing us trouble, I'll port the usb regs and comparator to twl6032.dtsi and it should work fine.

By Mighty on 2024-06-08T19:40:17

                                 enable  prepare  protect                                duty  hardware                            conne   clock                          count    count    count        rate   accuracy phase  cycle    enable   consumer
---------------------------------------------------------------------------------------------------------------------------------------- clk32kaudio                         0       0        0        32768       0          0     50000      Y   0-001a
                                                                                                           deviceless
 clk32kg                             1       1        0        32768       0          0     50000      Y   wlan-pwrseq
                                                                                                           deviceless

I'll push the changes to mr now, we can close these issues i suppose then

By Mighty on 2024-06-09T06:12:17

mentioned in merge request !5216 (merged)

By Mighty on 2024-06-08T12:43:27

mentioned in commit c9a1b901

By Mighty on 2024-06-10T17:59:22

closed

mentioned in issue #1224

linux-postmarketos-omap: duplicate drivers

Child items 0

Activity