PW2 Repair Code 0 after a few reboots

[varu]

The plan is to change how the display temperature is retrieved, having the adc_disp_gettemp_work function hard-set display_temp_c to a sane value, say 25c (which also happens to be the default sane value chosen: #define DISP_DEFAULT_TEMP 25).

This will ensure that a comparison for the battery temperature still takes place, while also disabling the constant ADC polls. Whether the display will come out unreadable remains to be seen; so far, the display temperature gets set to either 43 or 75C depending on the kernel and I've yet to see unreadable output on it.

Again, I'm open to any suggestions, minus relegating it to the spare parts bin - not giving up just yet

On kernel recompilation, I used the 5.4.3.2 sources, built the included toolchain, & got as far as the compile stage (using these instructions). I used the .config from a live 5.4.3.2 kernel, retrieved from /proc/config.gz. However, I'm now stuck at:

Spoiler:

  GEN     usr/initramfs_data.cpio
File /initrd/lib/klibc/lib/klibc-XXXXXXXXXXXXXXXXXXXXXXXXXXX.so could not be opened for reading
 line 61
File /initrd/initramfs/lib/arcotg_udc.ko could not be opened for reading
 line 69
File /initrd/initramfs/lib/g_file_storage.ko could not be opened for reading
 line 70
File /initrd/initramfs/lib/mxc_epdc_fb.ko could not be opened for reading
 line 73
File /initrd/lib/klibc/bin/ipconfig could not be opened for reading
 line 77
File /initrd/lib/klibc/bin/nfsmount could not be opened for reading
 line 78
File /initrd/lib/klibc/bin/run-init could not be opened for reading
 line 79
File /initrd/lib/klibc/bin/cat could not be opened for reading
 line 80
File /initrd/lib/klibc/bin/sleep could not be opened for reading
 line 81
File /initrd/sbin/recovery-util could not be opened for reading
 line 82
File /initrd/sbin/hotplug could not be opened for reading
 line 83
File /initrd/lib/klibc/bin/kinit could not be opened for reading
 line 84
File /initrd/lib/klibc/bin/sh.shared could not be opened for reading
 line 85
File /initrd/initramfs/bin/mkdosfs could not be opened for reading
 line 86
File /initrd/initramfs/bin/load_waveform.sh could not be opened for reading
 line 87
make[1]: *** [usr/initramfs_data.cpio] Error 255
make: *** [usr] Error 2

These seem to be the files from the statically linked initramfs. Extracting this and the files contained within from the 5.4.3.2 uImage is no problem thanks to this guide, but I'm not sure how to work it back into the kernel compile stage. The goal is to get a functional kernel compiled, after which I can make the necessary modifications to max77696-adc.c.

[varu]

The above turned out to be a deceptively easy fix: dump all the initramfs files in a folder, change initramfs.list locations to match, and voila.

Next up:

  CC      init/version.o
  LD      init/built-in.o
  LD      .tmp_vmlinux1
arch/arm/mach-mx6/built-in.o: In function `arch_idle_single_core':
usb_h1.c:(.text+0x874): undefined reference to `wdg_prep'
arch/arm/mach-mx6/built-in.o: In function `reduce_bus_freq':
usb_h1.c:(.text+0xfd0): undefined reference to `wdg_prep'
usb_h1.c:(.text+0x111c): undefined reference to `wdg_prep'
arch/arm/mach-mx6/built-in.o: In function `set_high_bus_freq':
usb_h1.c:(.text+0x13f0): undefined reference to `wdg_prep'
make: *** [.tmp_vmlinux1] Error 1

That references the watchdog, the function is imx2_wdt.c, which is being loaded:

cat .config | grep WDT
CONFIG_IMX_HAVE_PLATFORM_IMX2_WDT=y
CONFIG_IMX2_WDT=y

wdg_prep is a LAB126-specific option:

#ifdef LAB126

void wdg_prep(){
	imx2_wdt_ping();
}

#endif

.. which should also be enabled:

cat .config | grep LAB126
CONFIG_LAB126_PRINTK_BUFFER=y
CONFIG_LAB126=y

Now I'm stumped

[varu]

It worked!

Inspired by coplate's thread I decided to try compiling with 5.8.1 instead. I used the same process to retrieve the 5.8.1 kernel's config & initramfs. I then made one simple modification to drivers\hwmon\max77696-adc.c:

static void adc_disp_gettemp_work(struct work_struct *work)
{

    /*this is the only place display_temp gets updated, so no mutex needed! */
    display_temp_c = 25;

    schedule_delayed_work(&adc_disp_temp_work, msecs_to_jiffies(DISP_ADC_TEMP_THRESHOLD));
    return;
}

One successful compile & flash later and I have a successful boot. No more attempts to read the ADC, display_temp_c is always 25, the battery check passes, the display renders fine, and since I'm no longer in diags mode, the touchscreen works!

I'd still like to successfully compile a 5.4.x kernel with the 'fix' - once this is done I will revisit the thread and document the process start to finish. The build issue with the 5.4.3.2 kernel in the previous post remains open. This has now also been resolved, once the errors mentioned above are resolved the kernel compiles just fine & behaves just as the 5.8.1 does - no ADC hits, no temp check failures. That said, I'm not satisfied with the way I resolved it, so this step will be documented once I figure it out to my satisfaction. Now I'm happy with it, essentially looked at how they were doing it in the 5.8.7.0 kernel and adapted the 5.4.3.2 source, very few changes needed. Documentation coming soon.

[nitRA_M]

Hey varu, very nice work!

I studied your researches and as it looks i've exact the same problem on my PW3.
Could you mabye post your patched kernel for 5.8.1? I'm facing some serious troubles with my toolchain-installation so i'm not yet able to compile it myself...

[varu]

My apologies, it's been on my mind every now and then but I haven't gotten a chance to do a proper writeup. If you still need it, let me know and I'll write something up quick, else I'll do something a bit more thorough by the end of the year (I had this on weekly notification and the behaviour indicates a polling script that unfortunately ran the day before you posted).

I'm not sure of the differences between the PW2 and PW3, I only compiled mine for the PW2 so I have no idea if it'd work. I'm also not sure if posting a compiled kernel is allowed on the forum, if I remember correctly I had to extract various binaries & modules from the existing kernels & add them to the source for the compile process to work.

FYI, after leaving the thing on a non-smart charger for 4-5 days (the one I was using would cut off supply after the current draw sank below a certain threshold), the ADC stopped misbehaving! No more warnings with the diag or any other stock kernel, as well as successful boots every time (I tested a number of stock firmwares out of sheer disbelief.. worked fine every time) & an excellent battery life. Although I don't quite understand why this is the case, the evidence speaks for itself.. I should've listened to Cinisajoy and just left the thing plugged in

[knc1]

The kernel is GPLv2 - post away.
In fact, according to the license, you MUST make your changes available to the public.
(that is what the 'P' stands for in the name).

[eddie.t.h]

I know this is an old thread but maybe someone will be useful to what I discovered.

I had exactly the same problem as @Varu, i.e. battery error 22 after restart, and when the reader was sometimes able to start, the display worked very slowly. In the logs I found exactly the same errors regarding the lack of display temperature reading. In my case, the system set the default reading at -10 ° C (14 ° F), which I guess was the reason for the very slow operation of the display.
I found on the internet that the ED060XD4 display has a built-in 10k thermistor on pins 30-31.
When I looked at my display under the microscope, the cause of the problem appeared immediately - the temperature sensor was torn off. I don't know how it could happen because my reader was never taken apart.
After repairing the connections, everything works as it should, no errors.

[knc1]

That is a thermal stress crack created during the board's reflow soldering process.
I.E: Not the proper temperature curve at that point in the board.
Which also makes sense, that is near the edge of the board and your motherboard might well have been on the edge of that group of motherboards.

Translation: Shit Happens

[seventhoctave]

amazing! I found a kindle pw2 in the company e-waste with this error and thought it would be a simple battery replacement to fix. I've been "hacking" on it for days and was about to give up. Saw this post and sure enough the thermistor is long gone. I didn't want to order the part and wait a week to try it out so I just soldered a regular 10k resistor (the thermistor would be 10k at about room temp) and booted right up! Unfortunately it turns out there is other damage and the touch screen no longer works. But it's super cool to see it running again!