stendro / msm8996_lge_kernel Goto Github PK

Currently the phone's battery isn't detected at all and TWRP reports a fixed 50% capacity independent of how much capacity is left on it. The issue is related to of_batterydata_get_best_profile(), lge_get_power_by_name() and some missing code. Seems like the battery isn't being registered correctly and hence the other drivers can't get its profile correctly.

Camera doesn't use focus.

When i try to change drivers in drivers/media/platform/msm/camera_v2/ from V30A source (we use V30A camera blobs in our default kernel), the kernel doesn't build, showing me this error:
TXT.txt

G6 still needs fixes for proper basic usb functionality.

H870.txt

The anx7688 usb driver also had some bypasses used to help the kernel compile.

Same commit as #17.

For one, the following dmesg line has interesting information:
[ 7.280111] anx7688-usbc 6-0028: anx7688_ctype_work: Unknown Charger, 5000mV, 0mA

The driver appears unable to ascertain what kind of charger is connected to its USB interface and then disables charge current despite reporting that charging is enabled. Seems to be related to the chip->ccx variables, looks like they aren't being set/read properly or something, but a deeper investigation is warranted.
Edit: The entire anx7688_ctype_work() function is suspicious... gotta look around it.

For now the kernel is able to be compiled for the G6 with the ES9218P codec enabled. The compiler output however points to several rough edges that still need to be smoothed out. To be specific, the multiple defined but unused variables and functions now that the problematic ones have become static vars/funcs.

Edit: Not an actual issue, just got some files that were different from the ones in the remote tree.

Not much to talk about here, it's mostly related to smbcharger being unable to get the vbus regulator for it and then failing on its probe function.

Commit 10bc66f left some LG parts commented, to be merged later.

The original driver had 5 state functions and definitions in total: probe, remove, suspend, resume and id_table.
What was done to prevent the newer kernel from complaining was just turning the resume and suspend definitions into comments, preventing the driver from using them for the time being.

The file is located here: /drivers/video/slimport/Ctype_anx7816/slimport7816.c and those changes were done in lines 1705 and 1706.

Related commit: 3d1bcdb.

While not actually critical to the kernel, or even driver for that matter, those reads are important in case a firmware update is needed even if it is unlikely we'll ever update the anx firmware on that one.

Small update: After some testing, it seems that the moment the driver gets initialized has a bearing on those i2c errors. I might experiment with that later.

Hi, I need to compile version 1.5.5 (NOUGAT) for rs988, but I do not know what branch or until commit is version 1.5.5,
I need the full code to debug the bcmdhd driver for monitor mode support.
What branch and until commit is version 1.5.5? and what linaro version?

Hello,

On my quest to have the stock android 9 ROM rooted on my F800K, I came across a version of this kernel slightly adapted to allow installation on Android 9.

Overall it works pretty darn well (so does the root), except for the fact that the camera can't focus anymore.

Sadly I don't know enough about kernels to know why. Is it due to the changes made to allow installation on Android 9 ?

I believe you are not really following this anymore, could you point me in the direction of where to look at ? I am a developer but don't dabble in C very much.

Thanks !

Edit: This only applies when using the LG code for the USB and smbcharger handling.
The only change done there is that 2 lines were transformed into comments, forcing some LG definitions to be done at all times. The USB drivers are not completely implemented as is, so in the future we need to revert this change and then either add the drivers in a functional manner, or use some upstream/CAF code as a substitute like @stendro pointed out.

The related commit: c2bd411

The import from this commit: Import LGE touch, slimbus and panel drivers (latest oreo source)
72fa7e0

could not apply the change to: drivers/clk/msm/mdss/mdss-dsi-pll-util.c
because the file is missing, and no other suitable place was found.

By analyzing the kernel's output log when booting, we can find some log entries that might point to why the camera isn't working even with updated, and presumably compatible blobs:

msm_camera_pinctrl_init:1468 Getting pinctrl handle failed
msm_camera_power_up:1642 Initialization of pinctrl failed

The functions referenced on those entries can be found here:

The code on those functions appears to be correct, which makes me assume that we're either having some problems in the dts, or the camera blobs, with the former being more likely to be the culprit according to the logs.

When using LG's sources, one approach is to bypass some errors instead of actually fixing them, which takes a lot of extra time.
As far as the changes there go, they were mostly done by transforming the problematic lines into comments, so if needed, we might have to go back to it later.

Commit: c003254

Due to the sheer amount of changes in the code, this issue has been made mostly to keep track of it, as we might have to get back to it, despite the code being very clean and concise for the most part. Things to be noted are: I created a dummy_psy power supply struct to receive the result of power_supply_register() for now, it might be wrong depending on what the code now expects from it (before it only needed a bool variable and i have no idea what to do with that returned struct yet), and there's some really important changes in the power_supply files as well.

Related commits:
244c2e5 For the power_supply changes.
ac7ba6c For the driver update.
The others are relatively small and harmless.

There were also some fixes for ANX7418 following that same strategy, and so far they're both working, although we still need to fix some NACK warnings on that 7418 driver, so until then, this issue stands.

Despite increasing the array index to prevent it from accessing a negative one (C doesn't work like Python), making it access the zero index of the array might be incorrect and cause problems later. We won't know until we test it.

Related commit: 4dc0439

With the latest usb fixes that help the kernel handle the phone's layout better, sometimes smbcharger panics it when an usb insertion event is triggered. It happens randomly, but is guaranteed to happen after 4 or 5 usb insertions.

The call to dump_stack() looks like this in dmesg:

[ 57.068747 / 10-26 15:03:55.524][1] CPU: 1 PID: 587 Comm: mdss_fb0 Tainted: G W 3.18.71 #2
[ 57.068758 / 10-26 15:03:55.524][1] Hardware name: Qualcomm Technologies, Inc. MSM 8996 v3.x + PMI8996 MTP (DT)
[ 57.068765 / 10-26 15:03:55.524][1] Call trace:
[ 57.068782 / 10-26 15:03:55.524][1] [] dump_backtrace+0x0/0x130

There are no functions visible in call stack. What should I change, to see a complete call stack?

I'm building from source this kernel, because I need driver for external wifi adapter.
The file VERSION is not updated, could you update it?
Could you set tags? So I can checkout the same version announced on forum, for example for version 2.2 which is the commit?
I tried to build the latest commit from branch mko-v2, with gcc 9.1.1 aarch64 from arter97/arm64-gcc, but I have this error:

/home/xxx/arm64-gcc/bin/aarch64-elf-ld: unrecognised emulation mode: aarch64linux
Supported emulations: aarch64elf aarch64elf32 aarch64elf32b aarch64elfb armelf armelfb

I tried with version 9.2 and I have the same issue, I tried with kdrag0n/aarch64-elf-gcc 9.1 and 9.2 and nothing I can't compile this kernel. With gcc 8.3 from arm's site I can compile without issue.
I would like to compare with hackbench, the different versions of gcc.

Similar to issue #16 in application, but this one refers to the backlight driver located at:
drivers/video/backlight/lm3697.c

Apparently used by the V20, at least the H910 model.

Related commit: 9ba2c0d.